Say sorry, faggot. "They basically deactivated 1/2 of

File: 8Rch6JF.png (55KB, 1509x905px) Image search: [Google]

55KB, 1509x905px

>>59268032
Finally, good that somebody did that test I looked for that in day 1 benches and found nothing. Still disabling cores not improving casual overclocks, it's pretty much confirmed that is a silicon limitation unless you have exotic cooling.

File: 1488620340333.png (224KB, 1274x695px) Image search: [Google]

224KB, 1274x695px

What remarkable and trustworthy benchmarks showing a 50% ipc lead over kabylake. Wow! It's like AMD didn't even know they could do that, even after confirming with GN that Ryzen was 9% behind kabylake IPC.

How did they fix the SMT ?

>>59268195
Why do you think they did? Hypothetically disabling cores would limit windows scheduler moving threads between cores and lessen the effects of bad SMT management

>>59268154
>FAAAAAAAAAKE
>REEEEEEEEEE!!!!!!
Did you even read the image?

>>59268032
Wait, I am confused. Is Single-Thread better in this image or not?

>>59268032
anybody with a brain knew that low clock high core R7 wouldn't compete with 7700k in gaming. this was never the case.

the R5 at 4.0+ GHz is going to be the ass kicker. I really hope that the 6 core 12 thread can reliably clock high

>>59268372
Much faster in gayms.

>>59268295
all game tested use 8 core ? i would expected if game use 4 core or less it would be same problem as now

>>59268392
Well, Battlefield 1, aka one of the most reliant games on multi-cores. Say I wanted to play Yoda's Stories. Can AMD run it as well as Intel? These tests don't really tell me anything. Also, is Ryzen even stable with Windows 8.1, aka the last good version to play Yoda's Stories on?

>>59268457
>http://www.zolkorn.com/reviews/amd-ryzen-7-1800x-vs-intel-core-i7-7700k-mhz-by-mhz-core-by-core/3/

Read the fucking article.

File: Russia.png (41KB, 354x1091px) Image search: [Google]

41KB, 354x1091px

>>59268475
Holy Russian text Batman!

>>59268475
I don't get how this tells me if I can play Yoda's Stories. They keep talking about modern games with better core capabilities.

File: 23uyhjftg9fbiv3qqt.jpg (48KB, 228x130px)

48KB, 228x130px

>>59268475

If a cpu with 8 cores gets too hot with high clocks, why don't they separate the cores into two clusters of 4 and have a decent amount of physical distance between them?

>>59268154
>he posted the fake benchmark again

Joker literally had to redo that benchmark because it was inaccurate: https://www.reddit.com/r/hardware/comments/5xcnye/720p_r7_1700_vs_7700k/

>>59268154
in integer and fpu it leads on IPC that's what games use

also it confirms that L3 and SMT cache do not work at all yet

10 ip's ITT and only like 3 of them had read the image and or article...

>>59268032
Min fps
Average fps
Max dps????
Either shit benchmark or fake

File: AMD-Ryzen-7-1700-Delidded-1.jpg (73KB, 800x699px)

73KB, 800x699px

>>59268555
see this >>59268120
voltage vs frequency doesn't scale linearly after a point so going above 4.0Ghz isn't logical for performance given the wattage used. Also Ryzen has 2 core complexes that are already somewhat separate

>>59268599
It is hard when it is in Russian and talks about DX12/11 games. I don't see Yoda Stories or any equivalent I can compare it to.

>>59268608
Buy AMD, it makes you deal more DPS than Intel!

File: 1487098423024.jpg (52KB, 350x443px) Image search: [Google]

52KB, 350x443px

>>59268646
>all those bubbles in the tim

I'm expecting the 6c/12t Ryzen to be unusually strong compared to the 4c/8t and 8c/16t at single threaded tasks.

>>59268387

It won't.

It's the same chip as the R7, same process and everything.

4.1 will be the limit, and 1.45v will be the most it can take.

What it will do, is seriously undercut the 7700k in performance per dollar.

>>59268671
It's not TIM, Ryzen heatspreader is soldered to the die. They have to heat it to a certain temp and carefully to remove it in order not to destroy the die. Obviously the process won't end up in a smooth surface after removal

>>59268671
It's not an air bubble, it's soldered on and part of the solder broke off and stayed on the IHS when they removed it.

