i7 7800x = R5 1600 across 30 games

File: 7800xVS1600Average.png (371KB, 1277x718px) Image search: [Google]

371KB, 1277x718px

Here is the average which of the 30 games tested

File: 7800xVS1600cost.png (291KB, 1274x726px) Image search: [Google]

291KB, 1274x726px

>>61500249
Here is the cost

>>61500203
>How can intel compete
a $330 8700k

File: 1050$builds.png (256KB, 1900x780px) Image search: [Google]

256KB, 1900x780px

>>61500339
It wont compete
It will perform exactly the same as the 7800x
Even if its 330$ its still much more expensive than a 1600
199$ 1600 with stock cooler vs 330$ 8700k with 60$+ cooler
I would way rather spend 190$ on a better GPU than in a 1% faster CPU

Price gap is so extreme you can buy a r5 1600 with a GTX 1080 cheaper than a i7 7700k with a GTX 1060

>>61500388
Depends if the go with the ebin ringbus. That's the only reason the 7800 loses out, well on gaymin anyway.

>>61500424
I thought ringbus was out with the skylake X chips? Didn't intel make a new interconnect?

>>61500464
It's literally ringbus squared

>>61500488
its just a cpu interconnect
Infinity fabric/glue is the same thing

>>61500464
They don't have to implement it, the can just rebadge skylake or give it a few tweaks

From the leaked specs the 8700k has 12mb of L3(7700k had 8 for 4 cores). Meanwhile the mesh 7800x has 8.25mb.

It's still up ion the air though and the 1600 has until q1 2018

x299 was a mistake

>>61500537
I thought zen 2 is q1 2018?
8700k was pushed forward recently right?

>>61500555
I meant the 8700k was supposed to be q1? They might have brought it forward.

Dunno about zen 2 timescales either tbqh, still kinda up in the air with that one

>>61500572
8700k is in 6 weeks

>>61500572
zen on 7nm is early 2018
>>61500594
They really rushing these launches... Didnt they learn?

>>61500594
O weow, they really are in damage control.

>>61500605
I have my doubts of them getting mass produced 7nm chips out in early 2017, I'd probably put $10 on then not getting 7nm until 2019. Love to be proven wrong, we'll soon see I guess.

File: 5nmboys.jpg (81KB, 740x553px) Image search: [Google]

81KB, 740x553px

>>61500638
we have 5nm coming soon too

>>61500555
>>61500572
>>61500605
Pinnacle Ridge is early 2018, but that's just the platform name. Could mean anything from new chipset only to new CPU entirely, my guess is new steppings and a small SKU bump (1850X?). Zen 2 is just labeled 2018 in roadmaps, and I'll be fuckin amazed if it's before Q4. They might have to do quad patterning to fab the initial lot, though I've heard that EUV rollout is going quicker than expected.

File: 6ZFrv2R.jpg (43KB, 430x707px) Image search: [Google]

43KB, 430x707px

>>61500638
>I'd probably put $10 on then not getting 7nm until 2019

they not gonna leave Zen 14nm till 2019 its obvious
AMD will do their magic as always

>>61500525
the problem lies on the implementation..

amd has the uncore speed tied directly to ram speed
intel being intel and cares only for money doesnt so they wont get the free "oc" from it

File: Clocks.png (39KB, 512x446px) Image search: [Google]

39KB, 512x446px

>>61501818
You can still OC via BCLK on skylake chips

wow lad

>>61501836
pretty sure you know what i meant...

File: 1497883390389.jpg (18KB, 267x297px) Image search: [Google]

18KB, 267x297px

>>61500203
>>61500249
>>61500268
SHUUUUUT UUUUPPPPPPPPP

>>61500424
also mb shekels

>>61500203
Intel user here

They are gonna have to make a completely new CPU. Anything past 4 cores doesn't communicate worth shit.
Their square ring buss shit doesn't work.
It makes every SkylakeX chip run like shit.

SkylakeX preforms worse than broadwell E with twice the power usage.

>>61500268
Why is Ryzen so much cheaper? It's like same performance for half the price. That doesn't make any sense.

>>61502378
because Ryzen yield is so much higher.

>>61500268
Why's the Intel system been given twice the RAM?
>>61502378
Crazy yields. Boards have gotta be cheaper since the gold pins are on the CPU vs the motherboard for Intel. They've included good coolers for a while so they probably have that supply line down pat. Couple that with Intel using cum and the Blue team suddenly needs a good cooler to force the heat off the chip.

>>61500268
>16GB vs 32GB RAM
>Box cooler vs AIO

Nice try, AMD shill.

