GTX 580 vs 7970

[drunkenmaster]

AMD did exactly the same thing with shogun 2 and the Dirt series so both are as bad as each other.

AMD did what exactly with either games? They helped them code it and add DX11 features? Were these locked out to Nvidia, no, did they cripple Nvidia speed, no Nvidia was faster in Dirt 2 using the code AMD helped implement.

Ready what is being talked about Assassins Creed HAD dx10.1 in it, AMD was 15-20% faster with DX10.1 than plain DX10, Nvidia then asked Ubisoft to REMOVE DX10.1 as their hardware couldn't do it, thus Nvidia and AMD both had to use DX10 and AMD lose 15-20% of its speed.

Ah yes, the good old, "I don't understand X and my favourite card manufacturer is losing out cos of X so it must be NVIDIA's fault" argument.

Direct3D 11 deferred context command lists are a multi-threading feature in DirectX 11 defined by microsoft that NVIDIA added driver support for in CUDA 4.0 (iirc). If AMD is limping along on this front without driver support for a multithreading DX11 feature (and they claim to be all up in DX11!) it's their own damn fault. Sheesh! The level of fanboism here is ridiculous.

No, the level of your fanboyism is ridiculous, its absolutely and without question known that Nvidia did this, you're talking about DX11 while quoting and responding to a post about a DX10.1 game that had it removed in a patch after launch despite DX10.1 working perfectly on the only hardware to support it, AMD's.

As always you use one completely irrelevant example to have a go at another.

Sometimes Nvidia will be faster in a game, to pretend Nvidia don't actively try and kill AMD performance is to be utterly utterly blind. Or are you one of the Nvidia fans who thinks the tessellation in Crysis 2, which is done WAY beyond a point Nvidia cards can make any IQ benefit, was entirely for the end user, despite the performance drop on Nvidia's own cards while pushing super high tessellation its own users couldn't see?

 

[xsistor]

AMD did what exactly with either games? They helped them code it and add DX11 features? Were these locked out to Nvidia, no, did they cripple Nvidia speed, no Nvidia was faster in Dirt 2 using the code AMD helped implement.

Ready what is being talked about Assassins Creed HAD dx10.1 in it, AMD was 15-20% faster with DX10.1 than plain DX10, Nvidia then asked Ubisoft to REMOVE DX10.1 as their hardware couldn't do it, thus Nvidia and AMD both had to use DX10 and AMD lose 15-20% of its speed.



No, the level of your fanboyism is ridiculous, its absolutely and without question known that Nvidia did this, you're talking about DX11 while quoting and responding to a post about a DX10.1 game that had it removed in a patch after launch despite DX10.1 working perfectly on the only hardware to support it, AMD's.

As always you use one completely irrelevant example to have a go at another.

Sometimes Nvidia will be faster in a game, to pretend Nvidia don't actively try and kill AMD performance is to be utterly utterly blind. Or are you one of the Nvidia fans who thinks the tessellation in Crysis 2, which is done WAY beyond a point Nvidia cards can make any IQ benefit, was entirely for the end user, despite the performance drop on Nvidia's own cards while pushing super high tessellation its own users couldn't see?

You need to stop drinking. How's that hangover? You're back now are you?
Every post of yours makes me LOL. Clearly you lack all understanding. In this context the discussion was about Anandtech pointing out BF3's advantage in using driver command lists. Suggesting that it smacks of NVIDIA' crippling performance a la assassin's creed or whatever is just plain false.

I'm sure all that is well beyond your limited understanding of tech. Go talk about areas or something more your speed which you can manage with your tenuous grasp of grade 5 maths. LOL


EDIT: Let me make this very simple for you: My comment, if you had read and understood any of it, was NOT about whether or not NVIDIA crippled AC, but about whether NVIDIA was pulling a stunt with BF3 performance... sheesh, really. L2 read. NOt understanding a concept (deferred context command lists implemented on the driver) and jumping to the conclusion that NVIDIA is crippling performance indeed IS fanboism.

