Battlefield 3 total system resource benchmarks.

[mmj_uk]

Put simply the game is no longer using the paging file. I would imagine with it enabled the game uses it no matter how much vram you have. It looks like a set percentage to me in the game code, just to make sure that *most* of the textures are being loaded into vram, but with a certain percentage being laid off into the paging file just to make sure there is no memory overflow stack error.

The paging file you disabled has nothing to do with VRAM it's merely when the system runs out of system RAM it uses the HDD for 'virtual memory'. A GPU using system RAM used to be known as 'AGP Aperture' back in the AGP days and it's a completely different animal to the Windows paging file.

If anything all that your tests show is that either you have insufficient system RAM or Windows is poor when it comes to system memory management.

I'd also like to know how you managed to run these tests fairly? what BF3 benchmark mod did you use to ensure both runs were identical?

 

[steven_belfast]

That sucks

What hardware are you running?

Edit nvm just checked your validator.

All in all :

Corsair Hydro H100
Asus Maximus IV Extreme
Intel Core i7-2600K 3.40GHz
Sapphire ATI Radeon HD 6990
Corsair Obsidian 800D (6GB/s)
G.Skill RipJawsX 8GB (2x4GB) DDR3 PC3-12800C8 1600MHz
Corsair Force Series 3 120GB SATA 6Gb/s Solid State Drive
Corsair Professional Series AX1200 High Performance 1200W

 

[straxusii]

I'd love a way to be able to restrict vram on a card so we could effectively test with 1,1.5,2,3GB of vram on that same card.

I see a number of posts where people show vram usage in excess of 1GB, 1.5GB and 2GB but it doesn't necessarily follow that cards with less will experience a material degradation in performance.

 

[ALXAndy]

The paging file you disabled has nothing to do with VRAM it's merely when the system runs out of system RAM it uses the HDD for 'virtual memory'. A GPU using system RAM used to be known as 'AGP Aperture' back in the AGP days and it's a completely different animal to the Windows paging file.

If anything all that your tests show is that either you have insufficient system RAM or Windows is poor when it comes to system memory management.

I'd also like to know how you managed to run these tests fairly? what benchmark did you use to ensure both runs were identical?

Did you read the whole thread?

It basically answers all of the questions you have. Most notably -

1. Until the GTX 680 it seems that cards could not use physical memory when they ran out of vram. So, as Nvidia call it they "texture stream".

The tests I ran show that when the paging file is disabled the physical ram usage does not change. Therefore it is "texture streaming" from somewhere, and that somewhere is not system memory.

2. When you disable the paging file the game can crash. Why? And why doesn't it crash for me? Put simply the game must be using the paging file. We already knew this, however, and this is why we celebrated the 680 using system ram. Even though it's still a million miles from being ideal it's got to be better than using the page file.

As for insufficient system ram? it didn't change. I know it's been pointed out that windows memory monitor isn't very accurate, but the results were identical.

3. If you had read the opening post you would have seen how I ran the tests. However, as I pointed out a few times they are not accurate benchmarks and should not be taken as so. They are simply there to show what happened when I disabled my paging file and ran the game on the same level.

No BF3 benchmark is ever going to be completely accurate. Any one can work that out. There is no "set routine" benchmark that does exactly the same things during the FPS recording.

What I mean is when some one benchmarks BF3 they have to play it. And when they play it each benchmark is going to offer a multitude of scenarios and differences as it's not a set benchmark.

Metro 2033 would probably have been better. But, this whole thread was not about Metro 2033. So I simply benchmarked it using the only method available to me, which is to play the game.

In other words I am trying to make people understand that sometimes what you see isn't exactly what you get. All it takes is one spike in FPS or one extra explosion and you have millions of people declaring one product better than the other.

Stupid really.

 

[James J]

Hows this lol
Ok only 1 of them in over the 3gb mark but the rest are dam close to it.

snipped images

Sorry mate but that isn't 3gb of VRAM being used if you are using 2x 7970s. Crossfire memory usage is combined when being read by MSI afterburner, so you aren't even using 1.5gb per card mate.

If you are using 2x 580 3gb cards, I'll retract my statement, yet I own those cards and I get no where near 2gb per card, let along 3gb.

Biggest VRAM eating game I've played so far is serious sam 3 which used 2.1gb of VRAM per GPU, which is about 700 more than the max I've ever seen in BF3 on ultra.

 

[steven_belfast]

Sorry mate but that isn't 3gb of VRAM being used if you are using 2x 7970s. Crossfire memory usage is combined when being read by MSI afterburner, so you aren't even using 1.5gb per card mate.

If you are using 2x 580 3gb cards, I'll retract my statement, yet I own those cards and I get no where near 2gb per card, let along 3gb.

Biggest VRAM eating game I've played so far is serious sam 3 which used 2.1gb of VRAM per GPU, which is about 700 more than the max I've ever seen in BF3 on ultra.

No i am using a 6990

 

[andybird123]

you are running crossfire then, you only have 2GB per GPU so you can't be using 3GB of VRAM

 

[wolvers]

I see a number of posts where people show vram usage in excess of 1GB, 1.5GB and 2GB but it doesn't necessarily follow that cards with less will experience a material degradation in performance.

