GeForce GTX 460 1GB SLI vs. Radeon HD 5870 CFX @ [H]

[Blacklash]

HD 5870 Crossfire should not be around 20FPS slower than a 5970 in BC2. Looks like a bug of some sort.





If time permits you may want to test HD 5870 Crossfire using the 10.6 driver. At least, it seems that driver is fine for the 5970 in any case.

 

[pasta4u]

The review is useless.

What are the settings for each card.

Why are the radeon 5870s not overclocked ?

Is catalyst A.I enabled and what level is it at. What quality are u using on the geforce set up.

Why did you not test other drivers. You took the time to email Amd about the drivers knowing that your forum users are having problems with them. Why did you not use the older drivers as a control .



http://www.hardocp.com/article/2010/08/02/asus_ares_dual_5870_gpu_video_card_review/4

Here you guys are. We have the radeon 5870 1 gig in crossfirex at 2560x1600 with 12x CFAA/16x AF getting 38/92/64. Now in battlefield the same set up with 2560x1600 4x AA 16x af gets you 12/68/43 ? ????? I rather play at the higher fsaa then whatever is going on in your second review.

Metro 2033 gets you 24/70/48.9 with AAA and 16x AF with 24/70/48.9 and we see a reduction again with 24/67/44.6



I wouldn't mind if these are simple mistakes but you already knew there were problems with the drivers and you tested them anyway and you didn't give another refrence point. Ati puts out drivers every month. Are you going to update next month with the new drivers or will we not see another comparison for months at a time.

Real gaming benchmarks don't mean anything if they aren't using working drivers and don't tell a proper story.


Take the two 5870s and over clock them to 950-1ghz like your able to and use the 10.6 drivers and then compare them to the gtx 460 super OC set up and mabye we will have an idea of what the value of these cards are.

 

[Derangel]

The review is useless.

What are the settings for each card.

Why are the radeon 5870s not overclocked ?

Is catalyst A.I enabled and what level is it at. What quality are u using on the geforce set up.

Why did you not test other drivers. You took the time to email Amd about the drivers knowing that your forum users are having problems with them. Why did you not use the older drivers as a control .



http://www.hardocp.com/article/2010/08/02/asus_ares_dual_5870_gpu_video_card_review/4

Here you guys are. We have the radeon 5870 1 gig in crossfirex at 2560x1600 with 12x CFAA/16x AF getting 38/92/64. Now in battlefield the same set up with 2560x1600 4x AA 16x af gets you 12/68/43 ? ????? I rather play at the higher fsaa then whatever is going on in your second review.

Metro 2033 gets you 24/70/48.9 with AAA and 16x AF with 24/70/48.9 and we see a reduction again with 24/67/44.6



I wouldn't mind if these are simple mistakes but you already knew there were problems with the drivers and you tested them anyway and you didn't give another refrence point. Ati puts out drivers every month. Are you going to update next month with the new drivers or will we not see another comparison for months at a time.

Real gaming benchmarks don't mean anything if they aren't using working drivers and don't tell a proper story.


Take the two 5870s and over clock them to 950-1ghz like your able to and use the 10.6 drivers and then compare them to the gtx 460 super OC set up and mabye we will have an idea of what the value of these cards are.

The settings are explained in the bloody graphs.

Why would the 5870s being OC'd matter? Has H not proven time and again that OCing only effects frames by a little bit? Besides shouldn't a set up that costs almost double the 460 be able to beat it OC'd or not? Its a bullshit argument that only serves to undermine the rest of your post.

As for drivers, they specifically stated that AMD is aware of their numbers and did not mention ANYTHING about CF scaling being messed up in the drivers. So if the drivers are fucked its entirely AMD's fault, not the person writing the review. If the drivers suck AMD needs to get its shit straight its not up to reviewers to test every single bloody driver that comes out.

This was a very quick article, not even a fully indepth review, you act like this is simple benchmark testing where they'd have time to cherry pick drivers. Cherry picking serves no purpose, it only undermines the review process. H has always used the newest drivers at the time of testing. You're not new here, you should know that. Kyle has said that IF they are wrong they will report it, but unless AMD says something they don't know. I don't see a reason Kyle, Brent, or any H editor should waste their time and Kyle's money to test each and every driver that comes out to make sure its perfect. We should expect that EVERY driver released by both sides are perfect. If either fucks up then so be, they deserve to be kicked in the balls for it in reviews using those drivers.

