Announcing project 5900 and Core_14 on advmethods

[ihaque]

We've just released project 5900 to advmethods. It uses a new core, FahCore_14.

EDIT: As of Mar 9, this project and core are now accessible on GPU Folding@home in general, not just advmethods.

Please note, the behavior of this core is different from what you may expect. The following behaviors are NORMAL, and the points credit takes it into consideration:

• Varying time per frame
• Low GPU temperatures
• Fluctuating CPU usage (somewhat higher than GPU core 11).

Preferred deadline: 2 days
Final deadline: 3 days
Stats credit: 420 points/WU


(EDIT: added mention of OK behavior)

[soa_rru]

Hmm with this new Core i cant seem to run 2xconsole clients along side the gpu client in Vista anymore, seems like the gpu core works intermittently when i have 2xconsoles running, like it isnt getting enough core time? funny never had a problem before.

[MtM]

I know this core has more communications going on, what I don't understand is why this wasn't noticed in beta testing I thought cpu usage spikes where reduced already.

ihaque can I ask since I know some people are expecting this to be the fix for the NaN issues as well, if that is the intended use or not? Trying to prevent dissapointments if it's not totally true.

Soa_rru are you setting affinities with a program other then fah?

[Gadreel]

thx mate, gotten the fahcore_14 and running 5900 and the temp is doing pretty good lower then the other jobs



-- Neo2 --
GTX295
PPD 6762
Min. Time / Frame : 52s
Avg. Time / Frame : 52s

-- Azazel5 --
PPD 4593
9800GX2
Min. Time / Frame : 1mn 23s
Avg. Time / Frame : 1mn 25s

[soa_rru]

Soa_rru are you setting affinities with a program other then fah?

No ive never had to with Vista before as the gpu client would work along side 2xconsole clients without any probs before, only thing i used to have set was Core Priority set to slightly higher in the gpu control panel.

Just updated from 6.20 to 6.23 and now the gpu client is working intermittently again even without any other fah clients running

Dont think this WU /Core can makes its mind up, this on on my 8800gt

Project : 5900
Core : Unknown
Frames : 100
Credit : 420


-- Nvidia GPU --

Min. Time / Frame : 34s - 10672.94 ppd
Avg. Time / Frame : 55s - 6597.82 ppd
Cur. Time / Frame : 55s - 6597.82 ppd
R3F. Time / Frame : 55s - 6597.82 ppd
Eff. Time / Frame : 1mn 15s - 4838.40 ppd

[MtM]

I'm going through 3 threads trying to find a plausible cause but I can't find one for your issue. I did notify ihaque in one of them of your issue just trying to make sure it's under his attention.

[ei57]

Up to 40% spread in frame rates - from 64 to 89 sec.
10C to 15C lower GPU temp than usual.
Uneven CPU demand.
Running 2 GPU's and SMP on a quad - one core feeding the GPU's. Affinity changed to include new core.
I've never seen such variation!

[21:02:08] Starting GUI Server
[21:03:12] Completed 1% - 64 sec
[21:04:45] Completed 2%
[21:05:57] Completed 3%
[21:07:26] Completed 4% - 89 sec

[ihaque]

Up to 40% spread in frame rates - from 64 to 89 sec.
10C to 15C lower GPU temp than usual.
Uneven CPU demand.
Running 2 GPU's and SMP on a quad - one core feeding the GPU's. Affinity changed to include new core.
I've never seen such variation!

This is all expected behavior, not to worry.

[ihaque]

No ive never had to with Vista before as the gpu client would work along side 2xconsole clients without any probs before, only thing i used to have set was Core Priority set to slightly higher in the gpu control panel.

What kind of problems are you having, exactly?

[MtM]

No ive never had to with Vista before as the gpu client would work along side 2xconsole clients without any probs before, only thing i used to have set was Core Priority set to slightly higher in the gpu control panel.

What kind of problems are you having, exactly?

I think there been a bit of a correlation of speech confusion or something, I thought he got problems with his console's not the gpu client itself. As I understand it now, he's just seeing swaying frame times and it's not tied to the cpu consoles.

Sorry for probably adding to the confusion, Soa_rru sorry for not understanding before what you ment

[HD580]

The voltage regulator current is bouncing from 32 to 48 amps.

By HD580
Compared to Project: 5753 (Run 11, Clone 69, Gen 172) which is a somewhat constant line. The 384 and 353 pointers are constant too but at lower amps.

By HD580
Have there been changes to reduce the stress on the gpus?

[^w^ing]

Does this core do some different type of analysis than core 11 or is this just somewhat tweaked/newer core 11?
from some of the above reports it seems that this core does artificially throttle itself

[p2501]

All this high/low load just to lower GPU temps? Or what's behind this? Are we betatesting something here? Will you tell us what's behind these fluctuations and if it's not changes for science but for lower GPU temps or whatnot, can I get rid of this by removing advmethods?? Those fluctuations can't be good for the hardware in the long run.

Edit: Did you guys notice that this client/WU mix produces stuttering, both for desktop work and video (GTX295 on 185.20)? It obviously has been tested in the betatesters group which leads me to the thought that this behaviour shouldn't have gone unnoticed, so do you even care?

[soa_rru]

No ive never had to with Vista before as the gpu client would work along side 2xconsole clients without any probs before, only thing i used to have set was Core Priority set to slightly higher in the gpu control panel.

What kind of problems are you having, exactly?

I cannot run 2xconsole clients along with the gpu client now (vista) its like the gpu isnt getting enough cpu core time. Seems to work ok (but with reduced PPD) if i just use 1xconsole client with the gpu client. If i just run the gpu client by itself then i get full PPD from the gpu.

Mind you i suppose if the PPD stays as high as it is i wont need to run any other clients

[Gadreel]

All this high/low load just to lower GPU temps? Or what's behind this? Are we betatesting something here? Will you tell us what's behind these fluctuations and if it's not changes for science but for lower GPU temps or whatnot, can I get rid of this by removing advmethods?? Those fluctuations can't be good for the hardware in the long run.

Edit: Did you guys notice that this client/WU mix produces stuttering, both for desktop work and video (GTX295 on 185.20)? It obviously has been tested in the betatesters group which leads me to the thought that this behaviour shouldn't have gone unnoticed, so do you even care?

shuttering? i aint notiecing that on my gtx295 with 185.20 display drivers, what i notice is the HUGE PPD diffrence betwin them one is getting 6720.00 ppd the other one is getting 5852.90 ppd both are clocked the same