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

[sarinaide]

Im going to put this in bold and caps until people get this;

THE PURPOSE HERE IS NOT TO SHOW CPU PERFORMANCE BUT TO SHOW THAT AMD PROCESSORS FAVOR NVIDIA CARDS MORE THAN THE RADEONS

1) By going high end to high end you rule out the "but you used this and that is faster"
2) It is furthering the proof that Nvidia's Kepler better drives a CPU as a weaker 8350 is pushed further than when used in tandem with a HD7970 which requires a faster CPU to get its maximum performance.

Nothing here is new or really life altering, your full AMD setup will work just fine and dandy and you can go back to playing Crysis 3 on ultra. This article is not comparing Intel to AMD, because gawd we know how tediously boring they get. So lets keep this to what the article is about.

DOES FX ACTUALLY FAVOR GEFORCE.... (if you think I'm making this up, just read the title )

Thank you for your time.

 

[Onus]

I'd rather have 40 properly-rendered and timed FPS than 80 "half runts and drops" jittery FPS. Despite all the work that went into this, without FCAT measurements, translating the results into a gamer's experience is an exercise in futility.

 

[zolton33]

sarinaide A better processor is able to better utilize a card correct? So using a higher grade card then the other will show that the higher grade card is better then the other correct? Now here it only shows that a better grade cpu is better at processing then a lower grade cpu when it comes to using the same card. Really? Every layman with any knowledge of pc's already knows this.

So again the article is pointless. Here is why:

1.The i 7 and the 8350 are in entirely different leagues.
2.The i7 is rarely bought over an i5 cpu in gaming so not very many people who game actually buy this cpu.
3.The 8350 was made to compete with intel i5's not i7's so of course it out performs it.
4. Amd no longer tries to compete with intel at the high end every one with knowledge of cpu's knows this.

It is about whether or not amd cpu's favor one card or another. That is true. But how can you tell this by only testing 1 amd cpu and then comparing it to a cpu that is far above the cpu it is beintg compared with? How does this show whether an amd cpu works better with one gpu over another testing this way? If that was the full true reason for this they would have used an older amd cpu a less newer one an apu or 2 and the 8350 and the 6300.

If you wanted to throw a few intel cpu's into the mix to test as well that would have been fine but you need to compare it to a cpu that is closer to each of the cpu's. Did they do this? No. Instead they put an i7 against an 8350 with no other results to help further their hypothesis. All it shows is the result testing of one single amd cpu and then making a leap in logic based on one result on one amd cpu.

That tells us next to nothing about the amd architecture and its preference for one gpu over another. And then putting an i7 result into this added what exactly in answering the question of whether or not an amd cpu fairs better with one gpu over another? So again i say that its a pointless test.

 

[utroz]

Besides the fact that the results of this test were obvious to anyone that pays even a slight amount of attention to computer hardware, Tom's used old drivers that do not reflect current performance of either platform. Do your tests right or don't bother. This review would have been fine 3 month ago but at this point as data is not current.. You should have done FCAT testing as well. So this was a lets through something together from old data and act like it actually is what people will see for performance nowdays... You gotta stay current or do not bother publishing old info...

 

[ohim]

So except F1 and Skyrim bouth CPUs perform the same no matter what the GPU is be it AMD or Nvidia, now how the hell do people still recommend so badly the intel CPU so often given the fact that it is double the price ....for few frames advantage lots of gamers supports monopoly ... it is sad actually for everyone in the PC market.

 

[ojas]

Yes, yes, I was right. Less than 10% of the community actually reads articles these days, and are by and large AMD fanboys, even when it's not required.

Listen folks, here're a few pointers.

1) This is not about absolute performance, it's about scaling margins.

2) They're comparing highest-end intel to highest-end AMD. Well, technically not highest-end Intel, but it's good enough for the purpose.

3) The purpose: Does an AMD CPU hamper CrossFire performance.

4) What's Thomas is telling you is this: When the scenario is graphics limited, SLI performance is better on an AMD CPU than the equivalent CrossFire performance.

5) Now you see, to realise point 4, you have to be able to have a control group and a test group. The AMD CPU + AMD/Nvidia GPUs are the test group. The complementary i7 setups are the control group. When you see SLI outperforming CF on AMD, you check if it does the same on Intel. If it does, then yay, AMD doesn't hamper performance of its own cards. However, if CF wins on the Intel, and SLI performance is identical/similar, there's a problem on the AMD w.r.t CF.

