Battlefield 3 Performance Analysis

[harmony]

6950 2GB performance got trolled by 6950 1GB on 1920 res Think this pretty much confirm peoples' theory of "games will use more VRAM if available, but doesn't mean that the game NEEDS it".

And to think people spent like extra £50-£60 for the extra 1GB VRAM or upgrading from 5870 1GB to 6950 2GB for the sake of "preparing for BF3"

If I remember correctly, 'Going Hunting' does not use much vram. This review is useless on vram.

 

[harmony]

For vram issue, I recommend this review:

Bhavv and other followers of 1GB feel free to call this test BS.

 

[setter]

For vram issue, I recommend this review:

Bhavv and other followers of 1GB feel free to call this test BS.

From looking at the above ss, gtx 560ti's in sli are pretty similair to my overclocked 470's. On ultra preset, no blur, ssao on, high post aa, 16x af at 1920x1200, im seeing vram use of 1220mb ono, the only stuttering ive seen has been caused by lag, and the previous drivers i was using, updated to the latest betas, (285.79) and the game is running silky smooth.

 

[harmony]

From looking at the above ss, gtx 560ti's in sli are pretty similair to my overclocked 470's. On ultra preset, no blur, ssao on, high post aa, 16x af at 1920x1200, im seeing vram use of 1220mb ono, the only stuttering ive seen has been caused by lag, and the previous drivers i was using, updated to the latest betas, (285.79) and the game is running silky smooth.

Some users with 1GB vram have already reported lags that can be noticed when opening a sniper scope - very obvious lag loading textures of the objects far away from the player. They suspect that the driver has "optimized" vram usage by decreasing the "view distance". The cure is to use FXAA instead of MSAA to reduce vram usage.

 

[setter]

Yep, im using FXAA, the difference from using MSAA is so little its not worth the performance hit. Ill be holding out on any gpu upgrades till keplar so the 470's are more than enough to tide me over.

 

[tommybhoy]

Wait for next gen...not worth going from the 5870 to a 6950 2GB considering the cost and with only little improvement on performance.

A bit of apples and oranges?

As you often say, there is not much difference in them, which I usually agree.

As I have went from a 5870 to a 6950, I can compare.

Imo, there is far too much of a difference in fps, were talking as low as 15 fps with the 5870 max CCC settings in Witcher to minimums of 28 fps with the same settings using the 6950.

There is no way almost double the frame rate is down to core differences.

 

[harmony]

Yep, im using FXAA, the difference from using MSAA is so little its not worth the performance hit. Ill be holding out on any gpu upgrades till keplar so the 470's are more than enough to tide me over.

But I remember that by using FXAA instead of MSAA the screenshots taken would suffer from aliasing.

 

[drunkenmaster]

For vram issue, I recommend this review:

Bhavv and other followers of 1GB feel free to call this test BS.

Jesus, AGAIN.

Name one person who would switch on MSAA 4x to go from 80fps to 35fps? The answer is none, so again you've highlighted yet another card that ISN'T effected by memory in ANY SITUATION YOU WOULD ACTUALLY USE IT.

Top to bottom the only setup I would use 4xmsaa with would be the 570 sli, because its the ONLY setup that has enough performance out of those for a really good experience at that resolution........ and it isn't memory limited.

I wouldn't even run a single 580gtx with 4xmsaa from those results, it too has enough memory.

FXAA won't kill the 560gtx performance to unplayable levels, nor will it max out ram usage, this is why 0% of 560gtx would even think of using 4xmsaa, and the 1gb memory will not effect them when they don't.

You lose, AGAIN. As per always, you are NOT the first person, or the 100,000th person to know that higher settings uses more memory and there are situations more memory will help. The problem is your crusade is 1920x1080, and you have failed, again, to find a situation where a 1gb card is memory limited. You found a situation where performance is low on a 2gb, a 1.25 and a 1.5GB card that you WOULD NOT USE THOSE SETTINGS, so the fact that the 1gb card also suffers at those settings, is irrelevant.

Again and again you fail to find situations where using a 2gb version of a card gives playable experience but 1gb doesn't..... the 2gb 560gtx would ALSO give too low performance here.

 

[harmony]

Jesus, AGAIN.

Name one person who would switch on MSAA 4x to go from 80fps to 35fps? The answer is none, so again you've highlighted yet another card that ISN'T effected by memory in ANY SITUATION YOU WOULD ACTUALLY USE IT.

Top to bottom the only setup I would use 4xmsaa with would be the 570 sli, because its the ONLY setup that has enough performance out of those for a really good experience at that resolution........ and it isn't memory limited.

I wouldn't even run a single 580gtx with 4xmsaa from those results, it too has enough memory.

FXAA won't kill the 560gtx performance to unplayable levels, nor will it max out ram usage, this is why 0% of 560gtx would even think of using 4xmsaa, and the 1gb memory will not effect them when they don't.