>>59266666

These benchmarks may not match actual R3s and R5s.

Basically the 8c/16t dies have to be disabled in twos, 1 per core complex so each side of the die is in balance. Here the two cores that are left in each core complex have access to more cache than they would normally have so they perform unusually well. Also since there's fewer cores in each complex they have more headroom for overclocking while keeping the power under control. You can expect the 6c/12t variant to perform similarly to this but maybe not quite as fast. The 4c/8t variant which I believe is only one 1 core complex wouldn't allocate cache the same way that these chips do so it wont necessarily perform the same.

>>59268457
Unless someone can prove here that Ryzen allows for more copies of Yoda's Stories on your computer without lag than Intel, I see no reason to purchase it.

>>59268933
You can't purchase a 4c/8t Ryzen chip currently, there is nothing to prove.

>>59268933

Indiana Jones Desktop Adventures was superior, so no one gives a shit.

>>59268989
You can disable 4 cores... that is what the whole topic here is about.

>>59268032
delet this goyim

>>59268996
Yoda Stories was better. It had high scores so I felt I amounted to something when I stomped myself.

so when half of the cores are disabled the ryzen is faster than on 8 cores with the same 4ghz oc?
that makes no sense at all

>>59269007
You can't "prove" it, it's a simulation based on what is available. See >>59268843

It's possible that AMD can tweak microcode or something else to change how a 4core will work with caches etc

Even though this doesn't necessarily reflect the true performance of the R3s and R5s this is still a big deal because the lowest 8 core chip also happens to be around the same price range as Intel's 4 core chips so you have a chip that can either be for a workstation or a gaming chip depending on how you configure it. R7s are more versatile.

>>59269041
maybe it has something to with how the zen die configuration is since its two quad core modules.

>>59269041
It makes sense when you consider how bugged bioses and Windows are at the moment, especially concering Zen'a cache(win reads 170mb lol), disabling a single ccx makes it look more like a traditional Intel core complex

>>59269041
it somehow fix the Smt, and 7700k is not it real clock.

>>59268032
DELET THIS
This board is for comparison of 4c i7 and 8c r7 in single-threaded applications!

>>59269050
Hmm, I guess I will continue on Intel until R5 gets released. Then we shall continue this debate and find out what the superior CPU is.

>>59268032
Does the performance increase have something to do with the cache?

>>59269079
>>59269090
so why did literally no one make this test before? it's not like it was hard to disable 4 cores

>>59269121
It takes a couple days to install Gentoo on two computers.

>>59269121
Try to get fast review for more click ?

>>59268496
that's not russian you uncultured dipshit

>>59268843
This

Unrealistic as fuck

>>59269041
Two things to consider.

Less power drawn and less heat, silicon will perform somewhat better in terms of efficiency and keeping things stable.

Windows scheduler performance differences due to moving threads between cores/logical cores

>>59269121
Watch this video
Wendell and other youtubers. They have different priorities in testing
https://www.youtube.com/watch?v=bIbwuLdHbMg

Wendell also talks about hypothetical future 4 and 6 cores and windows scheduler issues in that stream

>>59269121
Dunno, just like no one tested a cTDP 35w 8 core ryzen, but it's still there, only 25% slower in mulithreaded work than the regular 65w 1700.


Neat, huh?

>>59269121

You expect competence from youtubers?

>>59268032
>that farcry benchmark
Jesus. Intel BTFO.

>>59268843
You cant know if the R3s will have a ccx disabled or 5wo cores per ccx disabled.
The latter is better for yields and power consumption might I add.

>>59269145
Ah, United Arab Emirates. Even better.

File: Screenshot at 2017-01-04 23:47:01.png (7KB, 383x399px) Image search: [Google]

7KB, 383x399px

>>59268032
4c and 6c Ryzens won't have 4 GHz base clock tho.

Basically 1600X is 1800X with 2 less cores.

Intel has more experience on the 14nm process and they can clock higher as a result. Removing cores won't make r5 & r3 clock as high as Intel.

>>59269172
Two CCXs like on the R7s would lead to performance exactly like in the OP but I don't know if AMD would do that. It's definitely possible but a single complex die would be smaller and have higher yields so it makes sense for AMD to go that route too. The 6c/12t is for sure going to be two complexes though so definitely pay close attention to that one when the benchmarks come out.