 

[Blacklash]

If AMD provided [H]ard with buggy drivers that's definitely on them. I've seen them break Crossfire in titles before.

 

[Mr. K6]

Something is off with those results, at least on the 5870 CF in BF:BC2. Hell, those framerates are what I get on my single 5850, nevermind the 5850 CF that I used to have.

 

[baker269]

The review is useless.

What are the settings for each card.

Why are the radeon 5870s not overclocked ?

Is catalyst A.I enabled and what level is it at. What quality are u using on the geforce set up.

Why did you not test other drivers. You took the time to email Amd about the drivers knowing that your forum users are having problems with them. Why did you not use the older drivers as a control .



http://www.hardocp.com/article/2010/08/02/asus_ares_dual_5870_gpu_video_card_review/4

Here you guys are. We have the radeon 5870 1 gig in crossfirex at 2560x1600 with 12x CFAA/16x AF getting 38/92/64. Now in battlefield the same set up with 2560x1600 4x AA 16x af gets you 12/68/43 ? ????? I rather play at the higher fsaa then whatever is going on in your second review.

Metro 2033 gets you 24/70/48.9 with AAA and 16x AF with 24/70/48.9 and we see a reduction again with 24/67/44.6



I wouldn't mind if these are simple mistakes but you already knew there were problems with the drivers and you tested them anyway and you didn't give another refrence point. Ati puts out drivers every month. Are you going to update next month with the new drivers or will we not see another comparison for months at a time.

Real gaming benchmarks don't mean anything if they aren't using working drivers and don't tell a proper story.


Take the two 5870s and over clock them to 950-1ghz like your able to and use the 10.6 drivers and then compare them to the gtx 460 super OC set up and mabye we will have an idea of what the value of these cards are.

The 460s used are factory OCed. To get a factory 950 core OCed 5870 @ newegg costs $500.

 

[cageymaru]

Well, that is Eyefinity+CF, which has always been broken and crap.

Just to quote a previous review: http://hardocp.com/article/2010/07/26/geforce_gtx_460_sli_performance_vs_amd_gpus/1


So first you tell us that 5850/5870 CF will smoke GTX 460 SLI, and then when you test it you find that 460 SLI is faster, and then you claim that that was expected? Well, no, according to you guys that was NOT expected. That's why I'm puzzled that you aren't investigating it, because it directly contradicts what you've previously claimed. And of course there is the whole issue of 5970 laying the smack down on GTX 460 SLI according to that eval as well - 5870 CF should easily best it. I get that benchmarks != gaming, and that's what I love about [H], but the difference here is simply staggering.

Well looking at it from my perspective, for this test they chose more demanding maps to test the prowess of the cards. The AMD drivers are enabling Crossfire according to their follow up tests. But they are only seeing 55% usage at times (if I remember right... getting sleepy). That is AMD Driver Team's problems; not the reviewer's.

I don't think Kyle or Brent or whomever has the ability to edit drivers to fix whatever is wrong. I mean I think they are very intelligent, but I don't think they know how to write drivers. And I don't hold it against them if they choose a more graphical demanding map to test GPU's with. That's their prerogative. Personally I like the fact that they chose a more demanding map for the follow up review.

I don't think you should expect them to know what lines of code needs to be changed to get more GPU usage in Crossfire. That's AMD's thing. They sent their proof to AMD, and AMD said they would see if something is amiss with the drivers. I mean what else do you expect them to do?

Yes reading the old test I can see why they thought the AMD cards would outpace the 460's. But when using a more demanding map, the AMD cards choked while the 460's kept producing. And that is a matter for AMD to rectify. I think that it's wonderful that Hardocp is that thorough that they caught it.

I guess you could argue that they should test the easier maps instead; but what's the point in that when you have to eventually play the more demanding ones. And when you're spending this type of money, the cards should be able to handle it.

Sorry for being long winded.

 

[artika23]

The review is useless.

What are the settings for each card.

Why are the radeon 5870s not overclocked ?

Is catalyst A.I enabled and what level is it at. What quality are u using on the geforce set up.