6) He doesn't have FCAT. And i checked this on the FCAT article, the delta b/w FRAPS results and FCAT results remain the same (or within 2%). Since this IS NOT A PERFORMANCE COMPARISON OR VALUE GUIDE, but a more academic article, it doesn't matter, we just want to see the scaling.

7) While current drivers *may* change the scenario a bit, he's using drivers from the same time period, so view this article as accurate till 2 months ago. Seeing that this "myth" has apparently been floating around for years, it's likely that the last two months haven't changed much. He's got certain constraints too.

8) Please FFS drop the i5 argument. If you're spending $800 on GPUs then you're sure as hell not getting an i5. But if you want AMD for whatever reason, this is an issue, and you should be aware of it. Whether you heed it or not, it's your problem.

Thank you for your time.


p.s. The news comments section has gone to hell, i sincerely hope that awful thing isn't coming to the regular articles. No it's not about being resistant to change, that system is just broken.

 

[rpmrush]

[citation][nom]sarinaide[/nom]It is true my A10 5800K runs slightly faster with a 680 opposed to a 7970, needless to say I just threw in some 6850's made life a lot simpler and enjoyable, had much better CFX scaling out of the 6850's than I did 7770's and 7850's, ol faithfuls.[/citation]

U have had the complete opposite experience compared to my own and your post is confusing. I had 6850's and scaling was decent in some games, but in most the stuttering was detestable. I moved to one 7950 and will never attempt dual cards ever again. Why in the hell would you settle for 6850's when you had a 680 and 7970? Do you just play one single game that happens to do well with them?

 

[sarinaide]

Yes, yes, I was right. Less than 10% of the community actually reads articles these days, and are by and large AMD fanboys, even when it's not required.

Listen folks, here're a few pointers.

1) This is not about absolute performance, it's about scaling margins.

2) They're comparing highest-end intel to highest-end AMD. Well, technically not highest-end Intel, but it's good enough for the purpose.

3) The purpose: Does an AMD CPU hamper CrossFire performance.

4) What's Thomas is telling you is this: When the scenario is graphics limited, SLI performance is better on an AMD CPU than the equivalent CrossFire performance.

5) Now you see, to realise point 4, you have to be able to have a control group and a test group. The AMD CPU + AMD/Nvidia GPUs are the test group. The complementary i7 setups are the control group. When you see SLI outperforming CF on AMD, you check if it does the same on Intel. If it does, then yay, AMD doesn't hamper performance of its own cards. However, if CF wins on the Intel, and SLI performance is identical/similar, there's a problem on the AMD w.r.t CF.

6) He doesn't have FCAT. And i checked this on the FCAT article, the delta b/w FRAPS results and FCAT results remain the same (or within 2%). Since this IS NOT A PERFORMANCE COMPARISON OR VALUE GUIDE, but a more academic article, it doesn't matter, we just want to see the scaling.

7) While current drivers *may* change the scenario a bit, he's using drivers from the same time period, so view this article as accurate till 2 months ago. Seeing that this "myth" has apparently been floating around for years, it's likely that the last two months haven't changed much. He's got certain constraints too.

8) Please FFS drop the i5 argument. If you're spending $800 on GPUs then you're sure as hell not getting an i5. But if you want AMD for whatever reason, this is an issue, and you should be aware of it. Whether you heed it or not, it's your problem.

Thank you for your time.


p.s. The news comments section has gone to hell, i sincerely hope that awful thing isn't coming to the regular articles. No it's not about being resistant to change, that system is just broken.

Its like I am reading my own posts

sarinaide A better processor is able to better utilize a card correct? So using a higher grade card then the other will show that the higher grade card is better then the other correct? Now here it only shows that a better grade cpu is better at processing then a lower grade cpu when it comes to using the same card. Really? Every layman with any knowledge of pc's already knows this.

So again the article is pointless. Here is why:

1.The i 7 and the 8350 are in entirely different leagues.
2.The i7 is rarely bought over an i5 cpu in gaming so not very many people who game actually buy this cpu.
3.The 8350 was made to compete with intel i5's not i7's so of course it out performs it.
4. Amd no longer tries to compete with intel at the high end every one with knowledge of cpu's knows this.

