CPU Single Thread Performance

File: multi-thread.png (96KB, 678x625px) Image search: [Google]

96KB, 678x625px

Actually LULZ Intel completely dominates in multi-thread performance too, although all the processors are like $2000.

Ayymd is low end shit for poorfags only

Because you're looking at pissmark, the most garbage simulated benchmark on the planet.

>Pissmark

xd

File: single-thread2.png (214KB, 1195x702px) Image search: [Google]

214KB, 1195x702px

>>60943702
It's literally the same on every benchmark I checked

File: quad-core.png (259KB, 1202x827px) Image search: [Google]

259KB, 1202x827px

>>60943745
Quad-core results here

File: ryzen.png (17KB, 599x238px) Image search: [Google]

17KB, 599x238px

Ryzen's quite good

>>60943845
It's only #86 and that's in multi-core

>>60943862
Why are you trying to shill passmark results when you've been told it's a terrible benchmark?

Why are you not looking at multiple benchmark and real program reviews? Are you 12? Are you retarded?

lol amd btfo

>>60943890
Uhh you might want to look above you, I already posted multiple benchmarks.

File: 1496608320532.png (299KB, 1100x1002px) Image search: [Google]

299KB, 1100x1002px

intel shills are this desperate with the i9 housefire benchmarks results?

top kek

File: intel i5 7600k.png (16KB, 599x240px) Image search: [Google]

16KB, 599x240px

>>60943862
It's doing a lot better than any similarly priced Intel CPU

>>60943907
It's almost like you're a fucking moron.

O wait

>>60943933
The Ryzen looks like a good value for multi-thread, but still bad value for single-thread

File: sth ryzen benches.png (2MB, 973x6115px) Image search: [Google]

2MB, 973x6115px

>>60943745
There you go, now you can stop your shilling collect your shekels and leave us in peace

File: AMD-Ryzen-7-1700X-vs-i7-7700K-IPC-single-threaded.png (10KB, 493x141px) Image search: [Google]

10KB, 493x141px

Because pissmark is a terrible benchmark for CPUs that has no bearing to real world performance.

AMDs zen chips now have a bit more IPC than intel kaby lake processors which is why the i7-7700K and skylake-x housefores exist at all.

At the same frequency ryzen outperforms kaby lake.

>>60943960
I guess. I like the Pentium g3258 for that reason.

>>60943983
These charts are missing a lot of processors, I don't see any E5 v4

>>60943999
>Because pissmark is a terrible benchmark for CPUs that has no bearing to real world performance.

This isn't true though, it lines up with all the other tests I see too.

>>60944020
That is immaterial to your fraudulent claims in your paid shill OP. The dishonest question you posed has been answered. The answer is that AMDs $300 mainstream processors for home users spanks the shit out of Intels server processors.

>>60943999
>Ryzen 7 1700X @ 4.5GHz
in what wet dream? You're damn lucky if you manage to hit 3.9-4.0, but forget about 4.5 until Zen 2 arrives

>>60944085
What are you even talking about, look at the charts. AMD is not even in the top 50 for multi-core and not even in the top 100 for single-core.

File: amd-naples-server-workstation.png (106KB, 1597x868px) Image search: [Google]

106KB, 1597x868px

>AMD pioneers multi cores
>Microsoft Devs and Program Devs don't implement multi core enabled processes
>OS's that utilize multi-core show AMD at huge advantage while Intel dominates single thread
>fast forward a decade
>Intel gets multi cores
>Softare catches up with advanced technology
>AMD/Intel at even ground again while Intel has all the capital and finacial momentum.

AMD creates
>freesync
> _64,
>keeping socket the same for multiple generations
>internalizing the north/southbridges
>APU's
>Mantle(now Vulkan)
>open capable GPU's
>adoption or creation of advanced and sometimes controvercial technologies
>willing to take the risk to advance the industry while undercutting the competition

yet they never got the recognition they deserved

File: 1491400572774.jpg (138KB, 804x395px) Image search: [Google]

138KB, 804x395px

>>60943573

v.happy with devils canyon, interested with threadripper for the best of both worlds

File: ny.jpg (62KB, 400x533px) Image search: [Google]

62KB, 400x533px

>>60943999
>IPC
How about you compare performance on base frequency, you disgusting ayymd shill?

File: chinese-present.jpg (35KB, 465x473px) Image search: [Google]

35KB, 465x473px

>>60944085
Who care about price/performance ratio?!?

