Ok, just to throw my two cents worth into the mix, I'm running Vista home premium X86 sp-2 with a 9500GT 195.62 2.3 dlls. After reading this thread I went to the Nvidea control panel and turned off PhysX. I didn't check the exact times but my run time went up by quite a bit from about 1 hour 40 some minutes to over two hours. I then went back in and turned PhysX back on and my next WU finished again back at 1:40 or so. Both WUs were angle range 0.396698.

---

PROUD MEMBER OF Team Starfire World BOINC

Ok, just to throw my two cents worth into the mix, I'm running Vista home premium X86 sp-2 with a 9500GT 195.62 2.3 dlls. After reading this thread I went to the Nvidea control panel and turned off PhysX. I didn't check the exact times but my run time went up by quite a bit from about 1 hour 40 some minutes to over two hours. I then went back in and turned PhysX back on and my next WU finished again back at 1:40 or so. Both WUs were angle range 0.396698.

Link to host and affected results?
Can it be that >2hours tasks were performed by CPU instead of GPU ?
(CUDA MB can fall back to CPU if something goes wrong with GPU initialization)

http://setiathome.berkeley.edu/result.php?resultid=1436082173 This is the two hour one. This is the 1:40 one http://setiathome.berkeley.edu/result.php?resultid=1436092616

Host http://setiathome.berkeley.edu/show_host_detail.php?hostid=4731257

---

PROUD MEMBER OF Team Starfire World BOINC

I needed to enable PhysX with 195.62 for to let run CUDA WUs (only) on the GPU.
If not, I had also 25 % CPU/ 100 % CPU-Core usage.

I uninstalled PhysX via App remove completely - CUDA MB still able to run, 195.62 drivers.
Now under benchmark will PhysX removal change times or not.

P.S. No changes in performance.

Hmm.. strange.. I know what I saw..

With my PC settings and WinXP 32bit I needed to enable PhysX for to let run well CUDA on GPU. If not.. like I mentioned on the top.
I made this 195.62 test two times.
I guess it's a BUG in the nVIDIA_driver_195.62 and some OSs.


Raistmer, you would like to make a 'officially' nVIDIA_driver test?

It will last long time for to let run 190.x , 191.x and 195.x (CUDA_V2.3 + .._V3.0_BETA). Also with PhysX enabled/disabled?
Uninstall/install drivers and so on..
But then we would know it for sure..

---

Raistmer, you would like to make a 'officially' nVIDIA_driver test?

In progress already.

I see some speed improvement with older driver:

CUDA 2.3, 195.62 PhysX uninstalled
App Name Task name AR CPU time Elapsed Wall-clock
MB_6.08_mod_CUDA_V12_double_triplet_fix_4_11_rebuild PG0395.wu 0.394768 18.58 427.459 424.4
MB_6.08_mod_CUDA_V12_double_triplet_fix_4_11_rebuild PG0444.wu 0.444184 17.597 370.963 368
MB_6.08_mod_CUDA_V12_double_triplet_fix_4_11_rebuild PG1327.wu 1.326684 16.567 196.145 193.2

04/12/2009 03:28:23 NVIDIA GPU 0: GeForce 9400 GT (driver version 19038, CUDA version 2030, compute capability 1.1, 512MB, 45 GFLOPS peak)
App Name Task name AR CPU time Elapsed Wall-clock
MB_6.08_mod_CUDA_V12_double_triplet_fix_4_11_rebuild PG0395.wu 0.394768 16.209 425.722 422.6
MB_6.08_mod_CUDA_V12_double_triplet_fix_4_11_rebuild PG0444.wu 0.444184 16.365 369.893 366.9
MB_6.08_mod_CUDA_V12_double_triplet_fix_4_11_rebuild PG1327.wu 1.326684 15.709 195.334 192.4

I use lowend GPU Instead of high-end one, but it seems performance of 9400GT altered in last drivers too.
More will follow.

http://setiathome.berkeley.edu/result.php?resultid=1436082173 This is the two hour one. This is the 1:40 one http://setiathome.berkeley.edu/result.php?resultid=1436092616

Host http://setiathome.berkeley.edu/show_host_detail.php?hostid=4731257

yeah, both done on GPU...
So, in your case PhysX matters for performance it seems...
AFAIK on some hosts with many GPUs installed PhysX required for correct operation of additional GPUs. That is, all this stuff pretty diverse one.

