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

[drunkenmaster]

http://www.anandtech.com/show/3621/amds-radeon-hd-5870-eyefinity-6-edition-reviewed/6

Biggest gap is Crysis minimums, 15fps vs 8fps, neither are particularly playable, basically they've run out of gpu power before memory became an issue, even then the average is indistinguishable, what does that mean, has it run out for the entire benchmark, no, it almost certainly indicates for 1-2 seconds it came borderline but for the rest they were evenly matched. The drop to 15fps from 30fps average would be as clear a drop as the one to 9fps, both would be visually obvious and for the rest of the benchmark you'd also be able to tell no difference.

Battlefield, marginal difference, but at that point the cards themselves are unable to provide high enough average for you to play anyway.

http://www.anandtech.com/show/3621/amds-radeon-hd-5870-eyefinity-6-edition-reviewed/7

No real difference, horrible performance in some situations anyway.

http://www.anandtech.com/show/3621/amds-radeon-hd-5870-eyefinity-6-edition-reviewed/8

Would have been great if they showed the 1gb card versions in with that set as you would almost certainly see max performance vs 1fps on the 1gb versions. But shows, look at some of these numbers, 6x 1920x1080 screens, on a 2gb card, some with 4xaa, no memory limits.


ANyway, what that shows is, at 2560x1600 max quality, 4 or 8xaa, incredibly marginal effect of memory, and where it happens, it offers no playable difference to a 2gb card.

Again I'll point out, CONFIRM YOUR READINGS, go e-mail Nvidia and ask them what gpu-z is reading, ask a reviewer to ask, a reviewer might know, drop anandtech an e-mail.

Or most importantly, show the difference in performance, with the same settings, where a 1gb card is unplayable and a 2gb card offers a completely different performance. No, one frame in a benchmark does not constitute that, not even close.

 

[drifting]

You are proving the point AGAINST you, AGAIN, and still miss it.

Low performance, low min fps in NO WAY automatically means running out of vmem and I really really wish you'd get that.

Look WHERE your min fps is, exactly the same place Bhav's minimum is........ because of memory right, no because thats the hardest part of the test.

You are getting 28fps here, at the LOWEST res on the LOWEST settings, how is 7fps, at a much higher res and full settings........ bad. Answer it isn't, its exactly where you'd expect the card to be. He doesn't have low average FPS due to memory he has low average fps because with max settings the card isn't capable of going much much faster.

I hope you could still read posts and pictures carefully. I have already said "Keep in mind that the min fps is not always due to vram shortage. I use Metro 2033 as partially an example of case (a) in the OP, where even the average fps is affected." Also, from the pictures you can tell that bhavv is really getting min fps as low as below 10, while for me the actual min fps was about 50, and the number 28 came from Frame 6 which was merely the beginning of the benchmark when the computer just finished loading.

This is my issue with every single one of your conclusions, you pick up ANY drop in performance and utterly utterly ignore any other answer and insist it confirms low memory.

THe simple fact is you are wrong, your own results that I've seen prove this.

"I used settings that SAID it was using 1.2gb, on a 1gb card and saw no performance drop".

Mind reading my posts more carefully? "You have misunderstood here. I mean GTA4 itself (the game engine) claims to need 1.2GB, while MSI afterburner says only 1GB. The numbers from MSI afterburner or GPU-Z are still trusty, and I made the list according to these. There is no way to get everything max'ed out if you have only 1GB - the game engine simply doesn't allow that. Therefore if you only have 1GB card, you can either bypass it with some 3rd-party cracks along with trojan, or use medium settings and be done with it." "I haven't said I've used settings on a 1gb card with the game claiming it needs 1.2gb with no drop in performance. When I was using 1gb cards, the game didn't even allow me to max out the settings. I have only said other people were using 1.5gb card to capture a screenshot of vram usage, when the game asks for 1.2gb to max the settings, msi afterburner measures only about 1gb."

Look up benchmarks I've linked to that catagorically show what happens when you run out of memory, 30fps to 0.5fps.

When you drop out of gpu memory to system memory, you're talking about magnitudes less in performance, a split second or low performance is not due to memory limitations.

Memory limitation if you're talking about gpu's, or a program in system memory, when you run out whatever you are doing, a game, or a tonne of big photo's open in photoshop, performance goes from ok to an absolute, painful, OMG crawl. You have failed to show this in anything you've said so far. You are completely confusing minor drops in performance, or simply games with settings too tough for that particular gpu with the EPIC, MASSIVE, unquestionable, unplayable drop in performance you get from memory limits.