>>61502378
>Why is Ryzen so much cheaper
Because it's just two 4 core CPUs glued together.
It's wayyyy cheaper than making a monolithic 8 core.
That and Intel requires a 60% profit on all chips sold.

TAhat madman Keller made that cheap ass CPU work. I was convinced the CCX latency would be a huge issue for AMD.
I'LL Admit I was wrong.

>>61502424
>Why's the Intel system been given twice the RAM?

According to the reviewer, anyone who is investing in X299 platform is going to be using 32GB.

File: 1371871249114.jpg (73KB, 1068x600px) Image search: [Google]

73KB, 1068x600px

>>61502428
My 1600 runs at 3.8GHz on the stock cooler without any additional voltage. I could definitely push it to 3.9GHz-4GHz. The 7800X needs an AIO minimum because Intel jewed out on the TIM again and it sucks power like a crack whore looking for a tip.

>>61502596
>Intel jewed out on the TIM again
Why are AMD fans so uninformed?

The Tim has nothing to do with the temps.
The tolerance between the chip and HIS is too large.
Even cheap TIM is still 5-7C behind the good shit.

>>61502704
Evidently not, Kaby Lake is a fucking fire hazard, and now so is Skylake X.

>>61502704
Then why does delidding a mayo Intel chip to cool the die directly causes a 20C decrease while delidding a soldered AMD chip to cool the die directly causes a 4-8 decrease in temps?

Synthetic thermal paste will never ever be a better heat conductor than gold ol' fluxless solder.

File: 1494977356687.png (315KB, 664x602px) Image search: [Google]

315KB, 664x602px

>>61500249
>7700k is still the best
All I needed to know.

>>61500203
I don't understand how the 7800X performs so poorly, I has to be something to do with the mesh interconnect

>>61502704
Oh so Intel as terrible quality control as well as just being cheap bastards.

File: Screenshot_2017-07-22-01-38-40-1.png (187KB, 997x753px) Image search: [Google]

187KB, 997x753px

>>61502741
>Evidently not
Are you retarded?
The large tolerance causes bad heat transfer from the HIS. Several tests have shown the Heat spreader on a 7800x does that really get hot while the chip gets 90c.

>>61502782
>Then why does delidding a mayo Intel chip to cool the die directly causes a 20C decrease
Because you remove the glue Intel uses to connect the HIS to the PCB when deLidding.
That causes the HIS to fully contact the chip.

WHY are AMD fans so fucking retarded?
Seriously, how do you all not know this?

>>61502886
Then why does delidding a soldered chip result in barely any noticeable temperature drops hmm?

>>61502542
>Because it's just two 4 core CPUs glued together.
That doesn't matter in this case since both complexes are part of the same 8-core die. The yields would be the same even if it was a single 8-core CCX on the die. It only has impact on Threadripper and EPYC since those use two and four dies respectively.
What it does, though, is allow a lot more modularity during the _design_ process - for example Raven Ridge for notebooks will probably be one 4-core CCX and an iGPU on a single die.

>>61502809
Double the price at that. And it's not like there's a drastic difference.

>>61502896
>Then why does delidding a soldered chip result in barely any noticeable temperature drops hmm?
Are you just pretending to be retarded?

>>61502907
Each CCX is separate and isn't part of a 8 core chip.
Their ccx has a 81% yeild so they are stupid easy 2 make.
I can't imagine Intel has a bigger than 40% yeilds on their 8 core house fires

>>61502964
Are you fucking retarded?
One Zeppelin die contains two compute complexes (CCX). It's 195mm^2. AMD is not chasing 60% margins, it's that simple.

File: 1453229998000.png (192KB, 501x445px) Image search: [Google]

192KB, 501x445px

>>61500203
>benchmarks

>>61500203
I like how this Aussie AMDrone pretends to ignore that Intel is competing with itself

File: intel cuck.png (648KB, 1150x629px) Image search: [Google]

648KB, 1150x629px

>>61500388
>1080 n the ryzen build
>1060 on the intel build

>>61500388
>>61503389
Fuck, Ryzen is better at pretty much everything.

>>61502932
>Double the price
wrong

>>61503620
>deliding kit
>cooling for a delidlake
>proper non-exploding motherboards

Yeah, he meant to say close to 3x the price

File: bz8l2tz1rxaz.png (897KB, 1920x1080px) Image search: [Google]

897KB, 1920x1080px

>>61500203
>same performance

>>61500203
>7700k on top of pretty much every benchmark
Feelsgoodman

>>61503650
This is some fantastic cherrypicking, it's almost like the first two images in the threads don't already have an average 30 game score and their breakdown.

