Just inform all guys, my 4850CF has poor performance in NFS:Shift

[Gothic3]

I use Cat 9.9 drv

[Destroy]

Can anyone else confirm ATI issues with Shift?

[DOXau]

do you have motion blur turned up high while using AA? That was killing performance for me... turning it down gave me a solid 60ish

[Gothic3]

Originally Posted by Destroy

Can anyone else confirm ATI issues with Shift?

pcgameshardware.de already confirmed it, it said AMD was aware of this problem.

one of my friend he found the GPU usage didn't reach 100% loading on his 4850 when running SHIFT.

[Gothic3]

Originally Posted by DOXau

do you have motion blur turned up high while using AA? That was killing performance for me... turning it down gave me a solid 60ish

all max setting, since the GPU usage didn't reach 100%, I can say this's another TWMTBP game.

[KAC]

Same problem. Crappy performance. Removing vsync resulted in fps in ranges of 30-45 but w/ tearing it looks crap.

Will there be a time when 4 different games are released and we not need 4 different driver sets...

[DOXau]

yeh i retract my earlier comment after playing it a bit longer, game has serious fps issues..60fps on one part of track..dropping down to 15fps a corner later :/ (with only 2xaa on 4870x2)

[kn00tcn]

Originally Posted by KickAssCop

Will there be a time when 4 different games are released and we not need 4 different driver sets...

isnt that the nvidia way? a flood of multiple driver versions, each better for 1 group of games than another driver & its other group of games



as for shift, remember how undercover got some patch that fixed a ton of what sounds like the exact same perf issues? too bad prostreet's patch just capped the game to 30 to hide fps drops
although blackbox didnt make shift...

EDIT: http://www.rage3d.com/board/showpost...8&postcount=38 ?

[Gothic3]

4850CF GPUs Usage，obviously not 100%, I''ll tried 4850 single card later

http://bbs.pczilla.net/attachments/m...174e501962.jpg

[KAC]

Originally Posted by kn00tcn

isnt that the nvidia way? a flood of multiple driver versions, each better for 1 group of games than another driver & its other group of games



as for shift, remember how undercover got some patch that fixed a ton of what sounds like the exact same perf issues? too bad prostreet's patch just capped the game to 30 to hide fps drops
although blackbox didnt make shift...

EDIT: http://www.rage3d.com/board/showpost...8&postcount=38 ?

I couldn't understand much of what was being said in that post. Care to explain in English to a layman? Thanks.

[Gothic3]

4850CF GPUs Usage in NFS: SHIFT，obviously not 100%





4850CF usage in Resident Evil 5, seems NV found a new trick to demolish ATi graphics cards performance.

[spyre]

I get an average of 75 FPS a low of 30 (for the grid lineup bits where the people are around the car) and a high of 110 FPS (although that part depends on the track) with my 4870X2 at 1920 x 1200 with no AA, with 8xAA it brings the average down to 65FPS.

GPU usage isnt 100% though so no doubt it can get a bit higher.

[kn00tcn]

that gpu usage pic is useless without fps (& possibly cpu) listed

if the game has an fps cap or vsync, gpu usage wont be solid
if the game is killing the cpu, gpu usage wont be solid

that said, FC2 dx10 had strange minimums, where gpu usage would start dropping, fixed in 9.8

[kn00tcn]

Originally Posted by KickAssCop

I couldn't understand much of what was being said in that post. Care to explain in English to a layman? Thanks.

i have seen many nv users using all kinds of different driver versions because each of these versions works best on certain games only

in need for speed undercover, there was a patch that fixed a bunch of performance issues that the stock game had, so it is possible similar is happening to shift (but the dev of shift is not the same dev as undercover)

in need for speed prostreet, it also had a patch, but it capped fps to 30 permanently, thereby hiding fps fluctuations

the link is someone tweaking nfs shift so that it stopped having as huge performance drops

better?

[Gothic3]

Originally Posted by spyre

I get an average of 75 FPS a low of 30 (for the grid lineup bits where the people are around the car) and a high of 110 FPS (although that part depends on the track) with my 4870X2 at 1920 x 1200 with no AA, with 8xAA it brings the average down to 65FPS.

GPU usage isnt 100% though so no doubt it can get a bit higher.

then tried 1920x1200 4xAA 16xAF Max setting( Motion Blur High) pls

[Destroy]

So Spyre is the only one that has no issues with running this game? Weird.