Why did you not test other drivers. You took the time to email Amd about the drivers knowing that your forum users are having problems with them. Why did you not use the older drivers as a control .



http://www.hardocp.com/article/2010/08/02/asus_ares_dual_5870_gpu_video_card_review/4

Here you guys are. We have the radeon 5870 1 gig in crossfirex at 2560x1600 with 12x CFAA/16x AF getting 38/92/64. Now in battlefield the same set up with 2560x1600 4x AA 16x af gets you 12/68/43 ? ????? I rather play at the higher fsaa then whatever is going on in your second review.

Metro 2033 gets you 24/70/48.9 with AAA and 16x AF with 24/70/48.9 and we see a reduction again with 24/67/44.6



I wouldn't mind if these are simple mistakes but you already knew there were problems with the drivers and you tested them anyway and you didn't give another refrence point. Ati puts out drivers every month. Are you going to update next month with the new drivers or will we not see another comparison for months at a time.

Real gaming benchmarks don't mean anything if they aren't using working drivers and don't tell a proper story.


Take the two 5870s and over clock them to 950-1ghz like your able to and use the 10.6 drivers and then compare them to the gtx 460 super OC set up and mabye we will have an idea of what the value of these cards are.

Looks like someone got there laced AMD panties in a bunch. What's funny is that if you read your post between the lines your essentially accusing the whole [H] review team of being biased and making shit up. Kyle championed the 5870 when it first came out and now nvidia has a better multi-gpu, multi-monitor, and driver solution and your now crying foul. Boy someone is about to get a good ole fashion Texan beatdown!

 

[cageymaru]

That's a different map, the map I'm using now, Refinery, is more graphically demanding, heavier texture/bandwidth, more aliens and tessellation, than the Ruin map. NV drivers have also come a long way fast since Fermi launch.

I have done a quick look and installed MSI Afterburner, turned on the OSD, and ran through my map in BC2, I am seeing between 55%-80% GPU usage on both GPUs as I play through, so CFX is working. What I notice is that it fluctuates, and often times there may be only 55%-60% usage on each GPU, meaning efficiency and CFX scaling is low, but both GPUs are very much being tapped for performance, just not anywhere as one would want, idealy you'd want at least 80% GPU usage each GPU and even better would be 100% usage across both GPUs, I think SLI comes to the closest to this. I can tell you guys that AMD is looking at both maps in AvP and BC2 that we test in. That's the best I can do for now, I have other evaluations to work on at the moment.

Something is off with those results, at least on the 5870 CF in BF:BC2. Hell, those framerates are what I get on my single 5850, nevermind the 5850 CF that I used to have.

I would assume that since the AMD drivers are only using 55%-60% of both GPUs at a time, that may be what's wrong with the driver. I mean if it fell to 50% of each GPU then that's the equivalent of just having 1 GPU in the first place.

AMD needs to fix it.

 

[Mr. K6]

I would assume that since the AMD drivers are only using 55%-60% of both GPUs at a time, that may be what's wrong with the driver. I mean if it fell to 50% of each GPU then that's the equivalent of just having 1 GPU in the first place.

AMD needs to fix it.

Which is odd, because I never had any issues like this when I had 5850 CF going about 3 months ago. I have no idea why they release drivers that actually break more than they fix, what a bunch of fuck ups.

 

[matrix563]

did you install the crossfire profile? its separate from the entire driver install.

 

[Derangel]

Which is odd, because I never had any issues like this when I had 5850 CF going about 3 months ago. I have no idea why they release drivers that actually break more than they fix, what a bunch of fuck ups.

Both sides have had some pretty crappy drivers the last few years. I miss the days when they both worked their asses off on the driver side.

 

[jeremyshaw]

Both sides have had some pretty crappy drivers the last few years. I miss the days when they both worked their asses off on the driver side.

That's when they had neck and neck products, and drivers meant the world to them. ATi got complacent after ruling for a mere few months, so now they have a lot of tough work ahead of them. Hopefully their new release plan also included a more effective development procedure, that would ensure this kind of... calamity never happened again.

Of course, now I care more since I ordered another HD5870... shortly before reading this review

 

[Notleh]

That Overclockers thread is friggin hilarious. Pure fun.

The point here is that the 460's are owning the 5870's. So what if it is the drivers? Hardware or software...it doesn't matter. ATI has had plenty of time to get either right and have not done it.

At the end of the day the mainstream consumer is going to Google "Best graphics card" and buy what pops up. Better performance for $300 less? pffft.


This is how you recapture market share.

 

[Notleh]

Oh, and just because it is late I will go ahead and point out the stuff of nightmares for ATI. The 460's are the cheap entry level card in the 4XX GTX series.

