** 2GB GFX RAM VS 1GB GFX RAM IN BF3 (560Ti 1GB VS 560Ti 2GB) TESTING RESULTS!

[MoRT489]

Those results seem weird - why doesnt the computer use the spare GPU power at lower settings to create more FPS? I noticed on mine that if I go from ultra down to high and even drop AA to 0 from 2, it makes very little/no difference to fps...(Q9550 @3.6, GTX 570 @ stock, 8GB 1066MHz RAM).

Socket 775 doesn't have the bandwith, it was never designed for a Quad core CPU, Intel just made it work.

 

[decto]

Those results seem weird - why doesnt the computer use the spare GPU power at lower settings to create more FPS? I noticed on mine that if I go from ultra down to high and even drop AA to 0 from 2, it makes very little/no difference to fps...(Q9550 @3.6, GTX 570 @ stock, 8GB 1066MHz RAM).

Socket 775 has a single front side bus (FSB) to link the CPU and the north bridge. The north bridge then provides the links to the memory, the PCI-E lanes and to the south bridge for other functions, Lan, Sata, audio etc etc.

Particularly for a quad core chip, this is an issue. Every bit of data for each of the four cores has to pass through the same narrow interface.

For later chipsets, e.g. P67 and Sandy Bridge, the north bridge is an integral part of the CPU. There are separate dedicated lanes for memory connection, PCI-E and to link to the south bridge. This allows a much higher overall bandwidth so the data the CPU needs to feed each of it's cores can be moved very effectively.

In my opinion, socket 775 cannot cope with the high bandwidth streaming of today and the sheer amount of data required to drive a modern mid to high range GPU or two. That's why it's been replaced.

The CPU is very capable for the right workload, I have both a Q6600 and a Q6700 in two different machines and they both show the same issue.

The computer can't use the spare graphics power to draw more frames as the CPU has to tell the GPU what to draw and with four occupied cores there simply isn't the capacity to shuffle that data around.

AD

 

[kissenger]

Well I turned down my settings to all medium last night with the 3.6Ghz Q6600 and GTX560Ti.

At High + some ultra FPS 42 GPU 90%
At High FPS 42 GPU 70%
At Medium FPS 43 GPU 55%

8GB DDR2 @ 800Mhz
FSB 1600Mhz
P5N-D

It's so bottlenecked in MP

Interestingly slowing the CPU down doesn't change it much either.

2.4Ghz (6x400) High 39 FPS
3.2Ghz (8x400) High 42 FPS
3.6Ghz (9x400) High 42 FPS

So I conclude this is a platform issue. Sticling an extra card in will only cause a lag fest as if the platform can only drive the card at 40ish FPS and you then SLI it to split between two cards ... it can still only do 40FPS except now you've made it stutter far worse across two cards.

Socket 775 doesn't have the bandwith, it was never designed for a Quad core CPU, Intel just made it work.

Have Sandybridge waiting in the wings for an upgrade once the motherboard arrives so I'll see what happends to the frame rates then.

AD

I have a stock Q6600 and when I paired it with a 560Ti I was getting an average of 55fps with everything on ultra (except MSAA) at 1080p in the campaign (only tried first mission though).

 

[decto]

I have a stock Q6600 and when I paired it with a 560Ti I was getting an average of 55fps with everything on ultra (except MSAA) at 1080p in the campaign (only tried first mission though).

The campaigns are better than the multiplayer, it's in the muliplayer maps that is suffers. I feels quite jerky to play. 3dmark etc all shows everything is performing as other comparable systems.

AD

 

[andybird123]

@decto: gtx580 arrived today
i was getting 90-100% usage on a 560ti, im now getting 90-100% gpu usage on a 580
BF3 1080p ultra preset with 4x msaa, 40-60 fps most of the time, 1480mb vram usage, couple of dips to mid-20's but didnt notice these in game, think these were at spawn

q6600 definitely not a bottleneck

 

[decto]

@decto: gtx580 arrived today
i was getting 90-100% usage on a 560ti, im now getting 90-100% gpu usage on a 580
BF3 1080p ultra preset with 4x msaa, 40-60 fps most of the time, 1480mb vram usage, couple of dips to mid-20's but didnt notice these in game, think these were at spawn

q6600 definitely not a bottleneck

The bottleneck may be the Nvidia 750i I was using in the P5N-D, performace is well below where it should be.

I'm just setting up my upgrade to sandy B.

Would be interested if you could post a graph similar to this for multiplayer
http://forums.overclockers.co.uk/showthread.php?t=18345642

I just can't get my S775 to perform. 3dmark says it's fine... It just lags around 40fps.

Can't wait to test SB hence why I'm installing now.... but have a gazillion more windows updates to install.....

AD

 

[andybird123]

will do later - just OC'd the 580 to 850mhz and am now getting a pretty consistent min of about 50fps with a couple of dips to 40ish

I have the same mobo btw, it was really unstable at 3.6ghz till I upped the HT and NB to 1.48V and did the vdrop mod

 

[decto]

will do later - just OC'd the 580 to 850mhz and am now getting a pretty consistent min of about 50fps with a couple of dips to 40ish

I have the same mobo btw, it was really unstable at 3.6ghz till I upped the HT and NB to 1.48V and did the vdrop mod

It's the most challenging mobo I've had in a long while. Did the vdroop pencil mod and use 1.44 HT and NB with the fan cooling. Core at 1.45V 400 x9.

