nVIDIA PhysX broken?

Magic Carpet

Diamond Member
Oct 2, 2011
3,477
232
106
4770K
GTX 670
8GB RAM
347.09
Clean Install, 8.1x64

Mafia 2, in-game benchmark (all high @ 1050p, PhysX high), I never see GPU usage above the 55-80% range, as a result FPS tanks. I get about 56 FPS. I am not CPU limited in that test.

What's up with that? Bad drivers or further Kepler owners negligence? :biggrin:

Pretty sure, last year I played that game, I saw my GPU loaded consistently (same game install).

EDIT: With PhysX off, I get 109 FPS, GPU @ 93-99% throughout the test.
 
Last edited:

WittyRemark

Member
Dec 7, 2014
119
0
0
PhysX uses CUDA,thus offloading physics calculations from the CPU.
So how can he be CPU bottlenecked? or am I missing something?
 

Magic Carpet

Diamond Member
Oct 2, 2011
3,477
232
106
Here is the graph taken exactly during the benchmark. The high CPU spikes represent the games menu, I have Vsync Off.



The point is, it's very easy to see in Afterburner when both GPU/CPU usage isn't near the max. So there is a problem.



Only towards the end, I see constant GPU usage.
 
Last edited:

Grooveriding

Diamond Member
Dec 25, 2008
9,108
1,260
126
GPU physx just stinks like that. Always performance issues and low gpu usage with it enabled. It's really bad if you're running SLI.
 

Magic Carpet

Diamond Member
Oct 2, 2011
3,477
232
106
GPU physx just stinks like that. Always performance issues and low gpu usage with it enabled. It's really bad if you're running SLI.
I wonder, if Maxwell cards are experiencing similar issues.

EDIT: Maybe if I add a dedicated PhysX card my problems will be gone. Thank you Nvidia. Good business.
 
Last edited:

Magic Carpet

Diamond Member
Oct 2, 2011
3,477
232
106




With a dedicated PhysX card, my framerate improved to 82 FPS, which is what I expected with a single card, to be honest.

So yah, pretty disappointed with nVIDIA. You pretty much need a dedicated card if you want PhysX to work more or less as it's designed (still far from perfect, as you can see from the graphs). If they can't figure it out, I would appreciate if NV at least, let AMD owners have a trouble-free life if they choose to run nVIDIA explicitly for PhysX operations (as far as I am aware, you need some hacks for this).

Don't have Maxwell to test this.
 
Last edited:

Deders

Platinum Member
Oct 14, 2012
2,401
1
91
I wouldn't take that benchmark as representative of in game performance. It's always been vastly different.
 

Magic Carpet

Diamond Member
Oct 2, 2011
3,477
232
106
I wouldn't take that benchmark as representative of in game performance. It's always been vastly different.
Looks like, the PhysX software is extremely poorly optimized even for NV GPUs. At least in Mafia 2.

Maybe it's different in other games / PhysX revisions. I've never seemed to pay enough attention until recently.

EDIT: Hell, I remember similar performance issues in BL2, but never really investigated it. And when I googled, everybody blamed the devs, not NV.
 
Last edited:

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
You are surprised you get lower FPS with physics on?

Physics calcualtions+more graphics to show.
 

Magic Carpet

Diamond Member
Oct 2, 2011
3,477
232
106
You are surprised you get lower FPS with physics on?

Physics calcualtions+more graphics to show.
I am annoyed that both my CPU and GPU are not used to their full potential :thumbsdown:

Not just in Mafia 2. BL2 had similar issues. Everything points to poor software optimization.
 

Headfoot

Diamond Member
Feb 28, 2008
4,444
641
126
Maxwell is supposed to have better context switching than Kepler if I recall correctly, so that running a GPGPU load like PhysX and normal GPU code carries less of a penalty.

It will always lower your FPS because its adding visual effects... If it didn't lower your FPS it isn't doing anything
 

Headfoot

Diamond Member
Feb 28, 2008
4,444
641
126
I am annoyed that both my CPU and GPU are not used to their full potential :thumbsdown:

Not just in Mafia 2. BL2 had similar issues. Everything points to poor software optimization.

Not really.

http://www.codercorner.com/blog/?p=1129

Fully utilizing all resources is not optimized... you want them to start throwing on random stuff so that your meter pegs at 100%? That's literally the opposite of optimization. Optimization is getting the same results for less work...

Utilizing all resources is scalability, which is very different from optimization
 

Magic Carpet

Diamond Member
Oct 2, 2011
3,477
232
106
Not really.

http://www.codercorner.com/blog/?p=1129

Fully utilizing all resources is not optimized... you want them to start throwing on random stuff so that your meter pegs at 100%? That's literally the opposite of optimization. Optimization is getting the same results for less work...

Utilizing all resources is scalability, which is very different from optimization
This situation, is rather simple. If you want decent PhysX performance, a separate GPU is a must (scalability, you reckon?)

But if I want to have good PhysX performance on a single GPU (granted my primary GPU isn't fully utilized), I call this optimization! Or lack of.

Maxwell is supposed to have better context switching than Kepler if I recall correctly, so that running a GPGPU load like PhysX and normal GPU code carries less of a penalty.

It will always lower your FPS because its adding visual effects... If it didn't lower your FPS it isn't doing anything
Yeah, I'd love to bench Maxwell to check that.
 