More confusion here......PhysX on or off for best performance......
I had been told a while back that PhysX had no impact on Cuda crunching, only on gaming applications........

Where are the fish?

---

"Freedom is just Chaos, with better lighting." Alan Dean Foster

PsychX enabled is good for multi gpu systems when u have only one monitor atached or gaming - enable it on the card that is not connected to the monitor.

It cant affect performance (if u arent playing games that uses it when crunching) because it is totally independent computational engine implemented in the drivers ...

---

PsychX enabled is good for multi gpu systems when u have only one monitor atached or gaming - enable it on the card that is not connected to the monitor.

It cant affect performance (if u arent playing games that uses it when crunching) because it is totally independent computational engine implemented in the drivers ...

My question applied to Cuda crunching only......
The 3 rigs that I have Cuda cards in do nothing else but crunch for Seti.
They have no other purpose in life.

I disabled PhysX on my 2 top rigs this morning....and saw no difference in the Killawatt readings, which is a good indicator of work being done....so I suspect it does not impact Cuda crunching at all.

There may be other implications in multi-card rigs to get the drivers to work right or to use all the cards, I suppose, but basically, I think the advice I first got was correct. Which was, it does not matter whether PhysX is on or off as far as Cuda is concerned.

---

"Freedom is just Chaos, with better lighting." Alan Dean Foster

(which is NOT public release BTW. nVidia Beta access is under closed NDA (registered developer program). Getting beta libraries yourself somehow through the backdoor, or from nVidia directly through their forums isn't in our consideration, since *we* agree to terms to have access... So nVidia & you can pretty much do what you want .. *we* can't go handing out libraries or beta DLLs etc.

If this is directed at me, then wtf. Firstly, if getting dlls from the main cuda zone page is a "backdoor", and not "agreeing to terms to have access" then fine, otherwise, I have no clue what the heck you're on about. Secondly, if beta in this sense does not mean public release, then it deviates from the unmarked version of beta and needs to be highlighted - I do not see this anywhere, in fact, I see on the same page that its available to the public, and there is nothing in the CUDA EULA that says I (public) cannot use it. But whatever, if you're right, then lock the thread - don't really care tbh. Simply wanted to know if anyone else was using them and if they saw a performance gain (which is what this forum is for, arguably).

In other words, by all means experiment and find what works best for yourselves, but please don't:
(a) expect special builds just for you ... (speaking for myself really) we don't have the time to tailor to each individuals needs/wants. This is why we make source code available, so YOU, can use the toolkit you have and tailor a build for yourself.
(b) expect us to release software based on beta libraries. beta code is at all times to be considered unstable, and so not usable for release, and most likely covered under NDA we have agreed to. (what nVidia or you do on the other hand is between you and them).

I expect nothing of you, nor did I ever say I expected anything of you o_O nvidia and I are doing nothing together, I really don't see what you're trying to say here.

Adding to that one more thing for clarification. If, adding straight new DLLs to existing Lunatic's builds, you happen to be seeing a 10% performance boost at all angles ranges, on a few different cards, then that is worthy of *us* taking a deeper look.

Which was the point of this thread -_-.. Either way, I've remembered why I never posted here in the past, sad to see this hasn't changed. Oh well, was fun while it lasted I suppose XD

Good luck to everyone, irrespective of whether or not you're using updated stuff, and happy crunching.

Good luck to everyone, irrespective of whether or not you're using updated stuff, and happy crunching.

Your input is welcome by most......

My curiosity still stands whether PhysX has any impact at all on Cuda.......

Don't let 'em get you down.

