Known/suspected games to eat more than 1GB video memory at 1920x1200

[Paracelsus]

Just done another quick test. 5 min playing Crysis Warhead, through dense jungle + combat scenes, I pick jungle as that's where I get lowest FPS so it makes sense to me this kind of scenery will use the most VRAM.

1920x1200 - 690MB
1920x1200 4xAA - 980MB
1920x1200 16xAA - 1026MB (why such a small jump from 4 to 16xAA I wonder?...)

So yeah I can believe that a lot of games with AA push cards over 1GB at 1920x1200. Without AA...they must be pretty badly coded to do so IMO, considering crysis/crysis2 don't and look better than every other game (metro 2033/stalker CoP excluded)

I wouldn't know as I never see the need for AA at 1920x1200 tbh

 

[Buckster]

^^ is this with 1 gb card though ? as if you thats why it won't go above 1026 vs 16AA ...

 

[Paracelsus]

Nope 1280mb GTX470, sorry

btw still only have 2GB RAM so not hitting the 3.5GB 32-bit limit either

 

[bhavv]

Regarding the comments saying that a game is choppy because its exceeding the Vram limit, try the same game again on two of the same 1 Gb cards in SLI or crossfire and I can guarantee you that it wont be choppy anymore, proving that the claim that it is the Vram causing choppiness is incorrect.

If a game is choppy when using a single card with details maxed out and 4x AA, well durrrrr? Its absolutely no evidence that it is caused by not having enough Vram, it just shows that this particular single card is not powerful enough on its own to run that game.

May I ask how much vram do you have on your card? If you only have 1GB, a simple example would be to run the ending of Crysis Warhead.

I have two 1 Gb GTX 560s in SLI running at 950 Mhz. They played Metro maxed out with 4x AA absolutely fine without any noticable lag for about 30-60 minutes, but I dont enjoy FPS games, I just got Metro included with the THQ pack on steam that was only £24 for loads of games.

One other thing I notice, people complaining about lag and slowdown only tend to have a single card, and they are trying to play games maxed out with lots of AA. Then they wonder why the game is slow and start blaming their Vram? LOL.

Blame your single card instead and buy a crossfire or SLI setup if you want to game at 1920x1200 with maxed everything and lots of AA.

 

[drifting]

Regarding the comments saying that a game is choppy because its exceeding the Vram limit, try the same game again on two of the same 1 Gb cards in SLI or crossfire and I can guarantee you that it wont be choppy anymore, proving that the claim that it is the Vram causing choppiness is incorrect.

If a game is choppy when using a single card with details maxed out and 4x AA, well durrrrr? Its absolutely no evidence that it is caused by not having enough Vram, it just shows that this particular single card is not powerful enough on its own to run that game.



I have two 1 Gb GTX 560s in SLI running at 950 Mhz. They played Metro maxed out with 4x AA absolutely fine without any noticable lag for about 30-60 minutes, but I dont enjoy FPS games, I just got Metro included with the THQ pack on steam that was only £24 for loads of games.

One other thing I notice, people complaining about lag and slowdown only tend to have a single card, and they are trying to play games maxed out with lots of AA. Then they wonder why the game is slow and start blaming their Vram? LOL.

Blame your single card instead and buy a crossfire or SLI setup if you want to game at 1920x1200 with maxed everything and lots of AA.

Good, so you have 1GB cards. I used to have 2 x 5870 1GB in CrossfireX and I had the choppy lag during Crysis Warhead ending. Try it out yourself.

A simple example would be to run the ending of Crysis Warhead. During the realtime animated cinema scene in which Psycho fights the Korean boss, every time when the camera switches, the game would lag for like 0.2 seconds or so, which is pretty obvious on 1GB cards. However 2GB cards doesn't have that problem.

 

[bhavv]

Good, so you have 1GB cards. I used to have 2 x 5870 1GB in CrossfireX and I had the choppy lag during Crysis Warhead ending. Try it out yourself.

You would probably get the same lag with 2 Gb cards of the same GPU, you have less lag now because 6950s are significantly faster than your 5870s were, not because they have more Vram.

The GTX 560s are on par with the 6950, so my performance wouldnt be much worse than yours.

