NV GTX 460 1GB SLI vs. ATI HD 5850 CFX Redux @ [H]

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,624
NV GTX 460 1GB SLI vs. ATI HD 5850 CFX Redux - We re-visit GeForce GTX 460 1GB SLI versus Radeon HD 5850 CrossFireX performance using new Catalyst drivers and new CrossFireX Application Profile 10.8a. We also have a new BETA ForceWare driver in play and we will once and for all see how these video card configurations really compete when software is updated
 
Love to see a 5870 redux! Thanks for showing us the differences too.
 
I'm not sure whether to be incredibly impressed with the 146% performance increase by a simple driver release or disgusted that there was some kind of 'bug' or 'issue' in ATI's drivers giving nVidia a 98% performance lead. I can't help but wonder how long it took to fix that bug versus how long it went unfixed for? It seems the performance issue was fixed within a few days of the issue being publically reviewed(some humiliation there) but otherwise went unfixed for ....3 months has it been that AvP has been out for?

Wes
 
I don't think the issue would have been fixed if the 460 SLI review hadn't made ATI crossfirex look so bad. A pair of $200 video card beating out the 5870 CF? So this is why honest review sites that report the facts, both favorable and unfavorable are important. So people can find out about the real performance and so manufacturers become alerted to problems with their drivers so they can get them fixed!
 
One heck of a performance jump for just a driver profile patch. Impressive, to say the least, but it's a head scratcher why the performance drop happened in the first place...

Thanks for the article [H].
 
I'm not sure whether to be incredibly impressed with the 146% performance increase by a simple driver release or disgusted that there was some kind of 'bug' or 'issue' in ATI's drivers giving nVidia a 98% performance lead. I can't help but wonder how long it took to fix that bug versus how long it went unfixed for? It seems the performance issue was fixed within a few days of the issue being publically reviewed(some humiliation there) but otherwise went unfixed for ....3 months has it been that AvP has been out for?

Wes

It just goes to show that IF you(we, all of us)don't speak up and demand, we won't get jack.
 
Crossfire performance is still pretty piss poor. ATI needs to step up, that last driver update still wasn't enough.
 
I am very impressed with the SLI and crossfire scaling and the amount of a performance gain through driver updates. I think my next PC will definitely have 2 6770's cross fired!:eek:
 
There is something awkward in the article I like to mention..

"What we mean by that is, the highest performance difference that existed in our testing was in AvP where HD 5850 CFX was 15% faster than GTX 460 1GB SLI, and that is with HD 5850 overclocked"

What do you have to emphasize 5850 is overclocked? It make it sounds like 5850 have to be overclock in order to beat GTX 460, while GTX 460 in this review is also overclocked.. Something is fishy here...
 
It's not a magical driver update. Why are people impressed? ATi fixed what they broke. That's all.

These clowns had to release a hotfix to restore proper MGPU performance to the old HD 4870 X2 for Christ's sake.

ATI Catalyst 10.8b Hotfix Features:

* ATI Radeon™ HD 4870 X2 and ATI Radeon HD 4850 X2 use both Graphics Processing Units (GPU) for high performance

http://support-prod.amd.com/us/kbarticles/Pages/ATICatalyst108bHotfix.aspx

As far as HD 5850s go they can often OC to the same levels as the 5870 provided they receive the same voltage. Below is 20 end users doing 1000MHz on the core or better with the HD 5850.

http://www.xtremesystems.org/forums/showthread.php?t=235693

IMO below 800MHz is a very conservative OC for the 5850, particularly if you're using MSI's Afterburner for said OC.

I'd still go with nVidia's SLi though because ATi are hit and miss with MGPU support. In one driver it works great and then it a later driver they go and break support. Completely unacceptable. Once you have Crossfire working properly in a title it should not vanish with newer driver releases.
 
Last edited:
It's not a magical driver update. Why are people impressed? ATi fixed what they broke. That's all.

I agree. It's quite telling that when [H]ardOCP first asked ATI/AMD if there was anything wrong with their crossfire drivers after the initial comparo, the answer from their marketing mouthpiece was "No". They didn't even know they had borked their own crossfire support!

Then a few weeks later they quietly release a 10.8a profile update the fixes the problem. Sheesh, let's hope this gets ATI/AMD to bulk up its regression test suite for crossfire profiles... I guess resting on their laurels for approx. 2 years with no real competition from nVidia has made them soft... so when the first half decent (not even fully decent) nVidia respin card comes out, ATI/AMD is caught with their drawers down in the multiGPU segment.
 
Hmm, to bad I can't use 10.8's. I really do hope that AMD address the other issues introduced with 10.7 and 10.8 in the 10.9 release (HDTV scaling, application of saturation adjustment across both cards when gaming). Right now I'm stuck on 10.5a, and even those are busted (5.1 and 7.1 audio channel scrambling).
 