Last edited:

Flapdrol1337

Golden Member
May 21, 2014
1,677
93
91
My CPU usage usage barely reached 60%. Let me recheck.

That means nothing. Games don't multithread perfectly. On haswell it's extra hard to spot because stuff keeps hopping from core to core, if I start a single thread of prime95 I will see it perfectly distributed over the cores in taskmanager.
 

Magic Carpet

Diamond Member
Oct 2, 2011
3,477
232
106
That means nothing. Games don't multithread perfectly. On haswell it's extra hard to spot because stuff keeps hopping from core to core, if I start a single thread of prime95 I will see it perfectly distributed over the cores in taskmanager.
You are, correct, but I don't think that's the issue here. With more clock speed, I only see single digit gains.

I am seriously GPU limited here, with the addition of a second gpu you can see how that effects my framerate in a very positive way. Yet, I'd like to see more, get my GPU to work harder.

The reality is, you need to have a second GPU for proper performance in PhysX heavy games. Maybe I'll change my mind when I get my hands on a Maxwell sample. But with a mid-range Kepler, it is like that. Sad, but true.
 
Last edited:

Deders

Platinum Member
Oct 14, 2012
2,401
1
91
Have you tried playing the game itself with PhysX maxed? I expect performance will be fine.
 

Magic Carpet

Diamond Member
Oct 2, 2011
3,477
232
106
Have you tried playing the game itself with PhysX maxed? I expect performance will be fine.
Of course I have. And it has been fine for the most part.

Except, I am having, occasionally, serious frame drops in Borderlands 2. When I have lots of action, going on (particles and what not). Pretty much what's happening in Mafia 2 benchmark. When it's PhysX heavy, frame rates just dive... now I could understand that, if my CPU/GPU were pegging close to full usage, but they were not. I always have Afterburner on, and can see that.

The only cure to this, apparently, is having a separate PhysX card, or a faster, newer GPU. Really, I don't give a damn, how this works, as long as it works. But when you buy a product with advertised features, and they don't work as you expect. You can be a little disappointed. I expect this to work at ~90+%, not just ~65%. This technology is flawed. Until fixed.
 
Last edited:

Flapdrol1337

Golden Member
May 21, 2014
1,677
93
91
You are, correct, but I don't think that's the issue here. With more clock speed, I only see single digit gains.

I am seriously GPU limited here, with the addition of a second gpu you can see how that effects my framerate in a very positive way. Yet, I'd like to see more, get my GPU to work harder.

The reality is, you need to have a second GPU for proper performance in PhysX heavy games. Maybe I'll change my mind when I get my hands on a Maxwell sample. But with a mid-range Kepler, it is like that. Sad, but true.
Strange, when I tested mafia 2 (demo) with my 260's there was barely any difference between everything on 1 card and one as dedicated physX.
 

hawtdawg

Golden Member
Jun 4, 2005
1,223
7
81
PhysX is borked. Try playing Borderlands 2 or the Pre-Sequal. I have dual OC'd 780's and an OC'd 3770K and I've seen framerates in the 20's in Pre-Sequal with max GPU PhysX
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
PhysX is borked. Try playing Borderlands 2 or the Pre-Sequal. I have dual OC'd 780's and an OC'd 3770K and I've seen framerates in the 20's in Pre-Sequal with max GPU PhysX

I did play the presequel with everything on max and mostly MP as well. Not a single issue. Maybe there is a SLI issue?
 

Magic Carpet

Diamond Member
Oct 2, 2011
3,477
232
106
Strange, when I tested mafia 2 (demo) with my 260's there was barely any difference between everything on 1 card and one as dedicated physX.
The problem is masked, when there is little going on, you just don't notice anything. The actual Mafia 2 game isn't as taxing on resources as that benchmark. In all the games I have played, it was just Borderlands 2 where it was noticeable and annoying (when you see your framerate dipping to 20's and below, from 60+ = yet GPU/CPU utilization isn't maxed out).

The thing is, they could just make a dedicated PhysX chip in hardware and eliminate this problem once and for all. Charge an extra 50 or something. I would pay for that. It is a better alternative than having to run an extra card, if there is no other practical way around this matter.

I did play the presequel with everything on max and mostly MP as well. Not a single issue.
But you didn't really monitor your rigs usage, did ya?

I suspect, with GM204 being ~50% faster than GK204, it's easier to mask this thing.

PhysX is borked. Try playing Borderlands 2 or the Pre-Sequal. I have dual OC'd 780's and an OC'd 3770K and I've seen framerates in the 20's in Pre-Sequal with max GPU PhysX
Did you try running it off one card and putting the second for PhysX duties?
 
Last edited:

Keysplayr

Elite Member
Jan 16, 2003
21,209
50
91
PhysX is borked. Try playing Borderlands 2 or the Pre-Sequal. I have dual OC'd 780's and an OC'd 3770K and I've seen framerates in the 20's in Pre-Sequal with max GPU PhysX

I have both games. Say where and at what settings you experienced this. I can try it with or without SLI.
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
145
106
But you didn't really monitor your rigs usage, did ya?

I suspect, with GM204 being ~50% faster than GK204, it's easier to mask this thing.

Thats true. The guy I played with mainly had a GTX780 or 780ti. Didnt hear any problems there either. But again, also faster as you say and could mask it.
 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |