First Ashes of the Singularity patch to improve Ryzen support

>>59648913

pcper is pretty biased against AMD, is that why they left out the fact that it now beats the 6900k at high preset?

is this the kuhrazy cpu benchmark?

>>59648975

just this:

https://www.pcper.com/reviews/Processors/Ashes-Singularity-Gets-Ryzen-Performance-Update

>>59648913
>>59648927
Probably.
It must have beat the 6900k at "high" so they left it out. Also odd that they didn't test the 6900k at 3200mhz. It surely would have done better on "extreme" there, but they didn't want to show the greater gains Ryzen gets with faster memory?

Pcper are known to be dumb shills.
And outside of their actual technical guys on staff, lazy, seeing as how they ommited showing the 7700k getting stomped.

31% performance increase. Pretty nice, when people were saying 30% performance increases with optimizes in games would be impossible.

Fuck pcper.

>>59648913
Now we really see why the 1400, 1500, 1600 were delayed.

The next batch of Ryzen reviews should be really interesting unless Intel has a magic trick to pull out of their ass.

All the gains are from the 3200MHz RAM which only 5% of mobos/chips can achieve btw.

Notice they are not running intel on the same RAM.

>>59650316
You're an idiot.

>>59650330

Go read about infinity fabric, dipshit.

>>59650098
exactly this
they completely omitted the 6900k

>>59650316

Intel gains pretty much nothing from RAM speeds you idiot.

The only reason Ryzen does is because of the CCX design.

>>59650316
Retard shill detected

>>59650352
>All the gains are from the 3200Mhz RAM
>picture clearly shows both 2400 and 3200 were re-tested
I stand corrected.
You're a cock sucking retard.

>>59650312
They don't

How does 7700K compete against 6900K and 1800X in that game?

>>59650369

I'm talking about the massive gains between RAM and pointing out most people cannot run at 3200 on Ryzen...

Are you retarded?

Now run 3200 on Intel and re-test, you won't see a massive increase but you will still see one. Now go to 4600 and you will see even more improvement.

>>59650394
combination of game devs only coding for a certain amount of cores and usage etc, and then intel's mediocre performance gains for their price

File: rjwocm[1].png (67KB, 700x739px) Image search: [Google]

67KB, 700x739px

>>59650394

Here's a launch bench for you to compare to. Should beat it handily now.

The new patch updated thread pooling to avoid CCX bottlenecks.

AMD paid $1.5 million for this patch btw.

>>59650403
you are a fucking dumbass, just kill yourself already

>>59650403

All boards support 3200Mhz. Dumb people will buy RAM without checking if it will work. 2 minutes of reading will get you 3200Mhz with 100% certainty.

On that note, every BIOS patch has made compatibility better.

Intel would gain 1~2% at most off RAM speed. Only exception is Fallout 4.

>>59650403
1800x = $500
6900k = $1000

1800x = 91.44% of 6900k performance in AotS
> 73.8 / 80.7 = .9144
1800x = 50% the cost of 6900k

Keep yapping like a slack jawed nigger.

>>59650438

Considering it's only about 3 weeks of work no they didn't.

>>59650430
Thanks. So that game clearly needs more than 4 core.

>>59650316
Pcper's benchmarks show the gains are 18-31% for a given RAM speed depending on configuration.

So... no.

The 1700 was performing worse than the 7700k in ashes before this update.

>>59650438
>AMD paid $1.5 million for this patch btw.

how do you know?

>>59650312
Intel has no magic tricks
They might have if they had competent people in charge, but instead they've got some retard who's been throwing millions of dollars away at funding SJW activists instead of Intel engineers and technicians.

For the past 5 years Intel's basically invested fuck-all into R&D because they didn't see AMD as a credible threat. Beancounters just don't see any reason to 'waste money' on research when you've got a de-facto monopoly. No problem wasting money on fucking identity politics activists tho.

So for the next 2-4 years Intel has literally no response to AMD. They prepared for continued monopoly with marginal revisions/iterations on their current chips rather than architecture redesigns or general computer hardware advancement. This is why you don't let beancounters and marketers run your company.

>>59650438
you really think AMD spent 1.5 million on a patch like this

good one

>>59650495

To be honest, I don't think ANYONE thought AMD could do it this well. This is the biggest shake up in PC hardware in a decade.

>>59650495
Not 2-4 years, at most it'll be 2
Intel's been pretty spooked by the hype from ryzen for some time, and the reveal scared the hell out of them
I'd expect major price cuts and bigger performance gains

>>59650521
Eh, Intel was supposed to enter WORRY mode the moment AMD started to hire old (and new) talent back.

>>59648913

Woah it took AMDs most loyal puppet company this long to patch Ryzen bias into its game?

That's worrying.

>>59650430
Thanks. Nice.

So in all likelihood it'd probably be around 95 fps in that test at 720p if they reran it again.

>>59650471
91.44% at DDR4 2400.
6900k improves with faster memory too, obviously, but Ryzen even more.

I bet the performance is more like 95% if both had DDR4 3200 but the shills at pcper didn't want to test that (or omitted the results).

>>59650438
>>59650492
Well 400 engineering hours according to the developer (which seems high. That seems like bullshit marked up hours to bill higher. I'd guess it was at least 50k, but far from 1.5 million.

>>59650495
>>59650524
Coffeelake is bringing 6 core on the mainstream socket. That's actually going to be tough competition for the 1600/X considering no CCX issue and games are already optimized for it pretty much.

I wouldn't be surprised if the "8600k" is 4c/8t for $250 as well.
Intel can't compete with this 4 thread garbage being sold at $160-$250.

>>59650495
>tho.
Opinion discarded.

>>59650521
uhh jim keller?
>>59650529
this
AMD hiring Jim Keller half a decade ago should have made intel drop all their money into R&D instantly.
the white man is so based he can destroy a company by himself

File: 1488693519789.png (84KB, 653x726px) Image search: [Google]

84KB, 653x726px

>>59650547
H-haha, yeah three weeks of optimizations only netting a 30% increase in performance... s-s-stupid AMD

>>59650438

AMD didn't pay anything, Oxide has been in bed with them for years, all their shit from their games to website have AMD logos plastered in them. It's like how Ubisoft has Nvidia shit all over their site and games.

Only difference being it's perfectly alright when AMD biased games are benchmarked.

>>59650492
>Ashes of the Singularity

AMD is directly supporting certain engines to prop up DX12 performance. They give them a ton of support and engineers.

AMD has called out a few devs for doing it right, because of their oversight obviously.

1.5 million is what I would assume the resources they need it to accomplish it, by giving them a few high end engineers for a few weeks, support, probably extra funds etc, might be even more.

"The statement begins with Taylor reiterating the momentum of AMD to support developers both from a GPU and a CPU technology angle. Getting hardware in the hands of programmers is the first and most important step to find and fixing any problem areas that Ryzen might have, so this is a great move to see taking place. Both Oxide Games and Creative Assembly, developers of Ashes of the Singularity and Total War respectively, have publicly stated their intent to demonstrate improved threading and performance on Ryzen platforms very soon."

https://www.pcper.com/news/Processors/AMD-responds-1080p-gaming-tests-Ryzen

>>59650563
And Clark, and Raja, and Paparmaster and a whole lot of other people. But we'll, that's what you get for letting bean counter run hardware company.

>>59650569

Just noting their tardiness, shill.

Oxide had their 980ti benchmark gimping patch out for AotS the day before the RX 480 launched, figured they'd have preemptive bench padding out for Ryzen before its launch.

File: firefox_2017-02-27_19-44-32.png (32KB, 973x347px) Image search: [Google]

32KB, 973x347px

>>59650521
>>59650524
>>59650563
>To be honest, I don't think ANYONE thought AMD could do it this well.
>ANYONE

<mfw I heard Jim Keller was being hired back

>>59650582
1.5 for the entire program sounds right, but not just for Oxide.

>>59650582
>by giving them a few high end engineers for a few weeks
>support
>probably extra funds etc, (??????)
>1.5 million

lmao

>>59650524
You have to keep in mind they didn't have any ongoing projects in CPU hardware/architecture design until maybe 4-8 months ago when they finally got wind of what AMD's up to. They've got iterations for kaby/coffee/whatever lake, they can keep doing that - But they don't have an architecture revision in the pipe the way AMD did with Ryzen.

>>59650553
>Coffeelake is bringing 6 core on the mainstream socket. That's actually going to be tough competition for the 1600/X considering no CCX issue and games are already optimized for it pretty much.