What happens when the hardware upgrades and improved drivers get applied to the new high end Nvidia cards you know are in the works?

 

[cannondale06]

Oh, and just because it is late I will go ahead and point out the stuff of nightmares for ATI. The 460's are the cheap entry level card in the 4XX GTX series.

What happens when the hardware upgrades and improved drivers get applied to the new high end Nvidia cards you know are in the works?

there is an edit button so no need for a double post.

if it is drivers then it does matter because someone smart enough to buy this stuff from looking at reviews will also be smart enough to come across these discussions.

 

[babochee]

Well i sold my 5970 and 5870's and went with nvidia before the boost in the drivers just because of driver and firmware hassles.. but ill be more than willing to go with amd when the 6000 series hits and we end up waiting a year for nvidia to follow up

What i would like to see are some benchmarks with a gtx480+460 then 480+470, ya know?

 

[Devistater]

As for drivers, they specifically stated that AMD is aware of their numbers and did not mention ANYTHING about CF scaling being messed up in the drivers. So if the drivers are fucked its entirely AMD's fault, not the person writing the review. If the drivers suck AMD needs to get its shit straight its not up to reviewers to test every single bloody driver that comes out.

This ^

Why should [h] have the test burden of trying every single driver to find the exact optimized one for a particular game (because often each game has a differant driver that is a few percent better for it), and in a particular mode (single, sli, CF, etc)? They should use the latest driver.
Now if AMD comes out as says "btw, that latest driver is really screwed up with CF, please no one running CF run this driver, instead run 10.x, everyone running single cards keep running the latest driver", then yeah, they should run differantly because the official source has suggested otherwise.

BTW, if someone wants to go benchmark a bunch of drivers and card configurations with a pile of differant games, feel free Its not going to be a 1 hour job, more like a 100+ hour job.

 

[Blkout]

Something still seems off about the two reviews. We've seen SLI vs. CF before, even by [H], and the gap was never this big.

I mean, look at 5850 CF vs. GTX 470 SLI: http://www.hardocp.com/article/2010/05/10/galaxy_geforce_gtx_470_gc_sli_review/3

GTX 470 SLI is across the board faster, but not by all that much. Yet suddenly GTX 460 SLI is mopping the floor with 5870 CF?

5850 CF vs. GTX 470 SLI in BFBC2 are neck and neck: http://www.hardocp.com/image.html?image=MTI3MzQ0MDEwMWh0SGNIbnFhTEdfNV83X2wuZ2lm

Yet 5870 CF is losing bad to GTX 460 SLI in BFBC? http://www.hardocp.com/image.html?image=MTI4MTAzMjM1MmFlWk5qclJDU2JfMV80X2wuZ2lm

Something is seriously wrong, and I'm shocked that [H] doesn't seem to have really investigated wtf is going on. It saddens me to see little more than a small paragraph talking about the *huge* discrepancy going on when previously there were entire articles devoted to such things. If it turns out that for whatever reason 460 SLI really is just that good compared to even 470 SLI, then that's great, but to just take it at face value with little apparent questioning of the results doesn't sit right with me.

I agree, the last two comparisons using the SLI 460's are WAY off the mark. Nowhere on any other credible review site are the 460's beating a pair 5850's across the board much less a pair of 5870's.

I will concede that a pair of highly overclocked 460's will come close to a pair of stock 5850's and may win SOME benches, but I think we all know the 5850 stock clock speeds are VERY conservative and will also overclock very well also and blow right past a pair of overclocked 460's.

A pair of stock 470's is slightly ahead of a pair of stock 5850's, overclocking keeps them close together with the advantage staying with the pair of 470's.

A pair of stock 470's are slightly behind a pair of 5870's, overclocking keeps the 5870's slightly ahead of a pair of overclocked 470's.

A pair of 480's are ahead of a pair of 5870's, overclocked 5870's are close to a stock pair of 480's, but overclocking the 480's keeps them ahead.

 

[yuplis3d]

In conclusion

Xfire is broke and needs to get fix asap has been more than 8 months and still not working correctly.

What is not broke? older drivers?

Fine.

SLI scales much better than Xfire.

Have tried, have been there, but the true is that sli delivers more performance.

 

[greenfrogman]

Xfire is an chat program

CF and SLI are there names (and CF is shorter)

 

[Derangel]

Xfire is an chat program

CF and SLI are there names (and CF is shorter)

CFX technically since AMD is now calling it Crossfire X.