Not quite fully prime stable, core 3 faults after 20mins of prime as it heats up. Would need a better cooler or a little more voltage but had only clocked to see if it would work before invested in more cooling so is perfectly game stable.

Got my SB installed last night but kept getting a punkbuster kick as I'd reused the old origin game directory on a second drive. Managed to fix that this morning but have then had to head to work. A quick blast on the single player was 95%+ usage. Will have another look later.

Out of interest, what GPU usage do you get a medium settings?

Cheers

AD

 

[andybird123]

i had loads of issues with mine too at first, had my HT and NB at 1.44 as per a guide I'd found, kept upping my VCORE trying to get it stable but nothing

mounted a fan pointing directly at the NB heatsink, upped to 1.48 HT and NB, VCORE is now at 1.41 (1.395 in CPU-Z - stable never fluctuates), and now it's 100% stable

I'm running a xigmatek HD-1283 cooler with stock fan and load temps are mid 60's

will do a run on medium tonight and let you know

 

[bhavv]

whats that got to do with anything? plenty of people are playing the game on ultra with 4gb of memory..

try opening task manager and checking how much ram bf3 is using even on 64 player maps

Regardless of how much ram BF3, you need additional ram for caching and shared memory.

If a game is fully using up 3 Gb ram, you arent going to have enough left over for the rest of your systems needs.

Also with only 4 Gb ram installed, obviously you will never see your ram usage exceed 4 gb because you dont have enough ram to see it.

 

[bhavv]

Interestingly slowing the CPU down doesn't change it much either.

2.4Ghz (6x400) High 39 FPS
3.2Ghz (8x400) High 42 FPS
3.6Ghz (9x400) High 42 FPS

So I conclude this is a platform issue.

Maybe try doing some research:

CPU doesnt affect BF3 performance.

 

[Buckster]

I find it hard to believe the notion that in BF3 - a 560Ti w/ 2gig would be smoother gameplay than a 570 1.25 or 580 1.5gig

 

[kissenger]

I find it hard to believe the notion that in BF3 - a 560Ti w/ 2gig would be smoother gameplay than a 570 1.25 or 580 1.5gig

It's not. They could put 5Gb on it and it wouldn't make a difference.

 

[decto]

Maybe try doing some research:

CPU doesnt affect BF3 performance.

For clarity... which of those is a socket 775 CPU ?


My discussion is about the limitiation of the socket 775 interface.

The graph you present is not relevent in the slightest

AD

 

[andybird123]

Out of interest, what GPU usage do you get a medium settings?

GPU usage drops slightly from 85-100% usage to 75-90% but frames go from 40-50 to 50-80FPS (stock clocks on the GPU as the missus is watching TV and hates the fans on too loud)

one thing to check on the P5N-D is to set the RAM timings manually as mine were defaulting to some really weird settings on auto (like 8-8-8-21-32 instead of 5-5-5-15-22) which was severely hampering my 3dmark scores initially

 

[bhavv]

For clarity... which of those is a socket 775 CPU ?


My discussion is about the limitiation of the socket 775 interface.

The graph you present is not relevent in the slightest

AD

Its relevant in every way, it shows you right ther that CPU type / cores / clock speed. do not affect performance in BF3.

S775 is not limited anymore than any of the other sockets in that comparison.

Going from 1 gb to 2 gb vram boosts minimum FPS from 17 to 22 - which makes no difference as the game is still unplayable. As I have said so many, increasing Vram doesnt provide any more than a 2-5 FPS increase, and in any case that it does the game still isnt playable.

No one here has yet shown me an unplayable game with 1 gb vram that becomes playable with 2 gb.

If you genuinely need a single card to play BF3, a 2 gb GTX 560 or 6950 are both useless, you need at least a GTX 570, or the new 560 448 core version.

The GTX 560 ti is too slow for 2 Gb vram. Even if you added 100 Gb vram to it, it would stll be slower than a 1.2 Gb GTX 570.

if BF3 actually was Vram limited, it wouldnt perform well on a 1.2 Gb GTX 570. There is never going to a significant performance difference between 1 Gb and 1.2 Gb vram, the difference is so small that shared ram would easilly negate any differences.

 

[Final8y]

http://benchmarkextreme.com/Articles/Radeon 5870 2GB/P1.html

 

[bhavv]

http://benchmarkextreme.com/Articles/Radeon 5870 2GB/P1.html

2560x1600 + 8x MSAA is completely irrelevant to this thead which is discussing 1080p up to 4x MSAA.

If you want to discuss higher resolutions and AA settings you should maybe make a seperate thread rather than derailling this one with useless propaganda.

 

[Final8y]

2560x1600 + 8x MSAA is completely irrelevant to this thead which is discussing 1080p up to 4x MSAA.

If you want to discuss higher resolutions and AA settings you should maybe make a seperate thread rather than derailling this one with useless propaganda.

The relevance is showing Vram limitation that some people believe cant make a playable difference and that GPU will always be the limiting factor before Vram..

The relevance is that shared ram didn't help in the slightest.

The relevance that some of the symptoms mentioned on some of the games are starting to appear at 1080p, people should read them.

BF3 maybe the first to pip things at 1080p but who knows what may follow.

But your right as far as the 2560x1600 goes but many 2560x1600 users got caught out by the same discussions.

 

[bhavv]

GPU is always the limiting factor before vram!

Or are you telling me that a 2 Gb 560 ti is going to outperform a 1.5 Gb 580 at 2560x1600?