Yeah it is, and I wouldn't be surprised if it ekes ahead of the 1600x, but it's a bandaid on a stab wound. The knockout punch is going to come next year when AMD releases their Zen+ revision. We've already seen some really easy low-hanging-fruit improvements that could be made to Zen. Zen+ is going to fix the overclock wall and CCX at the very least and you can screenshot me on that.

side note.

whats up with amd fanboys. amd releases a product that hardly overclocks, was over hyped with misleading tech demos, and had trouble working with faster ram. As necessary updates to Ryzen Roll out reviewers Test and Retest to be thorough, and Fanboys think everyone is Bias because Reworld Tests arent showing anything special. Especially when games and everyday tasks arent designed to take advantage of what Ryzen has to offer, like heavily multithreaded production software is. AMD releases a Bike with square wheels and Fanboys Moan that the Roads are bias for not being rounded.

https://www.youtube.com/watch?v=LgbWu8zJubo

>>59650553

Coffeelake isn't till end of year, which will be price cut territory for Ryzen.

Also it's been confirmed that there's a revision on the way which has the IMC running 1:1 with RAM instead of 1:2 meaning you'll get double the gain for RAM speed you get now (until you hit diminishing returns, but we know up to 3600Mhz still makes a difference). So even cheap RAM kits will run at the best expected current performance level.

File: 1489880038374.jpg (18KB, 267x297px) Image search: [Google]

18KB, 267x297px

>>59650593
>gimping
>intel performance hasn't changed

F-f-fucking AMD, cooking the books again

>>59650617

You never worked in the industry and are probably some stupid kid. Everyone does this. A few million is nothing to get better performance on an engine tested in 90% of mainstream benchmarks.

>>59650623
more like intel's been using square wheels all this time and charging people to use their shitty roads and now AMD has circular wheels but idiots are still using intel square roads.

>>59650650
Which video game developer did you work at last that spent 1.5 million dollars paying a handful of engineers to optimize your game for a few weeks?

>>59650438
less than the 6 gorrilian that intel pays for shills per day

BTW all of this is good news. Zen will be improved and many issues will be ironed out in the next year, before revisions or even Zen 2 comes out, I think this is where they will really become a competitor and make a dent.

Meanwhile Intel is pushing the R&D hardcore mode. Unless they fail, they will release something insane in a year or two or at least slash prices.

We will see 30-40% improvements in CPU performance in a few years for the same price. Not just task specific optimization, but overall.

CPU market has been stagnate for years and really fucking needs this.

>>59650650
Look it happens. My dad works at nintendo and he said they paid five software engineers $6billion to implement Zelda's hair physics in an afternoon.

>>59650653
you should understand that the example still works.
a company building a product that doesn't fit the standard market, and an Obsessed Fan base wondering why its not hitting its stride. the wheels and roads will be tweaked over time, so the Hysteria and conspiracy theories just look ridiculous. And when Benchmarks are now heavily dependant on Multiple Updates, its not surprising if different reviewers have different results before and after dozens of updates.

tl;dr: its a shit show, give it time.

>>59650316
>3200MHz RAM which only 5% of mobos/chips can achieve
day1 sure, now it's 100% of boards.

>>59650619
I'm sure it'll be faster than the 1600X.

But I'm also sure it'll be around $400 and require a $30 more expensive motherboard as well. So that's $180 more expensive.

So yeah. It's going to take more than Coffeelake and Skylake-X (especially when X390 is looking to shit on Skylake-X. And are people really going to buy a new $200 motherboard for the 7740k, even if the 7740k is only $250...?)

Intel has to drastically change their designs. Drop the wasted iGPUs on die. Stupidest fucking thing ever that does not justify the prices they charge.

They currently can't compete with how affordably AMD can make Ryzen. If Intel simply drops prices 30% to compete, they cut their margins in half and share holders sue them. They need CPUs that are even cheaper to make like Ryzen are.

>>59650637
Yeah I'm not saying Coffeelake is going to completely turn things around.
I'm just saying it'll compete a bit better while the i3 and i5 lineup now doesn't remotely compete at all in any way at all.

Intel will go from way behind and caught with their pants down to at least being justifiable to fanboys, maybe.

>>59650781
Still not 100%, but seems to be "most" as long as the RAM isn't garbage.

>>59650623
>Hardly overclocks.
R7 1700 can do +1000mhz from base on air.
i7 7700k can do 800
4000/3000=1.33
5000/4200=1.19
>being this bad at math on a tech board.
7700k would have to run at 5.6Ghz to match the 33% overclock.
>7700k only turbos 3 bins
turbolets will defend this

>>59650796
people have this misconception about 1700, they think it's 3.6 CPU
its 3.0 CPU that goes to 3.9 ffs

>>59650098
I'm actually ok here with the 7700k being left out, the 6900 though......no not at all. Considering that the R7s compete with the 6900 line up it makes sense to compre the two. What needs to happen is when R5 launches that those get extensively compared to the consumer i7 and i5 line up.

>>59650821
It goes to 3.75 stock.
It's all core turbo doesn't go to that. Only when only 2 cores are being used.

Most people don't reach 3.9 perfectly stable all core at reasonable 24/7 voltage, either. And if you clock too high you get worse performance from stability features kicking in. So people end up thinking their 3.9 is stable but get worse performance than if they ran at 3.8.
The same thing happens with Intel CPUs. Everyone thinks they have golden silicon and have 5.0 stable, but really it's not stable and would perform better if they dialed it back to 4.8 or 4.9 where safeguards aren't tripping.

>>59650740

Let's not act like there's no reason for AMD fans to be paranoid with the media, and the tech industry as a whole.

So when AMD comes out with a competitive product that is shooting WELL above it's weight, there's going to be a massive amount of support behind it. And a VERY critical eye towards a shitty media who operates as marketing department budgets. Watching for ANY shady shit. So if you're a reviewer and you don't like being second guessed and criticized? Get a new fucking job.

>>59650849
It's still priced around the 7700k. People see it as competition to it. That's absolutely reasonable. The 1700 is actually about $60 cheaper than the 7700k given the included decent cooler and the cheaper motherboards.

All benchmarks should have the G4560, 8370, 7500, 7600k, 7700k, 1700, 1800x, 1600x, 1500x, 1400.
Maybe throw in some other older CPUs like the 2500k and 4790k as well.

I think everyone can agree those would give relevant results.

>>59650882
You know what I learned from this launch? Average fps do not matter and that 90% of reviewers are hacks.

>>59650900

Double aughts confirmed.

>>59650789
>Still not 100%, but seems to be "most" as long as the RAM isn't garbage.

rip corsair

>>59650882
Once full product stack is out yeah include most every relevant CPU since sandy, but for right now the target use for the R7s is the same as the 69** series, and its not gaming. But of course I also understand the pricce point being similar, but to the same end a tractor trailer cost around 200-250k and so does a decent Ferrari, but obviously you dont seriously test the two against each other.

>>59650850

Hi I'm at 4Ghz @ 1.38v which is still within reasonable territory for a 1700 for 24/7 use. Yes it's not on the stock cooler but it's on an air cooler. I dialed it back down a bit from 4.050.

80% of 1700's can expect to hit 3.9Ghz stable around this voltage.

>>59650932
What's a safe voltage?

I'm on a 1700X with asus ROG at 1.395 or some shit at 4.0

>>59650980

Stay below 1.4 not counting LLC.

>>59651004
Alright that's what I was doing because intel was the same way usually

is 1.4 just the limit for 22nm and 14nm or something?

>>59650900
Yeah.

Even though Linus was being pretty shilly for AMD doing 4k and shit, it really opened my eyes how their "minimum frames" was THE SINGLE WORST FRAME.

And what a poor job reviewers do about not highlighting what is the most important with gaming: frame pacing, and 0.1/1% minimums.

>>59650932
>>59650980
On stock cooler it seems like 1.33 or lower is better.
I mean 1.38 is safe and not going to hurt the life, but the stock cooler doesn't seem to handle over 1.3-1.33 at high load. Which makes since sense it's rated at 95W TDP and that's over 95W.

When it comes to safety for the silicon itself, 24/7, 1.4v but you need better cooling than the stock 1700 cooler for that.

>80% of 1700's can expect to hit 3.9Ghz stable around this voltage.
Silicon Lottery says 70% reach 3.9GHz @ 1.408V Which sounds right from what I've heard from others that have made ACTUALLY stable overclocks and not "XD CPU-Z and games for a few hours works!!!"

File: 2%.png (254KB, 401x455px) Image search: [Google]

254KB, 401x455px

>>59650571
Oxides website has no AMD logos, the game website itself does.