I WANT DAT BIIIIG CPU SONN
U FEEL ME??!?

INTEL INSIDE(TM)

File: x265-2.png (582KB, 600x600px) Image search: [Google]

582KB, 600x600px

>>60944075
No, it doesn't. Look at real world usage like HEVC video encoding.

R7 1700: 8c16t @ 3.0 - 3.7 GHz
i7-6950X:10c16t@ 3.0 - 3.5 GHz

Yet despite more fucking cores the i7 is only 25% faster. AMDs advantage here isn't much but when you factor in the price of the i7 ($1,700) vs the R7 ($300) you'd have to be batfuck insane to consider intel at all.

>>60944187
how about you compare the shelf price to the performance you disgusting intel shill

>pay 2-3x more
>oh em gee iz festur

>>60944206
AMD seems to have good multi-core value, but your chart just shows the exact same thing as mine, but with fewer processors listed.

Intel still on top with performance.

>no i5-6600k

;_; is that bad? I believed it was a good cpu. I have it over locked at 4.4.

>>60944233
not in the same price bracket though

>>60944238
It's not bad, just not stellar single-thread

>>60944238
Although overclocked it should be a lot better, but then again you could overclock most of these CPUs

>>60944216
>being poor
oh look, maybe you should find a job and live on your own instead of draining your parent's money???????

File: Comb17062017122715.jpg (133KB, 720x1585px) Image search: [Google]

133KB, 720x1585px

>>60944233
>Intel still on top with performance.
Yes I never denied that but it doesn't matter any more because of the pricing. Intel barely gives you better performance for a higher price.

i7-7700K: 4c8t @ 4.2 - 4.5GHz
>$325 + $100 AIO

R5 1500X: 4c8t @ 3.5 - 3.7 GHz
>$189 + $0 stock cooler

So, is paying more than twice really worth 20% faster single thread performance?

>>60944141
AMD didn't create freesync, its a proprietary version of adaptive sync that didn't qualify as adaptive sync because they wanted to cheap out, its basically a hack to prevent vsync from causing frame drops.

G-Sync on the other hand was created before adaptive sync was part of the displayport spec and uses an ASIC in the monitor for true fps on demand low latency gaming.

File: value.png (88KB, 672x584px) Image search: [Google]

88KB, 672x584px

>>60944357
Yeah value is a different thing altogether, but it's kinda a stupid metric because low-end processors will always win in value. You can get a $5 processor that beats almost all other processors in value (pic related).

>>60944357
20% faster every single core. You got 4 cores there, senpai.
Not to mention you're talking about benchmark """performance""" while most programs are designed and compiled (optimized) for intel cpus. So this 20% will transform into 50% faster performance in real life.

>>60944426
Sure thing but that didn't answer my question I gave you.

Also how is the R5 1500X a "low end" processor? It has 4 physical cores and 8 logical cores just like the i7-7700K albeit running at lower frequencies.

>>60944463
source?

>>60944504
http://www.agner.org/optimize/blog/read.php?i=49

>>60944508
No, I want the source on:
>most programs are designed and compiled (optimized) for intel cpus. So this 20% will transform into 50% faster performance in real life.

>>60944481
It's not low end, which is why it's not even anywhere near the top in value

As far as whether it's worth it to pay more, I'd say it just depends on your budget. Generally the more you spend, the faster it's gonna be.

>>60944522
>Date:2009-12-30 10:22
This is almost a decade old. Do you have any recent proofs?

>>60943573

With the multi-threaded rush, is higher single-threaded performance realistically achievable since 7700K is already tapping out?

>>60944538
Can you please stop posting the pissmark scores, they're worthless m8. He just proved it to you with the hevc benchmark.

>>60944423
Kill yourself.

>>60944566
Hevc benchmark shows exact same thing as these benchmarks, just has fewer CPUs listed.

File: yeezy.png (152KB, 2440x595px) Image search: [Google]

152KB, 2440x595px

>>60944574
Relax, he's the official clown of /g/. He's just joking around.

>>60943573
Because AMD focuses on multi-core performance. This is the way to the future. Advances in transitor technology will slow down, meaning that we won't be able to improve cores and such. The answer? Multi core improvements and optimisations, both in hardware and software. AMD have the future, while intel are going to be stuck in the past.

>>60944423
Oh look, it's /g/'s mascot. At least you had the decency to put your trips back, faggot.

