F@H: new GPU2 NV Work Unit

[Insidious]

Probably don't want to lower your fans. You're still going to get some of the hot WUs from time to time.

-Sid

Have I mentioned I am seriously jealous of your hardware?

 

[Duvie]

So someone list cards...make and speed and ppd you are getting while running the work unit....

I like to get a feel for how some are getting such high ppd and Mark is not...

 

[Insidious]

Originally posted by: Duvie
So someone list cards...make and speed and ppd you are getting while running the work unit....

I like to get a feel for how some are getting such high ppd and Mark is not...

We've been trying to figure out what Mark's problem is for years! :laugh:''

-Sid

 

[geokilla]

PPD vary because the WU does not fold at a consistent rate. Sometime's it's faster, sometime's it's slower.

 

[IEC]

I've been getting about 1800PPD off my HD3850 rig. Currently folding a GPU2 p4752 (477pt)

Both my GPU2 client and my x86 client are running -advmethods -verbosity 9. That's what I used to run, no clue if that's the optimal or not. I tried out the SMP client but it's a pain and uses more resources.

 

[Markfw]

Originally posted by: Insidious

Originally posted by: Duvie
So someone list cards...make and speed and ppd you are getting while running the work unit....

I like to get a feel for how some are getting such high ppd and Mark is not...

We've been trying to figure out what Mark's problem is for years! :laugh:''

-Sid

I have 19 GLU clients (one currently broken). My PPD (except the wimpy 8800 GTS cards) ranges from 3700 to 4300 for various 8800 GT's and 9800 GT's all 600-740 mhz. Not much variance. The 9800 GTX+ cards range from 4600 to 5100 and the 260 (192 core) is 5300. All of these seem reasonable variance.

BUT a 260 getting 5900-7100 ppd just doesn;t sound right, way too much variance.

 

[Drsignguy]

I took a Screenshot of the differances of PPD the new WU's are getting. I have them fairly easy laid out.

The "Q6600 XBX2 GTX 260 2" is my only core 192 card. The "Q6600 EP45 GTX 260 1 and 2" are 65nm 216's and all others are 55nm 260/216's. 2 8800GTS that are 320MB and I set 1 for the larger client just to see the differance and it produces pretty well.

Sid, I never heard that from you about my hardware? :evil:

 

[Markfw]

what mhz is your 192 card running at ? I had to de-clock mine a little to be stable.

 

[Drsignguy]

Originally posted by: Markfw900
what mhz is your 192 card running at ? I had to de-clock mine a little to be stable.

The default speeds of that MSI card is 620/1080/1296. I have it set to 610/1060/1530.
My 8800 GTS cards are set to 1566 on the shaders and it seems to have a tiny bit better PPD with this new WU, so far....

EDIT: I too had to de-clock to become stable with F@H....

 

[Duvie]

Originally posted by: Drsignguy
I took a Screenshot of the differances of PPD the new WU's are getting. I have them fairly easy laid out.

The "Q6600 XBX2 GTX 260 2" is my only core 192 card. The "Q6600 EP45 GTX 260 1 and 2" are 65nm 216's and all others are 55nm 260/216's. 2 8800GTS that are 320MB and I set 1 for the larger client just to see the differance and it produces pretty well.

Sid, I never heard that from you about my hardware? :evil:

Those numbers are all over the place....I think you need to list the specs on your 2 Q6600's and Q9450....we need to see if all system variables take some effect.

I see a 4000 number on a 260 on 1 Q6600 box and like 6600 on another Q6600....

The questions would be for the systems...

CPU overclock
OS used
driver set used
video card core/mem/shader clock settings
how many smp units being ran.
Speed of bus for 2nd gpu....is it a 4x, 8x or 16x bus lane?

Marke may have many of his 2nd cards running in systems that only offer 8x if running 2 to 3 cards, or one specific slot not running 16x....

 

[Markfw]

It seems to make no difference what slot the card is in. In the other post where I link to my fahmon page, note the GPU1,2,3 on the 940. They are identical cards, and veru close in ppd.

 

[Duvie]

Have you verified that pci-express slot is running at 16x or 8x? Use GPU_Z to determine. I know we have talked about this in the past but this could be a different thing altogether. Usually FH has said they award more points as bonuses when work units take advantage of more resources.

Maybe this one is memory intensive, showing gains for people running 4 to 8 gigs of memory which may only benefit 64bit OSes....

You never know....but we should get to the bottom of this...

This reminds me of the time Mark and I found that certain work units were 2x faster if ran on a cpu with more L2 cache as related to the early core 2 duos...

 

[Markfw]

Yes, one at 16x, 2 at 8x, all three almost identical ppd, on identical cards. I don't think the 8x makes any difference.

 

[Drsignguy]

Originally posted by: Duvie

Originally posted by: Drsignguy
I took a Screenshot of the differances of PPD the new WU's are getting. I have them fairly easy laid out.

The "Q6600 XBX2 GTX 260 2" is my only core 192 card. The "Q6600 EP45 GTX 260 1 and 2" are 65nm 216's and all others are 55nm 260/216's. 2 8800GTS that are 320MB and I set 1 for the larger client just to see the differance and it produces pretty well.

Sid, I never heard that from you about my hardware? :evil:

Those numbers are all over the place....I think you need to list the specs on your 2 Q6600's and Q9450....we need to see if all system variables take some effect.

I see a 4000 number on a 260 on 1 Q6600 box and like 6600 on another Q6600....

The questions would be for the systems...