 

[xsistor]

As always you use one completely irrelevant example to have a go at another.

And this is just plain hilarious!!! Are you really serious? You probably glanced at the mirror while typing. If anyone on these forums goes off on random irrelevant tangents having a go at others by bringing up the most irrelevant arguments, then it's you.

In fact your arguments all have this format:


1. <WRITE 200 WORDS OF CRAP ABOUT HOW SOMEONE'S OPINION ON SOMETHING IS WRONG AND UTTER BULLCRAP>
2. <RESORT TO A FEW REDUCTIO AD ABSURDUMS AND MAKE A FEW ASSUMPTIONS/CONSTRUCT A FEW INTENTIONALLY FALLACIOUS ARGUMENTS AND ASCRIBE THEM TO SAID PERSON AND THEN DISCREDIT THESE ASSUMPTIONS -- THEREBY CREATING THE ILLUSION THAT YOU REFUTED SAID PERSON MORE THOROUGHLY>
3. <RANT OFF ABOUT 1000 WORDS ON A RANDOM TANGENT ABOUT SOMETHING IRRELEVANT>
4. <WRITE 500 WORDS ABOUT YOUR OWN OPINION, AND SOMEHOW ASSUME/PRETEND THAT YOUR OPINION ON THE MATTER HAS MORE WEIGHT THAN ANYONE ELSE'S>

 

[cancausecancer]

...so I play BF3 at 2560x1600, and with my GTX 580 1.5G I've been unable to run Ultra textures and any MSAA. I was hoping that the 7970 might have performance figures that would allow me to have the above config @ 60FPS, but it appears to hit that number I'm going to have to buy two of them and Crossfire.

Underwhelmed.

Thoughts?

Give some time for drivers. I waited for Fermi to come out with a lot of excitement, checking my bookmarks daily for any piece of news on it and when the benches finally hit the net I felt so underwhelmed I bought a 5970 instead. The reason I was underwhelmed was I thought (thanks Huang) that the Fermi was going to be powerful enough to fuel a Delorean but instead it was right where the natural next step for geForce would be.

The 580 since then is much more impressive.

I have a 7970 coming in Jan to replace my 5970. When I looked at the benches I wasn't blow away like I hoped to be. The card is great only because it will go thru driver maturing/tweaking and come out great. For now it's fast but it's where I'd expect it to be (except a little more solid than expected)

 

[ATIorNvidia]

Yes, but you can also OC a GTX 580 substantially - 900 out of the box is totally doable.

Let's compare apples with apples - at this point in time the 7970 isn't all that.

And you can OC a 7970 even more. The 7970 destroys a 580 with both at their max clocks.

But go ahead, spend almost the same amount of money on the vastly inferior 3GB GTX 580.

 

[opethdisciple]

Yeah if the 7950 comes in at about £300-£350 and it unlocks, good times. I don't think AMD will make the same mistake again though.

Thats if it was a mistake in the first place!

 

[Rroff]

The lack of performance effect is only in the case where the DX runtime software emulates the feature iirc... i.e. in not supporting it.

Again its off the top of my head but I thought even the GTX480/580 emulated most of it in software hence its not _that_ beneficial at this time, but could translate to decent performance gains, especially in geometry shader heavy games, on the 79xx cards if its supported there and definitely on kepler.

I have to agree its another case of people not understanding the technology and blaming nVidia.

 

[xsistor]

Again its off the top of my head but I thought even the GTX480/580 emulated most of it in software hence its not that beneficial at this time, but could translate to decent performance gains, especially in geometry shader heavy games, on the 79xx cards if its supported there and definitely on kepler.

I have to agree its another case of people not understanding the technology and blaming nVidia.

You're right they did emulate it software... Right upto CUDA4 which Added driver support for deferred context display lists for multithreading. I imagine is what anand tech meant when they said (somewhat cryptically), "DRIVER display lists." ... Does that sound like a reasonable interpretation?

