R9 280x versus GTX 760 4GB - battle of the drivers! FIGHT!

[Wolfur]

Hello!
 
Firstly, let me introduce you to my world...
World of Phenom II x6 at 3,2Ghz.
 
I know well that Nvidia drivers are far better for AMD CPUs due to their multi-threaded nature which results in less CPU overhead and thus much better frames per second than Radeon GPU of similiar power.
 
I know well that Radeon r9 280x is a bit more powerful than GTX 760.
 
I know well that GTX 760 4GB costs the same as R9 280x.
 
I know well that I don't want a 2GB card in my setup since I plan on expanding to SLI (which is excellent compared to Crossfire) in the near future and 2GB vRAM will be a bottleneck quickly.
 
I know well that PhysX and Mantle are both proprietary and both will not be widely supported, ever, so I am not really counting them in altho i preffer eyecandy over 10% more FPS (and Mantle for sure won't bring 20%+ performance, I wouldn't hold my breath), since, again, the drivers are multithreaded on Nvidia's side ALREADY.
 
Phenom II x6 (and in fact other AMD CPUs as well) are a bottleneck for today's GPUs compared to i5/i7, but it's much more of a bottleneck for the 7970/r9 280x than a bottleneck for GTX 770 (which is somewhat more powerful than GTX 760 of course).
 
I'm not speaking out of my ass, I have seen benchmarks that prove this and actually posted about this on linustechtips before:
 
 

Depends on the GPU you want to pair it with currently. I5/i7 will do the best work in my opinion, just look at the current drivers for Radeon and GeForce:
 
It's a particularly big problem with Battlefield 4's beta but this has been case for years now, GeForce works better with slower multicore CPUs due to multithreaded drivers, while Radeon 7970 needs a quick single thread.
 
CPUs + AMD Radeon HD 7970:
http://pclab.pl/zdjecia/artykuly/chaostheory/2013/10/bf4_beta/charts/bf4_cpu_radeon.png
 
CPUs + Nvidia GeForce GTX 770:
http://pclab.pl/zdjecia/artykuly/chaostheory/2013/10/bf4_beta/charts/bf4_cpu_geforce.png
 
Just look at FX 83xx! With GTX 770, at stock clocks, it pulls 7 (!!!) FPS more  than OVERCLOCKED FX 83xx with Radeon 7970!
 
OVERCLOCKED FX 83xx with GTX 770 gets 17 (!!!) FPS more than OVERCLOCKED FX 83xx with Radeon 7970!
 
We'll have to wait for Mantle before answering your questions but right now:
 
If you go AMD, go Nvidia GPU.
 
If you go Intel, you can choose either AMD or Nvidia GPU.

 
There is also this test which actually compares i3, Phenom II x4 and FX 6350 paired with high-end 7970/gtx 770 GPUs to further analyse the driver problem on Radeon+AMD setups. It sometimes leads to EXTREME differences, even in single threaded games.
 
http://pclab.pl/art53784-12.html
 
For example:
 

 

 
That leads me to conclusion:
 
GTX 760 + Phenom II x6 1045t will result in higher frames per second than R9 280x + Phenom II x6 1045t.
 
Should I be worried about the drivers?
 
I already ordered Sapphire's r9 280x VAPOR-X but I can still cancel the order and get Gigabyte GTX 760 Windforce 3x 4GB instead (+ Batman Arkham origins which i honestly do care about since I love Batman and it is a Steam key so i can add it to my huge library), and then SLI the mofokas in the future.
 
Do you think this would be a sane decision? I trust Nvidia drivers more than AMD and if they didn't improve them to work better with their own CPUs yet I doubt they will do it anytime soon.
 

PS on a sidenote, do you think single GTX 760 will give me Ultra on 1400x1050 resolution with PhysX ON? I will be upgrading to GTX 760 4GB SLI + 1080p monitor as fast as possible.

[Wolfur]

How can a GPU driver benefit CPU cores? I'm confused

 

More threads being used in GeForce drivers leads to less overhead leads to better FPS in games that are heavy on CPU.

 

Single thread being used in Radeon drivers leads to more overhead leads to worse FPS in games that are heavy on CPU (we're talking about AMD CPUs getting lower performance with Radeon vs GeForce drivers, because AMD has slow single thread, MUCH slower than intel's, thus the Radeon driver is running slower on AMD CPUs, but Intel has no problem chewing through this overhead).

[Str_Mape]

How can a GPU driver benefit CPU cores? I'm confused

I was thinking the same thing I've never heard of it before.

[Wolfur]

I was thinking the same thing I've never heard of it before.

 

Look at post #3.

[Str_Mape]

Look at post #3.

