AMD FX 8370 vs Intel I7 5960x [GTX 970 SLI 4K benchmarks]

[ZetZet]

amd cpu getting 60+ fps on gta v average has to be fake. No way it does. Unless someone patched gta v.

[SpaghettiCarbonara]

In no land does this make sense, the 5960x has lower minimums despite having more cpu horsepower in every aspect. It has more cache more single thread and more cores.

Not to burst bubbles but the 8370 does not have 8 true cores. They are 2x modules with massively shared resources, and indeed some shared computation. (like think where two Intel smt threads give 110-130% performance of one thread, two amd "cores" give 170% [made up number less than 2 greater than 1.5] the performance of one regular thread.)

http://www.anandtech.com/show/6201/amd-details-its-3rd-gen-steamroller-architecture

Any article will do. It's a quite well documented idealogy. Either you call them 4 super cores or 8 gimped cores.

 

Ah yes, I forgot about this. My bad.

[Curufinwe_wins]

Hi have you used an AMD CPU before?

I think he was missing a /sarcasm....

[Curufinwe_wins]

Um, they're 8 cores, but 2 cores per module. They are 8 "true cores" since there are 8 physical cores inside the processor. You still get to use all 8 physical cores. Just because they are smaller doesn't mean they aren't cores.

But each core doesn't do all the opperations we traditionally require individual cores to do. The module overhead does a not insignificant amount of the work including most of the fetches, which if you'd believe it is what most of cpu time is made of.

http://images.anandtech.com/doci/6201/Screen%20Shot%202012-08-28%20at%204.38.05%20PM.png

[Curufinwe_wins]

I will take the time to look this over, and talk to the other writer and try to figure it out. Unfortunately its almost 2:30 AM and I've got benchmarking I need to get to...

Like I said before you had geniune questions and I really appreciated it

Yea go for it! It'd be nice to know really. Perhaps the shared core module architecture actually is showing a bit of stretch in its legs with this over SMT.

I wonder what @patrickjp93 has to think about all of this.

[dizmo]

As the writer of this article, I must first thank the OP for posting it here.. appreciate the feedback and exposure!

 

 

 

AMD has not ever sponsored any work I've done, including this. They're not an advertising partner nor did they sanction this report. I did request the hardware from them (as well as a few other companies for other components) just as I would with any review or report. That being said, they were never promised the results would be favorable. They shared the link to it on their facrbook simply because they clearly liked the report. I never even asked them to, they found it on their own. 

That's good to know, I suppose.

[pascil]

Intel got #Rekt?

[Sintezza]

 

Intel vs amd on LTT.

[Agosto]

Well...

The review was shared by AMD on facebook and it could say a lot about it, since the post was like "hey our cpuz are good for gaming!!1!1!11!1!"

Moreover, not to be disrespectful towards the reviewer, his results are different from all the dozens of other reviews published during this years, where FX83xxs usually lose to i5s and even i3s ( in gaming )

[Tass]

Flashback to AMD's initial "own results of benchmarks" right before fury x released.   :rolleyes:

[SpaghettiCarbonara]

I know how they are made, but the problem is that the cores are there.  The controller is the slow part since everything is shared in each module.  However, the cores are still there, so they are still true cores.  They are true 8 cores, but they're just not used as well in a lot of tasks.  If the cores are there then they are true.  If they were not there then they'd be false.  That's my argument.

See:

Any article will do. It's a quite well documented idealogy. Either you call them 4 super cores or 8 gimped cores.

I don't think he's saying they're false cores, he's just saying they're cut down a bit if I'm understanding it correctly. In his and your words combined, they'd be "gimped true cores", wouldn't they?

 

Well...

The review was shared by AMD on facebook and it could say a lot about it, since the post was like "hey our cpuz are good for gaming!!1!1!11!1!"

Moreover, not to be disrespectful towards the reviewer, his results are different from all the dozens of other reviews published during this years, where FX83xxs usually lose to i5s and even i3s ( in gaming )

 

The difference now is they're using DX12, which has changed a lot.

[Agosto]

The difference now is they're using DX12, which has changed a lot.

No game currently supports DX12 since they need to be rewritten in order to do so, at least partially.

[SpaghettiCarbonara]

No game currently supports DX12 since they need to be rewritten in order to do so, at least partially.

Ah shit, I read the post wrong.

 

Yes and no.  When I use them for 3D modelling no they perform like full powered 8 core, but the controller slows them down compared to the 700+ dollars more intel option.  Where the toll really takes effect is in single threaded performance.  So, they are and they aren't gimped.  If you saw my cinebench post you'd see where they get gimped.  

Interesting...

[IMPERIUS]

Why are people saying that at 4k the cpus have less stress? Isnt it like that the CPU must do the same operations in a game no matter the res?

[IMPERIUS]

Because at 4k VRAM and power is more important than the CPU.

Yes but people saying "at 1080p it would be different"?