Seems to me amd takes longer with drivers, but gets it working just fine in the end.

That's exactly why I've stuck with Nvidia as long as I have even when they had their asses handed to them originally with their GTX 465/GTX 470 Fermi introductions by the 5000 series from ATI. I'm still using a GTX 260 OC from MSI as I guess I don't play all that many high end games right now as everything still runs fine with this card for me in 1920x1200. I'm still highly considering getting a GTX 460 1gb card though every single time I see a review of them. I just hope Nvidia doesn't have some super card coming out, value wise, right around the corner (although by the time I do get a 460 they probably will have already have released it).
 
That's exactly why I've stuck with Nvidia as long as I have even when they had their asses handed to them originally with their GTX 465/GTX 470 Fermi introductions by the 5000 series from ATI. I'm still using a GTX 260 OC from MSI as I guess I don't play all that many high end games right now as everything still runs fine with this card for me in 1920x1200. I'm still highly considering getting a GTX 460 1gb card though every single time I see a review of them. I just hope Nvidia doesn't have some super card coming out, value wise, right around the corner (although by the time I do get a 460 they probably will have already have released it).
why would you even consider a gtx460? its only about 15-20% faster than your gtx260 and in some cases its basically no faster at all.
 
While 10.8a CAP helped Xfire tremendously here, Nvidia still has just a huge driver issue with Fermi cards only using 50% GPU usage in certain games like BC2/Mafia II/Borderlands/etc.

I'd love to see Kyle report on the GPU usage along with the fps, especially in BC2. 1xAA 1xAF on both 5850 vs 460 @ 2560x1600 with a graph of their GPU usage. You'll notice the 460s not even being used as much as the 5850s.

460s can be pushed a lot harder with a proper driver. They can easily OC to 900-950+ as well with a minute voltage increase.
 
Last edited:
It could just be nVidia's metric for "100% GPU usage" being different from what it previously used to be.
 
While 10.8a CAP helped Xfire tremendously here, Nvidia still has just a huge driver issue with Fermi cards only using 50% GPU usage in certain games like BC2/Mafia II/Borderlands/etc.

I'd love to see Kyle report on the GPU usage along with the fps, especially in BC2. 1xAA 1xAF on both 5850 vs 460 @ 2560x1600 with a graph of their GPU usage. You'll notice the 460s not even being used as much as the 5850s.

460s can be pushed a lot harder with a proper driver. They can easily OC to 900-950+ as well with a minute voltage increase.
I had that issue in BC2 with my old 8800GTS 320mb, the GPU load didn't get above 60%, altough framerate was terrible. When I switched from C2D to C2Q, the GPU load jumped to 100%, so you can't simply say its drivers fault.
 
Great article! I have noticed that AVP uses up all of the 1GB of VRAM for 1600x1200 so it must be quite a limiting factor for larger resolutions.
 
If they released this sooner I probably would have kept my 5850s. They sucked in performance a month a ago.
Good to see AMD rebound.
 
Older ATI represent! Good showdown, surprised to see the brand new 460's SLI not winning against the crossfire.
 
Conclusion said:
... is that extra $40 going to buy you real tangible improvements to the gameplay?
No, but it might cover the difference between a motherboard supporting SLI vs one that doesn't... :(
 
lol at all of the ATi haters, they fixed it and it works.......not a big deal.......
 
There is something awkward in the article I like to mention..

"What we mean by that is, the highest performance difference that existed in our testing was in AvP where HD 5850 CFX was 15% faster than GTX 460 1GB SLI, and that is with HD 5850 overclocked"

What do you have to emphasize 5850 is overclocked? It make it sounds like 5850 have to be overclock in order to beat GTX 460, while GTX 460 in this review is also overclocked.. Something is fishy here...


Good post!

no one said anything about it...very good observation shansoft...hmmmm fishy fishy...
 
I had that issue in BC2 with my old 8800GTS 320mb, the GPU load didn't get above 60%, altough framerate was terrible. When I switched from C2D to C2Q, the GPU load jumped to 100%, so you can't simply say its drivers fault.

You can't tell me my 980x @ 4.5 system is holding the card back. It's a well known issue posted on every overclock forum and on Nvidia's official forums.
 
I've hung tough with CF for the last two series of GPUs......48xx and now 58xx.

Unfortunately, AMD just can't seem to harness this power correctly.

It really doesn't matter that some of the nvidia cards only use 60-70% if they do it better and the game is better.

All the games mentioned have beeen out for quite a while as have the 58xx cards, and AMD is just now waking up to the fact......ouch.
 
ATI better be rocking it on the driver side come their next release in a few months, or i might have 2 cry.

With all that coverage from the sites on how nVidia has shifted their focus away from enthusiast and mainstream gaming components it really made me mad -_- and I already had beef on them for many reasons before that.