>>59269201
are you sure?

>>59269228
It's either gonna be a summit ridge die or a raven ridge die(apu), their size won't be all that much different to affect yields.

>>59268572
This 720p tests just shows a very underused cpu, you can see percentage in every thread of Ryzen being always lower in 7700k, I ask why. It should at least use the first thread more, and if it was slower, it should be using its threads even more than 7700k while delivering less fps.

Ryzen have slightly lower IPC than Kabylake but its higher L2 cache helps get on kabys feet.

>>59269261
Now there's something I hadn't considered yet.

R3s might end up being Raven Bridge cut down while the R5s might end up being cut down R7s with two mostly intact CCXs

>>59269239
1600x has the exact same base and boost clock

>>59268295
> moving threads between cores
between CCXs.

switching between cores ain't a problem as long as it stays in the CCX.

>yet another thread dedicated to gaymers on /g/ arguing over some fucking hardware brands
losers tbqh

>>59269339
Most of the thread looks cautious, only the first few kneejerk posts are gaymen related.

>>59269339
>yet another "you can't mention gayms on g" post
you forget that a good chunk of software slows down because of this too

>>59269360
The most analitical guy strikes again
https://www.youtube.com/watch?v=ylvdSnEbL50

Look the FX 8XXX getting ahead of the 2500K on the years goes by. FineWine Technology, bitches.

>>59268032
https://www.youtube.com/watch?v=ylvdSnEbL50

File: intel-post-ryzen.png (243KB, 800x612px) Image search: [Google]

243KB, 800x612px

>>59268032
LMAO Intel corelets btfo!

>>59269396
>>59269392
No, this guy is clueless.
Surprisingly the only ones keeping some objective and technical outlook on Zen so far are the Stilt, Wendel and Patrick Kennedy

>>59269360
These results are still impressive I mean the R7s are pretty close to Intel's 4 chips so if you're willing to disable 4 cores you get at least equivalent performance if not much better with the ability to enable more cores when necessary. It at least can provide an end to those dumb comparisons to the 7700K I hope.

Now we wait(tm) for the BIOS and scheduler patches.

>>59269449
It'll end up being compared to i3s OCd to 5.0 in ST tasks and 2 cores will be the new 4.
You can't win.

>>59269442
How is he clueless, what the video, he only shows data from reviewers, nothing pulled out of his own ass.

>>59269442
I was really disappointed in Ian from AT in the last podcast.

i dont care about oc so i will just get whatever is better with default settings if i ever need to upgrade from my i5 2500k

>>59268843
>Basically the 8c/16t dies have to be disabled in twos, 1 per core complex so each side of the die is in balance.
no they don't. you can disable 1+1, 2+0, 2+2, or 4+0, or 3+3, or 4+2 in BIOS.

>>59268032
You realize that due to the arc you have specific cores to disable right? None this shows shit because you have no clue if the correct side is disabled,and windows think all smt cores are real

>>59269575
I think he meant from a manufacturing standpoint

>>59269601
In this case I meant from the BIOS standpoint. I didn't realize it was even possible for asymmetrical configurations like 2+0 to work.

I don't know how much this changes, I assume they probably configured it 2+2 because that's likely the best option.

File: stock-photo-businessman-with-thoughts-of-suicide-86941702.jpg (27KB, 298x470px) Image search: [Google]

27KB, 298x470px

>>59268032
DELETE THIS

DELETE THIS

DELETE THIS

File: 640x480.jpg (148KB, 680x486px)

148KB, 680x486px

>>59269442
No. he just btfo low res testing. I use to believe it back then. It absolutely true if 1 core vs 1 core during pentium era , that not how it work today. yet review still use it without question the validity of it

>>59269309
I shouldn't hope so. Raven Ridge will probably lack the level 3 cache, so any Raven Ridge sans IGP will probably stick to the Athlon branding we've seen them use.

>>59269685
unrelated to the thread but I think this is the best version of this comic I've seen yet

>>59269685
Best OC on /g/ in the last 3 years.

Sadly.

File: 1f602.png (25KB, 160x160px) Image search: [Google]

25KB, 160x160px

>>59269392
>The most analitical guy
Oh please, this shithead was on the "the rx480 will be between a gtx 980 and 1070 and will overclock really well" hype train.