>>60944550
>is higher single-threaded performance realistically achievable

I have no idea, I imagine yes?

What's really interesting to me is that a cheap $150 processor has the #2 best single thread score.

File: Screenshot_20170617-125001(1).jpg (251KB, 1280x556px) Image search: [Google]

251KB, 1280x556px

>>60944587
No it doesn't, at all.

i7-6950X is only 25% irl than the R7 1700 yet in pissmark it's 50% faster.

Fucking dumbass.

>>60944626
I don't know the exact percentages, I'm just saying the lineup looks the same, with all the Intel high-end multicores on top

>>60944588
One battery can have two different watt hour ratings depending on the voltage its discharged at moron.

>>60944657
>50% looks like 25%
Are you literally retarded?

>>60944705
I don't think you understand what I said

>>60944708
Can you elaborate, it sounded like you were defending pissmark scores.

>>60944685
lol

>>60944732
I'm just saying look at the order of the CPUs ranked here: https://www.cpubenchmark.net/high_end_cpus.html

vs Hevc benchmarks or any other highly threaded benchmark

Same order, just Passmark has more CPUs listed.

>>60944784
Oh ok. Well anyway stop mentioning pissmark, I can't see any legit use for it except 9gagers or reditors thinking they know anything about tech.

We should honestly have people reviewing processors by running them through cinebench/HEVC encoding benchmarks on stock settings for at least an hour. Then again most people flock to pissnark and geekbench scores because actually benchmarking the processors is just too much work.

File: untitled-2.png (18KB, 688x380px) Image search: [Google]

18KB, 688x380px

>>60943573
>pissmark

>>60944685
I like you

>>60944937
Those CPUs are not stock

>>60944357
>caring about value for something you own for at least 4 years when it only costs ~$350
Kids these days, tsk.

>>60944878
Passmark actually seems to match up pretty well with real world benchmarks, but it's not going to get the exact percentages the same.

Also it's separated into single-thread and multi-thread, while most programs are a mix of both.

>>60944982
No shit negro, he's comparing IPC. Are you gonna do that while the intel chips run at 5 GHz and the AMD ones don't?

>>60945014
Comparing non-stock CPUs is pretty useless.

>>60944982
All CPUs are clocked at 4.2GHz

>>60945030
AKA they're underclocked and overclocked

>>60945057
This is how you do IPC benchmarks.

>>60944342
le poor maymay xD

i think u shud use the austits maymay itds more epicc XD

>>60945084
IPC benchmarks are worthless for real world performance though.

>>60945110
Yes and in real world the i5s are dead and the locked 7700 is trash. You'll have to delid the 7700k and overclock it to hell and back in order to beat Ryzen.

>>60944995
>>60945011
>>60945028
dumbfucks

>>60945153
Uhh what? Look at the charts in OP

>>60945183
>pissmark

>>60945187
It's actually accurate, compare it to others

>>60945238
Just as accurate as CPU-Z

>>60945238
then explain >>60944206 and >>60944937 you dumb fuck

File: cpu-z.png (45KB, 1224x727px) Image search: [Google]

45KB, 1224x727px

>>60945271
CPU-Z seems to be missing a lot of processors

>>60945384
1. is the same, look at the order, same as Passmark

2. is underclocked and overclocked processors, not even sure why you'd post that

File: im_fucking_retarded.png (142KB, 600x600px) Image search: [Google]

142KB, 600x600px

>>60945418

>>60945440
Amazing argument

>>60945238
>>60943573
>>60943623
>>60943745
>>60943762
Surely if something's not running optimally it's always the fault of the hardware, not the software? You should use source based benchmarking so you can bisect the issue.

7700K clocks 13% higher on per core boost than the 1800X; it should come as no surprise that Ryzen is behind. It has more integer and the same floating point computational resources per core with Kaby-Lake, has the same decode width - Ryzen is practically prefetcher limited on single core. Showing quad core results shouldn't be surprising if single core results are greater when comparing two processors with the same number of cores. Many people also overclock the 7700K well beyond that which skews the Userbenchmark results. Linux benchmarks (check phoronix) show that single threaded performance trade blows in general applications excluding AVX2 accelerated programs and Ryzen is much more comparable with 6950/6900 in thread heavy applications e.g. C-ray, Blender, x264. 1700/1700X should correspond closely with those results as well. Some outliers are out there and I suspect unoptimized software (e.g. Apache, Redis).