 

[Raap]

Actually, this review( which of course I can't find at the moment) I read regarding the Asus Ares stated that they were getting strangely low BC2 fps with their 10.6 drivers. They investigated and found out that ATI had actually screwed up and dropped one of their main optimizations for the game out of the driver. They contacted ATI and got a hotfix which increased their fps by nearly 50% in that game. I suspect what has happened here is that ATI either didn't get to implement that hotfix into 10.7, or they simply messed it up again.

 

[greenfrogman]

CF is easier in any case (Terms should be CF and SLI as when he posted that comment made me think what was he talking about an chat program for lol, he is only 4 months old so he must be new to this any way)

seems an even match to me the 460 and the 5870 or 5850 (more so the 5850 in most cases, really depends on the game) when SLI or CF is used Nvidia scaling seems to be better and more consistent drivers (that do not take 6 months to bring fix's out)

the ATI 10.x drivers must be the worst yet they have ever made or its just there are more ATI users that have an ATI card now so more users to complain

do note HBAO was turned on this review so that would have heavily impacted the BC2 results towards nvidia 4xx cards (ATI cards do not work well with it enabled and for whats it does to FPS its not worth it at all)

i would use ATI if i knew that i May have to reload my pc every time an new ATI driver came out (also the currant 5850 i got and sold made so much noise from the chokes i could hear it throw my headphones)

 

[Hapatingjaky]

Something still seems off about the two reviews. We've seen SLI vs. CF before, even by [H], and the gap was never this big.

I mean, look at 5850 CF vs. GTX 470 SLI: http://www.hardocp.com/article/2010/05/10/galaxy_geforce_gtx_470_gc_sli_review/3

GTX 470 SLI is across the board faster, but not by all that much. Yet suddenly GTX 460 SLI is mopping the floor with 5870 CF?

5850 CF vs. GTX 470 SLI in BFBC2 are neck and neck: http://www.hardocp.com/image.html?image=MTI3MzQ0MDEwMWh0SGNIbnFhTEdfNV83X2wuZ2lm

Yet 5870 CF is losing bad to GTX 460 SLI in BFBC? http://www.hardocp.com/image.html?image=MTI4MTAzMjM1MmFlWk5qclJDU2JfMV80X2wuZ2lm

Something is seriously wrong, and I'm shocked that [H] doesn't seem to have really investigated wtf is going on. It saddens me to see little more than a small paragraph talking about the *huge* discrepancy going on when previously there were entire articles devoted to such things. If it turns out that for whatever reason 460 SLI really is just that good compared to even 470 SLI, then that's great, but to just take it at face value with little apparent questioning of the results doesn't sit right with me.

I want to apologize in advance if you feel insulted by my post here, it is in no way directed at you, but you should read through the articles before linking them:

5850 CF vs. GTX 470 SLI: http://www.hardocp.com/article/2010/05/10/galaxy_geforce_gtx_470_gc_sli_review/3

Looking at the link it is clear Nvidia is using a much more graphically intense setting then ATI, Nvidia is using 16xCSAA compared to ATI's 8xAA. Scroll down to the SLI setup and Nvidia has been able to leave the 16xCSAA and also enabled TRMSAA where as ATI has had to drop down to 4xAA with TRMSAA to stay playable. Clearly Nvidia surpasses ATI in this post regarding performance. This is why you see the difference, Nvidia is able to provide a much more rich gaming experience at the same performance as ATI at lower IQ level.

5850 CF vs. GTX 470 SLI in BFBC2 are neck and neck: http://www.hardocp.com/image.html?image=MTI3MzQ0MDEwMWh0SGNIbnFhTEdfNV83X2wuZ2lm

This is not actually Apples To Apples if I recall this article correctly. Durring this eval HardOCP uses Transparency AA on the Nvidia card and in BC2 DX11 there is no TRAA support from ATI. So again, Nvidia is providing a much more rich IQ then ATI at the same performance. I believe this is the article in question http://www.hardocp.com/article/2010/05/10/galaxy_geforce_gtx_470_gc_sli_review/5 It would be nice to see HardOCP go back and test this to make sure TRAA was enabled or not.