Them taking their sweet time from 2xx to 4xx was pretty hilarious.
 
As far as HD 5850s go they can often OC to the same levels as the 5870 provided they receive the same voltage. Below is 20 end users doing 1000MHz on the core or better with the HD 5850.

5850 has only 1440 stream processors vs. 1600 in the 5870 so it will take more than simply matching 5870 clocks with an OC on the 5850 to bring them up to par.
 
There is something awkward in the article I like to mention..

"What we mean by that is, the highest performance difference that existed in our testing was in AvP where HD 5850 CFX was 15% faster than GTX 460 1GB SLI, and that is with HD 5850 overclocked"

What do you have to emphasize 5850 is overclocked? It make it sounds like 5850 have to be overclock in order to beat GTX 460, while GTX 460 in this review is also overclocked.. Something is fishy here...

I see where you're coming from but I dont think that is what he meant, he was just pointing out that the 5850 was also overclocked because alot of people have been saying how in HardOCP's articles that the nvidia cards are oc models compared to stock AMD cards. In this case BOTH cards were overclocked and this was a moderate performance boost for cards that are supposed to be in another class and overclocking has little to do with anything here.

IMO below 800MHz is a very conservative OC for the 5850, particularly if you're using MSI's Afterburner for said OC.

I'd still go with nVidia's SLi though because ATi are hit and miss with MGPU support. In one driver it works great and then it a later driver they go and break support. Completely unacceptable. Once you have Crossfire working properly in a title it should not vanish with newer driver releases.


800 is a conservative overclock for either cards. GTX 460s and AMD 5850s have been hovering near or over 1000mhz core stable. The Shader and memory can be pushed much further either. A valid point but if both cards were pushed to the extremes I doubt anything would change here. In fact that may work to nvidia's favor if anything.

Good post!

no one said anything about it...very good observation shansoft...hmmmm fishy fishy...

As stated above, I don't think they were trying to bash amd I think you guys just took it out of context.
 
While 10.8a CAP helped Xfire tremendously here, Nvidia still has just a huge driver issue with Fermi cards only using 50% GPU usage in certain games like BC2/Mafia II/Borderlands/etc.

I'd love to see Kyle report on the GPU usage along with the fps, especially in BC2. 1xAA 1xAF on both 5850 vs 460 @ 2560x1600 with a graph of their GPU usage. You'll notice the 460s not even being used as much as the 5850s.

460s can be pushed a lot harder with a proper driver. They can easily OC to 900-950+ as well with a minute voltage increase.

GPU usage doesn't mean anything, it's a useless number. It doesn't even mean that there is any power left to squeeze out. For example, BFBC2 could be shader limited, leaving the tessellation, ROPs and TMUs under-utilized - but guess what? Can't do shit about that. GPUs are a lot more generalized now, sure, but there is still specialized hardware. You will never actually get 100% GPU usage unless the game is perfectly tuned for a specific architecture.

Not to mention we don't actually know what that number is measuring, and I assume it is more theoretical than anything. The power the card is theoretically capable vs. what it is currently doing. Which just comes back to making that number completely useless outside of seeing if the GPU is being used at all (say, to see if the second GPU in CFX or SLI is seeing any work at all). But outside of that, it's a useless number that may as well be pulled out of someone's ass.
 
I see where you're coming from but I dont think that is what he meant, he was just pointing out that the 5850 was also overclocked because alot of people have been saying how in HardOCP's articles that the nvidia cards are oc models compared to stock AMD cards. In this case BOTH cards were overclocked and this was a moderate performance boost for cards that are supposed to be in another class and overclocking has little to do with anything here.

If the statement doesn't trying to mean what I just point out, then perhaps the sentence should be change a bit to make it sounds like both card has been overclocked for fairness purpose.
 
God the people that try to find things to bitch about in every review thread is bullshit. Put the fucking tinfoil hat away guys.
 
Great job on doing another comparison! I have always liked AMD (ATI) but they really need to stop failing at software.
 
I'm not sure whether to be incredibly impressed with the 146% performance increase by a simple driver release or disgusted that there was some kind of 'bug' or 'issue' in ATI's drivers giving nVidia a 98% performance lead. I can't help but wonder how long it took to fix that bug versus how long it went unfixed for? It seems the performance issue was fixed within a few days of the issue being publically reviewed(some humiliation there) but otherwise went unfixed for ....3 months has it been that AvP has been out for?

Wes

They fixed it, that means that they listened. Good enough for me.


Also, My second rig has a 4870 in it and the new drivers helped speed up starcraft 2 for me which is nice. Before these drivers my laptop played it just as well and it has a 8800m.



God the people that try to find things to bitch about in every review thread is bullshit. Put the fucking tinfoil hat away guys.



Well said, I agree
 
Back
Top