Also, unlike AC:Unity, Fallout 4, and Watch Dogs (all Nvidia sponsored) the game doesn't run like fucking trash. It's incredibly well optimized for the amount of shit that goes on. Thousand+ unit battles with FX and projectile physics.

File: scyhle[1].png (8KB, 501x223px) Image search: [Google]

8KB, 501x223px

>>59651032

It's 77% on your source, and I doubt they're super accurate, so closer to 80% anyhow. My OC is stable. 24 hours of stress testing and a bunch of 3 hour 4k video encodes.

>>59651032
>When it comes to safety for the silicon itself, 24/7, 1.4v but you need better cooling than the stock 1700 cooler for that.

Eh, I'm not so sure. High end APUs could take 1.6v all day long as long as temps never hit 80 degrees.

With Zen, once you hit 1.4v your temperatures start getting insane and harder to control.

>>59650521
Every engineer in the industry worth their salt (and all the people with half a brain) knew exactly what was happening and realized it was literally only a matter of optimization since the hardware was there.

The only people who have doubted Ryzen are retarded gamers, fanboys, Youtube personalities with no actual expertise on the hardware they talk about, and the memespouting shitposters around here.

>>59651224

It's not just about AMD, it's also about GloFlo

>>59651255
GloFo is basically former IBM foundry business now. And poolaris showed they can get decent yields.

>>59651081
I made something with 1400 ships and tens of thousands of physical projectiles that runs on my 4 year old Nexus 4 phone, though.

I think Oxide has better programmers than 99% of game studios, but come on.

>>59651130
77% @ 1.392v for the 1700X. 70% @ 1.408v is for the 1700.

>>59651206
That was a different manufacturing process.
AMD says that over 1.4v will start shortening the lifespan. Or was it 1.45v?

But eh.. 1.3-1.4 really seems reasonable to stick at. They become hard to cool and do funny things above that it seems.

People that ran 1.8v+ overclocks on LN2 said the CPUs were perfectly healthy after that, but that was only after a few minutes. That's impressive, but still doesn't indicate that you should keep it at 1.5v or 1.6v all day long simple because you can cool it.

File: bdc.jpg (52KB, 326x315px) Image search: [Google]

52KB, 326x315px

>>59650495
>>59650521
>>59650524
Bingo, Intels ground up architecture won't hit till 2021 at the earliest, and that's only if they say "fuck PAO". All they have going forwards until then is Icelake in 2019, whose major feature set is "Dedicated Voice Processing", and "Computer Vision Engine".

Cannonlake might shrink power but 10nm desktop and server chips will have to wait for Icelake. Even for mobile Cannonlake might not be enough. Ryzen already uses 20% less power for the same performance, the APU's won't have the CCX problem, and GCN already uses much much less power than Gen does. NCU should widen that even further.

Lord knows what Pinnacle Ridge will have.

File: qufaen[1].png (34KB, 540x476px) Image search: [Google]

34KB, 540x476px

>>59651332

Can you not read the image I included? How stupid are you? Do I need to link you to the literal page?

>>59651373
https://www.youtube.com/watch?v=OjdJ1RU4ulU&t

https://www.youtube.com/watch?v=TSYuEopnCfI

sample of 5 chips
amd provided 1700 could go 4.0 not surprisingly

>>59651373
>voltage not listed
You're not smart.

100% of them will reach 5Ghz if you give them enough voltage. Since you're too stupid to understand that, please do not discus or argue anything with me further.

>>59651032
Alright thanks, I do have a custom loop so my cooling is hopefully good enough.. the temperatures are correct in the bios I hope and they don't exist for linux yet (not added to teh actual kernel yet as far as I know)

>>59650553
>Intel can't compete with this 4 thread garbage
fucking blinded by your own hubris. 6c coffee lake is going to best amds best handily and at similair price points

>>59651444
No unless Intel is about to kill these sweet 65% profit margins.

File: topfortykeks.png (582KB, 600x600px) Image search: [Google]

582KB, 600x600px

>>59650619
hahah if you think a 6c intel wont dominate 1800x and handily match zen2 at the expense of some wattage your drinking the wrong coolaid

File: mlgimc[1].png (45KB, 540x476px) Image search: [Google]

45KB, 540x476px

>>59651421

>voltage not listed

You really need to stop

Handle 0x0039, DMI type 17, 40 bytes
Memory Device
Array Handle: 0x0030
Error Information Handle: 0x0038
Total Width: 64 bits
Data Width: 64 bits
Size: 8192 MB
Form Factor: DIMM
Set: None
Locator: DIMM_A2
Bank Locator: Channel A
Type: DDR4
Type Detail: Synchronous Unbuffered (Unregistered)
Speed: 2666 MHz
Manufacturer: Unknown
Serial Number: 00000000
Asset Tag: Not Specified
Part Number: CMK16GX4M2B3000C15
Rank: 1
Configured Clock Speed: 1333 MHz
Minimum Voltage: 1.2 V
Maximum Voltage: 1.2 V
Configured Voltage: 1.2 V


should I push my ram higher if it's only 1.2volts?

I am scared because .. asus rog /10

>>59651474
Lul someone pasted from the wrong page. That's what they have listed for 1700X and the 1700 as lower.

>>59650694
>We will see 30-40% improvements in CPU performance in a few years for the same price.


hahahahahahah

ahahahahahahahahaha

hahahahahahha

yes because these assholes are int he business of giving extra just cuz they so cool

they wont give anymore than whats necessary to move the chips. if they can do a 50% bump theyll do a 25% and wait a year or price the 50% part a couple price levels up

>>59651444
not unless they can get cofee lake on all six cores to 4.5Ghz, not on 14nm
it's another bentium4 rushed housefire later this year
broadwell 6800/6850 is 140w tdp, 6850 goes up to 230w power draw on 6 cores.

File: 1484109919337.jpg (77KB, 640x638px) Image search: [Google]

77KB, 640x638px

>>59651444
>4 threads are the best haha silly corefag
>... but just you wait for 12 thread!!!!!!!!!! then it'll really be good!!! I mean it is already good now and 12 threads aren't needed but wait okay!!!!!!!!!!!! GRRRRRRRRR

>>59651484
I sincerely hope Intel forgot the secret of longer bibelines or else their new arch will be top arsonist.

>>59651484
7700k doesn't do 4.5ghz on all cores, just one or two, no? Unless you overclock.

I wouldn't be surprised if the Coffeelake 6 core is 4.5Ghz turbo on a core or two.

They'll rate it at 140w TDP but really it'll draw 200w+ under gaming alone just like how the "91w TDP" 7700k draws 140w.

Maybe they'll have optimized the process enough to make it work. I can't see their 6 core working if gaming performance on average is worse than the 4 core.

Just look at how Intel artificially gimps all the i3s and i5s so they don't compete in a single core performance with their 8 thread 7700k. In order to still offer strong 4c/8t, they need their 6c/12t to not be gimped.

>>59651483
yea, because why would you want to give people a large boost so they go 'fuck it, upgrade time' apposed to a 25% boost and make them think 'you know what, mine is good enough'

>>59651483
It's pretty well assumed by anyone technically literate that they're going to get at least a 10% clock speed bump when they move to LPU or IBM's 7nm process.

So that's 10% there. They only need an 18.5% IPC increase "over the new few years" to reach that 30% total performance increase.

That's really completely reasonable to assume that, in a few years, Ryzen will be at least 30% better performance for the price.

>>59650650
>You never worked in the industry and are probably some stupid kid. Everyone does this.

you sir never worked in the industry

Lol, AMD bribing devs again.

CPUs don't need optimizations, they either work or don't work, they aren't GPUs

What a retarded shill move from AMD never buying them again

>April 11, big w10 update.
>April 11, R5 comes out.

hmmmm.

>>59651848
What the fuck did I just read.

d-doesn't count!
Ashes is a AMD SPONSORED GAME

DOESN'T COUNT

>>59651870
R5 will be tested a few days before the update, so if there's some fixes you won't see them on launch day benchmarks.

>>59651848
POO IN LOO

>>59651848
This desu, such a dishonest company

>tfw Zen 2 will be on AM4 and I can just drop it in and sell the 1700 for like $250 in a year

Hoping for a nice 15% boost between clock speeds and IPC improvements.

>>59651848
Didn't you mean to write that with Trump grammar?

Good try, though. Was almost copy pasta quality.

>>59651848
>CPUs don't need optimizations, they either work or don't work, they aren't GPUs

There is actually something extremely fundamentally wrong here if a game has to optimize for a specific CPU. The entire point of having a higher level architecture that the uarch implements is to avoid this.