They won't, unless they are in a scenario where the available VRAM is all used up and then performance dies. Gibbo has proved it already with his 1gb and 2gb 560 test. It amazes me that people still argue against it (not saying that you are BTW, I don't remember seeing your posts on the subject).

 

[andybird123]

I'd love a way to be able to restrict vram on a card so we could effectively test with 1,1.5,2,3GB of vram on that same card.

I see a number of posts where people show vram usage in excess of 1GB, 1.5GB and 2GB but it doesn't necessarily follow that cards with less will experience a material degradation in performance.

here is your answer;

The Ultra Texture Quality setting for example is designed specifically for GPUs with 1.5GB or more of VRAM.

If you notice that your system is constantly stuttering or momentarily freezing, or you frequently see textures being streamed in while you're moving around, then lower Texture Quality to see if this helps reduce the issue.

http://www.geforce.com/Optimize/Guides/battlefield-3-tweak-guide/#5

 

[wolvers]

You can do it yourself. Get Skyrim running at high settings on a 1gb card and keep adding HD texture mods until it's used up all the VRAM = welcome to stutter city.

 

[mountaincow]

No i am using a 6990

The same applies, your maximum available VRAM is 2GB, afterburner will add up the VRAM used by both GPU's. Therefore if you halve the number you see displayed then you get your actual usage.

 

[Macro]

As you can see from these screens. This is 1080P @ full ultra (all the trimmings)I have the page file on also.

Ewwww... nasty image quality...

 

[straxusii]

here is your answer;



http://www.geforce.com/Optimize/Guides/battlefield-3-tweak-guide/#5

Hi Andy, I'm talking in general it would be useful. Also I would be interested to test ultra settings simulating 1GB vram on my card for example.

 

[Gregster]

Ewwww... nasty image quality...

lol I was thinking the same after I uploaded

 

[Gregster]

And of what relevance is that?

Please don't tell me that after nearly a full 24 hours you still fail to grasp what this thread is about?

Honestly is it really that? do you really not understand what is going on?




That's what people need to do here. We've now established how much vram it uses. What's hilarious is if he disables his paging file then he will highly likely see a boost in FPS.

I was showing how BF3 caches textures. Carry on insulting me with your "Oh I am so much better than you" attitude.

I get exactly what you were saying so please don't speak down to me. You seem to think your special or above everybody. You need to lighten up and stop trying to be a know it all.

If you don't like what I say, then thats your choice but trying to put me down will not work sunshine.

 

[andybird123]

Hi Andy, I'm talking in general it would be useful. Also I would be interested to test ultra settings simulating 1GB vram on my card for example.

yeah, it would be useful in terms of settling this type of argument much easier as people could see for themselves the effect of VRAM limitation

though from the tone of this thread, among others, even when faced with clear evidence that contradicts a particular viewpoint they will blindly ignore it and claim that their personal choice of graphics cards can play a certain game at certain settings when they don't even own a copy of the game and they refuse to believe people who have tried that game at those settings even with the other provisos imposed (such as system RAM requirements)

 

[straxusii]

yeah, it would be useful in terms of settling this type of argument much easier as people could see for themselves the effect of VRAM limitation

though from the tone of this thread, among others, even when faced with clear evidence that contradicts a particular viewpoint they will blindly ignore it and claim that their personal choice of graphics cards can play a certain game at certain settings when they don't even own a copy of the game and they refuse to believe people who have tried that game at those settings even with the other provisos imposed (such as system RAM requirements)

Totally, it's a shame this forum gets so inflammatory and abusive, it's got to the point where reasoned debate is not possible.

For me being able to restrict vram usage and run tests would be great for making informed buying decisions.

At the moment it seems we are all shooting in the dark and going with more vram just to be safe.

I'd love to be able to restrict my card and then you'd be able to tell what amount of vram makes performance fall off a cliff in a completely apples to apples environment.

 

[steven_belfast]

The same applies, your maximum available VRAM is 2GB, afterburner will add up the VRAM used by both GPU's. Therefore if you halve the number you see displayed then you get your actual usage.

you are running crossfire then, you only have 2GB per GPU so you can't be using 3GB of VRAM

I see.
Well to be honest i am not that fussed about these things as long as the game runs smooth thats the main thing. lol

 

[ALXAndy]

Totally, it's a shame this forum gets so inflammatory and abusive, it's got to the point where reasoned debate is not possible.

For me being able to restrict vram usage and run tests would be great for making informed buying decisions.

At the moment it seems we are all shooting in the dark and going with more vram just to be safe.

I'd love to be able to restrict my card and then you'd be able to tell what amount of vram makes performance fall off a cliff in a completely apples to apples environment.

It's a crying shame. Mostly because the whole idea of the thread wasn't to argue, but set out with a purpose to show some points.

Nvidia seem to know the exact sizes of the textures. I would imagine they got that info from DICE. So their guide is very interesting indeed, and very useful.

 

[Vir]

For every game I play I set the Disable Desktop Composition/Disable Visual Themes flags on the .exe Properties Compatibility tab. Even in Windows 7 that is an extra 150MB of VRAM saved right there.

It turns Aero On/off quickly and painlessly automatically no matter where I launch the game from.

I'll tell you what! Just found out you can't do that anymore in Windows 8! I think that is because of the whole metro user interface thingy. Baaaad bad bad.