THGCrossFire Versus SLI Scaling: Does AMD's FX Actually Favor GeForce?

[csbin]

http://www.tomshardware.com/reviews/crossfire-sli-scaling-bottleneck,3471.html

Does AMD’s flagship CPU really allow GeForce cards to run closer to their potential than Radeons? That conclusion is still a little difficult to reach, since the Radeon HD 7970 is faster than the GeForce GTX 680 in today's benchmark suite. We can still get there by comparing the results of a less-bottlenecked processor, however, and we can get even more information by comparing CrossFire to SLI. After spending way too much time poring over the data, the best illustration of CPU-to-GPU scaling occurred as I wrote this article’s introduction.

The above chart shows that, without significant CPU bottlenecks, two GeForce GTX 680s in SLI and a pair of Radeon HD 7970s in CrossFire offer similar performance. It also illustrates that this performance similarly comes not from having identically-performing cards, but from a slight scaling advantage favoring the slower GeForce GTX 680s when you put two of them together.
Our big revelation is that AMD’s FX-8350 performs 5% better in SLI than it does in CrossFire. When all else is made equal, AMD’s current flagship host processor really does favor Nvidia's graphics technology. Whoops.

 

[csbin]

In fact, this test there are still many open to question, first of all CPU overclocked to 4.4GHz Although it seems both sides are equal, but for FX-8350, the fundamental frequency is 4.0GHz, accelerating frequency is 4.2GHz,AMD's Turbo Core technology or quad-core acceleration, Intel's Core i7-3770K basis of frequency is 3.5GHz, single-core maximum speed up to 3.9GHz, 3.6GHz quad-core acceleration. Therefore, if you restore the default frequency, AMD FX-8350 performance decline than the Core i7-3770K less, the gap will narrow.

 

[ViRGE]

I rarely say this, but good heavens that's a terrible article. Both with respect to what exactly THG is looking for, and how they go about testing it.

Measuring multi-GPU scaling by varying the CPU doesn't make any sense. Whether SLI scales any better than CF on an AMD CPU has no significance when their own data clearly points to there being a CPU bottleneck on said CPU. Once you're CPU limited it doesn't really matter what GPU you're using; it's no longer the GPU that is the deciding factor for performance.

 

[R0H1T]

In fact, this test there are still many open to question, first of all CPU overclocked to 4.4GHz Although it seems both sides are equal, but for FX-8350, the fundamental frequency is 4.0GHz, accelerating frequency is 4.2GHz,AMD's Turbo Core technology or quad-core acceleration, Intel's Core i7-3770K basis of frequency is 3.5GHz, single-core maximum speed up to 3.9GHz, 3.6GHz quad-core acceleration. Therefore, if you restore the default frequency, AMD FX-8350 performance decline than the Core i7-3770K less, the gap will narrow.

Umm okay but I still don't get what you're trying to say here, disable turbo aye :hmm:

 

[inf64]

Umm okay but I still don't get what you're trying to say here, disable turbo aye :hmm:

He is saying that both CPUs @ stock would perform much closer in SLI/CF ( the difference between i7 and FX with SLI or CF would be lower).
What is interesting is that when both are OCed to 4.4GHz and using SLI with 680, 100$ pricier 3770K holds just a bit lower than 20% performance advantage. The OC, percentage increase from stock, is not equal though.Still shows that FX is a viable multi GPU platform in present and probably more so in future games.

 

[frozentundra123456]

I rarely say this, but good heavens that's a terrible article. Both with respect to what exactly THG is looking for, and how they go about testing it.

Measuring multi-GPU scaling by varying the CPU doesn't make any sense. Whether SLI scales any better than CF on an AMD CPU has no significance when their own data clearly points to there being a CPU bottleneck on said CPU. Once you're CPU limited it doesn't really matter what GPU you're using; it's no longer the GPU that is the deciding factor for performance.

I agree, reading that article almost made my head explode. Seems like they are testing two variables at once, and you never know which one is limiting. Even the supposed difference they make their conclusion on is only 5%, which must be close to the margin of error of the tests.