AMD has a history of incompetence though, so anything is possible.

>>59651908
no!!! buy 7700k so you can buy a whole new motherboard and CPU for a 3% increase next year

>>59651932
It's minor threading priority changes, dude. One of the simplest changes you can make. It's nothing like Bulldozer. Fuck off.

>>59651932
>new arch with new topology requires target-specific optimizations
Oh wow I guess it's okay when Nehalem sucked raw cock in gaymen on launch.

>>59651932
No it's not, because not all CPUs are made in the same fucking way.
Intel didn't have problems because their shit over the last 5 years was practically the same shit, but in Zen's case to some games it looks more like two CPUs than a CPU with two cluster of cores and the game doesn't know how to deal with that.

>>59651950
>Nehalem sucked raw cock

fixed that for you m8.

>>59651954
AMD's fault, they should have made a normal CPU like Intel then

>>59651988
No they shouldn't, because this method brings about much higher scalability, yields and improved power consumption, than any monolithic core complex.
At the cost of gaming performance in some cases, which are obviously fixed by software optimizations.

>>59651954
>Zen's case to some games it looks more like two CPUs than a CPU with two cluster of cores and the game doesn't know how to deal with that.

that's false, the OS (at least winblows) properly reports the correct number of physical and logical cores.

userland applications also don't usually have any control over scheduling, they just spawn threads and the kernel handles everything else.

What are other games besides Fallout that show really weird performance on Zen? GTA5 I remember at launch but recent benches I've seen Ryzen obliterating Intel at min frames in GTA5 so that's out.


Far Cry Primal was problematic IIRC, maybe Tomb Raider?

>>59652046
Fallout 4, GTA5, FarCry and Tomb Rider are 4 games that everyone benches that prefer Intel.

>>59652036
When dealing with two CPUs, even Windows is smart enough to keep as much threads as it can on the first CPU before jumping to the other, because going through QPI incurs quite a latency hit.

It's similar with Zen, Windows should optimally fill up one CCX before moving to the other

>>59652062
AMD are working with Bethesda so FO4 will eventually get fixed.

Dunno about the other two, and honestly I don't give a fuck, shit games I don't play.

>>59652036
lmaaaaooo you have no idea what >>59651954 is saying there, but you respond anyone with something makes absolutely no sense in the context of a reply to it.

>>59652046
Far Cry Primal, Tomb Raider, Deus Ex, Doom, Total Warhammer, Watch Dogs 2.

That's mostly it, from what I've seen. Basically any game where the 6900k outperforms the 7700k yet the 1800X does way worse than the 7700k like Ashes was.

>>59652062
With 3200Mhz RAM, FO4 gets a 50% minimum framerate improvement
FIFTY PERCENT.
Goes from like 46fps to 70fps 1% minimums.

GTAV benefits a lot from faster RAM too, iirc.

In some games where it looks really bad like those, just faster RAM seems to bring them at or above 7700k performance.

>>59652099
Interestingly enough, AMD entered a long term partnership with bethesda, so maybe we'll get some interesting results in the future.

>>59652099
>FO4
>Fixed
Not in your lifetime, they probably already destroyed games code at this point not to ever see that mess ever again.

>>59652117
>ayymd will fix bethshit oldass engine
Woohoo, it took some 16 years.

>>59652146
I think "Fix" is being abit too generous.
Maybe, improve. So better multithreading support (Which benefits everyone you shitposting plebs) is going to be the main thing we can expect out of it as neither AMD nor Bethesda have released any statement declaring their intentions in this partnership.

People are really weird, it was shown time and time again that 1800X has enough power as a 6900k and sometimes even a 6950X when all its cores is under full load, clearly indicating its core performance is good, but for some reason shat the bed in gaming.

Why did anyone think there wasn't something fishy here? Hello, you don't suddenly have the same performance as your equal core competitor and suddenly drop by 30% out of nowhere, clearly it's a software issue.

>>59652166
Well Beth stated they are going to adopt Vulkan. Prey will use it.

>>59652168
Any decent human being smart enough to look at the die and core layout would tell you what's the issue.

This is very different than any previous AMD or Intel design, the closest comparison would be a dual socket system, software optimizations are necessary.

>>59652146
They'll probably mostly only improve the rendering pipeline. So 25% improvement in framerates for the given quality at best.

Still will be riddled with bugs and physics locked to framerate, I bet.

>>59652168
>Why did anyone think there wasn't something fishy here?
They just put their fingers in their ears and didn't want to believe it's a powerful CPU because they're shills.
Simple as that.

File: 1490275503916.png (981KB, 1726x968px) Image search: [Google]

981KB, 1726x968px

>>59652210
Speaking of something new, POWER9 looks have somewhat a similar design.

>>59652242
>17 metal layer design

You could use this chip as a wrecking ball

>>59652242
Absolutely fuckhuge core layouts with an enormous shared L3 pool seemed to have hit diminishing returns, especially on the scaling side.

>>59652250
zen got 11

>>59652294
And Kaby 12, shows what kind of monster IBM is making, this thing will have enormous performance and the power consumption to back it up

>>59652274
Massive L3 means massive latency.

While splitting them up means less shared overall, it means lower latency to use the adjacent L3.

Ryzen has nearly HALF the L3 latency within a CCX of Broadwell-E.

Most software likes this lower L3 latency, so long as it doesn't need something from another L3 (which it shouldn't, so long as the application doesn't fuck up and move a software thread to another hardware thread that doesn't share the cache, which is a programming and not hardware issue)

>>59652410
Speaking of L3, zen's L3 seems to have pretty enormous bandwidth as well.

The only thing on the memory side where AMD loses is DRAM and L1 bandwidth, the L1 bandwidth on the Intel's are simply insane.

>>59652438
I don't see how insane L1 bandwidth is useful when it's only 64KB.

Sure it wasn't one of those things on the architecture where it didn't cost hard anything to have so it's just there even though it does nothing?

>>59652438
Intel is still the king of performance SRAM cells, no arguing there.

>>59650623
>Especially when games and everyday tasks arent designed to take advantage of what Ryzen has to offer

This might come as a surprise to you but some of us are adults and don't spend all our free time on playing video games

File: ARMA.png (70KB, 1398x438px) Image search: [Google]

70KB, 1398x438px

>>59650357

>Intel gains pretty much nothing from RAM speeds you idiot.
Yeah no.
All CPUs benefit from faster RAM.

>>59652543
Depends on the game.

>>59652543

>cherry picks benchmarks

>>59652494
L1 latency is under 1ns, L1 bandwidth is over 1TB/s

How much 64kb calls can you fill in one second with 0.7-0.9ns ?
How much power are using it to do it? ICs are more power efficient when they're running under 100%, so there's always gonna be need for more bandwidth

>>59652618
Huh. I thought L1 latency was faster in Ryzen than Broadwell-E and it's only the (2x the size) L2 that's slower?

There's some independent test graphs somewhere.

>>59652543
Finland

>>59652725
I'll find some AIDA graphs when I get out of bed but it went like this IIRC.

L1 lat and bandwidth - Ryzen slightly slower in lat but much slower in bandwidth
L2 lat and bandwith - Ryzen equal in lat but more bandwidth, but its twice the size so per MB it does better
L3 lat and bandwith are both wins for Ryzen

>>59652749
AIDA was initially not working right for Ryzen. It did get updated recently, I think.

>>59652776
Yeah, I'm going by the updated numbers released 2 days ago.

Both DOTA and AOTS were updated recently and got some 30% increase in performance on Ryzen.

I think I'm seeing a pattern here

>>59648913

Look how he didn't dare to include the 200$ cheaper 7700K : ^ )

>>59652815
Same happened with every new arch ever.

>>59652815
>shitty coded games got updated

I wonder why it isn't needed for Sniper Elite, Mafia or Watch Dogs?

>>59652854
All of those are shitty, it's just that some are coded less shittier than the others.

File: 1490733957280.png (92KB, 539x538px) Image search: [Google]

92KB, 539x538px

>>59652784
>>59652749
Here's Zen with the updated AIDA test.

File: 1490737715421.jpg (144KB, 640x456px) Image search: [Google]

144KB, 640x456px

>>59652946
And BDW-E

I can't seem to find the 6900k in this test for comparison, only a overclocked one.

>>59652946
Ah I see. Thanks.

File: 1490736617053.jpg (74KB, 800x574px) Image search: [Google]

74KB, 800x574px

>>59652996
Here's the 6900k OCd

>>59652946
>>59652996
AMD's L2 and L3 are amazing, it should focus on L1 latencies and DRAM latencies next gen and hopefully increases the data fabric's data paths to 64B so it's less reliant on DRAM speeds.