This is fairly interesting, cheers.

[Wolfur]

@LinusTech

 

What's your opinion on this problem/question of mine? GTX 760 is approximately 15% slower than R9 280x but the overhead of Radeon drivers can and will result in much lower FPS in my case.

[Wolfur]

Help me, I still have ability to cancel the order on 280x ;x

[Wolfur]

I think I will cancel my Sapphire R9 280x Vapor-X order and buy GTX 760 4GB with Batman Arkham Origins instead, the price is virtually the same.

[xAcid9]

Any other source beside PCLAB.pl? Would love to see either Linus/Slick to cover this.

[Wolfur]

Any other source beside PCLAB.pl? Would love to see either Linus/Slick to cover this.

I'd love it as well.

Besides pclab there was some benchmark on Tom's hardware that showed similiar things but it seemed to give strange results. It was about crossfire vs SLI on amd vs intel.

Well on pclab it is common knowledge now that amd works better with GeForce.

[ShadowRaven]

Any other source beside PCLAB.pl? Would love to see either Linus/Slick to cover this.

Your wish is at hand

 

 

 

 

 

 

 

[GPUXPert]

pclab.pl's results contradict what the other reviewers are showing.
 

Here the 770 & 760 perform within margin of error, probably due to a driver anomaly.

Even In Tom's results the 270X was faster than the 760 in minimum frame rates.


280X is smoother than both the 770 and 760 and has the least variance in Skyrim.

[Wolfur]

snip

You didn't understand, it isn't about performance of gpus because I know the GTX 760 is a bit slower than r9 280x, its about drivers that work better with low-end amd CPUs. GeForce drivers are multithreaded, which gives the low-end multicore CPUs a break.

Radeon needs strong single thread.

the pclab test was testing i3, phenomm ii x4 and FX 6300 paired with high-end GPU 770 and 7970 to prove that GeForce drivers are better on amd systems.

[MetroP]

You didn't understand, it isn't about performance of gpus because I know the GTX 760 is a bit slower than r9 280x, its about drivers that work better with low-end amd CPUs. GeForce drivers are multithreaded, which gives the low-end multicore CPUs a break.

Radeon needs strong single thread.

the pclab test was testing i3, phenomm ii x4 and FX 6300 paired with high-end GPU 770 and 7970 to prove that GeForce drivers are better on amd systems.

That's a load of drivel, the draw calls made by the CPU are requested by DirectX not the graphics driver and the DirectX code is highly inefficient and is pushed on a single throughput thread.

That's why AMD made Mantle, which is 9 times more efficient than DirectX and can use as many threads as there are available on the hardware.

[Wolfur]

That's a load of drivel, the draw calls made by the CPU are requested by DirectX not the graphics driver and the DirectX code is highly inefficient and is pushed on a single throughput thread.

That's why AMD made Mantle, which is 9 times more efficient than DirectX and can use as many threads as there are available on the hardware.

 

You're wrong, DirectX is not the only thing going in the background. GPU drivers are really important and happen to be singlethreaded on Radeon setups and multithreaded on GeForce setups.

 

Actually GeForce GPU + dual core ends up with a HUGE GPU bottleneck because Nvidia drivers are made to work with more than two threads. For example i3 vs pentium, i3 wins a lot when running nvidia GPU.

 

Take a look:

 

 

 

Look at the differences in Frames per second between pentium dual cores and i3 paired with Titan, and between pentium dual cores and i3 paired with 7970.

[Humbug]

@GPUXPert

I think you missed the point of the discussion.

 

[Wolfur]

 

@GPUXPert

I think you missed the point of the discussion.

 

 

 

Yep, seems so.

[Kuzma]

Even though you'll have a slightly bottleneck with your CPU because of the raw power of the r9 280x you'll most likely still out perform the 760 4GB fyi I recommend against the 4GB model unless you're going 3 way SLI straight away and if you plan to get some amazing overclocks

[Wolfur]

Even though you'll have a slightly bottleneck with your CPU because of the raw power of the r9 280x you'll most likely still out perform the 760 4GB fyi I recommend against the 4GB model unless you're going 3 way SLI straight away and if you plan to get some amazing overclocks

 

4GB seems plausible if I was to mod Skyrim. Also I heard Nvidia GPUs are far better for ENB packs in Skyrim.

 

and i'd be scared to get just 2GB vram with a new GPU, better be safe than sorry. otherwise I would just get 670 2GB (1 more SMX unlocked, 192 CUDA cores more compared to GTX 760) for the price of 760 4GB. But I'm afraid some games may pull a bit more than 2GB of vram.

 

Well I wouldnt be sure if Radeon r9-280x will indeed grant me higher FPS.