 

[Final8y]

AMD tell me that they are actively working on this but cannot disclose a timeframe.

This year is about as specific as I am allowed to be but since there isn't that much left of the year you don't have long to wait.

http://www.rage3d.com/board/showthread.php?p=1336734728#post1336734728

I would be surprised if they make it happen in the days left but you never know.

 

[Rroff]

You're right they did emulate it software... Right upto CUDA4 which Added driver support for deferred context display lists for multithreading. I imagine is what anand tech meant when they said (somewhat cryptically), "DRIVER display lists." ... Does that sound like a reasonable interpretation?

Ah I've not caught up with CUDA 4 yet, tho find CARMA interesting CUDA on ARM could have quite wide spread implications.

 

[TheRealDeal]

There is another interesting fact on anand about those deferred context command lists. Nvidia use these in civilisation and amd's performance with the 6 series was so bad that anand were not gonna use it in there bench sweet but when they seen how well the 7970 done on amds reviewer guide they wanted to check out the numbers as they did not believe them. Here is what anand said.

"Because of the fact that Civilization V uses driver command lists, we were originally not going to include it in this benchmark suite as a gaming benchmark. If it were solely a DCL test it would do a good job highlighting the fact that AMD doesn’t currently support the feature, but a poor job of actually explaining any hardware/architectural differences. It was only after we saw AMD’s reviewer’s guide that we decided to go ahead and include it, because quite frankly we didn’t believe the numbers AMD had published.

With the GTX 580 and the 6970, the 6970 routinely lost to the GTX 580 by large margins. We had long assumed this was solely due to NVIDIA’s inclusion of DCLs, as we’ve seen a moderate single-GPU performance deficit and equally moderate multi-GPU lead for AMD melt away when NVIDIA added DCL support. The 7970 required that we rethink this.

If Civilization V was solely a DCL test, then our 2560 results would be impossible – the 7970 is winning by 12% in a game NVIDIA previous won by a massive margin. NVIDIA only regains their lead at 1680, which at this resolution we’re not nearly as likely to be GPU-bound.

So what changed? AMD has yet to spill the beans, but short of a secret DCL implementation for just CivV we have to look elsewhere. Next to DCL CivV’s other killer feature is its use of compute shaders, and GCN is a compute architecture. To that extent we believe at this point that while AMD is still facing some kind of DCL bottleneck, they have completely opened the floodgates on whatever compute shader bottleneck was standing in their way before. This is particularly evident when comparing the 7970 to the 6970, where the 7970 enjoys a consistent 62% performance advantage. It’s simply an incredible turnabout to see the 7970 do so well when the 6970 did so poorly.

Of course if this performance boost really was all about compute shaders, it raises a particularly exciting question: just how much higher could AMD go if they had DCLs? Hopefully one day that’s an answer we get to find out."

Its pretty interesting stuff which could lead to performance gains if amd gets this in action.

 

[xsistor]

Ah I've not caught up with CUDA 4 yet, tho find CARMA interesting CUDA on ARM could have quite wide spread implications.

I only recently heard about CARMA when I got an email from one of the CUDA mailing lists. I hear the name was suggested by a community member. And yeah it is extremely interesting from an embedded systems engineering perspective. I can think of a number of applications that are now possible thanks to it. Mainly in robotics and machine vision ... Matter of fact there is an OpenCV port to GPU, though I'm not sure how far along that project is. And something like Particle Swarm Optimization or a highly-connected massively parallel neural network for real-time AI applications could be very interesting indeed.

 

[Rroff]

There is another interesting fact on anand about those deferred context command lists. Nvidia use these in civilisation and amd's performance with the 6 series was so bad that anand were not gonna use it in there bench sweet but when they seen how well the 7970 done on amds reviewer guide they wanted to check out the numbers as they did not believe them. Here is what anand said.