>Don't include 3200 memory on Intel
>Don't include 7700K
>No 'Ashes' test in their first Ryzen review to compare to
>No retest of Hitman
PCPer are useless lazy shits.

>>59653028
I think the biggest thing is the DRAM latency. The best I've seen it at is low 70s ns while intel is usually in the 40-60ns range.

If ALL they did with Zen2 is change process to get 10% higher clocks, and got memory latency down around to what Intel has been at for a while, that'd be such a big performance increase.

>>59653307
>PCPer are useless lazy shits.
Yes.

File: Screenshot from 30-03-2017 13:17:41.png (84KB, 1201x599px) Image search: [Google]

84KB, 1201x599px

>>59648913
Here's the Guru3D version

>>59650456
The question is why you need ((special memory)) to run ryzen.
Any good quality ram on 3000+ league should work without issues

File: Screen Shot 2017-03-30 at 13.24.31-fullpage.png (760KB, 1900x5566px) Image search: [Google]

760KB, 1900x5566px

I thought the higher memory would benefit mostly games but it seems even multithreaded stuff gets a decent increase.

I wonder what the situation with 3600MHz memory will look like

>>59653443
It should, but motherboard vendors fucked up the launch and there's compatibility issues with certain brands of memory until the BIOSes get fixed.

>>59653443

Blame motherboard vendors. The compatibility is pretty good by now already as most sets can hit 3000Mhz on any board, but it'll get even better with time.

>>59653447

i saw a test of it running 3200mhz CL12 and the results where amazing. but i dont remember where i saw the benchmark.

it would be great to see how 4266mhz CL19 will perform with ryzen

Notice how the Intel shills are pretty silent. Not as many 'poo in loo' posters as there normally would be.

>>59653584
Scrambling up to find Digitalfoundry and GamersNexus benchmarks.

http://www.legitreviews.com/amd-ryzen-performance-update-released-ashes-singularity_193137

I don't care about the 22-33% performance increase so much as seeing how it actually compares to the 7600k, 7700k, 6900k.

>>59653723
>I can't compare average FPS numbers
Are you american? Because it seems so.

>>59653723

It's 4~% behind the 6900k now, so well ahead of the 7700k.

OK I cannot let this pass. Here is AdoredTV's response to criticism of his previous video. Love him or hate him he just makes too much sense.

http://www.youtube.com/watch?v=gFBKFz9n2hc

>>59653866
He's like almost entirely on point there, especially with how big of shill faggots those aussies are without directly saying it.

But he was really off about something when I watched it. Something about resolutions increasing in the future...?
That's not why future gaming performance equalizes. It's just because games get better optimized for something.
Games got better optimized for 8 threads because the PS4 and Xbone were both 8 threads. Not because we moved from 720p to 1080p or some shit.

He's smarter than those Aussie fucktards, but he's no Wendell. He often tries to find answers to questions driven by his bias which makes him wind up with the wrong answer even when his original premise was correct.

>>59651466
>if you think a 6c intel wont dominate 1800x
The 6900k is an 8c intel, and it's doesn't dominate the 1800x now, and won't in the future.

File: 4500 CL16.jpg (277KB, 853x1280px) Image search: [Google]

277KB, 853x1280px

>>59653560

I want to see Ryzen run with this monster RAM.

http://www.youtube.com/watch?v=lN5mxFfkr7g

>>59650850
>It goes to 3.75 stock
yeah, on only one core at specific times, while being able to push all 8 1000MHz more on air if you're not a dumbass who doesn't know how to overclock

>>59651017

Anything higher than 1.4v is going to start electromigrating the doping in the silicon. Basically shits gonna get fucked.

>>59651940
why not buy the 7740k anon? It has got a whopping 100MHz higher base clock, no iGPU and you get to need better cooling, because it now has more than 110W TDP and to top all that off, you will need a new *professional grade* motherboard

>>59648913
I wonder how the 1700 performs now?

>>59654161
This is very educational.

>>59651279

The problem with 14nm LPP isnt yields, its the fact it hits a non linear voltage wall at higher clocks. The architecture can clock pretty high, its the silicon that cant handle anything more than 1.4v for daily use.
You can clock a 1700 to 3.8ghz with 1.25-1.3v, from a stock clock of 3.2 and ~1.2v. Once you hit 3.9-4.0 you gotta feed an extra 0.1v at least for that extra 100-200ghz.
The test for AMD and GloFo isnt fixing CCX or microcode optimizations, its going to be to get the dam silicon to clock higher. If, IF, they can squeeze out another 200-400ghz without increasing voltage+power draw too much, while still having some OC headroom, Intel is gonna be BTFO completely untill their next new architecture.
Also, anyone remember if they will have a Ryzen revision on GloFos new 7nm?

>>59651466
>top shekel

>>59651444
>>59651466
Fucking Intelshills are so fucking delusional I swear to fucking god. Intel does not have anything up their sleeve, their arch is at the end of it's lifespan and they have no follow-up. You're all retards.
>>59651483
>Assuming all other companies are as jew-happy as Intel
Kill yourself.
>>59651848
This is the dumbest post in this thread, even dumber than the fanboys.

>>59654076
Not that guy, but the Coffeelake 6 core is almost certainly going to be higher clocked. So it should perform like a better 6850k, which should often match or exceed the 6900k in games.

But who knows. Since that'd be an actual 200w TDP chip.

Housefires again seems like it'd be Intel's likely strategy to compete since they can't compete on efficiency nor cost.

>>59655426
>they can't compete on efficiency
What have they been doing the past 5 years? kek

>>59655540

Sitting on their ass, picking their nose and replacing American tech worjers with H1-b visa workers.

>>59655582
And replaced the TIM with cum.

>>59655540
What are you getting at?
5 years ago isn't today.

The EIGHT CORE 1800X uses less fucking power for everything except AVX2 than the 4 core 7700k.

The 4c/8t 1400 is looking like it'll use less power than the 2c/4t G4560.

Intel can't suddenly make a completely new architecture and manufacturing process (and drop their clocks and performance) to compete in efficiency NOW like they could against 32nm CPUs. Holy fuck you're dumb.

Hell, even when AMD was on 28nm and Intel was on 22nm, Intel couldn't compete with the power efficiency of AMD's APUs until they had HALF the manufacturing process at 14nm.

Intel rested on their laurels relying on node shrinks while AMD was locked in a 6 year contract for 28-32nm and now they have nothing.

>Was the "they'll patch it" meme real for once?
That's pretty much par for the course for AMD. The hardware is good(with the exception of Bulldozer family), but it takes time for the software to catch up, making them look worse than they are at launch.

>>59648927
I don't think they are that bad. They gave Ryzen a gold award on launch, IIRC their conclusion was more positive than say Tom's Hardware, Anandtech, TTR.

>>59655658

If any site gave Ryzen anything less than their top award it would be a blood bath. That's not saying much.

Their recent test on Ryzens CCX latency was a fucking joke.

>>59655658
Uh Anandtech's was super positive.

And Tom's Hardware is an obvious biased shill as they've been for 15+ years running.
They must literally have deep family ties to Intel or something to be consistently shilling for Intel, especially during those times when AMD completely shit on the joke that Pentium 4 was.
In their Ryzen review, they ran all the games where Ryzen did better than the 7600k with a GPU bottleneck to keep them even and make it look like the 7600k was just as good as the 1700 and 1800X (and 7700k for that matter).

It's a fucking joke. All these high end CPUs and they're only getting 55-75 FPS. Half or less than what they got in other benchmarks.

>>59655766
It's funny. Their test is good, and their technical expert that created it understood what was going on... but every other person there tells a totally different story and conclusion from it.

File: 14nm.jpg (186KB, 1529x836px) Image search: [Google]

186KB, 1529x836px

>>59655426
Coffeelake will clock even higher than Kabylake.

>>59655842
Probably wont.

File: 5.2uncore.jpg (442KB, 1920x1080px) Image search: [Google]

442KB, 1920x1080px

>>59655426

File: 1267027605405.jpg (120KB, 640x427px) Image search: [Google]

120KB, 640x427px

>>59650729

>proof or gtfo

File: Road-to-Zen_Architecture-Brief-6-1440x811.png (225KB, 1440x811px) Image search: [Google]

225KB, 1440x811px

>>59652618
In the case of Ryzen, L1D bandwidth is on the order of 100 GB/s you idiot.