You lose, AGAIN. As per always, you are NOT the first person, or the 100,000th person to know that higher settings uses more memory and there are situations more memory will help. The problem is your crusade is 1920x1080, and you have failed, again, to find a situation where a 1gb card is memory limited. You found a situation where performance is low on a 2gb, a 1.25 and a 1.5GB card that you WOULD NOT USE THOSE SETTINGS, so the fact that the 1gb card also suffers at those settings, is irrelevant.

Again and again you fail to find situations where using a 2gb version of a card gives playable experience but 1gb doesn't..... the 2gb 560gtx would ALSO give too low performance here.

What a load of tosh that I no longer bother to read.

 

[kissenger]

FFS! Harmony, WE ALREADY ESTABLISHED THAT THOSE BENCHMARKS ARE ******** YOU CRETIN!

http://forums.overclockers.co.uk/showthread.php?p=20492894#post20492894

Yet you just keep posting this ****! Why are you still allowed to troll the forums? I mean, really, it's just ridiculous!

Nobody is, as you put it "a follower of 1GB ram". It's not a team sport. We are just pointing out some pretty obvious facts.

 

[stulid]

Harmony, not satisfied talking trash in his threads now decides to talk absolute tosh in other threads.

 

[kissenger]

Harmony, not satisfied talking trash in his threads now decides to talk absolute tosh in other threads.

We can't be too harsh on him. He's obviously an aspy.

 

[harmony]

FFS! Harmony, WE ALREADY ESTABLISHED THAT THOSE BENCHMARKS ARE ******** YOU CRETIN!

http://forums.overclockers.co.uk/showthread.php?p=20492894#post20492894

Yet you just keep posting this ****! Why are you still allowed to troll the forums? I mean, really, it's just ridiculous!

Nobody is, as you put it "a follower of 1GB ram". It's not a team sport. We are just pointing out some pretty obvious facts.

Again, if you can post 10 screenshots with a framerate of local maxima (like 60 fps), it does not necessarily mean you can get constant 60 fps in every area in the game.

 

[Marine-RX179]

A bit of apples and oranges?

As you often say, there is not much difference in them, which I usually agree.

As I have went from a 5870 to a 6950, I can compare.

Imo, there is far too much of a difference in fps, were talking as low as 5 fps with the 5870 max CCC settings in Witcher to minimums of 28 fps with the same settings using the 6950.

There is no way that's core differences.

Seriously...you are still basically doing what harmony does...you points to the 6950 2GB performing better the 5870 1GB (which most people would see and agree, despite extra performance is not much), and then declare it is due to the extra VRAM despite the lack of evidence.

As for the example you given, typical responses by people would be "driver issue?" You can't prove it wasn't a driver issue, just like you can't prove the better performance on the 6950 2GB over the 5870 1GB on the scenerio you mentioned is due to the extra VRAM.

As I said, the only RELIABLE way of identifying if VRAM really make much a difference is to compare same card on the 1GB and 2GB...any other ways for comparison (such as using comparing different cards) is simply flawed methodology.

 

[harmony]

As I said, the only RELIABLE way of identifying if VRAM really make much a difference is to compare same card on the 1GB and 2GB...

It is still flawed. The drivers may actively and aggresively throttle vram usage in certain popular games, which is an optimization for 1GB cards only, ignoring the possible benefit that can be obtained from the 2GB cards. For example, the vram usage of BF3 and Metro 2033 have been intentionally limited to around 1GB on the GTX560 Ti 2GB despite the fact that a GTX580 can easily use 1.5GB vram. This has introduced unnecessary lag spikes and stuttering on the GTX 560 Ti 2GB, which is sad.

 

[harmony]

Quote again (for those who is not smart enough to understand what is lag spikes / stuttering):

 

[harmony]

The whole 1GB vram vs 2GB vram is like comparing 4GB system memory vs 8GB system memory: do you see much performance improvement by going from 4GB to 8GB? Obviously no in most cases, not even a single improvement in any benchmark scores, if you pick the "right" tests. However, are there cases where 4GB system memory will suffer? Definitely yes.

This sentence is for those who's smart enough to understand: A decrease of the average fps / adhoc min fps is not the only symptom of minor vram shortage.

 

[harmony]

Harmony, not satisfied talking trash in his threads now decides to talk absolute tosh in other threads.

This forum is full of people like you with poor ability to understand and analyse problems. Talking about quantitative analysis here is like teaching pupils about calculus.

 

[stulid]

This forum is full of people like you with poor ability to understand and analyse problems. Talking about quantitative analysis here is like teaching pupils about calculus.

There's nothing to understand, it really is simple,

Your trolling and everyone apart from yourself knows you are taking rubbish.

Why haven't you figured it out yet that there are no 1.5GB VRAM at 1920x1200 resolution issues.

 

[harmony]

There's nothing to understand, it really is simple,

Your trolling and everyone apart from yourself knows you are taking rubbish.

If I had the time to fill the forum with 26k posts of tosh within a year or so then I'd better stop thinking about how to earn money to buy more decent graphics cards with affluent vram.