Yet 5870 CF is losing bad to GTX 460 SLI in BFBC? http://www.hardocp.com/image.html?image=MTI4MTAzMjM1MmFlWk5qclJDU2JfMV80X2wuZ2lm

I am not certain as to what form of TRAA was used in this as its not listed throughout the article, was the Nvidia CP set to TRAA, was anything forced on or off in the CP, we just don't know. The other thing we need to look at is comparing the two resolutions, in #2 its running 1920x1200, and in #3 its running 2560x1600. From what this says is that the Nvidia 460 cards scale much better at higher resolutions compared to ATI. Also it is a very well known fact that HBAO causes the 5xxx cards to falter and choke, again Nvidia providing a much more rich gaming experience then what ATI can offer at the same performance.

Now if you`ve read over at Rage3D most would know that I do not agree with Kyle and Hardocp on many points, but this article is very well done and thought out. And I give them a thumbs up regarding this. I`ve been both an owner of Crossfire 5870`s ( nothing but trouble ) and I am presently an owner for SLI 480GTX Superclocks from EVGA. The fact of the matter is Nvidia provides a more rich gaming experinece then ATI, Nvidia has working Vsync, working AA, working Anisotropic Filtering, working TRAA in most DX10 and 11 software, excellent developer relations. The only issue is that the cards use an insane ammount of power and produce alot of heat ( not for me, mine run at 80c load same as my old 5870 Crossfire setup and idle at the same as well ).

The problem ATI faces is the fact that their driver are poor, extremely poor and this is what holds them back. We just had the fiasco with Starcraft 2 over AA support, ATI clearly refused to add it stating that they won`t support it till Blizzard adds it, but an outcry from ATI`s consumer base changed all that in the form of Catalyst 10.7a Hotfix. The Crossfire issue is a little more complicated then that. The Crossfire issue has existed for years, this is not something new, ATI`s support for Multi-GPU has always been behind Nvidia. Some complain that its because Nvidia bought 3DFX who had SLI since 1996 ( Obsidian 24Meg BABY! ) and that ATI had to scramble to add it. Crossfire has been around since 2003 so going on 8 years and its in no better shape then its inception those many years ago.

Honestly I can not see why anyone would recommend purchasing an ATI 5xxx card unless your budget is tight or performance per watt is what you are looking at. Even then the price for the 460 would be better:

http://www.memoryexpress.com/Products/PID-MX29652(ME).aspx

http://www.memoryexpress.com/Products/PID-MX25831(ME).aspx

For the price of one 5870 you could go SLI 460 and destroy the single card setup, you would actually save a couple bucks. I would like to know the power draw of an SLI 460 compared to the power draw of a single 5870 and Crossfire 5870.

Edit: The only good thing that came out of the 5xxx lineup was Eyefinity, when I went Eyefinity I could not look back, now I am using Nvidia Surround and of course everyone has read the articles regarding performance in that Category and Nvidia trounces ATI again. The only win for ATI here is the fact that they can do it on a single card whereas Nvidia requires SLI. But even looking at that two 460`s = the price of one 5870 so even there the price really is the same + much better performance and IQ.

 

[Brent_Justice]

did you install the crossfire profile? its separate from the entire driver install.

http://hardforum.com/showpost.php?p=1036037972&postcount=46

 

[Brent_Justice]

We use the latest driver, people complain, we don't use the latest driver, people complain. Our stance has always been to use the latest available driver that you can download from NV or AMD publicly, the only reason we'd use a lower driver version is if AMD or NV specifically told us that there was a bug or issue and that a previous driver was better to use, we had no such indication with Cat 10.7a of this, AMD obviously feels the driver is fine if they have publicly put it on their website for download. We just call it like we see it.

 

[Digital Viper-X-]

Quick Q Brent, do you guys plan to use Starcraft 2 in your future testing?

 

[FrgMstr]

Well, that is Eyefinity+CF, which has always been broken and crap.

Just to quote a previous review: http://hardocp.com/article/2010/07/26/geforce_gtx_460_sli_performance_vs_amd_gpus/1


So first you tell us that 5850/5870 CF will smoke GTX 460 SLI, and then when you test it you find that 460 SLI is faster, and then you claim that that was expected? Well, no, according to you guys that was NOT expected. That's why I'm puzzled that you aren't investigating it, because it directly contradicts what you've previously claimed. And of course there is the whole issue of 5970 laying the smack down on GTX 460 SLI according to that eval as well - 5870 CF should easily best it. I get that benchmarks != gaming, and that's what I love about [H], but the difference here is simply staggering.