As for lag in games like Metro 2033, it will happen regardless of having 1 or 2 Gb of Vram because the GPU itself is actually not powerful enough for the game:

http://www.hardwarecanucks.com/foru...s/40165-amd-radeon-hd-6950-1gb-review-12.html

1 Gb 6950 minimum frames = 19, 2 Gb 6950 minimum frames = 23, the difference is minimal and you would still notice lag on the 2 Gb card, maybe just not as much.

That is easilly overcome if you simply go crossfire or SLI.

How about you put your examples of 1 Gb vs 2 Gb Vram in a video for everyone to see, making sure to use cards with the same GPU and test environments?

 

[bhavv]

Also, as for Dragon Age II, heres the notes from the latest 270 Nvidia drivers:

GeForce GTX 580:

Up to 516% in Dragon Age 2 (SLI 2560x1600 8xAA/16xAF Very High, SSAO on)
Up to 326% in Dragon Age 2 (1920x1200 8xAA/16xAF Very High, SSAO on)

GeForce GTX 560 Ti:

Up to 461% in Dragon Age 2 (SLI 1920x1200 8xAA/16xAF, Very High)
Up to 241% in Dragon Age 2 (1920x1200 4xAA/16xAF, Very High)

 

[drifting]

You would probably get the same lag with 2 Gb cards of the same GPU, you have less lag now because 6950s are significantly faster than your 5870s were, not because they have more Vram.

The GTX 560s are on par with the 6950, so my performance wouldnt be much worse than yours.

As for lag in games like Metro 2033, it will happen regardless of having 1 or 2 Gb of Vram because the GPU itself is actually not powerful enough for the game:

http://www.hardwarecanucks.com/foru...s/40165-amd-radeon-hd-6950-1gb-review-12.html

1 Gb 6950 minimum frames = 19, 2 Gb 6950 minimum frames = 23, the difference is minimal and you would still notice lag on the 2 Gb card, maybe just not as much.

That is easilly overcome if you simply go crossfire or SLI.

How about you put your examples of 1 Gb vs 2 Gb Vram in a video for everyone to see, making sure to use cards with the same GPU and test environments?

Again, for the last time, I would say, try it yourself, not just quoting numbers from reviews. There are many things that fps data cannot tell, especially when reviews do not necessarily cover stressful scenes.

Making a video is not applicable, because the video would be recorded at 30 fps or even 24 fps, which is far too low to make the choppy lag obvious. It's also time consuming. It's not a bad idea to doubt about my OP, it can never be a bad idea not to trust someone, however I strongly recommend you to try it out yourself use your own settings. Only your practice can tell the truth, right?

 

[bhavv]

I have tried it myself with Metro 2033 and TDU2 (which was shown to use over 1250 Mb Vram by others on this forum).

Neither had any noticable lag from which I could conclude that 1 Gb of Vram isnt enough.

Hence in my experiences, your opinion that 1 Gb Vram isnt enough at 1920x1200 is completely false.

You yourself didnt even notice any performance increase from going from 1 Gb Vram to 2 Gb, you got better performance because you went from 5870s to 6950s. In this case, it has nothing at all to do with Vram, you have faster cards giving a completely invalid comparison of 1 Gb vs 2 Gb vram.

 

[Kevin1234]

Just done a bit of testing with crysis benchtest dx9 - 5850x3 - 59fps 11.3 drivers vram 1023MB smooth but with dx10 1423 vram stuttering gpus at 99% usage so these drivers needs fixing.......

 

[drifting]

I have tried it myself with Metro 2033 and TDU2 (which was shown to use over 1250 Mb Vram by others on this forum).

Neither had any noticable lag from which I could conclude that 1 Gb of Vram isnt enough.

Hence in my experiences, your opinion that 1 Gb Vram isnt enough at 1920x1200 is completely false.

You yourself didnt even notice any performance increase from going from 1 Gb Vram to 2 Gb, you got better performance because you went from 5870s to 6950s. In this case, it has nothing at all to do with Vram, you have faster cards giving a completely invalid comparison of 1 Gb vs 2 Gb vram.