 

[R0H1T]

He is saying that both CPUs @ stock would perform much closer in SLI/CF ( the difference between i7 and FX with SLI or CF would be lower).
What is interesting is that when both are OCed to 4.4GHz and using SLI with 680, 100$ pricier 3770K holds just a bit lower than 20% performance advantage. The OC, percentage increase from stock, is not equal though.Still shows that FX is a viable multi GPU platform in present and probably more so in future games.

Alright.

 

[frozentundra123456]

He is saying that both CPUs @ stock would perform much closer in SLI/CF ( the difference between i7 and FX with SLI or CF would be lower).
What is interesting is that when both are OCed to 4.4GHz and using SLI with 680, 100$ pricier 3770K holds just a bit lower than 20% performance advantage. The OC, percentage increase from stock, is not equal though.Still shows that FX is a viable multi GPU platform in present and probably more so in future games.

I knew someone would derail the thread by immediately saying the test was unfair because the 3770 was overclocked further (editercentage wise), and try to argue that the 8350 is equal in gaming. That was not the point of the article. Nowhere did they say that the 8350 was not suitable for multiple gpus. It was not a comparison of cpus per se. It was a comparison of gpus on two different cpus.

Personally, I dont think the conclusion was valid because the GPU DIFFERENCE was so small, but the fact that one cpu was overclocked more is not the problem. I cant say why they didnt test at stock, or why they did not OC the 8350 further, but it is irrelevant to what they are trying to show. Maybe they got a bad example of the 8350 and it would only do 4.4, I dont know.

BTW, I am not going to get into which CPU is faster in multiple gpu set ups, because I dont want to get the Intel vs AMD argument going, but I will say that I dont think anyone who purchases two high end gpus, an expensive motherboard, and the power supply to run them is going to be particularly concerned about a 100.00 difference in price between two cpus, and I cant imagine that they would not overclock either.

 

[inf64]

I merely responded to a question posted by another user. But you are so blinded with hate that you have to say "I tried to derail" the thread. What a surprise huh.

 

[KingFatty]

Could someone explain the chart in simple terms? What does it even mean about those percentages exceeding 100?

 

[frozentundra123456]

I merely responded to a question posted by another user. But you are so blinded with hate that you have to say "I tried to derail" the thread. What a surprise huh.

Is the thread about relative cpu performance or not? No it isnt.

You are so intent on defending AMD that you immediately get defensive and try to refute a point the article wasnt even making.

 

[Kalessian]

Could someone explain the chart in simple terms? What does it even mean about those percentages exceeding 100?

From my point of view, they are saying that in a CPU bottleneck situation for intel chips, SLI and Xfire scale similarly, hence the 130% tie. However for AMD FX, the SLi setup is able to overcome its single card card deficit and actually see a larger increase than the Xfire setup.

 

[inf64]

Is the thread about relative cpu performance or not? No it isnt.

You are so intent on defending AMD that you immediately get defensive and try to refute a point the article wasnt even making.

I was replying to the question man. Do you speak English or not? :sneaky:

 

[Idontcare]

I rarely say this, but good heavens that's a terrible article. Both with respect to what exactly THG is looking for, and how they go about testing it.

Measuring multi-GPU scaling by varying the CPU doesn't make any sense. Whether SLI scales any better than CF on an AMD CPU has no significance when their own data clearly points to there being a CPU bottleneck on said CPU. Once you're CPU limited it doesn't really matter what GPU you're using; it's no longer the GPU that is the deciding factor for performance.

That's a QED right there folks. Regardless whether or not there is anything to the basic premise of the article, the data they generated and the conclusions drawn from the data are entirely questionable.

 

[VirtualLarry]

That's a QED right there folks. Regardless whether or not there is anything to the basic premise of the article, the data they generated and the conclusions drawn from the data are entirely questionable.

It's Toms, what do you expect?

 

[R0H1T]

I guess they didn't take the readings at stock, didn't realize it till now, but how do you folks take in context of this ?