"Because of the fact that Civilization V uses driver command lists, we were originally not going to include it in this benchmark suite as a gaming benchmark. If it were solely a DCL test it would do a good job highlighting the fact that AMD doesn’t currently support the feature, but a poor job of actually explaining any hardware/architectural differences. It was only after we saw AMD’s reviewer’s guide that we decided to go ahead and include it, because quite frankly we didn’t believe the numbers AMD had published.

With the GTX 580 and the 6970, the 6970 routinely lost to the GTX 580 by large margins. We had long assumed this was solely due to NVIDIA’s inclusion of DCLs, as we’ve seen a moderate single-GPU performance deficit and equally moderate multi-GPU lead for AMD melt away when NVIDIA added DCL support. The 7970 required that we rethink this.

If Civilization V was solely a DCL test, then our 2560 results would be impossible – the 7970 is winning by 12% in a game NVIDIA previous won by a massive margin. NVIDIA only regains their lead at 1680, which at this resolution we’re not nearly as likely to be GPU-bound.

So what changed? AMD has yet to spill the beans, but short of a secret DCL implementation for just CivV we have to look elsewhere. Next to DCL CivV’s other killer feature is its use of compute shaders, and GCN is a compute architecture. To that extent we believe at this point that while AMD is still facing some kind of DCL bottleneck, they have completely opened the floodgates on whatever compute shader bottleneck was standing in their way before. This is particularly evident when comparing the 7970 to the 6970, where the 7970 enjoys a consistent 62% performance advantage. It’s simply an incredible turnabout to see the 7970 do so well when the 6970 did so poorly.

Of course if this performance boost really was all about compute shaders, it raises a particularly exciting question: just how much higher could AMD go if they had DCLs? Hopefully one day that’s an answer we get to find out."

Its pretty interesting stuff which could lead to performance gains if amd gets this in action.

Gotta be honest that loses that reviewer ALL credibility if they don't know what the performance difference in Civilization V is due to... its NOTHING to do with "DCL" directly (tho threaded command queues do give a benefit to geometry manipulation performance - which should be a dead give away as to why there is such a performance difference)......

EDIT: Theres actually 2 reasons for the big performance difference - the 2nd one is the use of GPU compute for transcoding custom texture compression but thats not the bigger performance hit of the 2.

EDIT2: For those that still think "DCL" is some nVidia trick to retard AMD performance:

Unfortunately, because no other games have used this feature yet, neither Nvidia nor AMD have publically released threaded drivers, so users may not experience all the benefits just yet. We decided to keep threading enabled for Civilization V, however, because we are continuing to work closely with Nvidia and AMD on their support for multi-threading. We expect publically available threaded drivers shortly.

 

[TheRealDeal]

I dont think its there to give amd a disadvantage. I think its there to boost nvidias performance and does a good job by the looks of it. Nvidias performance in civlization was near amds before they brought out a good driver if i remember correctly. The fact that amd are free to do the same means to me theres nothing going on here.

 

[J.D]

Ah yes, the good old, "I don't understand X and my favourite card manufacturer is losing out cos of X so it must be NVIDIA's fault" argument.

Direct3D 11 deferred context command lists are a multi-threading feature in DirectX 11 defined by microsoft that NVIDIA added driver support for in CUDA 4.0 (iirc). If AMD is limping along on this front without driver support for a multithreading DX11 feature (and they claim to be all up in DX11!) it's their own damn fault. Sheesh! The level of fanboism here is ridiculous.

Wind your neck in with calling me a fanboy!!. What, so because I've not had any Nvidia cards in my system since the 8800GTX that means I'm a fanboy?. I don't know how you even reached that conclusion from my post unless you're the one with invested interest in maybe the green team?. Otherwise why would your attitude drop to a level where you will act like this?. If you're trying to tell me the £200 I paid for my 6950 last year was a fanboy move when today my card is more expensive or on the same level and when unlocked and overclocked, they are even better. Sorry I went with AMD this time . Yes, I'm confused at your fanboy statement. Please clarify?.