>>61500203
>DX12 Deus ex AMD loses
>DX11 Deus Ex AMD wins

That's one fantastically implemented render path

I like that hardwareunboxed does through testing and isn't biased, but averages and min frames still in 2017? frametim graph is far more important, and basically tells you everything from highs, lows and microstutters

>>61503829
YEAH BUT I CAN'T JUST GLANCE AT A FRAMETIME GRAPH AND COME TO A HASTY CONCLUSION LIKE I CAN WITH A BAR GRAPH SAYING BIGGER IS BETTER

Fucking catering to normalfaggots

>>61503650
>posts one game
>op image has dozens of gayms
Your desperation is showing.

File: 1494078532776.png (19KB, 500x330px) Image search: [Google]

19KB, 500x330px

>>61503650
Hey I can do that too.

>>61504464
>AMDshillPowerUp
>>61500203
>AMDunboxed

>>61504464
...the fuck?

File: 1496164491711.jpg (43KB, 630x630px) Image search: [Google]

43KB, 630x630px

>>61504473

>>61504473
t. intel engineer

File: 1497565895125.jpg (112KB, 691x771px) Image search: [Google]

112KB, 691x771px

>>61500865
>5nm
DELETE THIS

Should I attempt to delid an i7-6850K?

File: 1364815917765.jpg (41KB, 400x398px) Image search: [Google]

41KB, 400x398px

>>61500203

>6 core ryzen costs 250 bucks
>6 core intel costs 600 bucks
>same performance

KEK

>>61500594

wasn't 7700K released in fucking February this year?

Jesus.

>inb4 requires new Z370 socket

>>61502704
Uh, no. The difference between core temps and heatspreader temps are as large as 30C.

>>61502428
>quad channel memory
>needs water cooling

Wew such shilling amirite

>>61500203
>>61500249
>>61500268

holy shit, releasing a 6c/12t at this point was a huge mistake. they themself destroyed the 'intel is for gaming' meme by showing everyone how shit the multicore performance of games is

>>61502886
This image shows that Core temp is not being efficiently transferred to heat spreader.

>>61500268
Holy fuck rekt

Also WTF I HATE LINUS NOW

>>61500555
>>61500946
AMD seems to have confirmed that that "2000 series" Ryzen CPUs will be 7nm. I don't think desktop ones would be here as early as q1 2018, though.

I have a feeling that 7nm is going to come to servers before desktop this time around.

>>61504796
AMD has confirmed nothing, only that "Zen cores" will be on 14nm+

Which probably means higher clocks, but it could only be for Raven Ridge and not Desktop

>>61504796
Early 7nm on duv I'm guessing is ryzen mobile refresh, get some of that quick and easy low power savings on low complexity chips and also a decent pipe cleaner for zen2 and euv.

File: Capture.png (498KB, 1943x617px) Image search: [Google]

498KB, 1943x617px

What's the difference between all of these? Logical Increments suggests the Krait one, but it looks like it's just painted differently and costs more.

Is it best to get the gaming plus one for a r5 1600?

>top boffin at Intel quits suddenly after every media outlet tears memelake-x a new rectum

Hmmm....

There is another thread about Coffee Lake S leaks. >>61496664

Looks like it will be slightly (10-15%) faster than a 1600X but cost $100 more.

>>61502585
you're not gonna say that the intel system performs the same as the ryzen one when the first one has twice the RAM. If you're gonna do a benchmark on this CPU you're not gonna use more expensive hardware just because

>just another amd biased review

>>61504856
You'd be looking at $514 vs $904. Not exactly looking good for Intel still.

I'm willing to bet most Intel fans are just waiting for Coffee Lake anyway.

>>61504829
You probably don't need x370 for a ryzen 5 1600 unless your heart is set on SLI. Go with something like the MSI B350 Tomahawk

>>61504870
>MSI
He should get B350 Strix.

>>61502809
>All I needed to know.
Same here.

>>61504870
Is that the only difference? For x370 and b350?

>Skylake-X will crush ryzen, 7700k with 6 cores at 5.0GHz
Look how that turned out

>CoffeeLake will crush Ryzen, imagine 7700k with 6 cores at 5.0GHz

>>61504885
That's the main difference. Going down to B350 you also lose some USB ports I think. That and since all the high end boards are X370 you get better features on the side, like better NICs and audio chipsets. Not inherent to the platform, but the way it tends to be

>>61504878
It's $130 though why not get x370 at that point?