CPU overclock
OS used
driver set used
video card core/mem/shader clock settings
how many smp units being ran.
Speed of bus for 2nd gpu....is it a 4x, 8x or 16x bus lane?

Marke may have many of his 2nd cards running in systems that only offer 8x if running 2 to 3 cards, or one specific slot not running 16x....

Q9450 @ 3.4Ghz running 2 Smp, Vista 64, 182.06,(1) EVGA superclocked 55nm GTX 260/216, Default clocks 626/1053/1350. Set to 586/999/1548. (2) EVGA 55nm GTX 260/216, default clocks 576/999/1242. Set to 576/999/1512.


Q6600 @ 3.3Ghz running 2 Smp, XP 64, 182.06, (1) EVGA 65nm GTX 260/216, default clocks 576/999/1242. Set to 576/999/1512. (2) EVGA 65nm GTX 260/216, default clocks 576/999/1242. Set to 576/999/1512. <--- this card doesn't like going over this setting.

Q6600 @ 3.1Ghz running 2 smp, XP, 182.06, (1) EVGA 55nm GTX 260/216, default clocks 576/999/1242. Set to 576/999/1548. (2) MSI GTX 260/192, default clocks 620/1080/1296. Set to 610/1060/1530.

E7200 @ 3.6Ghz running 1 Smp, vista 32, 182.06, (1) EVGA 8800 GTS 320, default clocks 576/850/1350. Set to 576/850/1566.

E2160 @ 2.4Ghz No Smp, Vista 32, 182.06, (1) EVGA 8800 GTS 320, default clocks 576/850/1350. Set to 576/850/1566.

First cards are in 16x slots.
Second cards are in 8x slots.

And as mark has said, no difference.

I am wondering if the actual project has something to do with some of the fluctuations. Yes it's 5902 but each are different ie; 5902 ( R2,C796,G0) , 5902 (R11,c800,G0)

 

[Markfw]

Here is a new one. I tried to go to the new 182.08 drivers to get better ppd. My first machine was my 260/192@677/1080/1415. It WAS 5200 ppd. Then for the first few frames it said 5450 ppd. Great, so let put it on all the cards. Now look at this log:
[22:20:05] Verified work/wudata_07.log
[22:20:05] Verified work/wudata_07.edr
[22:20:05] Verified work/wudata_07.xtc
[22:20:05] Completed 87%
[22:20:05] Starting GUI Server
[22:25:03] Completed 88%
[22:29:33] Completed 89%
[22:33:59] Completed 90%
[22:38:48] Completed 91%
[22:43:46] Completed 92%
[22:48:41] Completed 93%
[23:01:23] Completed 94%
[23:15:32] Completed 95%
[23:29:49] Completed 96%
[23:44:18] Completed 97%
[23:51:10] Completed 98%
[23:55:54] Completed 99%


Notice its 4-4 1/2 mines for a but then goes to 15 minutes per step ! 1600 ppd ! Then back to 5 minutes...Is this a new driver issue ?????

And whats the ppd on youyr 260/192 ??

 

[Insidious]

The frame rate is NOT constant as a 59xx WU is processed.

It's been discussed a lot at fah and the devs. assure us that is normal. (It's scared the hell out of a lot of people)

-Sid

 

[Drsignguy]

I actually got better results with my 260/192 when running with vista as opposed to XP, and just by itself. When I get my next 55nm 260, I will be removing the 192 core and replacing it back in the vista rig.

 

[Drsignguy]

WU 5903 just came into play with a couple of my cards and PPD unknown, Credit unknown, ETA 5 3/4 hours and it's already a 5% on a GTX 260/216 55nm. As soon as something gives, I will update.

 

[Insidious]

Originally posted by: Drsignguy
WU 5903 just came into play with a couple of my cards and PPD unknown, Credit unknown, ETA 5 3/4 hours and it's already a 5% on a GTX 260/216 55nm. As soon as something gives, I will update.

It's just like the 5902

You can see a project listing here

More goodness for the fire!

Intro thread from fah (I feel dirty.... )

-Sid

 

[Drsignguy]

Nice, Thanks Sid. :thumbsup: I now saved that list for future reference.

 

[Denithor]

Sorry, hadn't looked in here in a few days.

Right now have two 1680 point WUs running on my rig.

Running 185.20 beta drivers with Vista 64 Ultimate.

9800GTX+ (756/1836/1123)
Project: 5903 (R13, C50, G0)
FahMon reports 5604 ppd right now for this WU.

8800GTS 512MB (670/1674/972)
Project: 5902 (R5, C687, G5)
FahMon reports 8744 ppd right now for this WU.

 

[Duvie]

Originally posted by: Denithor
Sorry, hadn't looked in here in a few days.

Right now have two 1680 point WUs running on my rig.

Running 185.20 beta drivers with Vista 64 Ultimate.

9800GTX+ (756/1836/1123)
Project: 5903 (R13, C50, G0)
FahMon reports 5604 ppd right now for this WU.

8800GTS 512MB (670/1674/972)
Project: 5902 (R5, C687, G5)
FahMon reports 8744 ppd right now for this WU.

INteresting....

My 9800's GT superclocked have much higher shader clock speed then yours and I only see like 4400ppd max....using latest drivers....I am using latest drivers but for winxp pro

 

[Markfw]

Mine are getting 5582.77 for my 9800 GTX+ @ 756.. 1836 shader.

 

[Insidious]

FahMon reports 8744 ppd right now for this WU

I'm pretty sure this number is borked.

-Sid

 

[Drsignguy]

Have you noticed that the Wu's 59xx are 1888 points now? My 260's average over 7400+/- with these babies!