Simply upon the basis of some arbitrary benchmark that something runs slower than the other doesn't show particular insight on whether or not the software is poorly optimized or if the hardware has inherent flaws.

>>60945418
The point of posting that is to show that IPC is roughly the same (it's not a perfect benchmark due to nonlinear scaling and obviously Zen can't make it to 4.2GHz reliably), so it provides an accurate gauge of why an processor with a 500MHz clock advantage has a performance lead? Regarding the first point, the i7 6950X has 2 more cores and a 12% and is achieving the same performance as a 4GHz 1700X. It's also 4x the price and requires a 50% more expensive platform (assuming X370).

File: qNyq2Dh47IKBnKit-Q8uYy_j5QvfqNt457AeX8IddWU.jpg (132KB, 725x768px) Image search: [Google]

132KB, 725x768px

ah passmark a totally unbiased and well written benchmark...

you know when sysmark and passmark bias explodes for being anti consumer the vw scandal will be as nothing compared to this

File: gpus.png (93KB, 658x721px) Image search: [Google]

93KB, 658x721px

>>60945910
>2015

>>60946029
missing the point by some light years
or
you are the usual intel shill

>>60946029
>780ti that high
Lol

>>60946088
Go ahead and explain the point.

File: Doom_01.png (71KB, 1304x1616px) Image search: [Google]

71KB, 1304x1616px

>>60946029
>1060>480
Dam, I knew pissmark was bad but not this bad.

>>60946139
780 above a z
970 above a fury z and a titan black
a 980ti which is a cut down slower titan x is above an actual titan x and above an m6000 which is a full maxwell chip...

>>60945910
>passmark is biased anti consumer
>cpu-z is biased anti consumer
>userbenchmark is obsolete

AMDrones' truth shields deflect all facts that don't align with their ideals

File: perfrel_1920_1080.png (29KB, 500x690px) Image search: [Google]

29KB, 500x690px

>>60946163
Umm 1060 is faster than 480

>>60946163
1060 is faster in anything but vulkan, all benchmarks reflect this

>>60946206
maybe on launch cause now 480 is certenly a lot faster...

>>60946236
Not true, both have had a lot of driver improvements

File: Deus_Ex_Mankind_Divided_DX12_1080p.png (34KB, 678x858px) Image search: [Google]

34KB, 678x858px

>>60946206
uhhhhn...

File: GTX-1060-UPDATE-100.jpg (66KB, 649x221px) Image search: [Google]

66KB, 649x221px

>>60946248
>lots of drivers improvements

File: 1470344503672.jpg (134KB, 614x442px) Image search: [Google]

134KB, 614x442px

>>60946248
No driver update can give the 1060 full hardware async compute support.

File: Average.png (50KB, 1295x1331px) Image search: [Google]

50KB, 1295x1331px

>>60946301
1060 already had full async compute in hardware

Here's a benchmark from May from 27 different games

1060 is definitively faster than a 480

>>60946353
>nvidia cards had full hardware async support

there is no picture to show how much the world is lauhging right now

>>60946353
>1060 already had full async compute in hardware
HAHAHAHAHAHAHAA.

>>60944540

He doesn't because one of Agner's most recent posts is his architecture optimisation guide in which he makes the claim that ryzen has higher IPC in anything other than AVX2 than any currently released Intel architecture. To wit anything that loves wide cores and/or is tuned to ryzen (doesn't mean excluding kabylake) will have ryzen equal a 7700k in single thread performance even with the difference in clock speed.

Now if and when such software exists is a different matter entirely based on many, many factors but even in the vidya space we have seen some of the more problematic titles being patched to reduce the lead kaby lake has over the equivalent ryzen chip. RoTTR being the most infamous example.

>>60946380
>>60946393

All Pascal cards have hardware async

>>60946410
HAHAHAHAHAHAHAHAHAAAAAA STOP IT

File: bf1-benchmark-1440p-dx12.png (33KB, 712x609px) Image search: [Google]

33KB, 712x609px

>>60946353
lolnope, the only games that the 1060 is faster in right now are the games nvidia paid the devs to gimp amd graphics cards.

File: 1334329164853.jpg (18KB, 366x380px) Image search: [Google]

18KB, 366x380px

>>60946353
>1060 already had full async compute in hardware

>>60946431
You can see the results right here, 27 games tested just a month ago:

http://www.techspot.com/review/1393-radeon-rx-580-vs-geforce-gtx-1060/page8.html

1060 is faster than a 480 in both average frame rate and minimum frame rate.

