Hi there,

meanwhile i've upgraded the system to 8GB and Windows 7 professional. I'm happy with the system overall, but the crunching throughput of the system is "not amusing" compared to Windows XP.

So i tried again with using the -use_sleep and the -hp parameters. See for yourself at these WUs:

GT 430:
http://setiathome.berkeley.edu/result.php?resultid=4554904725

Name ap_02my11aa_B3_P1_00362_20151127_00804.wu_1
Workunit 1981389296
Created 27 Nov 2015, 21:35:03 UTC
Sent 27 Nov 2015, 21:35:04 UTC
Report deadline 22 Dec 2015, 21:35:04 UTC
Received 1 Dec 2015, 14:12:00 UTC
Server state Over
Outcome Success
Client state Done
Exit status 0 (0x0)
Computer ID 157931
Run time 6 hours 38 min 17 sec
CPU time 2 min 7 sec
Validate state Valid
Credit 564.83
Device peak FLOPS 692.35 GFLOPS
Application version AstroPulse v7
Anonymous platform (NVIDIA GPU)
Peak working set size 52.46 MB
Peak swap size 54.25 MB
Peak disk usage 0.04 MB

So, so, Ok

GT 640:
http://setiathome.berkeley.edu/result.php?resultid=4554970759

Name ap_02my11aa_B5_P0_00173_20151127_27222.wu_1
Workunit 1981420855
Created 27 Nov 2015, 22:05:35 UTC
Sent 27 Nov 2015, 22:05:38 UTC
Report deadline 22 Dec 2015, 22:05:38 UTC
Received 1 Dec 2015, 14:53:28 UTC
Server state Over
Outcome Success
Client state Done
Exit status 0 (0x0)
Computer ID 157931
Run time 5 hours 8 min 45 sec
CPU time 20 sec
Validate state Valid
Credit 569.77
Device peak FLOPS 692.35 GFLOPS
Application version AstroPulse v7
Anonymous platform (NVIDIA GPU)
Peak working set size 40.71 MB
Peak swap size 41.02 MB
Peak disk usage 0.04 MB

http://setiathome.berkeley.edu/result.php?resultid=4554812578

Name ap_02my11aa_B4_P0_00033_20151127_00807.wu_0
Workunit 1981345400
Created 27 Nov 2015, 20:52:20 UTC
Sent 27 Nov 2015, 20:52:21 UTC
Report deadline 22 Dec 2015, 20:52:21 UTC
Received 1 Dec 2015, 9:44:42 UTC
Server state Over
Outcome Success
Client state Done
Exit status 0 (0x0)
Computer ID 157931
Run time 4 hours 58 min 41 sec
CPU time 1 min 31 sec
Validate state Valid
Credit 539.61
Device peak FLOPS 692.35 GFLOPS
Application version AstroPulse v7
Anonymous platform (NVIDIA GPU)
Peak working set size 40.73 MB
Peak swap size 41.02 MB
Peak disk usage 0.05 MB

Both units were 0.00% blanked!
Only disgusting... :/

Both:
http://setiathome.berkeley.edu/result.php?resultid=4554812446

Here the taskmanager:


Here SIV:


On XP this system was straight on the way to >10000 RAC.
Now RAC falls like a rock, although only crunching APs.

All in all i am disappointed and will try a newer Nvidia driver tonight.

---

Aloha, Uli

Before jumping to a newer driver read this thread:
http://setiathome.berkeley.edu/forum_thread.php?id=75633

You will have a real nightmare trying to use a GTX430 and a GTX640 for SETI in the same machine, they are a generation apart in terms of their driver requirements, and their configuration settings.

Your GTX430 is a "pre-Fermi" card, so is affected by this. Also make sure you are using drivers from the Nvidia site and NOT those that Windows decides you should have installed.

I would expect an Astropulse run time (on a PC not doing much else) for a GTX430 to be in the region of a an hour or two.

You WILL have to free up a core to get anything like sensible performance on that PC. Fiddling with "sleep", "hp" and other flags will only have minimal gains compared with making sure you have screen savers turned OFF, power management set to "High performance".

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

Thanks for the tip, but the GT 430 (without X) is a Fermi class GPU and the GT 640 (without X) is a low end Kepler, so they both require the same driver and the Nvidia site always offers the same drivers, regardless of GT 430 or GT 640, namely the latest. Tonight i tried the latest version from Nvidia, 359.00. That version sucks big time, the first time i start video streaming running BOINC, the driver restart about every other minute, so down the null device with it! :þ

Sorry, but don't you all get it, there always was at least one core free idling for feeding the GPUs? :?

I'm now back to driver version 337.88, which is the last driver doing the job done.
No -use_sleep, no -hp, nonetheless the system again does what it is supposed to do:



So everything is fine now and i will not change the system.

These GPUs are GT 430 and GT 640 and both are crunching at a rate, that is absolutely adequate for their class. Times for APs are ~2 hrs 10 mins for the GT 640 and ~6 hrs 30 mins for the GT 430. That is nowhere near the "an hour or two" for the GT 430 (without X!).

---

Aloha, Uli

Ull I am right now trialling a setup to have different setting on both my cards .

I have a 650 and 680 so there is a mismatch and setting that work on 1 don't work to well on the other .

It's a little complicated to tell you how to do it now . I have had some advise from William and Tbar to help me do this so if you can wait a few more days i'll put what I have done up in a thread I have made called

Different Setting for more than 1 GPU

If all goes well you will be able to tune each card separately and get a good boost to your RAC so please let me finish testing as I've only just completed the changes and started the client so I don't have any ap's at the moment .

If you look at my main computer don't let it worry you I aborted heaps of units I had a bad day and got a bit impatient and detached from the project to abort all the units

---

[quote]...

Sorry, but don't you all get it, there always was at least one core free idling for feeding the GPUs? :?

...{/quote]
Seeing as you're doing AP's on both GPU's you'll want at least 2 cores reserved on that old CPU. ;-)

Cheers.

Seeing as you're doing AP's on both GPU's you'll want at least 2 cores reserved on that old CPU. ;-)

Cheers.

75% CPU usage on a 4 core system = 1 core absolutely free!
Please have a look at the SIV graphs. If i would free 2 cores, i would have just another core idling:



BTW: I think here is a major misunderstanding in interpreting the SIV graphs. Red doesn't automatically mean "task swap", it simply means, the thread is "inside the operating system" (=inside OS itself and/or drivers) and not inside the user process - that's all!

---

Aloha, Uli

Ull looking at your siv graph I would say your doing MB's . Or are you now using -use_sleep in the cmd_line . As that would explain why there is no red bar on the siv graph .

My Quaddy is doing 5 Ap's now and I have 1 core free to feed the GPU and it's only using aprox 8% doing 2 units on the GPU.

Just wait I have only just started to get AP's on the other machine witch is the 1 that has 2 GPU's .

You can't have different setting with just putting different setting in the cmd_line you have to create another file and then edit the appinf to include the new file you have to make to control each GPU seperatly and split the commands and put certain ones in the new file and other commands in the cmd_line.

Let me finish testing and as I said I put the whole thing up on that thread I have created

---