Yet again you ignore my description and suggestion of tests about Crysis Warhead ending scene, and you claim you are even satisfied with the performance of 560 SLI in Metro 2033. My understanding is that you don't need any high-end graphics card for fps over 30, and you are insensitive to choppy lags. If you disagree with this, then could you please run the official metro2033benchmark.exe at 1920x1200 with everything max'ed, using "MSAA 4X" instead of "AAA", except PhysX, and post your results? Sample results for my 2 x 6950 2GB CF:

By the way, 560 SLI is by no means superior than 6950 CF under high resolutions. http://www.chiphell.com/en/graphics/hd-6950-crossfirex-vs-gtx-560-ti-sli/3/. Note that this review didn't even use "MSAA 4X", because it would kill 1GB cards. You could also try to list which games a GTX590 can beat an HD6990 under 2560x1600 4AA 16AF, even when GTX590 has 512MB vram per GPU than your 560 SLI. The counterpart of 6950 is supposed to be GTX570.

 

[bhavv]

Yet again you ignore my description and suggestion of tests about Crysis Warhead ending scene, and you claim you are even satisfied with the performance of 560 SLI in Metro 2033. My understanding is that you don't need any high-end graphics card for fps over 30, and you are insensitive to choppy lags.

By the way, 560 SLI is by no means superior than 6950 CF under high resolutions. http://www.chiphell.com/en/graphics/hd-6950-crossfirex-vs-gtx-560-ti-sli/3/ You could also try to list which games a GTX590 can beat an HD6990 under 2560x1600 4AA 16AF, even when GTX590 has 512MB vram per GPU than your 560 SLI. The counterpart of 6950 is supposed to be GTX570.

Im never going to play Crysis all the way through the ending scene, so yes it is an ignored game for me. Why should I base my purchase opinions around FPS games that I'm never going to play?

As for your GTX 560 link, its yet another lame comparison of a 560 clocked at 822 Mhz while most come overclocked out of the box around 900-950 Mhz. Heres a better link comparing a GTX 560 that has been available since launch at a lower price than most others:

http://www.bit-tech.net/hardware/graphics/2011/01/27/msi-geforce-gtx-560-ti-1gb-review/3

An overclocked 560 outperforms a GTX 570 (see page 8 of the above review), without the same risk of exploding that the 570 has. I also never claimed that the 560 is superior to the 6950, I said it is equal, which it very definitely is unless you are looking at bogus results with them downclocked to 822 Mhz.

The 570 is the equivalent to a 6970, not a 6950 as you think it is, while the 560 is the equivalent to the 6950, but with much better overclocking headroom than most other cards out there.

Nvidia purposefully downclocked both the GTX 460 + 560 well below what the chips are capable of and allowed AIB's to make custom overclocked versions from the day of release, something which is never found on higher end Nvidia cards or on ATI cards.

And thats all completely offtopic ofc, but again, 1 Gb Vram is not a limitation at 1920x1200 in current games, any current PC is able to cope if a game exceeds 1 Gb Vram at that resolution.

 

[drifting]

Im never going to play Crysis all the way through the ending scene, so yes it is an ignored game for me. Why should I base my purchase opinions around FPS games that I'm never going to play?

As for your GTX 560 link, its yet another lame comparison of a 560 clocked at 822 Mhz while most come overclocked out of the box around 900-950 Mhz. Heres a better link comparing a GTX 560 that has been available since launch at a lower price than most others:

http://www.bit-tech.net/hardware/graphics/2011/01/27/msi-geforce-gtx-560-ti-1gb-review/3

An overclocked 560 outperforms a GTX 570 (see page 8 of the above review), without the same risk of exploding that the 570 has. I also never claimed that the 560 is superior to the 6950, I said it is equal, which it very definitely is unless you are looking at bogus results with them downclocked to 822 Mhz.

Nvidia purposefully downclocked both the GTX 460 + 560 well below what the chips are capable of and allowed AIB's to make custom overclocked versions from the day of release, something which is never found on higher end Nvidia cards or on ATI cards.

And thats all completely offtopic ofc, but again, 1 Gb Vram is not a limitation at 1920x1200 in current games, any current PC is able to cope if a game exceeds 1 Gb Vram at that resolution.

Post your results of metro2033benchmark.exe then.