>>61504973
Lowend x370 are junk

>>61504921
>imagine 7700k with 6 cores at 5.0GHz
What.

File: IMG_0012.jpg (201KB, 1280x819px) Image search: [Google]

201KB, 1280x819px

>>61500203
wtf, this proves Intel's architecture is only capable of making fast quad-cores firehouses, anything beyond that doesn't scale well and becomes an abortion.

>>61504921
lmao most 7700ks cant even make it to 5ghz. how is a 6 core variant going to manage that?

File: 1499876433358.jpg (166KB, 1312x1014px) Image search: [Google]

166KB, 1312x1014px

>>61503620

>10% higher perfrormance
>50% more expensive
>requires delidding and a non-stock cooler to OC
>still the best

>>61505200
Dunno, ask fanboys. They're the one constantly going on about how the next Intel CPU will have a zillion cores and hit 10 GHz on air.

>>61505268
>10ghz on air
B-bibelines!

>>61505248
Same logic as the Pentium 4 EXTREME Edition days.

File: sempron_3400-quake31.gif (14KB, 434x446px) Image search: [Google]

14KB, 434x446px

>>61505350
FUCK ATHLON
BUY BENTIUM

>>61505465
MOAR GIGAHURTZ

File: IMG_3227.png (933KB, 630x945px) Image search: [Google]

933KB, 630x945px

>>61500268
DELET

>>61505465

How much more expensive was the Athlon from Pentium 4 and the Extreme?

>>61500946
What is with all this bullshit FUD about pinnacle ridge/Zen2 lately?

Su herself has said Zen2 is pinnacle ridge. Which is a die shrink to 7nm with improvements to IPC.

>>61502428
You can't oc 7800x without aio, also the oc in those benchmarks required a 300 dollar loop, while ryzen oc results are on the stock cooler.

>>61505529
No she hasn't, Pinnacle Ridge is in the older leaked roadmaps as a early 2018 product.

Zen2 aka 7nm Zen is coming soonest Q4 2018 and maybe Q3 2018 in very limited qualities for either EPYC/mobile, certainly not for desktop

>>61505632
Since Zen2 is earlier than expected, i think they canned Pinnacle Ridge.
APUs will still use 14+.

>>61504807
Lol. Current Zen is on 14nm+. Pinnacle Ridge/Zen2 will be on 7nm. Zen 3 will be 7nm+.

>>61505660
Even AMD's roadmaps from the FAD say it's fucking not, and you don't even know what 14nm+ is, likely 14nm HP

https://forums.anandtech.com/threads/globalfoundries-7lp-7nm-leading-performance-finfet-process-and-fx-7-asic-platform.2508687/page-4#post-38975888

File: 1496181076435.png (973KB, 801x1500px) Image search: [Google]

973KB, 801x1500px

>>61505510
it was cheaper. a lot cheaper
also, notice the IPC
>Athlon 64 3400+ is 2.2GHz - 405fps
>Pentium 4 3.4GHz - 373.3fps
AMD basically adopted that naming scheme so people could have a comparison which CPU corresponds to higher clocked intel chip, hence that 3400+. And as you can see, it exceeded it's counterparts performance anyway.
bless the bibelines.

>t. proud owner of a 2800+

>>61505698
So a specific AMD node? Not exactly unexpected, they've done this before and even TSMC does this for Nvidia at 12nm

>>61505738
14HP was always planned, this isn't like the Nvidia only node.

>>61505698
In that case Zen+/PR or whatever you call it in Q1 2018 might be even better than expected, if it comes with some low hanging IPC fruit fixed it will be stellar.

>>61505910
>if it comes with some low hanging IPC fruit fixed it will be stellar
Nah, it's a refresh on an improved node.

>>61500203
How does it compare in the professional market for people that don't necessarily require xenon or ands equivalent

>>61505940
You don't know that, B2 was fixing PCH/SoC bugs, PR might actually see some core improvements, AMD polished bulldozer every year.

>>61505960
They definetly need to fix L1 bug though.

>>61505960
Even without any core fixes the added clockspeed will be more than welcome.

>>61505984
Yeah, but AMD should really build momentum and prove it can gain decent improvements every year, god knows there's plenty of stuff left in zen to expand upon

>>61506003
They were doing decent job at polishing the turd that was Bulldozer.
I doubt they'll struggle at polishing core as good as Zen.

>>61506003
Apparently, the branch predictor on the zen architecture use a baked neural network to do the predicting instead of trying to make a super planned from the ground up solution that intel uses.
And the more AMD train this neural network, the smarter it gets and more it gets the prediction right, which in turn gives it a higher IPC.
They can also give it more "data" about the code running which well, makes the NN smarter at predicting shit, or maybe even being able to give the cache fetch thing a help.