Looks like that statement is totally wrong on our part. I need to change that.

Edit: And just to be to the point, I made that statement based on assumptions before we did these articles. Bad on my part, I thought I was smarter than I actually was. I usually refrain from making assumptive statements, but let that one slip.

 

[Derfman]

Wow... Makes me want to switch over to two of these puppies from my 5870... Too bad life and it's priorities will probably push back any such plans until the 6000 series is out, then maybe I will have another choice "maybe"...

 

[kllrnohj]

Looking at the link it is clear Nvidia is using a much more graphically intense setting then ATI, Nvidia is using 16xCSAA compared to ATI's 8xAA. Scroll down to the SLI setup and Nvidia has been able to leave the 16xCSAA and also enabled TRMSAA where as ATI has had to drop down to 4xAA with TRMSAA to stay playable. Clearly Nvidia surpasses ATI in this post regarding performance. This is why you see the difference, Nvidia is able to provide a much more rich gaming experience at the same performance as ATI at lower IQ level.

Sounds like you aren't familiar with CSAA. 8xMSAA is arguably higher quality than 16xCSAA. Just like 8xMSAA is higher quality than ATI's 12xCFAA. Nvidia talks about it here: http://developer.nvidia.com/object/coverage-sampled-aa.html Note that even at 16xCSAA - the number of Color/Z/Stencil Samples is still just 4. Basically, 16xCSAA is closer to 4xMSAA than 8xMSAA. 16xCSAA is certainly less demanding than 8xMSAA, so Nvidia doesn't have nearly as much of a lead as you think.

This is not actually Apples To Apples if I recall this article correctly. Durring this eval HardOCP uses Transparency AA on the Nvidia card and in BC2 DX11 there is no TRAA support from ATI. So again, Nvidia is providing a much more rich IQ then ATI at the same performance. I believe this is the article in question http://www.hardocp.com/article/2010/05/10/galaxy_geforce_gtx_470_gc_sli_review/5 It would be nice to see HardOCP go back and test this to make sure TRAA was enabled or not.

It is Apples-to-Apples, which is why it SAYS Apples-to-Apples.

I am not certain as to what form of TRAA was used in this as its not listed throughout the article, was the Nvidia CP set to TRAA, was anything forced on or off in the CP, we just don't know. The other thing we need to look at is comparing the two resolutions, in #2 its running 1920x1200, and in #3 its running 2560x1600. From what this says is that the Nvidia 460 cards scale much better at higher resolutions compared to ATI. Also it is a very well known fact that HBAO causes the 5xxx cards to falter and choke, again Nvidia providing a much more rich gaming experience then what ATI can offer at the same performance.

None. If it isn't listed, it isn't on. And the 460 cards don't scale better at higher resolutions - the opposite is true, at least for single cards. the 5850 increases the gap between it and the 460 as resolutions increase.

The fact of the matter is Nvidia provides a more rich gaming experinece then ATI, Nvidia has working Vsync, working AA, working Anisotropic Filtering, working TRAA in most DX10 and 11 software, excellent developer relations.

I've got a single 5870, and all that stuff works for me, too - just with 3 monitors

There is *zero* difference in IQ between ATI and Nvidia, btw. You claim that later on in your post, but either back it up or shut up.

 

[Tudz]

The point here is that the 460's are owning the 5870's. So what if it is the drivers? Hardware or software...it doesn't matter. ATI has had plenty of time to get either right and have not done it.

It matters because a month ago or a month from now with new drivers the situation could be completely different, and THAT is what matters to the reader, not this "Its ATI's fault for not bring it to the table when we did the review".

It has nothing to do with supporting ATI or nVidia, we just want the results which are most representative, which if this month bad drivers are present, it wont be representative. You could pick a particular driver for ATI or nVidia in a given month where its broken some sort of compatibility, destroyed a feature you previously used or damaged performance. Its why many people dont bother upgrading drivers every month unless they actually have a problem. You want the good compared to the good and the bad compared to the bad.

Of course I'm not suggesting [H] should go testing every bloody driver set, but when results are out of the norm like this it'd be good to see some comparison with a different driver to check that its truely the accurate result or just an anomoly. And these results are out of the norm...

http://www.anandtech.com/show/3809/nvidias-geforce-gtx-460-the-200-king/18

I know I know, different drivers, different games, different reviewers, canned benchmarks vs real gameplay... but the difference in their results compared to the [H] review is large enough to warrant investigation.