>>59268154
91% of Kaby Lake would still be too damn high to be real

>>59269685
That signboard "muh ghz" lol

>>59269685
400 × 240 is the next big gaming hit resolution?

>>59269799
Gorillion FPS

>>59269702
Bulldozer had pretty bad latency in the L3 cache and it didn't get utilized much so maybe it made sense there but I think the L3 contributes heavily to Zen's general performance so I don't think they would cut it out completely. Maybe they would cut it down.

>>59269744
and whats your consensus about it?

when the game is properly optimised, like Doom, the rx480 shits on the 980

>>59269702
Not worth the effort, the core complex already has its own photomask developed, no need for a new one.

>>59269948
>when the game is properly optimised
I don't buy hardware because it theoretically has better performance than the competition if you tailor your use case scenario to the hardware. I buy hardware because it fits my use case scenario.

To me, the rx480 is a gtx970 with more vram.

>>59269991
i buy hardware for the act of future proofing the games i wanna play.

going amd is always better than intel or nvidia in this regard because they develop the future as we've seen in the past

>>59269912
You say that but apparently even carrizo couldn't compete with piledriver in terms of IPC because of the lacking l3 cache. Also, it's very possible that it might be a result of giving more space to the IGP as the l3 cache is rather large.

>>59270004
By the time Bulldozer became usable, it was outdated and a shitty value. I don't spend my money based on speculation and faith.

>>59268032
so they underclocked the intel chip and overclocked the ryzen chip, another bullshit benchmark.

not sorry.

>>59270027
>but going to churche is perfectly fine

>>59270039
>implying I'm not a fedora tipper

Too bad the R5 runs at 3GHz

>>59270030

Furthermore the R5 line will have a base clock around 3 GHz and the same architecture as the R7. The 8 core 4 GHz R7 can't beat Kaby Lake and the 4 core 3 GHz R5 won't as well.

>>59270030
Theyre showing performance clock per clock just for comparison. Ryzen has 4 cores and 8 threads disabled.

>>59270167
If it did it would use 25 watts.

>>59270335

>What is scaling

Do you honestly believe real word values scale linear ?

File: Screenshot from 05-03-2017 23:38:48.png (62KB, 957x532px)

62KB, 957x532px

>>59270475

File: Screenshot from 04-03-2017 03:47:54.png (54KB, 1489x889px) Image search: [Google]

54KB, 1489x889px

>>59270614
Zen's voltage scaling IS linear until 2.2GHz

>>59268154
>considering a stupid gaymen benchmark ""ipc""

kys

File: amd_shills_on_suicide_watch.png (148KB, 1643x853px)

148KB, 1643x853px

>>59268032
You were so disappointed in your own shilled product, you had to fucking PHOTOSHOP THE RESULTS.
pic related, it's the real one.

You didn't thought somebody might check out the source, did you?

Well, you were almost right, 110 post in and nobody had a fucking clue.

File: ryzen-die-guru3d.jpg (2MB, 1920x847px) Image search: [Google]

2MB, 1920x847px

>>59268032
Why the hell did you alter that Far Cry result? Theres no reason for it.

>>59268555
I can't tell if this post is satire or not.

>>59268032
>They tested the 7700K at 4Ghz for IPC reasons only.
Yeah let's test a cpu that does 4.8 ghz on 4ghz, look how competitive Ryzen is now

File: wewladyousureshowed'em.png (225KB, 1628x1559px) Image search: [Google]

225KB, 1628x1559px

>>59268032
>>59271330
>>59271418
Here's a not fake version of the tests.

Warning: it might trigger seizures if you're an amd fan and cannot understand how ryzen 7 is not made for gaming (even though that's what the hype was about, but maybe after 10 fails, you will learn not to trust amd's marketing hype)

>>59271848
So when the actual 4c/8t versions come out that are significantly cheaper than the 7700k you're going to start shilling for those, right? Since based on these tests they will be an objectively better value.

File: ryzen_delid.png (542KB, 1280x720px) Image search: [Google]

542KB, 1280x720px

>>59268646

>Also Ryzen has 2 core complexes that are already somewhat separate


Errrr, nope.

>>59272044
You can't see them on the naked die by eye, you silly.

>>59271418

It's the two big yellowish parts separated by the middle.
That's a core complex.