File: debicki cute.webm (579KB, 990x1070px) Image search: [Google]

579KB, 990x1070px

>>61500865
>mfw remembering all the "well never get below 14nm or so shit"

>>61506187
The whole "twice the transistors every 18 months" is at risk, but you still can get improvements going, but at a slower pace.. until quantum physics fucks everything up.
Then shit starts to get 3D.
Actually, shit already started to get 3D.
Finfet, trigate etc are about making components that are smart than dumb transistors, rather than just trying to make the dumb transistors smaller.
We're probably going to have shit like "a whole XOR gate that takes the space of a transistor because its literally a mini tower"

File: 1500514601432.png (84KB, 653x726px) Image search: [Google]

84KB, 653x726px

>>61504473
M-MY CHERRYPICK IS BETTER!

>>61500203
Shouldaused a real language :/

>>61504829
>>61504870


VRMs

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

>>61503063
t. intel

>>61504540
yes and if you fuck up, buy ryzen

>>61504540
I'm pretty sure that's soldered.

>>61504553
I'd be really surprised if Coffee Lake S actually launches this year. It would show Intel is rushing shit out the door, and it would probably have a lot of issues at launch if that's the case.

>>61502428
>buys a $600 CPU
>buys a $200 Mobo
>only wants 16GB of cheapest ram
kek

File: the chernobyl of CPUs.png (214KB, 681x429px) Image search: [Google]

214KB, 681x429px

>>61502428
>Implying X299 won't explode in flames if you even think about using an air cooler

>>61504881
>>61503644
>>61505248

>>61500268
oy

vey

>>61502424
>Why's the Intel system been given twice the RAM?
Because it would just be stupid to only use two memory channels for a quad-channel CPU, and meanwhile 4 GB DIMMs aren't even that much cheaper than 8 GB ones.

>>61506244
>We're probably going to have shit like "a whole XOR gate that takes the space of a transistor because its literally a mini tower"
While I can see that being what the manufacturers would like to do, I don't think there are any plans for how to realize that. You'd basically have to deposit new monosilicon (or some other semiconductor) on top of already etched transistors, and there's no known way to do that, even in experimental phases.

>>61507619
IBM Nazi science will make it happen.

>>61507619
Well, we've already got vertical GAAFETs in the late prototype stage. If they can make those its certainly possible for them to extend the construction process further to actually build one vertically.

Actually doing it though is, like you said, a whole nother ball game.

>>61500203
Who buys Xeons for gaming? Retard pooloos?

>>61504848
10% faster but 50% higher power usage?

>>61507709
>ho buys Xeons for gaming?
Used xeons and chink s1366/2011 mainboards became pretty popular in Russia. I think it should also be popular on other poor countries.

File: 1499799036743.jpg (281KB, 1599x899px) Image search: [Google]

281KB, 1599x899px

>>61507709
X299 was marketed as being more than a rehashed Xeon, shill damage control squad.

>>61504856
It's quad channel, retard.
You wouldn't buy 4 sticks of 4GB of RAM. Almost no one buys 4GB sticks of RAM except for Pentiums.
2x8 and 4x8 are most common.

>>61507709
>Who buys Xeons for gaming?
Poor ruskies.
t. ruskie

>>61507774
Depending on funny business with TDPs it could end up drawing more than 95 watts. Not 50% more, though.

>>61505712
ah the good old days of thoroughbred b where amd just softlocked the 2000+ and you could oc it to 3200+ level that chip was the definition of insane

>>61507679
But even GAAFETs still depend on an underlying monosilicon substrate for the doped regions. Stacking transistors would require putting a similar substrate on top of already finished ones

>>61505529
>>61505656
This is my understanding as well.
Either pinacle ridge was canned, or moved back, since 7nm is coming sooner than expected.

The desktop market isn't that big, anyway. It makes sense that they'd focus on larger iterations rather than "release Skylake again".

>>61500203
delet

>>61507709
>not buying xeon, save $30 over i7 and use a desktop mobo
this was literally it until intel being a jew and shut it down.

>>61500203
>>61500249
Why is the difference between the 7800X and 7700K so large, when there's only 200 MHz difference in clock speed? Did the cache rebalance hurt performance?

File: GRBENCH.png (1MB, 2153x665px) Image search: [Google]

1MB, 2153x665px

>>61507709
Depends on the platform. X58 was a great platform to pick up used Xeons on, you could have started with a lowly quad core i7 920 @ 2.6 and ended with a six core Xeon X5675 @ 3.06 or higher. The Xeon, even when not OC'd, still performed very close to an i5 4690k in most games.