>>60945399

CPU-Z gimped their benchmark specifically for ryzen because ryzen's branch predictor is smarter than the average bear.

>>60946427
>>60946442
All Pascal cards have hardware async, look it up.

>>60946029
>780ti
>higher then 390X and all Fury variants

Yeah passmark is fine, nothing too see here, AMDrones deluded, buy intel and nvidia only etc.

>>60946467
HAHAHAHAHAAA DUDE STOP IT.

File: 1795t4.jpg (81KB, 500x616px) Image search: [Google]

81KB, 500x616px

>>60946467

>>60946476
Keep laughing, it's the truth.

>>60946495
1.7 the fuck outta here, Huang.

>>60946410
so wait 1060 is faster by 2% than the 480 but 580 which is 2% faster than a 480 is 5% faster than a 1060?

that doesnt even make sense considering how well 4xx scales with oc LOL

but then again forgot to see that this is techspot and not some serious site

>>60946507
They're benchmarks, the numbers don't lie

Techspot has some of the most comprehensive ones I've seen

File: ZbvhEzwzWnyFDrShFT4JCo.jpg (86KB, 1024x574px) Image search: [Google]

86KB, 1024x574px

>>60946483
This driver?

>>60946410
No, it doesn't. It has lower L3 latency which reduces the async bottleneck because when it is pre-empted for a compute and graphics it must purge the L3 cache between each warp, so now it does not take as large of a penalty when switching. It has software level load balancing and which works concurrently with hardware preemption to provide "async" like effects, so Pascal, combined with the above changes, still gets certain performance advantages when subjected to mixed workloads.

>>60946519
benchmarks dont lie https://www.geek.com/games/futuremark-confirms-nvidia-is-cheating-in-benchmark-553361/

>>60946537
>all 27 games tested are clearly cheating benchmarks

Quite an imagination you have

>>60946531
Yes it does

http://www.anandtech.com/show/10325/the-nvidia-geforce-gtx-1080-and-1070-founders-edition-review/9

Even Maxwell had it, but it wasn't enabled

File: shrugs.png (12KB, 560x407px) Image search: [Google]

12KB, 560x407px

>>60944342
>maybe you should find a job
Paying more for less performance is not a sign of having money, anon. It's a sign of clinical retardation.

>>60946613
You actually pay more for more performance

>>60945011
Why do you continue to be a retard?

This >>60944626 shows that benchmark is not accurate at all. With 100% difference between realworld metric and Passmark benchmark.

What's usable is the single thread rating in Passmark, the actual final score seems highly inconsistent with real world metric used by other benchmark.

>CPU Mark 20014 vs 13785
45% difference

>
>http://cpu.userbenchmark.com/Compare/Intel-Core-i7-6950X-vs-AMD-Ryzen-7-1700/3604vs3917
Here's a more accurate bench.

SC = ~ 14% better
Multi = ~ 30% better
Effective speed = ~23% better

So 23% vs 45%

Passmark is inaccurate. It overinflates Intel CPU. I mean come on, 7350K beating 1700X or even other i7? How is that accurate or being close to real world metric?

>>60946627
I don't see how you can do that when AMD has the performance crown on consumer CPUs and is cheaper at the same time.

>>60946640
You might want to look at the charts in OP

>>60944105
>You're damn lucky if you manage to hit 3.9-4.0
Most chips will hit that, the luck begins over 4GHz

>>60946633
The order is the same, you get the same ranking of CPUS

I already posted the userbenchmark ones, same results

>>60946575
if nvidia was so good on parallerism every single cryptocurrency miner would have bought ONLY nvidia and not piss on them cause they have poor multithreading capabilities..

before you talk about a technology you know nothing about go https://forum.beyond3d.com/threads/dx12-performance-discussion-and-analysis-thread.57188/page-5#post-1868421
and start reading the level of async we see on games dont surpass 3 maybee 5% while on consoles the minimum they use is 30%

>>60946645
The picture OP posted is from Passmark. Are you blind?

ITT /g/ forgets passmark only ever updates its benchmark when new Intel architectures are released.

>sawako weeb tripfag still bothers posting when he has become literal meme tier

>>60946677
What's your point?

File: 1497727579490.png (12KB, 722x194px) Image search: [Google]

12KB, 722x194px

INTEL WINS

>>60946699
nice graph

>>60946699
AMD is finished and bankrupt!!!