File: wtf.png (3MB, 1440x7381px) Image search: [Google]

3MB, 1440x7381px

>>59271330
>>59271418
>>59271848
Huh? Look at the browser caches, chink blog silently changed their results

http://cc.bingj.com/cache.aspx?q=www.zolkorn.com%2freviews%2famd-ryzen-7-1800x-vs-intel-core-i7-7700k-mhz-by-mhz-core-by-core%2f3%2f&d=1579601384125&setlang=en-US&w=CGWV6fDMCupgMHeH0v_uXQ0K_IKUsQHk

https://archive.is/JLOCH

>>59271915
I don't think I'll care at that point. I'm using the best processor on the market today and have been for over a month. If they can finally play catchup good for them. They will also have cucked their 500 dollar part, that is amusing.

I'm hoping the new Intel X's are interesting a value. If not maybe I'll grab that cucked 1800x, I could actually use second box that can do faster 4k rendering.

>>59273031
Wow, it's literally the same shit at that clockspeed with Zen wining some points due its bigger cache.

I still don't know what's the use of a 8 core CPU with 4 disabled cores, just get the 4 core chips then, they'll probably cost some 150$

File: AMD-Ryzen-7-1700-c-ray-Benchmark.jpg (67KB, 849x514px) Image search: [Google]

67KB, 849x514px

>>59273286

I think the point is to estimate what the 4-core Ryzens will perform like.

Hell, if the Ryzen 3 lineup at 65W (or maybe even lower TDP) overclocks anything like the R7 1700 (from 3GHz to 4GHz) you'll have 1800X gaming performance at a price of $100-$130. THAT would be value for money if all you do is gaming.

The R7 1700 is already great value if you do ANY server workloads at all. Look at this shit

File: AMD-Ryzen-7-1700X-OpenSSL-Sign-Benchmark.jpg (77KB, 858x586px) Image search: [Google]

77KB, 858x586px

>>59273586
>>59273286

File: AMD-Ryzen-7-1700X-OpenSSL-Verify-Benchmark.jpg (82KB, 854x588px) Image search: [Google]

82KB, 854x588px

>>59273286
>>59273586

>>59273586
Can't estimate because we're still in fresh platform software hell thanks to MS and motherboard vendors on top of zero support for Zen's unique arch from the dev side.

File: AMD-Ryzen-7-1700X-Python-Linux-Kernel-Compile-Benchmarks.jpg (56KB, 884x568px) Image search: [Google]

56KB, 884x568px

>>59273660
>>59273640
>>59273586
>>59273286

>>59268496
that's the language of the street shitter

>>59273674

It's really AMD's fault that motherboard vendors were not given enough time though. Whether or not they told MS about the SMT and CCX issues I have no idea. Maybe they did, and MS is just slow. It is possible MS is just slow given that Linux 4.10 already fixed the Zen core topology.

>>59273586
I already saw those STH benches, proved quite enlightening on the design of these chips.
I mean shit, the chip has got multi-socket links on (GMI), it has fixed function blocks for AES, it scales down to 2.2GHz in a linear voltage curve..

This is as much as a gaming CPU as a Xeon-D is.

>>59273725
The thing is this seems to only affect ASUS and MSI.
Asrock and Gigabyte don't seem to have issues.

And as for MS being slow, nothing really new.

>>59273735

Yeah, absolutely. Zen was designed as a server ship and they said as much themselves when they announced they were working on it.

http://wccftech.com/breaking-amds-gen-x86-high-performance-core-code-named-zen-debut-k12/

>There’s very few people who know how to create server chips, you know Jim Keller has a lot of experience in that space.

Ryzen IS a server chip sold to enthusiasts. It was much later that they announced that their roadmap changed, and they'd launch consumer chips before Naples (maybe for validation? recover brand name?)

>>59273735
I've been telling people this is mostly a way for AMD to do validation faster for the upcoming Naples launch, if they managed to get some nice Revenue out of it it's a double win.


From what I've read these things are selling quite well, it seems everyone's waiting for motherboard vendors to pick up the slack since AM4 boards are out of stock everywhere

>>59273754

Well, programming relies on assumptions being made. It is quite possible that the Gigabyte and ASRock teams were just better at reading the specs and determining their assumptions from that. ASUS and MSI just plain fucked up.