It is about whether or not amd cpu's favor one card or another. That is true. But how can you tell this by only testing 1 amd cpu and then comparing it to a cpu that is far above the cpu it is beintg compared with? How does this show whether an amd cpu works better with one gpu over another testing this way? If that was the full true reason for this they would have used an older amd cpu a less newer one an apu or 2 and the 8350 and the 6300.

If you wanted to throw a few intel cpu's into the mix to test as well that would have been fine but you need to compare it to a cpu that is closer to each of the cpu's. Did they do this? No. Instead they put an i7 against an 8350 with no other results to help further their hypothesis. All it shows is the result testing of one single amd cpu and then making a leap in logic based on one result on one amd cpu.

That tells us next to nothing about the amd architecture and its preference for one gpu over another. And then putting an i7 result into this added what exactly in answering the question of whether or not an amd cpu fairs better with one gpu over another? So again i say that its a pointless test.

As has been stated a few times now, this is not about ultimate processor performance, this article is about whether the FX 8350 being the flagship and likely resonate through the family fundametally runs SLI better than Crossfire.

The long and short is yes Nvidia cards in single or multi card setups are faster on a FX part than AMD single or multi card equivilants. This is not a terrible thing as it has been shown that the AMD cards require more muscle to be pushed to their potential while NVidia is better at extracting CPU performance. This is not another AMD FX is bad thread so don't confuse it as one, the i7 is just a control test and by using the i7 you rule out whether you are using a top end control or not, this isn't about which part the fX competes against this is purely a scaling test to prove whether SLI out performs CFX on a FX platform.

 

[sarinaide]

[citation][nom]sarinaide[/nom]It is true my A10 5800K runs slightly faster with a 680 opposed to a 7970, needless to say I just threw in some 6850's made life a lot simpler and enjoyable, had much better CFX scaling out of the 6850's than I did 7770's and 7850's, ol faithfuls.[/citation]

U have had the complete opposite experience compared to my own and your post is confusing. I had 6850's and scaling was decent in some games, but in most the stuttering was detestable. I moved to one 7950 and will never attempt dual cards ever again. Why in the hell would you settle for 6850's when you had a 680 and 7970? Do you just play one single game that happens to do well with them?

Most of my equipment is ES samples, or sponsored so for the most part I don't really own a system I just test random hardware and provide feedback. My only purpose using the 680/7970 on the A10 was to see how far the APU can push them and where there is the divide between CPU limited and GPU limited, no sane person will ever pair those cards with an APU as the APU represents the balance between price and performance instead of performance at price.

 

[ojas]

Side note: No Bioshock: Infinite benchmarks?

 

[ericjohn004]

I don't see how anyone can not understand this article. It's very simple. And the last sentence in the article says it all. And Tom's is correct yet again.

You can't read this article and tell me that Tom's wasn't completely scientific and did these tests with any biased. If they did, they wouldn't be Tom's Hardware.

I was just reading a comment and this guy says that Tom's is are only using games that only use 1 core, and that they should change this so results wouldn't favor Intell. Come on now, All of these games including Skyrim use at least 4 cores. So just so they don't use 8 cores(some of them I'm sure do)Tom's shouldn't use them because AMD wouldn't run them well enough? And not to mention none of those games only use 1 core. Should Tom's cherry pick games just so AMD can MAYBE run as fast as Intel? Or should they pick the appropriate games? It makes absolutely no sense to go cherry picking games that are 8 core. They're are probably only 5 of them out anyways. They wouldn't even make up the whole suite of benchmarks. And they probably aren't demanding enough or aren't well optimized anyways. So the results would be the same anyways.