Look up reviews on eyefinity, on triple screen resolutions, on 2gb vs 1gb cards in these situations, the difference is patently obvious, and completely contradictory to your "oh no, performance has slightly dropped" levels.

Again, for memory limits you aren't looking for ONE frame in an entire benchmark, especially at the repeatedly confirmed toughest point in the benchmark which even showed up in your low res/low settings result, you're looking for massive, massive completely game playing experience destroying performance drop.

Please read the notes in the OP again. If you don't care about category (b) specified in the OP then this list is not for you. Why insisting your theory here?

 

[drifting]

Again I'll point out, CONFIRM YOUR READINGS, go e-mail Nvidia and ask them what gpu-z is reading, ask a reviewer to ask, a reviewer might know, drop anandtech an e-mail.

Again think of it, how would people struggle to develop CUDA programs under Windows if they don't have the readings of vram usage? It would be like walking blindly!

Just a quick look into the CUDA Toolkit Reference Manual, on page 189:

CUresult cuMemGetInfo (size_t free, size_t total)
Gets free and total memory

And on page 219:

4.31.2.30 CUresult cuMemGetInfo (size_t  free, size_t  total)
Returns in free and total respectively, the free and total amount of memory available for allocation by the
CUDA context, in bytes.
Parameters:
free - Returned free memory in bytes
total - Returned total memory in bytes

That's crystal clear I think.

 

[MOOGLEYS]

 

[arc@css]

All we need now is for Big Wayne to turn up.

 

[bigjimmyauk]

All we need now is for Big Wayne to turn up.

Is he actually still around? last time i saw him he was going to sue easyrider for internet bullying

 

[arc@css]

Is he actually still around? last time i saw him he was going to sue easyrider for internet bullying

I'm not sure, haven't seen any of his posts in a while. I'm thinking he got a bit fed up and went elsewhere. I never read any of the posts where those two clashed, bet it was amusing, amongst other things.

 

[bigjimmyauk]

I'm not sure, haven't seen any of his posts in a while. I'm thinking he got a bit fed up and went elsewhere. I never read any of the posts where those two clashed, bet it was amusing, amongst other things.

Yeah they were quite the pair, like a old married couple at times, i think half the posts in his overclocking thread was big vs easy LOL

good times

 

[pv123]

I play all my games (quiet a high percentage of those mentioned on post #1) and I have never EVER seen the 1GB mark being passed, and I have never seen issues because of high vram usage. I actually used to have constant lans next to a 6970 card (I have a 570GTX) and the performance were similar, we never came across a game which ran crap because of VRAM.

I would like to see real evidence of games running and suffering because of VRAM, even if you say that it uses more than 1GB, do you get performance issues? I have never seen that happen, I think that 1GB is enough for 99% of games at whatever resolution you want. Whoever says that VRAM is an important issue is 99% probably an ATI user trying to justify his investment.

 

[pv123]

You are proving the point AGAINST you, AGAIN, and still miss it.

Low performance, low min fps in NO WAY automatically means running out of vmem and I really really wish you'd get that.

Look WHERE your min fps is, exactly the same place Bhav's minimum is........ because of memory right, no because thats the hardest part of the test.

You are getting 28fps here, at the LOWEST res on the LOWEST settings, how is 7fps, at a much higher res and full settings........ bad. Answer it isn't, its exactly where you'd expect the card to be. He doesn't have low average FPS due to memory he has low average fps because with max settings the card isn't capable of going much much faster.

This is my issue with every single one of your conclusions, you pick up ANY drop in performance and utterly utterly ignore any other answer and insist it confirms low memory.

THe simple fact is you are wrong, your own results that I've seen prove this.

"I used settings that SAID it was using 1.2gb, on a 1gb card and saw no performance drop".

Look up benchmarks I've linked to that catagorically show what happens when you run out of memory, 30fps to 0.5fps.

When you drop out of gpu memory to system memory, you're talking about magnitudes less in performance, a split second or low performance is not due to memory limitations.

Memory limitation if you're talking about gpu's, or a program in system memory, when you run out whatever you are doing, a game, or a tonne of big photo's open in photoshop, performance goes from ok to an absolute, painful, OMG crawl. You have failed to show this in anything you've said so far. You are completely confusing minor drops in performance, or simply games with settings too tough for that particular gpu with the EPIC, MASSIVE, unquestionable, unplayable drop in performance you get from memory limits.