[MegaDave91]

Actually, PCPer found that one of these games (at least), GTA V, had an improvement from Sandy Bridge to Skylake at 1440p (not quite 4k, but still more HD than 1080p). Considering the most dramatic results in this post were in the GTA V benchmarks, I think PCPer's findings settle the "BUTT 4K IZ NAT CPU INTENZIV" argument. Yes, the majority of the work is on the GPU, but there is a fair amount of work for the CPU to do in GTA V at >1080p. I can't speak for the other games, however, because I'm too lazy to look up other benchmarks.

 

http://www.pcper.com/reviews/Graphics-Cards/Skylake-vs-Sandy-Bridge-Discrete-GPU-Showdown/Grand-Theft-Auto-V

https://youtu.be/F2DTWZoO1Ck?t=30m45s

 

Again, not 4k, but 1440p is still more GPU intensive than 1080p and there was an improvement, even though the difference between Skylake and Sandy Bridge is much less dramatic than the difference between the 5960x and the 8370.

 

 

"Not quite 4K"? It's significantly less and much closer to 1080p than UHD. And as per-usual; a bunch of presumptions.

[Nielips]

As an FX9590 user, anything AMD seems to struggle to give a solid frame latency, I always see random spikes in games.

[Majestic]

5960X lower min. framerate? heh, hehehe. 

Yeah no.

[MegaDave91]

5960X lower min. framerate? heh, hehehe. 

Yeah no...

...And the winner of the most non-biased objective view on the topic award goes to: @Majestic 

[Muatez]

For those saying that is fake another source and not only at 4k, even at 1440p.

 

http://www.tweaktown.com/tweakipedia/56/amd-fx-8350-powering-gtx-780-sli-vs-gtx-980-sli-at-4k/index.html

http://www.tweaktown.com/tweakipedia/55/amd-fx-8350-powering-gtx-780-sli-vs-gtx-980-sli-at-2560x1440/index.html

[Majestic]

...And the winner of the most non-biased objective view on the topic award goes to: @Majestic 

 

Sorry for not being elaborate in my response. But I feel like this type of disingenuous benchmarking doesn't deserve much of my time.

This is clearly pandering to an audience, fabricating results to get instant karma with the AMD users that fall into his trap.

 

There is just no way these results are conducted in a proper way, or not doctored.

 

 

Take this result for example. You can see the AMD only gaining 2 extra fps by overclocking it to 4.6ghz, except the 5960X beats it by 5 fps. Meaning the game is not totally GPU bottlenecked, and yet it doesn't scale with overclocking.

 

And it's not that the 5960X has a more consistent fps to reach a higher fps, it supposedly has a lower min. framerate. It doesn't add up at all, and just proves he made the results up or botched the testing. Willingly or unwillingly.

[othertomperson]

This just in SLI 970s are the limiting factor in 4K gaming. Who knew!?

 

I mean they cope admirably but they are not going to give you 144 Hz even if displayport could.

 

 

Because at 4k VRAM and GPU power is more important than the CPU.

 

GPU power definitely, the 970's vram is fine. Whether the slow vram part, or the 4GB total limit might become an issue were the GPU itself more powerful, who can say? But the GPU itself is the limiting factor here.

 

Which makes sense -- 970 SLI and a 980 Ti perform very similarly. Would they if memory were the 970's problem?

[MegaDave91]

Sorry for not being elaborate in my response. But I feel like this type of disingenuous benchmarking doesn't deserve much of my time.

This is clearly pandering to an audience, fabricating results to get instant karma with the AMD users that fall into his trap.

 

There is just no way these results are conducted in a proper way, or not doctored.

You can pander to me anytime with your ever so eloquent sentence execution.

[Donny Stanley]

Sorry for not being elaborate in my response. But I feel like this type of disingenuous benchmarking doesn't deserve much of my time.

This is clearly pandering to an audience, fabricating results to get instant karma with the AMD users that fall into his trap.

 

There is just no way these results are conducted in a proper way, or not doctored.

 

 

Take this result for example. You can see the AMD only gaining 2 extra fps by overclocking it to 4.6ghz, except the 5960X beats it by 5 fps. Meaning the game is not totally GPU bottlenecked, and yet it doesn't scale with overclocking.

 

And it's not that the 5960X has a more consistent fps to reach a higher fps, it supposedly has a lower min. framerate. It doesn't add up at all, and just proves he made the results up or botched the testing. Willingly or unwillingly.

 

Clearly, you don't know the difference between "minimum framrate" and "frame time variance" the frame time variance is the amount of time it takes for the GPU to render the frame, when it is low the result is a noticeable stutter. 

[Majestic]

Clearly, you don't know the difference between "minimum framrate" and "frame time variance" the frame time variance is the amount of time it takes for the GPU to render the frame, when it is low the result is a noticeable stutter. 

 

This is frames per second, it says so right in the top-right corner. It means the 0.1% low of the 5960X was 14-17 fps, and for the 8370 it was 32-35 fps.

It means at the 0.1% variance level and the 1% variance level, the 5960X, according to this benchmark, did worse in both occasions.

 

That means the numers are very strange, and it doesn't account for the non-existant scaling with overclocking eventhough CPU seems to matter in this benchmark. 

 

Scaling at GPU bottleneck

 

Scaling at CPU bottleneck

 

The graph from the benchmark in question shows both happening, which makes no sense.