 

[ojas]

[citation][nom]sarinaide[/nom]Its like I am reading my own posts As has been stated a few times now, this is not about ultimate processor performance, this article is about whether the FX 8350 being the flagship and likely resonate through the family fundametally runs SLI better than Crossfire. The long and short is yes Nvidia cards in single or multi card setups are faster on a FX part than AMD single or multi card equivilants. This is not a terrible thing as it has been shown that the AMD cards require more muscle to be pushed to their potential while NVidia is better at extracting CPU performance. This is not another AMD FX is bad thread so don't confuse it as one, the i7 is just a control test and by using the i7 you rule out whether you are using a top end control or not, this isn't about which part the fX competes against this is purely a scaling test to prove whether SLI out performs CFX on a FX platform.[/citation]
Lol yeah i read yours too...i just didn't want you (and Thomas) fighting the masses alone

 

[Onus]

It doesn't matter what FRAPS results you get for Crossfire; Crossfire is BROKEN; even AMD admits it, and they're working on a fix.
I agree that this is not another AMD vs. Intel issue, at all.

 

[potatoos]

Why no Crysis 3?

 

[ericjohn004]

sarinaide, I completely agree. I don't see why people are coming here and bitching about what benchmarks were used and arguing whether or not Tom's used the right games. Can't they see it's just about which graphics card performs better on the FX?

 

[Wisecracker]

It's a great start but with questionable conclusions, and you need another 600 man-hours to make it right

 

[ibn gozzy]

So basically AMD processors suck, but suck less when using Nvidia graphics. Interesting.........

 

[sarinaide]

[citation][nom]ojas[/nom]Lol yeah i read yours too...i just didn't want you (and Thomas) fighting the masses alone[/citation]

Well some don't get it, they want a i5 3570 and in the end the exact same thing will be shown. That the FX favors GeForce to Radeon, go figure

I don't mind fanboism preference if perfectly normal, hell I am probably the biggest AMD fanboi here but when a simple article gets turned into another but thats not a 3570 which its priced to compete against thread.

So just for troll value "DOES FX PREFER GeFORCE TO RADEON" heck I got that from the title.

 

[bigpinkdragon286]

So you have a control group of 1 and a test group of 1, and your control group effects it's own test results? That hardly sounds scientific. With that much logic and reason, why not consider the 8350 the control and i7 the test? The article could easily have been titled, "CrossFire Versus SLI Scaling: Does Intel's Core i Actually Favor AMD's CrossFire?"

For all of the back and forth about how the FX should have been compared to an i5, I actually don't see why that's relevant, but not for the reasons already proffered. If this is an article about SLI versus CrossFire performance on AMD's FX platform, why do we care what it does on Intel's platform? Mere curiosity, or is it a "feel good" thing?

Why not just do an article about scaling on both platforms, instead of trying to shoehorn a conclusion that isn't very scientific onto AMD's FX platform? Saying the FX platform actually favors SLI is as though the FX branded equipment has some sort of choice in the matter. It makes more sense to say that SLI performs better on an FX platform than CrossFire, but that still doesn't answer the question of value. I may still get better performance per dollar on an FX platform using CrossFire, versus whatever other combination, so why not look at that aspect?

I do not dispute the test results, I just find the article could have been constructed in a less biased and more useful fashion.

 

[potatoos]

[citation][nom]ericjohn004[/nom]sarinaide, I completely agree. I don't see why people are coming here and bitching about what benchmarks were used and arguing whether or not Tom's used the right games. Can't they see it's just about which graphics card performs better on the FX?[/citation]

I am just wondering why they didn't use Crysis 3 as it ought to be in every benchmark article. It is the latest/greatest game that everyone wants to see. Just seems like they should have included it.

 

[master9716]

Its to bad the the 7970 Fps are Fake not real Full Frames - Toms can you guys redo Amd Fps to only show only fully drawn fps. ' thanks techreport'

 

[bigpinkdragon286]

[citation][nom]potatoos[/nom]I am just wondering why they didn't use Crysis 3 as it ought to be in every benchmark article. It is the latest/greatest game that everyone wants to see. Just seems like they should have included it.[/citation]
As it's been brought up before, the test results used in the article are not brand new. Were benchmark results for Crysis 3 available when these other benchmarks were done? I don't know either way, and am just suggesting a possible reason for the lack of those benchmarks.

 

[gamerk316]

Really sad article:

http://www.pcper.com/reviews/Graphics-Cards/Frame-Rating-Dissected-Full-Details-Capture-based-Graphics-Performance-Testin

CF doesn't work, period.

 

[buzznut]

I think it probably has just as much to do with the titles used for testing.

 

[the great randini]

whoops... I wonder if this problem is not another symptom of lackluster drivers cropping up again.