>>61508313
Mesh and cache changes.

>>61500388
and don't forget that a 7nm ryzen 2600 will be out by then

>>61502428
if you want to use quad channel you need does gb, also if you want to oc you need aio.

>>61500388
you will never see actual test from youtubers with systems like this, since according to them pc gamers have infinite budgets and it would destroy argument that 7700k is best gamer pc.

Anyway gonna wait for 7nm ryzen and see how that performs, if its good gonna buy that

>>61500249
>7800X ties 1600 at 4.7 GHz vs. 4.0 GHz
It's almost as if Zen has higher IPC than Kaby Lake, isn't it?

>>61509597
No, it's just BINGBUS

>>61509633
It's multiple things.

Low L3 per core
Low single core bandwidth
high core to core latency
higher memory latency

shitty power draw on top of that just adds coal to the fire.

>>61509633
Are you implying that the memory architecture is not an IP factor?

>>61509683
>IPC*

>>61509681
>Low single core bandwidth
How do you reckon, when Kaby Lake has 64B datapaths?
>high core to core latency
>higher memory latency
Still lower than Ryzen, though.

>>61509681
What a fucking disaster

File: Screenshot_20170722_211751.png (97KB, 1182x835px) Image search: [Google]

97KB, 1182x835px

>>61509739
There's massive regressions in low thread bandwidth.

>>61509774
Ah yes, I remember reading that. Has there been any explanation of that since then? 12 GB/s really just sounds ridiculously slow. For a 2.1 GHz Xeon, that's less than one 64-bit read per core cycle, and almost half of what just one DDR4-2666 channel should be able to provide.

>>61504870
>MSI B350 Tomahawk
why that mobo has hdmi/dvi etc out? Will there be a zen with an APU?

>>61509885
>Will there be a zen with an APU?
You mean Zen-based APUs?
Yes, Raven Ridge for AM4 is Q1 2018.

>>61509876

I think it has something to do with the partioned memory controllers, 1 core can probably only hit 1 channel for god knows whatever reason.

Honestly, who fucking cares about the relative performance of CPUs when they can reach 60fps in games. It only really matters if you have a 144hz monitor.

>>61509919
Well that's the point, though. As I said, it's almost half of what a single channel should be able to deliver. One channel of DDR4-2666 should give ~21.3 GB/s.

>>61509964
Future proofing?

>>61509980
There's no futureproofing hardware besides PSUs

>>61509998
Of course there is. If you have a CPU that's running current gaymes at right on 60 FPS, then it's not going to be able to handle future gaymes, whereas one running current gaymes at 120 FPS will.

>>61509980
MOAR threads is future proofing, not higher average fps dumbo.

>>61509998
No, motherboards can be reasonably future proofed as well. If you buy into a motherboard with a new socket, you'll be able to buy a much better CPU on that socket by the end of that sockets life span.

>>61510015
If you're buying a CPU for 120fps, you're probably buying it because you have a 144hz monitor. In which case you'll want to upgrade your CPU later anyways when it can't do 120fps in all games.

>>61510024
For gaymes, higher FPS sure is futureproofing.

>>61510039
>you're probably buying it because you have a 144hz monitor
Either that, or you're buying it for futureproofing.

>>61510040
For sub 80 iq people maybe. Ryzen has a 10% disadvantage in avg fps compared to 7700K, albeit at 50% of its load. Right now games cannot utilize 8+ threads, guess what happens in the future genius.

>>61509973
Sorry if I didn't make myself clear enough.

Skylake-EP has 6 memory channels partitions into 3 channels on the left and right edge of the die.
That means each edge of the die has a 192bit memory controller
What's likely happening is that 1-2 threads can only hit a single 64bit region of the controller, since it makes no latency sense for 1 thread close to the memory controller to hop to the other side of the die.
Problem is why is it only accessing a single 64bit PHY out of the 192bit available? Probably some kind of multiplexing you got in NAND controllers, one core can't make enough requests to parallelize the memory controller enough.


This is all speculation though

>>61510081
>Right now games cannot utilize 8+ threads, guess what happens in the future genius.
You don't actually know what happens in the future. Gaymes won't magically support n-ary parallelism just because you want them to.

I'm not saying to not buy higher-core-count processors, but it's not like relative performance of CPUs -- especially at equal core counts, as in the article -- doesn't matter.

