Additional..

If you use Raistmer's app_info.xml entry for AP..

You should delete or change the

<flops>30987654321</flops>

entry.

If you use already flops, then you should take maybe first the half flops of your MB CUDA entry.

If this flops entry is too big, then you could get -177 errors.


- Best regards! - Sutaru Tsureku, team seti.international founder. - Optimize your PC for higher RAC. - SETI@home needs your help. -

@Sutaru Tsureku
I don't use flops so I will delete the line
Bye Paul

---

On my GTX460, I changed my MB count to 0.49 and the AP count to 0.51. That way I can run either 1MB and 1AP or 2MB but not 2AP.

---

perryjay wrote:

...
Just for information, I have mine set to .33 for both the AP app and the MB apps but the per device set at 2 on the AP app so that I can run either one AP and two MBs or two APs and one MB or three MBs. This way it might try to start a third AP but it won't go anywhere. It will just count up the elapsed time on the third AP without any work actually being done. When it finishes one of the other APs it will start up the third and elapsed time will reset to zero and start again.

I'm skeptical that the elapsed time will reset to zero, BOINC thinks applications obey its orders on when to run. I think you'd be better with .34 for AP and .31 for MB. Then BOINC would never try to run more than 2 APs or 3 MBs at once, but could run all the combinations you want to allow.

                                                                 Joe

I think you'd be better with .34 for AP and .31 for MB. Then BOINC would never try to run more than 2 APs or 3 MBs at once, but could run all the combinations you want to allow.

                                                                 Joe

Would that not also run 2 MBs & 1 AP?

I think you'd be better with .34 for AP and .31 for MB. Then BOINC would never try to run more than 2 APs or 3 MBs at once, but could run all the combinations you want to allow.

                                                                 Joe

Would that not also run 2 MBs & 1 AP?

Yes.

---

With each crime and every kindness we birth our future.

I think you'd be better with .34 for AP and .31 for MB. Then BOINC would never try to run more than 2 APs or 3 MBs at once, but could run all the combinations you want to allow.

                                                                 Joe

Would that not also run 2 MBs & 1 AP?

Yes, and perryjay indicated he wanted to allow that combination. There will always be far fewer AP tasks than MB, so that's sensible in my estimation.

                                                                  Joe

Thanks for the hint Joe. I went ahead and made the change to.31/.34. I have managed to snag three APs but they are way down on the food chain. Probably wont get to them until late tomorrow. I'm not even sure my machine can run two APs at a time, it tends to wait for me to go to bed before it tries silly things like that.

---

PROUD MEMBER OF Team Starfire World BOINC

I did my first cuda ap wu unit in a litter over 1 hour. Now that is fast.

---

Looks good Paul. Should be no problem validating. Now you get to sit back and wait for some more to roll in.

---

PROUD MEMBER OF Team Starfire World BOINC

Looks good Paul. Should be no problem validating. Now you get to sit back and wait for some more to roll in.

@perryjay
Thanks for your help. I am trying out the boinc reschudler v26. I expect my rac to improve due to the speed of the cuda ap app.

@raistmer
Thanks for your app.

---

I was following Joe's example, and set AP for .34, and MB for .31. I wanted to be able to run 2 AP units and a MB for a total of three. 2 AP units are running now on each GPU with no MB, but I get this message.

8/1/2011 10:33:25 AM SETI@home [coproc_debug] Insufficient CUDA for ap_26fe11ac_B5_P0_00340_20110722_19139.wu_1: need 0.340000
8/1/2011 10:33:31 AM SETI@home [coproc_debug] Insufficient CUDA for ap_26fe11ac_B5_P0_00340_20110722_19139.wu_1: need 0.340000
8/1/2011 10:33:33 AM SETI@home [coproc_debug] Insufficient CUDA for ap_26fe11ac_B5_P0_00340_20110722_19139.wu_1: need 0.340000

Do I have this set correctly?

Steve

---

Warning, addicted to SETI crunching!
Crunching as a member of GPU Users Group.
GPUUG Website

I was following Joe's example, and set AP for .34, and MB for .31. I wanted to be able to run 2 AP units and a MB for a total of three. 2 AP units are running now on each GPU with no MB, but I get this message.

8/1/2011 10:33:25 AM SETI@home [coproc_debug] Insufficient CUDA for ap_26fe11ac_B5_P0_00340_20110722_19139.wu_1: need 0.340000
8/1/2011 10:33:31 AM SETI@home [coproc_debug] Insufficient CUDA for ap_26fe11ac_B5_P0_00340_20110722_19139.wu_1: need 0.340000
8/1/2011 10:33:33 AM SETI@home [coproc_debug] Insufficient CUDA for ap_26fe11ac_B5_P0_00340_20110722_19139.wu_1: need 0.340000

Do I have this set correctly?