>>59273754
It didn't help that AMD found out late about ASUSs fuckups AFTER it sent the review kits(with ASUS boards), leading to last minute calls and panic.

x99 was pretty bumpy when it launched too, give it a month or so.

>>59273801
>From what I've read these things are selling quite well, it seems everyone's waiting for motherboard vendors to pick up the slack since AM4 boards are out of stock everywhere

The Ryzen 1700 was stocked today by many internet vendors, and it immediately sold out. The only Ryzen still in stock here is the 1700X as the 1800X also sold out.

The Ryzen 7 1700 is by far the best purchase though.

>>59273841
Completely agree for the 1700, I'd only get the 1700X if I never wanted to overclock by buying some cheapo board, but that's a waste.

Also the 1700/1800X are half the perf/watt of the 1700 due to their base clocks being higher than Zen's sweet spot on the voltage curve.

All in all 1700 is the best chip, overclocks to it's X brethren level and if you want a efficient running chip again just turn off the OC.

>>59269145
Yes, it's not russian runes. Russian language sounds like oink-oink.

>>59273754
>just ordered a MSI motherboard

I dun fucked up, right ?

It was the only one available anywhere ;_;

>>59273937
MSI has mostly fixed its issues but it's still lagging in performance (some 5-7%) compared to Gigabyte, that was some 20 hours ago I haven't checked if that changed.

The only real fuckups are with ASUS boards.

>>59273963
Sasuga ASUS


>fucks up X99
>fucks up AM4

These guys hate HEDT

>>59274005
ASUS? More like ANUS amirite

>>59273963
Dang, hope it's only a software fuck-up, I'll be bummed if the hardware is defective.

>>59274033

If it is a hardware issue, then they'll definitely send you an RMA'd board like with other troublesome platforms in the past.

>>59268496
Thats thai, the exact opposite of russian.

>>59274073
Oh yeah, but it'll still bum me out, disabling my main computer for a while and basically making me do a full re-build, which is somewhat tedious.

>>59268572
The 720p benchmarks are fake too. He used a 7700k at stock clock as opposed to what's shown in the video vs an OC'd 1700.

File: 1422055730444.jpg (265KB, 955x951px) Image search: [Google]

265KB, 955x951px

I really want to buy a 1200X and overclock the fuck out of it.

>>59268032
so based on this
ryzen is 0-30% better
but everything else seemed to show worse/equal/minimally better gaming performance

so amd once again didn´t manage to release a product that is not gimped at the beginning?
acquiring ati was truly a mistake seeing how their practice of
>it´ll get better later
seems to fall onto their cpus now

>>59274149
4.2 max

File: samefag.jpg (221KB, 600x799px) Image search: [Google]

221KB, 600x799px

>>59269716
>I made the best OC on /g/
>2 days later I become sick and tired of /g/
That's just my luck senpai

I don't expect the 1600X to be good at going past 4Ghz at all.

And Q2 release most probably means end of June.

The problem with this OP is that 4ghz is basically the max of ryzen while it's stupidity to run the i7 7700k at 5ghz.

If you care about performance you should care about overclocking. Ryzen overclocks like cocks and until they fix this, the enthusiast chips will always be worse for games.

>>59274955
Base stock clocks are important here, not how much it can turbo on one core.
It's a win if its stock clocks are close to 4.0, even if it means having a 95W TDP

>>59271915
The Ryzen 3 series will be released in H2.

>implying AMD is not a profit driven company and will try to get the most money with the 8 cores first

Problem is that by the time they are releasing it, Intel will have released their own.

>>59274993
>not buying the $130 4/8 SKU and overclocking it to 4.0

Beat that value.

>>59275013
SKylake-X has already been delayed, but most people put it somewhere between Q2 and Q3

>>59273219

>their $500 part is cucked
>while it outperforms Intel's $1000 offering in the tasks that both are intended for

Sure thing

>>59275104
No such thing exists.. tasks that are done on more cores can be done much faster on GPUs

>>59268032
>Literal poo page

>>59275170
Better dial up Intel and tell them they should shut down their 5 billion USD revenue Xeon business because Teslas are locked and loaded.

I'm sure they'll call Nvidia after, them that fairy tale and both will have a hearty laugh.

Jesus you'd think people on the technology board were above youtube-tier comments.