 

[samurai1999]

Since these results are so impressive, and contraversial, it would be worth including a few more cards for comparison
e.g. HD5850 CF & HD5970
- this would show if there was just a problem with HD5870 CF
- also, single HD5870
- maybe that's too many - could exclude the HD5850 CF

- but many people think the HD5970 is much better than the GTX460 SLI, so that would be worth including

Then also do the tests for a few more games, as these results could just be for a few games, and not the general case...

Anyway, interesting stuff ....

 

[jeremyshaw]

Since these results are so impressive, and contraversial, it would be worth including a few more cards for comparison
e.g. HD5850 CF & HD5970
- this would show if there was just a problem with HD5870 CF
- also, single HD5870
- maybe that's too many - could exclude the HD5850 CF

- but many people think the HD5970 is much better than the GTX460 SLI, so that would be worth including

Then also do the tests for a few more games, as these results could just be for a few games, and not the general case...

Anyway, interesting stuff ....

HD5850 CF test results were THE reason why this was brought out... as per the introducton of that article.

 

[LeninGHOLA]

10.7 broke (severely reduced) crossfire scaling for me on 5850's in several games. Blame AMD, not the review sites. I rolled back to 10.4 and not seeing any problems.

 

[MMAOmegaRED]

Jesus...well I'm just glad my 5970 seems to avoid the CF scaling issues. I figured it was the same as Crossfire but I guess not.

 

[adonn78]

SLI has always scaled better than crossfire so it was a smart chocie on their part. Personally I like AMD GPU's. I usually buy a system with an nvidia GPU and upgrade to ATI later becuase they cost less. I am vendor agnostic. I want the most performance for my money.

 

[samurai1999]

HD5850 CF test results were THE reason why this was brought out... as per the introducton of that article.

So, don't include that one....
HD5970 would be interesting...
- as some claim CF is broken, but not HD5970...

 

[Raap]

Jesus...well I'm just glad my 5970 seems to avoid the CF scaling issues. I figured it was the same as Crossfire but I guess not.

It is, but it doesn't seem like everyone is experiencing it; it depends on both which drivers and which crossfire profiles you have installed. Basically though, the 10.5a drivers should give you the best crossfire performance possible. It's said that 10.6 with a specific crossfire profile package installed will also work, at least for BC2.

Here is a review showing it even affecting the Asus Ares card: http://www.bit-tech.net/hardware/graphics/2010/07/27/asus-ares-review/8

Things went startlingly wrong for the Ares in Bad Company 2, as its frame rate results were about 30 per cent below where we expected them to be. After a little investigation however we found that ATI accidentally left out a key optimization in its Catalyst 10.6 release, hobbling HD 5970 performance in Bad Company 2. We should clarify that this was a mistake by ATI, not Asus, and affects all HD 5970s running the 10.6 version of Catalyst.

We did eventually find a hotfix for the problem but it wasn’t readily available on either ATI or Asus’s site which is disappointing. The hotfix did the trick though and once we'd applied it we saw frame rates jump by around 50 per cent. At 1,920 x 1,200 with 4x AA enabled for example we saw average framerates rise from 65fps to 101fps and minimum framerates jump from 40fps to 61fps. The hotfix also extended the gap between the Ares and a stock card with the Ares pulling out a lead of 8fps at 2,560 x 1,600 with 4x AA.

 

[Hapatingjaky]

Sounds like you aren't familiar with CSAA. 8xMSAA is arguably higher quality than 16xCSAA. Just like 8xMSAA is higher quality than ATI's 12xCFAA. Nvidia talks about it here: http://developer.nvidia.com/object/coverage-sampled-aa.html Note that even at 16xCSAA - the number of Color/Z/Stencil Samples is still just 4. Basically, 16xCSAA is closer to 4xMSAA than 8xMSAA. 16xCSAA is certainly less demanding than 8xMSAA, so Nvidia doesn't have nearly as much of a lead as you think.

You clearly did not read through the article you linked again.

In summary, CSAA produces antialiased images that rival the quality of 8x or 16x MSAA, while introducing only a minimal performance hit over standard (typically 4x) MSAA

It states it right in the first paragraph. Basically it improves the IQ while keeping the performance hit minimal or the same as 4xMSAA. Clearly you do not understand CSAA.