Steve

I would say yes, but i have no clue why no MB is running.
2 APs are running like the settings indicates.

The message only says it couldfinnish faster while running less instances.
What we already know.

---

With each crime and every kindness we birth our future.

That is the way mine is set and right now I'm running one AP and two MBs. Did you remember to set the instance per device to two?

It looks like mine tried to run two APs at a time this morning but judging from the times it ran one to completion before it actually started running the second one. I don't have the coproc_debug set so I don't know if I would have seen that message. Has anyone successfully managed to get two APs to run with one or more MBs at one time?

---

PROUD MEMBER OF Team Starfire World BOINC

I'm trying 2 APs & 1 MB now, so far it looks OK, using 1000mb vram out of 1024...

I was following Joe's example, and set AP for .34, and MB for .31. I wanted to be able to run 2 AP units and a MB for a total of three. 2 AP units are running now on each GPU with no MB, but I get this message.

8/1/2011 10:33:25 AM SETI@home [coproc_debug] Insufficient CUDA for ap_26fe11ac_B5_P0_00340_20110722_19139.wu_1: need 0.340000
8/1/2011 10:33:31 AM SETI@home [coproc_debug] Insufficient CUDA for ap_26fe11ac_B5_P0_00340_20110722_19139.wu_1: need 0.340000
8/1/2011 10:33:33 AM SETI@home [coproc_debug] Insufficient CUDA for ap_26fe11ac_B5_P0_00340_20110722_19139.wu_1: need 0.340000

Do I have this set correctly?

Steve

I would say yes, but i have no clue why no MB is running.
2 APs are running like the settings indicates.

The message only says it couldfinnish faster while running less instances.
What we already know.

Ill bet it's running the cache in FIFO and the next wu is an AP. Has to work through the cached AP to get to the next MB wu

edit - has to do also with all the discussed wu are for the GPU so boinc won't jump down in the cache to get a MB wu.

---

Boinc....Boinc....Boinc....Boinc....

Ill bet it's running the cache in FIFO and the next wu is an AP. Has to work through the cached AP to get to the next MB wu

edit - has to do also with all the discussed wu are for the GPU so boinc won't jump down in the cache to get a MB wu.

Can't be why, with cuda set to .34 on APs it won't try running a 3rd AP, it could go to a MB with cuda count .31.

This has been my experience.
Boinc refuses to jump down into the cache to get a smaller (.31) GPU wu but will jump down into the cache to get a CPU wu.

---

Boinc....Boinc....Boinc....Boinc....

This has been my experience.
Boinc refuses to jump down into the cache to get a smaller (.31) GPU wu but will jump down into the cache to get a CPU wu.

I think Geek has the right idea. Steve, have you checked to see if it has a third AP sitting there just running up time without actually doing anything? I caught mine doing that on one pair of APs.

The ones that ran this morning look like that is what happened but I can't be sure. This second one is still running so I can't tell if maybe it's just heavily blanked or didn't start until the first one finished. I also have a couple that came out with 30 repetitive that I think may have been seconds of a pair. I've reduced my unroll from 10 to 8 to see if that will help but I don't get enough to check very fast.

---

PROUD MEMBER OF Team Starfire World BOINC

This has been my experience.
Boinc refuses to jump down into the cache to get a smaller (.31) GPU wu but will jump down into the cache to get a CPU wu.

Not true.

I was running 1/1 over 2 weeks now and the MB units had a much longer deadline.
Even the AP tasks was downloaded 7 days earlier.

---

With each crime and every kindness we birth our future.

Mike, are you running 50/50 or 51/49? At 50/50 mine tried to start a second AP if it was next in line and count up time without any progress but when I changed to 51/49 it would run one AP and one MB or two MBs depending on what I had in fifo order. It would jump any other AP I had to get to another MB if it was already running an AP. this is what we are trying for with the 34/31.

Okay, finished the second one http://setiathome.berkeley.edu/result.php?resultid=2019447019
got an interesting message in the stderr I hadn't noticed before...

WARNING: Can't follow BOINC's proposal, will try to find free device..

That must be why it took so long.

Went back and checked the ones I got with the 30 replicating, all three of them also had that message. Those I've done after switching to unroll8 are finishing clean though even with that message.

---

PROUD MEMBER OF Team Starfire World BOINC