>>61510111
Proper coded DX12 and Vulkan allow for api based parallelism in draw calls and what's not, and it's only uphill from there.

>>61510098
Still, though, one 64-bit channel at 2666 MT/s should be able to provide ~21.3 GB/s, so that doesn't explain why one core can't be provided that much.

Also, it's not like a core can just choose which memory controllers to use, as particular physical addresses are be necessity bound to particular memory controllers. I don't know exactly how Intel CPUs allocate addresses over the controllers, though.

>>61510052
Whatever floats your boat I guess. I'd rather buy an i5/Ryzen 5 today and again in 3 years instead of spending double the cost of an i5/Ryzen 5 today so it lasts a few more years.

>>61510137
Doesn't automatically mean the actual game logic lends itself to parallelization, though.

>>61510154
>I'd rather buy an i5/Ryzen 5 today
I don't even disagree, but that doesn't mean that it wouldn't be foolish to not overprovision some certain amount, unless you're really keen on not even being able to run next year's gaymes.

>>61510157
We went through the same dance when dual cores hit, when quad cores hit and when octo cores hit.

It was always the same thing.

>games can barely use more than 1 core
Now we got plenty of games using 6 cores and some even 10, those are rare though

>>61510157
You cannot achieve perfect parallization in a task as complex as a video game, what you can however is utilize more than 8 threads, something that hasn't been done yet. You're just engaging in demagogy.

>>61510191
>It was always the same thing.
It's a very different thing to expand parallelization from one thread to two, than from 4 to 16. It's not just "the same thing".

>>61510204
>what you can however is utilize more than 8 threads
whether you can do that or not obviously depends on the algorithms you have to implement for your game. For many algorithms, paralllelization isn't a trivial task, especially as the degree of it increases.

>>61510244
Give an example of unparallizeable task in a video game logic that fits in Amdahl's law. You're just talking out of your ass.

>>61510218
I already told you we went through this phase 3 times.

How much cores can game hit depends on programming languages and consoles, pretty much nothing else.
The rise of programming languages that give major focus to messaging protocols and IPC aren't for kicks.
Sooner or later games will start normally using 8 cores, it's just a matter of time when.

>>61510276
>You're just talking out of your ass.
So are you, though. Neither of us knows exactly what it is the developers of any particular game needs to do. I'm not even saying that it won't happen; all I'm saying is that it's naive to just assume that parallelization will be extended much beyond what they currently implement, especially in somewhat short order.

>>61510282
>How much cores can game hit depends on programming languages and consoles, pretty much nothing else.
Also the algorithms in question, though. If all algorithms were as embarrassingly parallel as graphics work, games would run on GPUs only.

>>61510327
I see no viable reason to assume efforts at parallization will abruptly halt given the current developments. Ergo, judging future proofness of a cpu by avg fps disregarding avg load is borderline retarded. That's all there is to say really, you can demagogue me all you want.

>>61510327
And programming languages define said algorithms, they're evolving faster than ever.
What was once possible to only run on 1 thread is possible to run on multiple, this is exactly what happened with Vulkan and DX12(shitty implementations aside)

Naturally not everything can be parallelized obviously, either for security or concurrency issues, there will ALWAYS be serial code, but that serial code is only part of a large codebase, it's not choking the CPU.

All in all, we're in for a uncertain future, nobody can predict market interest in new programming languages, nobody can predict what the next console SDKs will be like, but one thing is for sure, it's not gonna stay still.

>>61510363
Assuming perfect scaling is also borderline retarded, though. The reality is obviously somewhere in between, but there's also the fact that the serial performance of a CPU is a known factor whereas future utilization of gaymes is just hope.

>>61510400
>it's not choking the CPU
Except when it is.

>>61510450
Serial code by definition cannot choke a CPU, it's running on 1 thread out of a dozen

It's choking the game, but for a few miliseconds.
Either way you can go kicking and screaming about single core performance all you want, but you're not gonna get much if any improved single thread performance on a von neumann arch using silicon, that's a known fact not even AMD and Intel can easily overcome, some 10% difference between them which is wowfuckingnothing in the grand scheme of things is nothing but nitpicking.

>>61510530
>but you're not gonna get much if any improved single thread performance on a von neumann arch using silicon
It is at least, obviously, possible to compare the serial performance of CPUs, as evidenced by the video in OP.

>>61510550
And then AMD next year closes that small gap and where are you then? Praying for more clockspeed?

Nope, you're already at 5.0Ghz, silicon acts funky over those frequencies, physical issues start hitting.

>>61510585
Life of an intel fan must be pure suffering right now.