[Falcon1]

I'll test tomorrow... it's 4:21am and I'm going to bed now... *yawn*

[spyre]

I think I know why I was getting better performance than everyone else now, the more AI opponents you add the worse FPS you get. If there are no cars in front of you then you get good FPS. With 15 AI opponents the FPS was 30. Once you pass them all it goes upto 60+ So 10 more AI opponents than normal drops the average FPS by 15 when they are behind you and by 35 fps when they are visible in front of you.

I guess it depends how much CPU strain the extra cars require vs amount of extra detail that needs drawing etc.

I'd try my 8800GT but it died.

I guess you would have to measure the average CPU load vs number of AI cars to see if that plays a significant part or not.

[Gothic3]

Originally Posted by spyre

I think I know why I was getting better performance than everyone else now, the more AI opponents you add the worse FPS you get. If there are no cars in front of you then you get good FPS. With 15 AI opponents the FPS was 30. Once you pass them all it goes upto 60+ So 10 more AI opponents than normal drops the average FPS by 15 when they are behind you and by 35 fps when they are visible in front of you.

I guess it depends how much CPU strain the extra cars require vs amount of extra detail that needs drawing etc.

I'd try my 8800GT but it died.

I guess you would have to measure the average CPU load vs number of AI cars to see if that plays a significant part or not.

never meet this problem even in DIRT1, tens of cars colide toghter, but fps is still high

[DOXau]

Yeh i was noticing the other ai cars lowering fps, dont think its the ai itself, but the rendering of their cars/tire smoke.. or perhaps even their physx set too high?

[Gothic3]

Originally Posted by DOXau

Yeh i was noticing the other ai cars lowering fps, dont think its the ai itself, but the rendering of their cars/tire smoke.. or perhaps even their physx set too high?

according to some nv boys, this game uses PhyX CPU

[becco]

Originally Posted by Gothic3

according to some nv boys, this game uses PhyX CPU

Link please, thanks. I dont think this game use CUDA.

[LordJuanlo]

DOXau said the game uses PhysX CPU, not GPU, this means equal PhysX performance on ATI and nVidia cards and no CUDA

[Gothic3]

9.8 cat seems better

[Riou]

Spyre might be correct, the game runs pretty quickly once your jump all the way up to first place on my 4870 crossfire.

Is everyone using the default cockpit view?

[Destroy]

Originally Posted by Riou

Spyre might be correct, the game runs pretty quickly once your jump all the way up to first place on my 4870 crossfire.

Is everyone using the default cockpit view?

But that doesn't explain why NV guys are having no frame rate issues.

[Lurk]

Originally Posted by Gothic3

never meet this problem even in DIRT1, tens of cars colide toghter, but fps is still high

Untrue. DiRT ran like dirt, and multiple AI was one of the reasons. The coding has been fixed miraculously on GRID.

[ECH1]

Here is what I noticed in a 15 car race at 1680, 4xAA, all max'd:

GPU usage is around 90% (however, it depends on the track) when you start the race. As you battle for 1st place it seem to be hoover around 90%. However, if contact is made with AI cars or other objects GPU usage seems to take a dip in the 80% range.

...For less complex physics calculations (low-level), like collisions with the track limits, Nvidia'sPhysx API is used...

PC Games Hardware. The GPU usage is 100% when you are in the lead with the AI opponents are far behind. CPU usage is around the mid 60% throughout the race. It seems to load 2 cores evenly.

Can someone from AMD confirm if the decreasing GPU usage can be fixed via new driver or driver patch?

[ECH1]

Originally Posted by Lurk

Untrue. DiRT ran like dirt, and multiple AI was one of the reasons. The coding has been fixed miraculously on GRID.

Actually he's right it wasn't the cars that was the problem in Dirt it was ridding on the grass and other none track aspects which caused frame rates to dip back then. Yes, the multi car events were a tad more taxing then the single car events but it wasn't the same as in this game. However, that was all fixed when CM improved on their ego engine.

[Gothic3]

Originally Posted by Lurk

Untrue. DiRT ran like dirt, and multiple AI was one of the reasons. The coding has been fixed miraculously on GRID.

u tried 4000 seires on DIRT before? or 2900XT?