 

Just take a look at this summary, where Phenom II x4 (l have 6 core Phenom II but let's keep it simple) + Radeon 7970 is 100%... Phenom II x4 965 + GTX 770 is 15% faster (15% higher frames per second).

 

 

even more apparent when comparing after overclocking the CPUs:

 

 

Do you think GTX 670 2GB is better buy than GTX 760 4GB even tho i might run into a huge vram bottleneck on modded Skyrim?

[Hellsrage]

I would need another source before I blindly follow this. 

[ShiverMePenguins]

your meager understanding of the English language coupled with your lack of knowledge in the tech field (primarily among cpu's and software) leads me to believe that:

 

 

Don't care, you're wrong.

[Wolfur]

your meager understanding of the English language coupled with your lack of knowledge in the tech field (primarily among cpu's and software) leads me to believe that:

 

 

Don't care, you're wrong.

 

You are talking about me?

 

Would you also say the same about @Kuzma ?

 

And why so rude? I have proof that this is a case... do you have proof that Radeon works at least as well as Nvidia with slower, but multi-core AMD CPUs?

[ShiverMePenguins]

You are talking about me?

 

Would you also say the same about @Kuzma ?

 

And why so rude? I have proof that this is a case... do you have proof that Radeon works at least as well as Nvidia with slower, but multi-core AMD CPUs?

firstly all the "benchmarks" you are showing are known to run better on nvidia cards. secondly they are the 770 versus the 7970GE, not the 760. thirdly all gpu's are "multi-threaded". why? because they have fucking thousands of shaders not just one. they don't render one object at a time, they render hundreds. so even though you should really use the terms "parallel processing" you are somewhat (but not really) correct.

fourthly, if you are using a processor that is almost half a decade old, you SERIOUSLY have some messed up priorities to get a gpu instead of upgrading the rest of your system.

[maazster]

There is something fishy with the graphs from PClab.pl, they really seem to be bias.

[Wolfur]

firstly all the "benchmarks" you are showing are known to run better on nvidia cards.

 

The ones that run better on Radeons like Tomb Raider or Max Payne 3 wouldn't prove anything because those games barely consume any CPU power.

 

How do I know that?

 

Take a look:

 

Without overclocked CPUs:

http://pclab.pl/zdjecia/artykuly/chaostheory/2013/06/i3_p2_fx_770_7970/charts/maxpayne3_1920.png

With overclocked CPUs:

http://pclab.pl/zdjecia/artykuly/chaostheory/2013/06/i3_p2_fx_770_7970/charts/oc_maxpayne3_1920.png

 

Notice how overclocking the CPUs doesn't change anything. These games like Tomb Raider happen to be purely GPU dependant and any CPU can feed the GPUs if there is no other load put on the CPU, obviously.

 

secondly they are the 770 versus the 7970GE, not the 760.

Yep, that's why I made the thread, to discuss 7970Ghz (15% faster GPU with single-threaded drivers) vs GTX 760 (15% slower GPU with multi-threaded drivers that help a lot when a game needs a lot of CPU power - Nvidia drivers are 'lighter' on multi-core CPUs because the load is spread between many cores.

 

thirdly all gpu's are "multi-threaded". why? because they have fucking thousands of shaders not just one. they don't render one object at a time, they render hundreds. so even though you should really use the terms "parallel processing" you are somewhat (but not really) correct.

 

I never said GPUs aren't multithreaded. In fact, you mistake terms.

 

GPU = graphics card.

Drivers = software that puts load on your CPU in order to 'drive' GPU.

 

Radeon's Catalyst Drivers run on a single thread, thus they need a fast single thread (i.e. Intel CPU) to be fed.

GeForce Drivers run on multiple threads, thus they run fine on multi-threaded CPUs, even if they are slower per core, but run slower (!) on dual core CPUs like pentiums compared to Radeons.

 

fourthly, if you are using a processor that is almost half a decade old, you SERIOUSLY have some messed up priorities to get a gpu instead of upgrading the rest of your system.

 

I'm not going to comment on this... Would you say FX 63xx is bad CPU? well check your facts, Phenom II x6 was released 27th April 2010.

Once overclocked it can easily match the raw power of FX 6300, sometimes even exceeding it (6 full Phenom cores vs 3 Piledriver modules with 2 ALUs, 1 FPU per module), as long as the newer instructions are not in use. Games don't use newer instructions.

 

Do you think that a newer CPU = faster CPU? Not always.

And in my case a newer CPU wouldn't improve gaming performance as much as a new GPU.

 

There is something fishy with the graphs from PClab.pl, they really seem to be bias.

 

What exactly do you mean?