>>61510652
Right now they're praying the 8700K is going to let them OC to 5 GHz (it won't) and will be released this year (7700K was released in February so not likely)

>>61510585
>And then AMD next year closes that small gap and where are you then? Praying for more clockspeed?
What makes you think that? I wasn't ever arguing that single-threaded performance would be seeing any huge jumps in the future. I was only arguing that a CPU with better single-threaded performance is better than one with worse.

>>61510909
All else being equal, of course.

>>61510759
Seems like they're praying that 8700k is 6-cores 12-threads, clocks to 5GHz, and doesn't light their system on fire.

>>61509964

Nigga most gamers buy 144hz now. I'm even waiting on 144/4k

>>61511115
I doubt 144hz even has 10% market share.

>waiting
Exactly

>>61503035
each die costs them roughly 100$ to manufacture (total cost).

Even the lowest of the low end R3 costs 117, that's a 17% margin on that alone, much less on the high end R5 (100% margin) or R7 (300% margin).

And then there's the server chips which they are going to sell an obscenely greater number of. The 7601 is marked up 900%, even the lowest end chip has a 75% markup.

So what's that about not chasing 60% margins? The only one fucked in this scenario is Intel, who's struggling to manufacture Xeon dies and has to sell them for massive prices to make it profitable.

>>61512537
>that EPYC margins
AMD is going to roll over in dosh veeeeeeeeery soon.

>>61512537
there's the multi-billion dollar introductory cost but once you get over that it's just 80-100 dollars per die produced

ryzen mobile looks to be much cheaper to fab since it's 2/3 the size (cost of fabbing a die is exponentially proportional to area)

>>61513369
Ryzen Mobile is actually bigger afaik.

File: AMD_Threadripper_RyzenMobile_800.jpg (65KB, 800x533px) Image search: [Google]

65KB, 800x533px

>>61513405
it's around 2/3 from what i can see
image is ryzen threadripper (same socket size as epyc) on left, ryzen mobile on lower right

File: AMD-EPYC-Four-Die-with-Infinity-Fabric.jpg (36KB, 957x539px) Image search: [Google]

36KB, 957x539px

>>61513405
bare epyc die shots (same socket size as threadripper)

>>61513504
>>61513515
It's definetly looks *wider* than Zeppelin.
Probably the same size actually.

>>61500203
Who cares man if you're playing at 60hz get the bentium. Can do 4k for $1k with a g4560 and 1080ti. If you're at 144hz like a real nigga you delid a 7700k with a good cooler on it and buy some fast fucking ram.

>>61513927
Or you just buy a ryzen 1600 with stock cooler and be done with it for a long time
Pentium g4560 is one of the lowest spec CPUs you can buy today. Its decent for up to 60 fps gaming of today. But as we saw with the G3258 dual core it will require an upgrade much sooner than other chips.
Spending ~130$ more on the 1600 is well worth it for longevity. Considering consoles use 8 cores these days you will start to see game engines requiring 4+ cores

>>61513927
>G4560 and 1080ti
That piece of shit falls well under 60 FPS in multiple titles. Just because it can hit 60 FPS average doesn't mean jack shit if it constantly dips under or drops frames massively because it hits 100% CPU. What a fucking joke. It's only good in an absolute budget build with a 1050ti or something.

File: 2600k-revisit-gtav.png (69KB, 806x544px) Image search: [Google]

69KB, 806x544px

>>61513927
>you delid a 7700k with a good cooler on it and buy some fast fucking ram.
Everything you say is wrong and/or retarded.

1. OCing a 7700K, even to crazy high amounts, does NOT give enough of a FPS increase to justify the risk and cost of delidding and watercooling
2. Fast RAM barely makes a difference for Intel compared to AMD. You're just wasting more money.

File: 7ae8ef32c12e7072019d6ea3d9d08319.png (131KB, 1449x845px) Image search: [Google]

131KB, 1449x845px

>>61514446
whatever you say poly the parrot. stay mad
>>61514474
well here i made this it's just labeled wrong. Bottom is stock 6600k ddr4 2133. top right is stock 6600k ddr4 3600 cl16, top left is OC 6600k ddr4 3600 cl16.
it's okay to be upset that other people might want the best possible hardware for high refresh but it's a very specific use case don't let it get to you. A nice 7700k at 5.3ghz with ddr4 4266 cl19 on an asus apex with a 1080ti on an aio is the way to go for high refresh anything else is inferior and it's not even that expensive

>>61514524
>WoW
Oh no, it's retarded.

>>61514596
STOP LIKING WHAT I DONT LIKE