---

"Freedom is just Chaos, with better lighting." Alan Dean Foster

Raistmer, thanks for the test!
I'm curious what will be the result.

mimo, PhysX enabled for multi GPUs?
Hmm.. to now I disabled PhysX on my both PCs.

Only at the QX6700 (1st PCIe 1.0 (x16 electric) OCed GTX260-216 (no display) + 3rd PCIe 1.0 (x4 electric) 6200 LE - only here display connected) I needed to enable PhysX only in nVIDIA_driver_195.x for to let run CUDA. (I uninstalled/installed two times 195.x . Every time the same. Also this bad result with ~ 2x more CPU support time and increased GPU calculation time, compared with older nVIDIA_drivers)
Windows (XP 32bit) Desktop to both GPUs. If only to 6200 LE no CUDA on OCed GTX260-216.

My GPU cruncher (4x OCed GTX260-216, WinXP 32bit). Display connected only to 1st GPU.
The others no display or VGA dummie connected. Only Windows Desktop on 1st GPU. The other 3 GPUs no Windows Desktop enabled.

So if I would enable PhysX, only the first would profit from it.
If I would extend the Windows Deskotp to all 4 GPUs and then enable PhysX.. I don't know if this would speed up the calculation time. I guess no Windows Desktop would be faster.

Both running now with 190.38 .


I guess this - which driver ? - PhysX enabled/disabled ? - Windows Desktop YES/NO ? - with which OS - everybody must look and test his own system.
There isn't an answer for all - I worry.

---

My GT240 has finally cleared customs. With a bit of luck, I will be putting the system together over the weekend. It will be interesting to see how the 195.62 driver behaves/performs with a "new" generation card - even if it is relatively weak.

My GT240 has finally cleared customs. With a bit of luck, I will be putting the system together over the weekend. It will be interesting to see how the 195.62 driver behaves/performs with a "new" generation card - even if it is relatively weak.

Good luck on your system I am getting a new card also. Sure is fun.

---

Official Abuser of Boinc Buttons...
And no good credit hound!

If this is directed at me, then wtf. ...

Careface,
Errm no. It was directed toward the people that keep PM'ing me wanting special builds and access to beta DLLs (before they were released ...'beta IS NOT release' ) Sorry you felt that it was directed at yourself. If I was addressing you, I would have used your name and addressed you directly.

Regards, Jason

---

"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.

My GT240 has finally cleared customs. With a bit of luck, I will be putting the system together over the weekend. It will be interesting to see how the 195.62 driver behaves/performs with a "new" generation card - even if it is relatively weak.

Good luck on your system I am getting a new card also. Sure is fun.

Fun it is. Always a bit nervous about firing up a new machine. Never know if the mobo is going to "beep beep beep"... Anyway, the GT240 was the last component required before I could start it up. Decided to wait and see what Fermi looks like, though XFX have been tempting me badly with their new HD5970 BE.

And I will startup my GT 240 in few hours :)
MSI with GDDR5, and I also need to see how this new generation is fast ( or not)

---

I am cruncher :)
I LOVE SETI BOINC :)

GT240 - 96 shader cores?

I wouldn't expect so/too much performance..
195.x and CUDA_V3.0_BETA ?

Maybe (I guess) like a 9800 with 190.x and CUDA_V2.3 ..

---

GT240 - 96 shader cores?

I wouldn't expect so/too much performance..
195.x and CUDA_V3.0_BETA ?

Maybe (I guess) like a 9800 with 190.x and CUDA_V2.3 ..

Yes I expect just that: something like 9800GT in performance

---

I am cruncher :)
I LOVE SETI BOINC :)

And I will startup my GT 240 in few hours :)
MSI with GDDR5, and I also need to see how this new generation is fast ( or not)

Mine is apparently (haven't seen it yet) a Galaxy 512mb GDDR3. The specs are in many aspects the same as the 8800GS I am currently running. I will be happy if I get performance similar to a 9800GT! We can compare notes on Monday...