>>60946729
He mastered in Graphworks

>>60946575
Load balancing is not "hardware async compute" in the same sense that AMD's GPUs do async. GCN can provide a queue with mixed workloads with virtually NO context switching latency, and can preempt with virtually no latency, AND can also schedule individual CUs (perhaps even individual SIMDs if a wavefront was single queue) to different tasks. nVidia can only do the latter two with Pascal; it's not known whether or not it's a hardware dynamic scheduler or if it's just flashed on top of their static scheduler, and there are generally fewer SMs than CUs so there's still more flexibility in GCN as far as async shaders go.

>>60946780
Pascal literally has asynchronous computations in hardware, there's no ifs or buts

>>60946692
My point is you don't know what you're talking about if you trust Passmark. There's loads of benchmarks you can check where even a 1600X wipes the floor with the 7700K.

>>60946816
nvidia hasnt had a hardware sc on their cards since the 7xx era stop talking shit you dont know nothing

>>60946816
Pascal does not have dedicated async shaders.

NVIDIA stopped trying that after the disaster that was Fermi.

>>60946842
>shaders
schedulers

>>60946836
>>60946842
This is widely available information, you can look it up yourself

Pascal 100% has hardware async

>>60946855
>literally lying
Nobody believes you already, do you really think this is helping convince anyone

>>60946855
bring us a single piece of evidence that nvidia has a hardware sc on any maxwell card

>>60946825
Passmark shows the 1600X beating the 7700k in multi-threaded, not sure what you're arguing against.

You mean single-threaded?

>>60946875
>Maxwell
>SC

You are very confused friend.

>>60946904
>pascal
>not literally maxwell on a new process node
lol fuck off retard

>>60946871
You don't have to believe me, there's tons of sources:

http://www.anandtech.com/show/10325/the-nvidia-geforce-gtx-1080-and-1070-founders-edition-review/9

https://www.eteknix.com/pascal-gtx-1080-async-compute-explored/

http://www.tomshardware.com/reviews/nvidia-geforce-gtx-1080-pascal,4572-3.html

Pascal has hardware async 100%, this shouldn't even be a discussion

>>60946951
why don't you post a die shot showing the async compute engines on pascal

>>60946934
You're still very confused.

>>60946855
>>60946816
No, they don't. They can handle "async" but it is not truly asynchronous; it must provide context switching with latency. There is nothing in hardware that provides Pascal with the power to provide multiworkload queues that can be run without context switching, albeit that the penalty is much lower now. As you can see in the article YOU linked, the maximum mixed queue is 1 graphics, 31 compute. AMD provides 64 (32 for GCN1) execution slots for both graphics AND compute (and presumably DMA as well) and can be seamlessly interleavened without latency. No need to stop a SIMD, to offload compute to another, or wait for L3 to purge.

>>60946963
you're confused if you think pascal isn't just the latest rehash of g80 and cut down version of fermi.

pascal is 14nm maxwell, maxwell is kepler with tile-based rasterization, kepler is a cut down fermi and fermi is tesla with hardware scheduling

you can believe all the marketing buzzwords you want but you don't know what you're talking about

>>60946973
Pascal is literally doing hardware async by definition, there's no ifs or buts.

>>60943573
>>60943660
Knowing how shady intel is I wouldn't be surprised if they pay people to shill on technology boards 24/7 about their shitty spyware processors.

File: confused.png (20KB, 673x201px) Image search: [Google]

20KB, 673x201px

>>60946983
Sigh

>>60947007
die shot or fuck off

>>60946951
they are all reposting the same shit nvidia provided which in any case isnt evidence of a hardware sc at all
if they had any hardware sc on the pascal cards their tdp would have gone even beyond amd's one we know they are using the cpu to do some of the dx12 thus why they are pushing hard for the abolishment of async in general and why 99% of the ports are a complete garbage

>>60947020
Completely false, part of the reason why Pascal is so good at DX12 is because of the hardware async

File: LRM_EXPORT_20170411_145818.jpg (1MB, 1293x1730px) Image search: [Google]

1MB, 1293x1730px

>>60944423
Kill yourself you perma-BTFO'd tripfaggot

>>60947052
Pascal is not good at DX12, it always performs better in DX11

>>60947007
please stop it you are tech illiterate
you dont have a single clue about asynchronous workload
stop pretending you know anything about it

File: maxresdefault[1].jpg (594KB, 1920x1080px) Image search: [Google]

594KB, 1920x1080px

>>60947052
>Pascal is so good at DX12
you keep telling yourself that nvidicuck

>>60947127
oh yeah and how the 1080 had a magically superb aos perfomance.. only to realise the nvidia "forgot" to render the most heavy snow shaders on the game!
literally the same shit they were pulling with the fx 59 when they "forgot" to render textures on 3d and they were 2d

>>60946983
Pascal is also finer in dividing SMXs; now its 64 just like a CU. It's supposed to be in line with their plan to improve concurrency.

>>60947052
Not all DX12 games even use async.

>>60947173

Not excusing Nvidia but AOS has a lot of graphical settings that change based on what cpu the gpu is paired with. On lower thread count chips the game flatout does not render some effects and there is no way to enable them which makes cpu comparisons rather tricky.

File: 1468965293887.gif (4MB, 298x150px) Image search: [Google]

4MB, 298x150px

>>60947095
kek

>>60947197
its not the first time nvidia has done shit

look at the mirrors edge catalyst.. like the reviewers realised how the 970 was able to beat 390 and x on ultra(which normally needs 7+ gb of ram)it was too late...

the devs burried the setting to disable the memory quick sync so deep that whoever found it got named king arthur and very few actually re did the test...

>>60944423
>G-Sync on the other hand was created before adaptive sync was part of the displayport spec and uses an ASIC in the monitor for true fps on demand low latency gaming.

you dumb nigger, do you even know how the G-sync wire protocol works?
it's a per-frame busy wait done by the GPU communicating bidirectionally.
VESA Adaptive-sync is just a connection-time agreement of display input frame interval tolerances, which only requires a simple low-precision timer on the GPU in order to avoid all the wasteful and inelegant synchronization communication.

G-sync's wire protocol exists only to create vendor lock-in between displays and GPUs, not because it's even remotely a good idea.

The only good parts of it are the internal display-side scaler implementation details, mostly just that it has a dedicated frame buffer with time-compressed row write-out, which trades a few ms of input processing latency for reduced pixel transition times/frame ghosting.

>>60947253
>AMD - low precision
>Nvidia - high precision

Nothing new to see here.

>>60947253
reminds me when linus bought a freesync monitor and tested it with a nvidia card and it was crying about stutter

>>60947279
huh? A freesync card knows by itself when it can send a frame, where a g-sync card has to keep asking the display and waiting for the response like a brain damaged toddler.

> can I give you my new frame yet?...
> ...no.
> can I give you my new frame yet?...
> ...no.
> can I give you my new frame yet?...
> ...no.
> can I give you my new frame yet?...
> ...no.
> can I give you my new frame yet?...
> ...yes.
> OK! .........
> ......
> can I give you my new frame yet?...
> ...no.
> can I give you my new frame yet?...
> ...no.
> can I give you my new frame yet?...
> ...no.
> can I give you my new frame yet?...
> ...no.
> .........

and so on.

>>60947455
That's called redundancy for smoothest performance, meanwhile AMD sends frames whenever it wants and it leads to tearing

This is some next level shitposting.

>>60947487
>amd creates freesync to prevent tearing
random drone says amd freesync is actually causing more tearing
cant make this shit up

>>60947506
I think it's time for the obligatory "bring back Snacks" post.

>>60947575

There are no originalfags left and oldfags only know the legend of Snacks. That said I wonder if he still does that radio stream thingy he used to (midnight snacks).

>>60947614
He does.

>>60946889
No, that's precisely what I was saying. Then why are you saying Intel is a better buy if it's more expensive and slower?

>>60947487
>it leads to tearing
Are you implying Freesync has tearing?

>>60946353
Older games skew the stats pleb. As newer games come out the 1060 will fall behind. You'll see. There is no full hardware async compute on the 1060. It does it in the driver.

Somebody post that Nvidia/AMD graph that shows AMD in the lead by a huge advantage but it's only a 1 fps difference

>>60948714
desu that sounds like something straight out of Nvidia arcane slides

File: ChzG_zrUUAEhfx8.jpg (27KB, 600x328px) Image search: [Google]

27KB, 600x328px

>>60943623
>Posting multi-threaded performance benchmarks with Xeon SKUs literally 3 days before the NDA lifts on Epyc and AMD dumpsters every Xeon on that chart
>None of the idiots here notice they're being baited after 187 replies.
Holy fuck you dumb fucks are gullible.