>>59652494
You're also an idiot. The other guy's reasoning is still mostly right. Also, L1Ds are almost always 32 kB due to a nuance with virtual address aliasing. L1Is can be bigger since they're read-only and don't need to worry about a written cache line having multiple locations in the cache block.

>>59655858
But it has the same incremental upgrade from 14+

>>59655954
>>59655858
>>59655842
It might clock a little higher(especially on the hexacore parts and better), but around 5 GHz seems to be maximum practical clockspeed for CPUs.

File: AOTS.png (80KB, 681x140px) Image search: [Google]

80KB, 681x140px

>>59648913
I suppose AMD would optimize their benchmark tool for Ryzen. Let's see if actual games that people play get optimizations.

>>59656010
If you don't see developers of your preferred games in first statement about optimizations, don't keep your hope up.

File: 1490715798902.png (21KB, 900x350px) Image search: [Google]

21KB, 900x350px

>>59656008
>clocking higher on hexacore when the 4 cores are already going supernova at 4.7GHz
Please tell me why you think that's the case.

>>59656008
>for CPUs
for Nehalem*

>>59655842
>same power usage
>but we're going to clock it even higher
hm...

I bet they're only talking about laptops (again).

>>59656008
5Ghz is not very practical, especially above 4 cores.
The 7700k at stock clocks typically draws 110-140w.

>>59656070
Higher compared to current Intel hexacores, which top at 3.6 GHz.

>>59656221
let's assume they increase IPC by whooping 5%(nope)
they would have to clock them at 4.5 to lead against 4.0 r5 6c
that's besides power limitations, temp limits will kill it if they don't solder the thing

>>59656389
Huh? No a hexacore Coffeelake would presumably only need to be at about 3.9-4Ghz to beat the 1600X at 4Ghz.

I don't understand how you can argue otherwise.

The only thing that would be an argument is the price. The Coffeelake 6 core is going to require yet another new motherboard and will probably cost almost twice what the 1600X costs.

>>59653963
>He's smarter than those Aussie fucktards, but he's no Wendell. He often tries to find answers to questions driven by his bias which makes him wind up with the wrong answer even when his original premise was correct.

that *was* essentially his premise, he was arguing that bulldozer had better longevity than a sandy bridge i5 because games would be better optimized for higher core/thread count in the years after sandy/bulldozers' releases. his point about the low resolution benchmarks was that they were a poor indicator of future performance in hindsight. did you really misunderstand the point of his videos this badly?

>>59656644
No, you are missing the part I'm talking about, and I don't disagree with that. It's toward the end, but I can't bother going through it again to link to a timestamp.

But basically he actually, specifically, said that in the future games would be commonly played at higher resolutions.
Which is... probably true. But we're still at like 95% 1080p and doesn't change how well a current game runs on a given processor. It's still perfectly fucking reasonable to test CURRENT cpus on CURRENT games without a significant GPU bottleneck.

>>59656692
Other dude here, but if I'm buying a 1080ti and a 1800X I'm not gonna be playing at 1080p, no way in hell.
Even at 1440p it seems slightly overkill.

>>59656692
Dude, he's just saying that games are generally played at gpu bottlenecks today, and that the same will probably hold in the future. The low rez testers are missing the point because these bottlenecks will not disappear in the future; people will be playing at higher resolutions.

>>59656010
You know the most popular games are the ones that still run on ancient toasters and intel integrated laptop garbage, right?

File: 9eff5f0750abd8d6a8b6220b9ef92e47.png (305KB, 799x634px) Image search: [Google]

305KB, 799x634px

>>59656599
It's going to be an overpriced housefire no matter how you look at it. Intel is now officially on the back foot.

>>59656784
Sure, but that has nothing to do with AOTS being the only game with an optimization and also the fact that literally nobody plays it.

>>59656692
Nobody is spending $300+ on a CPU and even more expensive GPU to play anything at 1080p. Only people who go for 1080p 144hz are eSports fags.

>>59651514
It will be housefire no matter what they do. They're in the exact same position AMD was years ago when they had a dud arch that could only be improved by pushing MOAR WATTS.

>>59656813
The optimization was nothing more than some core scheduling tweaks. They did not massively rewrite the game engine to support Ryzen.

>>59656746
>>59656815
You can conveniently ignore 'esports fags' all you want, but the fact of the matter is 1080p, or in some cases, even lower (think consoles) is the target for all games even in 2017. You might think your 1080ti is a 4k card today, but in a year or two, you're going to have to keep dumping endless amounts of money on new hardware to keep that up because consoles kids prefer muh graphix over frame rate. There is no such thing as a resolution=x card.

>>59654443
>The architecture can clock pretty high
How do you reckon?

>>59654443
>they can squeeze out another 200-400ghz [...] Intel is gonna be BTFO completely
I'd think so, yes.

>>59656599
>only need to be at about 3.9-4Ghz to beat the 1600X at 4Ghz.
compare skylake at 4.0 to ryzen at 4.0 beats it about 15% and ryzen loses to kabylake at 4,5-4.8 same 15-18% (on low speed ram at least)

all kaby got for it in gaming is high clocks

>>59651370
>Intels ground up architecture
Assuming they need one.

>>59656900
1080p 60fps, sure, which Ryzen CPUs can easily hit, and with better frametimes due to the increased core count and better SMT.

>>59656938
How much smack do you have to be taking to think they can really do much more with a 6 year old arch that's literally been milked to death?

>>59656889
>nothing more than some core scheduling tweaks

The result of 400 developer hours of work, the Nitrous Engine powering Ashes of the Singularity received an update today to version 26118 that integrates updates to threading to better balance the performance across Ryzen 7’s 8 cores and 16 threads. I was able to do some early testing on the new revision, as well as with the previous retail shipping version (25624) to see what kind of improvements the patch brings with it.

I'm sure it's beneficial to spend 400 hours on 'just some tweaks' for AMD, but I doubt other developers are going to spend this much time fixing AMD hardware.

>>59656599

I think this is going to be a big problem for Intel with the new Hexacores. ~5.1 is as high as you can OC with delidding, water cooling etc, and its producing a quite a bit of heat. I can't imagine that the 6 core will reach 4.5ghz (source my ass) without being thermally limited, though maybe insane cooling solutions might push it. So alright, you got problem 1, your gonna need a fuck huge air cooler or an expensive AIO/custom loop to get good overclocks, while people can get a 500ghz OC on a 1700, with the stock fucking cooler. Intel housefires etc etc.
Second problem, price. So now you need a beefy cooler, +$$. New socket, most likely more expensive than AM4, +$$. Next you have the actual chip itself. Now, Intel might drastically cut prices, but I dont think that the "political will" is there to price their 6 core lower than Ryzen. I expect 7700k maybe $40 off or something, while the 6 core will still be more expensive than a 1700 (the 1700 being the R7 chip to get). So in total you might spend a couple hundred bucks more for an Intel 6 core vs an R7.
performance is the next issue, that I think Intel will win on with their 6 core. It wont be as good as an R7 in multi threaded work loads, but it will be close enough. It will beat single threaded for sure however. It wont clock as high as a 7700k, but it will clock higher than a 1700x. Fine. The problem Intel will have is that by the time they release coffee lake, AMD could be ready to release a revision of Ryzen, which even an Intel like perf gain of 5% would still keep Intel on the defensive. However being the first revision, there could be anywhere from 5-15% increase. What the fuck is Intel going to do if AMD releases Ryzen 2 with say 5% increased IPC but 15% more clocks? AMDs SMT is better than Intels, at that point whats the point of buying an Intel chip for servers or workstation? Intel will be regulated to single threaded applications untill their new arch and 10nm. Shit man best timeline.

>>59648913
It is true most of the time, but news sites don't report it because it would make AMD look good.

They don't make tests of how cards perform 1 year after launch either, or if so, only to little fanfare.

>>59656987
Well, the real question is what they'd do with a from-scratch microarch that they aren't already doing. It's not like Zen is doing anything substantially different from what Intel is doing; in many ways it just fixed the performance bugs that Bulldozer had.

>>59657001
If they put two devs on it, it would take ~5 days.

>>59657001

400 hours is nothing in enterprise software. Thats 3 guys working less than 3 weeks, assuming 40 hour weeks. No doubt AMD tossed some cash their way but it was worth it for them.

>>59656977
Pretty much any CPU can hit that in an overwhelming majority of games currently.

Writing off high framerates and saying 'esports fags' because you're fine with a low, console-esque target *now* is simply denial of poor future performance. I mean I get it, you play games on a 60hz monitor and that's good enough, but using your low standards as a defense mechanism against buying inferior hardware is straight up lying to yourself.