Look up reviews on eyefinity, on triple screen resolutions, on 2gb vs 1gb cards in these situations, the difference is patently obvious, and completely contradictory to your "oh no, performance has slightly dropped" levels.

Again, for memory limits you aren't looking for ONE frame in an entire benchmark, especially at the repeatedly confirmed toughest point in the benchmark which even showed up in your low res/low settings result, you're looking for massive, massive completely game playing experience destroying performance drop.

Well said.

 

[bigjimmyauk]

does anyone have a 2gb 5870 they can pit against my overclocked 5850 (which = 5870 speeds)

 

[drifting]

now re-run the bench at 1920x1080 and look at the minimums

Oops sorry I missed your post. Now it's the bench at 1920x1080 with min settings:



Uploaded with ImageShack.us

My min fps is still around 50 as you can see. (You should ignore the reading at frame 6 which is at the beginning of the session, because my computer was loading all the game contents, which is exactly what happens when people run out of video memory.)

 

[drifting]

I play all my games (quiet a high percentage of those mentioned on post #1) and I have never EVER seen the 1GB mark being passed, and I have never seen issues because of high vram usage. I actually used to have constant lans next to a 6970 card (I have a 570GTX) and the performance were similar, we never came across a game which ran crap because of VRAM.

I would like to see real evidence of games running and suffering because of VRAM, even if you say that it uses more than 1GB, do you get performance issues? I have never seen that happen, I think that 1GB is enough for 99% of games at whatever resolution you want. Whoever says that VRAM is an important issue is 99% probably an ATI user trying to justify his investment.

If you have a GTX 570, could you post your result of metro2033benchmark.exe, with max settings except PhysX to disable?

To be specific: DX11, 1920x1200, Very High, MSAA 4X, AF 16X, DOF Enabled, Advanced PhysX Disabled

 

[Mr Krugga]

 

[Buckster]

^^ wheres the minimum frame rates ?

 

[Mr Krugga]

^^ wheres the minimum frame rates ?

There isn't any being discussed in the OP where he talks about average framerates.

Anandtech doesn't test that either as minimum framerates are dependent on number of things and as most educated blokes would tell you, they are irrelevant in this case.

Enough with the bull****, you can have all the vRAM you want but that won't change the fact that it's barely useful in any games and certainly not necessary to have more than enough - which 1GB is by the way.

EDIT: just realised, the second result is minimum framerates for Crysis Warhead. There goes your answer.

 

[Marine-RX179]

That's pretty much it. Of all the comparison I saw between 6950 1GB and 6970 2GB, the frame rate (both minmium and average) for most games at 1920x1200 16xAF 4xAA are no more than 0-1 fps difference, and even the worst case scenerio such as Metro2033, the difference is like only 4fps even at 2560 res. I mean let's face it...how many game are going to be anywhere as demanding as Metro2033?

I seriously don't get peoples' sudden panic worrying about 1GB cards not good enough for 1920 res. And people should really be worrying more about the GPU not having enough grant, rather than 1GB being not enough for 1920 res.

 

[bhavv]

EDIT: just realised, the second result is minimum framerates for Crysis Warhead. There goes your answer.

There we go then, its proven.

2 Gb over 1 Gb gives a whopping 1 FPS improvement to minimum frames in Crysis Warhead. Wow, what an amazing improvement .

My min fps is still around 50 as you can see. (You should ignore the reading at frame 6 which is at the beginning of the session, because my computer was loading all the game contents, which is exactly what happens when people run out of video memory.)

Actually your min frame rate is 24. Why should anyone ignore your minimums when you dont ignore anyone elses?

 

[Scougar]

Amzing the amount of crap that gets flung in here. Everytime it's a mud flinging contest RATHER THAN actually investigating and drawing conclusions from it.

Well done to the OP for looking into this possible issue and those that are providing results.

What you REALLY need is to get a measurement of how much memory is being used in a scene rather than just the total being currently stored in memory. Knowing this figure will give you a much better reference point, and you will know the average memory usage. If this can be provided by afterburner then great, if not look for an application that can measure it, or write one.

 

[MOOGLEYS]

Don't quote me on this but I remember reading that Skyrim running max settings @ 1900x1200 uses around 1500mb of V-Ram.. I'm trying to find where I read this but not having much luck...