I could point the finger at you for your reaction to something that actually happened in the past regarding DX10.1 and Assassins Creed but I won't drop to your level. Hopefully you won't keep jumping the gun on people like you have with me as it's attitudes and mannerisms like yours that ruin discussions.

From the article:

The big question of course is why are we only seeing such a limited lead from the 7970 here? BF3 implements a wide array of technologies so it’s hard to say for sure, but there is one thing we know they implement in the engine that only NVIDIA can use: Driver Command Lists, the same “secret sauce” that boosted NVIDIA’s Civilization V performance by so much last year. So it may be that NVIDIA’s DCL support is helping their performance here in BF3, much like it was in Civ V.

If you look at the time of posting (yes nearly 4am) and the short read of that article regarding only Battlefield 3 I apologise not knowing what I was talking about but when a review site states that only Nvidia can use the Command Lists plus what happened with Assassins Creed, I have a right to feel frustrated if the card in my system is or could be affected. I'm sorry I didn't understand what the Driver Command Lists were apart from reading that Nvidia are the only ones who can use them. I must just be a complete and utter fool to believe what was written in an article. You happy now?.

 

[xsistor]

I dont think its there to give amd a disadvantage. I think its there to boost nvidias performance and does a good job by the looks of it.

That's precisely what it is.

EDIT: I can't vouch for it being the cause for the particular game's performance increase, but the rest of it anyway NVIDIA supporting a core DX11 feature like DCDL does not impact what AMD wants to do.

 

[Rroff]

I can't vouch for it being the cause for the particular game's performance increase, but the rest of it anyway NVIDIA supporting a core DX11 feature like DCDL does not impact what AMD wants to do.

Tessellation maybe? hate to state the obvious but thats the big performance difference between nVidia and AMD in that game... and suprise suprise the 7970 with upto 4x the geometry throughput of the 6970 manages a lot better.

 

[Owenb]

it'll all come down to pricing - similarly if the 7950 is unlockable that will also be a very interesting card to watch

28nm is a new process so expect a lot the 7950 to be defective 7970 which would make unlocking, if possible, a complete lottery and one that most will lose for the first few months.

 

[J.D]

AMD did exactly the same thing with shogun 2 and the Dirt series so both are as bad as each other.

Could you provide me with some detail on this as I find it very difficult to believe?. AMD had DX10.1 and will have DX11.1 so are you saying that AMD used technologies they had but Nvidia didn't have which provided AMD cards with a boost?. I know it sounds one sided but hear me out, if Nvidia asked Ubisoft to patch out the benefits for Assassins Creed for AMD cards then what did AMD do in return with Dirt and Shogun?. Was it the same thing as with Assassins Creed?. Nvidia had the lead in an AMD sponsored game so AMD asked the developers to patch out the ability to use the full potential of the Nvidia card?.

If AMD have done this, then yes, I'm disgusted. All I stated was that if it happened to me regardless of what card (AMD or Nvidia) I had in my system that I would be annoyed. Why is everyone jumping to conclusions that this is an attack on Nvidia when all I'm doing is relating to something that HAPPENED in the past?. Yes I was stupid to assume it was anything like the Assassins Creed fiasco but I only read a tiny little bit. Luckily enough, a polite member took the time to correct me in a friendly fashion (not you James J) .

 

[TheRealDeal]

Tessellation maybe? hate to state the obvious but thats the big performance difference between nVidia and AMD in that game... and suprise suprise the 7970 with upto 4x the geometry throughput of the 6970 manages a lot better.

I am not sure its just tesselation as amd could easily sort that out in there drivers. I dont have the game but i can easily turn tesselation levels up or down or use the amd optimised option which i think means use amds driver level amount. It probably has some impact but other games with heavy tesselation dont gimp the 6970 quite so much to the point where nvidia has such a lead.