File: confused_little_chibi_by_xiorin-d6v4ofz.jpg (36KB, 636x1254px) Image search: [Google]

36KB, 636x1254px

>>59268032
So does this suggest that maybe Ryzen is massively memory bottlenecked? Or is it more to do with how the two CCX's communicate? I'm not sure what else what explain such a jump in IPC just by disabling cores. Something doesn't add up here.

File: 1488475707479.png (133KB, 317x277px) Image search: [Google]

133KB, 317x277px

>>59268154
R7 1700 vs 5GHZ 7700k on equal footing
Holy shit.
That's pretty rad.

>>59275170

Not every task can be offloaded to GPUs you fucking idiot. The only tasks that can be offloaded to GPUs are non-branching tasks. The only use for SIMD on a GPU is if you're feeding it a huge amount of data that all needs to be operated on in exactly the same way. It's fine to do branching on a GPU, but the one moment that there's a single conditional that modifies the execution path for a small subset of the data you fed, you have branch divergence and that is a problem. In such workloads (and there's a HUGE amount of those, probably like 60% of all HPC workloads) you are far better off with a GPU, and THOSE workloads what Ryzen 7 is targeting.

>>59275319
>>59275170

s/you are far better off with a GPU/you are far better off with a CPU/

>>59275319
GPUs are much faster at productivity than CPUs and this is a fact

>>59275499

No, they aren't. I just told you where they're not. GPGPU is extremely powerful when used correctly, but a lot of workloads simply don't allow for SIMD parallelisation the way a GPU demands. In these cases, CPUs are far better. The moment you have branch divergence, you have performance cut in half, and when branch divergence is a constant of the workload, a GPU is just not viable anymore. You'd do far better with a heavily threaded CPU in that case.

What I am saying is a fact, your bullshit is not.

File: tesla-k80-performance-acceleration.png (63KB, 511x416px) Image search: [Google]

63KB, 511x416px

>>59275543
So desperate shill

>>59275499
>>59275585
>>59275543

>>59275543
>>59275499

Also, some GPUs don't allow for branch divergence AT ALL and the program will crash if it happens. Both AMD and Nvidia have implemented serial execution of all branches, but that comes at a significant performance loss. If there are two branches of execution, you'll already have 50% performance loss, and if there are three, well, fuck, you're down to 33% performance etc.

Any workload with massive branching will perform MUCH better on a CPU. The majority of HPC workloads need good branching performance and so they're done on CPUs and CPU clusters. If that wasn't the case, why in the fuck would people still build new CPU clusters?

>>59268032
APOLOGIZE

>>59275614
Shitty coders can't handle GPUs, in a few years GPUs will do everything CPUs do and much faster

>>59275632

Go ahead and write an OS using GPGPU then and then tell me that it's viable.

You don't know what the hell you're talking about.

>>59275665
see
>>59275585

>>59275676

>cherrypicked workloads that just happen to do well on GPGPU because they require minimal branching

ok

>>59275695
No argument? No problem, I specialize in owning AMDcucks

>>59268032
>Say sorry, faggot.
Die in a car crash, you stupid son of a bitch

Yo, dude.
Nvidia sells under 16k Teslas a quarter, Intel sells that many Xeons with a equal $6k pricetag in a few days.

You're literally retarded.

>>59268032
What this indicates to me is that if Ryzen drivers and windows Scheduler were working together properly, Ryzen could turbo to 4.1ghz on 2 cores and should perform at least within like 3% of the 7700k worst case scenario in almost every game.
It's clearly not an architecture issue or clock speed issue. 4Ghz is lower than the 4.1Ghz the 1800X will turbo up to on 2 cores.

Instead it's like 10-20% behind in some of those games.

File: 1484111197324.jpg (263KB, 800x601px) Image search: [Google]

263KB, 800x601px

>>59268154
>he doens't know what IPC means and how it translates to games

>>59276337
IPC isn't static, it's workload dependent.

>>59268032
NOOOO NOOO FUCK YOU OP.

My 7700k still good for 2-3 years r-right?

That fact that Ryzen is going toe to toe with Intel on the same cores and clock shows just how well this will perform once optimizations are put in place for the extra cores and memory speed issues are addressed.

http://www.youtube.com/watch?v=ylvdSnEbL50

>>59269744
The RX480 is pretty damned good now after driver and game updates. It beats the GTX1060 in most instances which is the market it was aiming for.