>>59657109
You aren't going to future proof yourself by getting a 4c/8t with beefy single core performance that's already practically obselete.

>>59657001
This is not some repeated effort that must be brought again and again. Devs are optimizing their games anyway, AMD is just had to teach them how to do it for this new architecture.

>>59657052
In any case it wouldn't even be out until 2021.

>>59657039
Expect to hear
>muh single threaded gaymes
until the end of time.

>>59657104
>>59657108
Of course, AOTS is AMD's pet project, it's entirely AMD sponsored. If they don't take care of that, it makes their hardware look even worse.

400 hours is not in any way insignificant for game developers. Especially not for a fix for CPUs that less than 1% of the market is currently using and especially not for already-released games that have very low, if at all, post release developer budgets.

We'll see optimizations for future titles, sure, especially after lower end AMD CPUs are released, but it'll probably be too little too late at that point.

>>59657163
watch this
>>59654161

this is hilarious on its own

>>59657041
Good thing they'll be forced to retest Ryzen once the R5 series releases.

>>59656930
Uhhh you're thinking of the 7600k right? And how it loses to the 1500X at 20% lower clock speeds?

That's because the 7600k only has 4 threads.
The 7700k, however, beats the 1500X by a good 20%.

A 6c/12t at at least 4ghz Coffeelake CPU that OCs higher is going to beat the 1600X just the same.

It's very simple... I don't understand why I'm needing to explain this.
Why are you so defensive about the facts? Like the 7700k costs double the 1500X but only beats it by about 20%. The 7600k costs more and is slower on average, too. That's still a win in performance/$.

>>59657109
Kay, I mostly agree. But if you say that, I hope you never recommend i3s or i5s to anyone when a G4560 should clearly be sufficient, right?

>>59657039
Yeah. Ryzen is obviously better performance/$. The Coffeelake 6 core will cost more. You'll need a cooler that's almost twice the size. You'll need a more expensive motherboard. All for maybe a 10-20% performance increase on average.
But the 1600X at its way lower cost is still perfectly adequate to game at 90 or 120 FPS minimum for the vast majority of games, so there's little point in the (probably) double the cost for that Coffeelake 6 core for that 10-20% more fps while your house burns down.

>>59657204
>7600k right?
no, 6700=clock for clock kaby

>>59657187
That was pretty funny.

>>59657142
Again, assuming they need one. If all they actually need is, say, something akin to AMD's NN branch predictor, then there's no reason whatsoever why they couldn't retrofit that onto their existing architecture.

It's stupid to go around and say that "they need to rebuild it from the ground up" if there's nothing to be gained from doing that.

>>59651480
Ddr4 is safe up to 1.4v, a lot of the higher clocked dimms go up to 1.35. But it's more likely your timings that's the issue. If you purchased a pre-overclocked dimm, which is then xmp should have all the settings built in.

>>59657280
If it were really that simple, don't you think they would have bolted it on already? And do you really think it would be as efficient as if they had designed it with that in mind? Of course it won't.

>>59657280
>Again, assuming they need one. If all they actually need is, say, something akin to AMD's NN branch predictor, then there's no reason whatsoever why they couldn't retrofit that onto their existing architecture.
Intel has what, ten or twenty times the research budget AMD does(and that's the entire AMD, not just CPU division). It would be pretty strange if there was any simple way of significantly boosting the IPC of their existing architecture left after so many years spent refining it.

>>59657342
You don't get it man, Intel's been keeping all this top secret tech in their back pocket just to bust out when AMD is making a comeback. BAM! 5 GHz base clock CPUs outta nowhere! HELLO BIPELINE!

>>59657330
It was just an example.

The real point being that it's not like Zen is ahead of Core in terms of IPC or anything; AMD has only just come back to being about equal. There's no reason to believe that AMD will be racing far ahead of Intel any time soon either.

More likely, both Zen and Core are about as good as you can get with an OoO architecture, and to see real architectural gains again, some more radically new innovation has to be made. Right now, there is no known such innovation for either Intel or AMD to benefit from. If there were, then it would make sense to say that Intel should create a new architecture from the ground up to incorporate that innovation, but as things stand, they'd likely not have much to gain from it.

>>59657371
>There's no reason to believe that AMD will be racing far ahead of Intel any time soon either.
Zen+ will pretty much do exactly that while Intel is struggling to release 6 cores with decent clocks that don't spontaneously combust under load.

>>59657386
>Zen+ will pretty much do exactly that
Citation needed, really. While Zen has proven really good, in most scenarios it does still seem to be below Core in terms of IPC, if only slightly. Zen+ may well bring them much closer into parity (perhaps even nudging out Core if it turns out really, really good), but you won't see AMD leapfrogging Intel into all-new IPC territory.

>while Intel is struggling to release 6 cores with decent clocks that don't spontaneously combust under load
You have to keep in mind that the reason Ryzen is so energy-efficient is because they're using a process that can hardly clock higher than 4 GHz.

>>59657364
>>59657386
Is there any reason a company with one competitor would want to blow their load, releasing their best possible product, if that only competitor has very weak competition?

Wouldn't it be logical for this company to slowly trickle out small improvements as needed instead to maximize revenue?

I'm not saying Intel does or doesn't have a trump card in their back pocket, but if they did, I'm having a hard time with the argument that they would use it when it wouldn't make financial sense to do so.

>>59651908
This desu. Not having to switch motherboards and everything on every upgrade is probably the strongest argument for AMD. Given that Zen is at least remotely competitive with Intel, that alone is enough reason to go AMD.

>>59653447
/pcbg/ earlier had an OP image from a video I haven't been able to find after first seeing it. But a guy used bclk overclocking to get 3600mhz memory running and did a comparison video showing improvements with ram speed at 3.95ghz, and a comparison to a 7700k with 3200mhz memory. With 3600mhz memory it was matching the 7700k in the games he tested. Only one in particular I remember being tested off hand was tomb raider.

>>59657469
>one competitor
That's only in the desktop market. ARM has been pushing into microservers and IBM offers pretty competitive performance in high-end servers.

And of course, you also need to incentivize people who own previous gen products to upgrade. So I would say that there are plenty of reasons for Intel to not hold back if they indeed had a trump card in their hand. Whether those reasons would be SUFFICIENT is something I don't have the background to speculate.

>>59657671
>IBM offers pretty competitive performance in high-end servers
Competitive performance, yes, but at like 10 times the price.

>>59657309
why is my timings an issue? I just have it at 2666 cuz I haven't set it higher

>>59657671
>ARM has been pushing into microservers
I'm probably looking forward more to K12 than I did to Zen. It'll be really interesting to see if they gain anything substantial by not "paying the x86 tax", as Jim Keller said himself.

>>59657671
>you also need to incentivize people who own previous gen products to upgrade

Right, like they have been doing. I wouldn't call any of Intel's core line gens 'blowing their load'. They've been slowly trickling small improvements because AMD failed with Bulldozer.

Intel totally has some secret sauce cooked up in the back room. Source: My dad works at Nintendo.

>>59657800
So why is the i7-7740K such a fucking dogshit sandwich?

>>59657758
It might be more expensive, but given that high-end Xeons costs $6000+ a piece, it's a lot closer than you seem to think. Also keep in mind that power8 has super-wide cores with 8-way SMT, so a Power8 octocore isn't really comparable to Broadwell-E octocore.

>>59657800
>Right, like they have been doing.
They REALLY haven't. There's remarkably little reason to upgrade from Haswell unless you want DDR4 RAM or really need the features offered by new motherboards.

>>59657819

Kek, I honestly thought Intel had some shit cooking out back too, but I can't believe they didn't. Coffee lake is next year for fucks sake, if they had anything they had plenty of warning to release it to nip Ryzen in the bud, but nope. 7740k with 1% perf increase for more money, disapointing. You bet AMD will have a Ryzen revision ready for 2018, the server chips are a new stepping already (B2 for Naples vs B1 Summit Ridge).

>>59657943
Zen2 is already on the calendar for Q1 2018.

However, I wouldn't be surprised if Naples or Snowy Owl comes before Ryzen 7 this time.

Ryzen 7 came first this time as a test to iron out bugs for Naples. So we might not see consumer chips until Q2.

>>59657875
>given that high-end Xeons costs $6000+ a piece
Sure, but IBM's own POWER servers are more like $50,000+ a piece. There are cheaper alternatives from other vendors, though, but they are very unestablished.