 

[bhavv]

Post your results of metro2033benchmark.exe then.

Ok, I just found it, give me a few mins.

 

[Mr Krugga]

OMG, this thread again

I won't even try to argue the point about games using more than 1GB because some are obviously capable of doing that.

That doesn't change the fact that 99% of the new titles see no benefit of having more than 1GB of vRAM and excessive amount of it rarely increases FPS count by more than 5%. There is also no hidden smoothness to this subject and no minimum FPS differences.

All that can be drawn from multiple reviews that compared HD5870 1GB to 2GB and, I believe, HD6950 1GB to 2GB.

Games can use more vRAM but don't need it (unless you game at uber high resolution with multiple AA applied).

Also genuine LOL at the scientific comparison done by the OP, well done for taking HD5870 1GB and HD6970 2GB into your research

 

[Buckster]

^^ I probably agree ....

but AA is nice even at 1900x1200 - I disagree with thoughts that at 1900 you don't need AA - in fact at higher res's it makes more of a difference IMO

try something like New Vegas at 1900x1200 no AA - then with 8xAA - and 8xAA looks far far better

 

[bhavv]

Ok I did the Metro bench, and no, I honestly cannot see how having 2 Gb Vram on the same GPU would make the benchmark completely smooth and remain over 30 FPS:



And it only used up 932 Mb Vram, so its not like its slowing down because of Vram usage.

Also, max temp on GPU1 at 1 Ghz = 81 degrees. Eat that GTX 570, 480 + 580

but AA is nice even at 1900x1200 - I disagree with thoughts that at 1900 you don't need AA - in fact at higher res's it makes more of a difference IMO

I wouldnt play any game without AA, but dont forget that theres the new DX11 morphological AA that doesnt cause any performance impact. Only AMD have it atm, but Nvidia are working on a form that doesnt blur out text.

 

[drifting]

Ok I did the Metro bench, and no, I honestly cannot see how having 2 Gb Vram on the same GPU would make the benchmark completely smooth and remain over 30 FPS:



And it only used up 932 Mb Vram, so its not like its slowing down because of Vram usage.

Also, max temp on GPU1 at 1 Ghz = 81 degrees. Eat that GTX 570, 480 + 580



I wouldnt play any game without AA, but dont forget that theres the new DX11 morphological AA that doesnt cause any performance impact. Only AMD have it atm, but Nvidia are working on a form that doesnt blur out text.

So, thanks for your nice play, to use 4AF instead of 16AF, and disabling DOF. If using 16AF and DOF like I did, and you still see so little vram usage, it means this benchmark doesn't cover the stressful scenes, such like the stage "Child" from Chapter 4, and you can throw away all the numbers from all the reviews. If you see increased vram usage with 16AF and DOF, you would even get worse min fps. Do I have to stress again that the vram usage is already pushing your little 1GB to the edge, even if you only use 4AF without DOF?

 

[bhavv]

AF doesnt even create that much of a performance hit, I simply forgot to change it in the settings.

But you can clearly see that Vram use does not cause performance issues, nor would having a 2 Gb card increase the minimum frame rate in cases where only 932 Mb is being used and the game is still lagging.

Who could care about seeing worse or slightly better minimum FPS than 9 with a 2 Gb card? Its not like the game is going to be any more playable if 2 Gb of Vram boosts it up to around 12.

 

[drifting]

AF doesnt even create that much of a performance hit, I simply forgot to change it in the settings.

But you can clearly see that Vram use does not cause performance issues, nor would having a 2 Gb card increase the minimum frame rate in cases where only 932 Mb is being used and the game is still lagging.

So is it because you are afraid to use the settings I was asking for, for an apple-to-apple comparison? Even when you cheat with 4AF while turning of the expensive DOF, my min fps is 44% higher than your best min fps, while it's just pushing your vram to the edge.

Well, I understand that you might be having problems producing appropriate results with exactly the "max" settings I have been asking for all the time, because according to my previous experience maybe it's not easy for your tiny 1GB If you dare not even run this benchmark (which cannot be called stressful at all, compared with the stage "Child" from Chapter 4) with max settings, how can you claim that 1GB vram is sufficient?