>>59276356
Exactly. That's what the person doesn't understand.

>>59273881
There's a lot of people getting the 1700 to 4 GHz on B350 boards
This is the new 2500K

File: ysdesert.gif (40KB, 525x348px) Image search: [Google]

40KB, 525x348px

>>59268457
>Yoda stories
>performance mattering
You're gonna be at however many frames your monitor can do m80

This is my favourite star wars game

>>59275733
What the fuck, man.

>>59268032
I don't get you, OP.
Either shoop the image and not link the source or shoop the image and link an altered source with the shooped images.

>>59281274
Nothing is shooped though, source changed their charts.
>>59273031

>>59281328
Alright. Even the fixed ones look nice for the 1400X considering it's cheaper than a 7600k.

If games were using AMD's SMT as well as HT, things would be even closer.

>Games.
call me when you test something relevant.

>>59275253
>Or is it more to do with how the two CCX's communicate?
This. I'm pretty sure this is also what causes the memory and L3 latency issues but I don't think anyone has tested it yet.

>>59275585
>Throughput
So completely irrelevant to any workload then.
>>59275676
Fun fact, if we're just measuring e-peen with throughput, AMD is king.

>>59271915
I'm not shilling for either company. I simply stated the facts: OP's image is fake (or the source changed it later), and Ryzen 7 is definitely not for gaming.
If I'd work in the media, I wouldn't think twice buying a 1800x for rendering. price/performance ratio is almost double of intel's.

Buying hardware based on brand is the stupidest thing anyone can do, I know most intel and amd shills on /g/ are just baiting, it's very obvious.
For us, consumers, the best thing can happen is ryzen 5 beating intel, that just means better price and more performance for us. At least intel will finally start doing something, and we might seem better improvement in performance than in the last couple of years.

When I'm in need of an upgrade, I check every option out there. I have a 6700K right now, and I expect to use it for at least 6 years. At the time it was the best option.

6 years later I will check my options, and will either go with AMD or INTEL, based on which suits me better.

>>59283746
>(or the source changed it later)
see >>59273031

This is an extremely complex CPU release, the only thing that's agreeable universally is that Intel is shaken. That's more than enough for me.

>>59283960
Intel's enthusiast market is ded. Their server/enterprise market is probably disrupted because AMD is set to release Naples in a few months. Probably similar deal on the OEM front AMD has cheaper and competitive chips. More computers being sold is always nice for them.

>>59284044
The only reason to buy intel soon will be the G4560 and 7700k. Two CPUs out of a lineup of like 200.

Kek.

>>59284159
>G4560
If those 4 core R3's cost around $120 or lower there won't even be that.

>>59284044
>>59284159
Intel will still be in almost every Dell, Hp, and Lenovo pre-built computer since Intel will cut prices in half to get rid of unwanted inventory.

>>59284180
Eh. $120 is still more than $60 even if it's double the amount of real cores. I'd still get a G4560 for my mommy.

>>59284193
AMD will probably place APUs on the lowest end of the market so you get a good CPU and a good GPU. Intel doesn't really have a competitive part that can do both.

AMD also licensed their GPU tech to Intel so even when Intel moves chips they'll get a cut of it.

>>59284244

their plan is a cpu with gpu with HBM2 on it.

they might actually have a cpu with the gpu power of a 480. according to AMD.

>>59268572
Right, because I always game at 720p

>>59276716
well yeah it is lol

File: Intel_inside.jpg (250KB, 1032x774px) Image search: [Google]

250KB, 1032x774px

INTEL ON SUICIDE WATCH

>>59274955
>And Q2 release most probably means end of June.
pretty sure they halted it just to iron out everything that is wrong with platform right now, because it's shared throughout all the chips

R7 line is one giant beta test. Might be end of april.

>>59275253
>CCX's communicate?
yes, windows makes, due to thread juggling, software think that ryzen has 138mb L3 which makes in turn dump it into RAM increasing latency skyhigh

>>59273586
>a fucking 350$ CPU beating TWO CPUs that cost 1500$

holy fucking shit; where are my dual cpu mobos????

I hope the Zen-based Opterons will have the same price.

>>59284671
AMD's aggressively pricing these parts so they'll probably be significantly cheaper than Intel's but I bet they'll still be kinda expensive. Probably affordable though compared to Intel were you need to sell an arm and a leg.