>Also keep in mind that power8 has super-wide cores with 8-way SMT, so a Power8 octocore isn't really comparable to Broadwell-E octocore.
Of course, but then again Intel has 24-core processors where IBM has at most 12-core ones (at the extremely expensive end), and they don't seem to gain much from running beyond 4-way SMT.

Don't get me wrong, POWER8 is cool stuff and I'd love to see it gain more ground, but it's not really competitive with Intel at all if you factor in price, and especially not if you factor in software compatibility. There are lots of JIT compilers for popular languages that currently support x86 and ARM, but not POWER. To wit, even OpenJDK doesn't support POWER.

>>59657914
I agree, but that doesn't stop consumers from buying them. People will buy all sorts of stupid shit, that's the entire idea behind marketing and consumerism.

>>59658014
BUT IT'S THREE THOUSAND NUMBERS HIGHER

4nm 6Ghz CPU when?

>>59658072
ln2 cooled bipeline plz

>>59658088
ln2 AiO in every house by 2021!

>>59658121
Intel will pay a guy to refill it for you while you game. I mean, work.

>>59657998
>Sure, but IBM's own POWER servers are more like $50,000+ a piece. There are cheaper alternatives from other vendors, though, but they are very unestablished.
http://www-03.ibm.com/systems/power/hardware/index.html
I'm guessing the most expensive ones are the the ones that say "contact us for price", but for $6000 you can get a server with two 8-core CPUs.

>>59657967

Naples is to come out this year lad, Q2 or Q3 in fact. Snowy Owl might release at the same time or later in the year. Intel is gonna hurt.

If AMD manages to fix their broken DDR4 support I'll consider buying a 1700x.

Otherwise I'll wait and see what the 8700k 6 core has to offer

>>59656010
>WOW AN RTS IS NICHE IN 2017
great observation

>>59658154
Uhhh $50k sounds incredibly reasonable if you know what comparable tech usually costs.

>>59658167
I'm talking about Naples being the first new release for Zen2. Sorry if that wasn't more apparent.

>>59658180
Not just a shitty RTS that nobody plays, but AMD's own tech demo.

>>59658154
That was certainly better than I thought, but even so, in the best case you get a server that's kind of about the same performance as you get at a similar price from Intel, and I don't think there are many who'd willingly throw away any compatibility advantages they may get with x86 for what is, at best, a tiny advantage in performance.

>>59657967
>>59658167

I misunderstood you. Could be possible, but I dont think they would have such a small gap between Naples 1 and 2.

>>59658220
If they do move process to LPU, giving a 10% more clock improvement at a given voltage, I don't think they'd have trouble selling such an updated lineup so soon unless yields are worse. (reportedly they're around the same as LPP for other chips)

Otherwise, yeah, they'll probably wait until zen3 to update Naples on 7nm.

>>59658207
Not an argument, fanboy

>>59658251
Those were facts. You're welcome to provide an argument, though.

>>59658210
Even small performance per watt improvements are kind of a big deal in datacenters, supercomputers and other markets IBM targets. So depending on how much you need that compatibility, it might be entirely worthwhile to sacrifice it.

It's not exactly a huge target market, but IBM is by all accounts doing pretty well in it. And they'll probably do even better when they release Power9 later this year.

>>59658251
Well he's not wrong.
AOS has basically no player base.
It's just a game optimised by AMD

I'd like to see actual games people play get be benchmarked with some workstation stuff.

>>59658293
>Even small performance per watt improvements are kind of a big deal in datacenters
Ah yes, but POWER8 is considerably worse than Xeon in that metric.

File: power.png (21KB, 676x182px) Image search: [Google]

21KB, 676x182px

>>59658306
Adding source.

>>59658393
That's perf/watt at that particular test. Is that what it's actually designed for?

>>59658305
Please don't pretend that tons of games weren't benchmarked with Ryzen.

>>59658305
what games that can run on a potato like dota 2? no one plays gameworks games and no one seems to be complaining about those

>>59658496
>That's perf/watt at that particular test. Is that what it's actually designed for?
I haven't seen a test where it's better, at least.

I think the main selling point for POWER right now is that some of the companies with really large datacenters want to get out of the Intel lock-in, and IBM is trying to offer them such a way with the OpenPOWER foundation, not so much that it's more efficient. It's a laudable effort, though, and I wish them all luck, but they're not there yet.

>>59658530
Hm. I don't feel convinced though.

Like.. IBM sometimes misses the mark, but they're highly capable.
Like with Cell? It missed the mark. We did not need a CPU with that high TFLOPS/$, that could do so much vector math... that's GPU work.
But... they DID make a CPU that was absurd TFLOPS/$ and was the best performance in that regards even for servers that needed that.
It did take more than a year, IIRC, for GPUs to be better TFLOPS/$ than just buying PS3s.

So I'd wonder if it's software optimization or something else making it suffer? It does use 4-way SMT, after all. Are there not even synthetic tests that its perf/watt excels at?

Wait.... that's also Power8, not Power9. Oh, I missed that you guys were talking about Power8.

>Ryzen shows promise
> Intel "Guys, get this, guys! A Bentium with Hyperthreading AND, get this, an UNLOCKED i3. Oh you bet that pretty ass there's more. Price REDUCTIONS."
>drops mic
Coffee better bring some real heat to the party. Once Ryzen matures it's going to go for Intels head while they keep worrying about different ways they can be a more inclusive company. Competition is so exciting.

File: 1490643982495.jpg (256KB, 2140x1214px) Image search: [Google]

256KB, 2140x1214px

>>59658496
>Please don't pretend that tons of games weren't benchmarked with Ryzen.
They were
Pic related

>>59648913
>ryzen
>can get intel processor with the same performace for cheaper
>"just wait for it to be patched!! Ryzen will recover!!"
when will you guys stop?
stop embarassing yourselves.

>>59658590
>Are there not even synthetic tests that its perf/watt excels at?
There are tests where POWER8 excels at single-threaded performance, which is pretty cool.

>>59658739
>can get intel processor with the same performace for twice as much money

FTFY

>>59658178
They already did, I literally grabbed some random RAM (evga) from Micro-center that was 3200 and the XMP worked without having to adjust anything. It wasn't even on the list of supported ram for my mobo.

In fact I tightened the timing to 16-15-15-36 and I haven't even tried going lower yet. The original was 16-18-18-38.

>>59658709
You do realize those aren't recorded the same day they are published right? They would likely begin at least a week before they post them.

>>59658599
All I'm seeing is a lot of shills, and a lot of panic moves.

>>59658739
Kill yourself, retard.

>>59659037
what motherboard?

>“We intend to issue updates to motherboard partners in May that will enable them, on whatever products they choose, to support speeds higher than the current DDR4-3200 limit without refclk [reference clock] adjustments.”
http://www.pcworld.com/article/3186443/computers/amd-preps-for-ryzen-5s-launch-with-a-helpful-am4-motherboard-update.html

>>59657441
>Zen+
I wonder whether zen+ CPUs will be compatible with the motherboarda we have right now.

>>59659403
Yes. Same socket.

>>59659403
It's AMD, not Intel. I'd be very surprised if they won't stick with AM4 for the foreseeable future.

File: zbkshIM.jpg (200KB, 800x512px) Image search: [Google]

200KB, 800x512px

>>59659403
>mfw people are so used to getting jewed by intel they assume AMD is going to do the exact same thing

>>59658599
Coffee Lake is just another die shrink. That's assuming it doesn't just get delayed again (we're almost at 10nm we swear!)

Ashes of the Singularity is irrelevant and its fucking hilarious that they are using it for PR for the second time.

>>59659529
Oops was wrong, it's still 14nm. So basically no real improvement.

>>59659071
Those numbers are with the most current Bios drivers for AM4 and the fastest ram they could run.

>>59659544
Drink bleach.

File: dumb faggot looking stupid as fuck.png (563KB, 547x822px) Image search: [Google]

563KB, 547x822px

>>59648913
DELID THIS

>>59648913
Intel BTFO

>>59659458
>>59659510
I know that AMD doesn't change socket when someone sneezes in their HQ like Intel. With the leaks about AMD releasing a dual socket HEDT platform, which is not a thing they have done, I thought that they could go full intel for some reason. I didn't know that the socket wouldn't change and thought about am3 and am3+ for some reason. Thanks.

>>59660147
>socket
Sockets*

>>59659544
Weren't you BTFO'd awhile ago? Why are you still using that trip, tripfag?

>>59659510
I too wish AMD was a competitor rather than an alternative. Imagine how low CPU and GPU prices would be.

>>59659563
Imagine a world where people who believed the earth wasn't flat were fed bleach. Your precious company wouldn't even exist, fanboy.