My GTX 590 normaly run GPU1 usage 70-85% Temp 90c GPU2 usage 80-95% Temp 80c .
After intalling luntic and editing app_info MB 0.33 AP 0.5 and save it and restart Computer.
Cheking out Boinc, it seem to be fine 6x0.04cpu+0.33nvidiaGPU.
Hmm...estimate time 35-40 min, before it was 10-15 min??
Start aftreburemer and seen at GPU usage 40-45%??????
WU:s was only 2% after 60 min?? and i editing back MB 1 AP 1 and now it runs like a hell=) and ofcourse hot like a hell=). Is it because dual GPU or what?

---

Now you're running six Cuda tasks at once, did you up the processpriority in the mbcuda.cfg file?

Claggy

YMMV but when have the 590 the best number was 2 WU at a time (0.5) not 3 (0.33), and yes it runs hot as a hell, so you need to use a program to increase the GPU fan speed. I use EVGA precision (free to DL form the EVGA site).

And don´t forget each AP WU needs a free core to feed the GPU do not use 0.04 with AP, you need to set to 1.0 (some very fast CPU´s could handle 0.5 for the CPU too, but not the I5 for sure). So if you run 2 WU at a time in a 2X GPU like the 590, you can´t do any CPU work (the I5 only have 4 cores) at the same time since you actualy run 4 AP GPU WU at a time. But don´t worry, that realy don´t matter, the 590 is a lot faster to crunch than any I5.

With Mb only you could try to start some CPU task, but not to many, maybe 2 or 3 at max, but you need to test what is the best number in your particular configuration, in most of my cases (i use I5 too but the old slow ones) more than 1 MB task slow downs all the system, that´s why i dont do CPU work.

---

No. What should editing on that file set priority high and pfblockspersm 4? and should edit those MBCuda42.aistub files too? This Computer is dedicated seti.

Juha

---

Thanx that 0.5 solve the problem. Now GPU usage are 95-98% bouth GPU:s Temp is 2 celcius higher and fan run 95%. Should i go to my work and get a bit LN2=)

---

Test with 1.0CPU and 0.5GPU you will see the WU will be crunched in a lot less time.

---

Test with 1.0CPU and 0.5GPU you will see the WU will be crunched in a lot less time.

just wondering doesn't just leaving 1 core free do the same as editing the file from 0.04 to 1 ???

and yes 2 units not 3 is better slows down to much at 3 . it will do 4 but toooooooo slow

---

Not exactly, you need to free the core only when crunch AP not with MB, edit the file do that for you automaticaly.

And not forget the sugested number is 1 core per AP WU who is crunching (4 in his case) not 1 core for all.

---

(Crossposted)

Just to be ABSOLUTELY clear, the recommendation of assigning one CPU core per AP task is properly accomplished by setting:

ngpus .5
ncpus 2

or

ngpus .33
ncpus 3

or does setting ncpus=1 reserve one core per AP launch independent of the ngpu setting?

---

Member of the 20 Year Club

(Crossposted)
...
or does setting ncpus=1 reserve one core per AP launch independent of the ngpu setting?

Yes, both ncpus and ngpus settings are per task.

                                                                   Joe

Not exactly, you need to free the core only when crunch AP not with MB, edit the file do that for you automaticaly.

And not forget the sugested number is 1 core per AP WU who is crunching (4 in his case) not 1 core for all.

Ok. I should do that too. Im now set manualy 2 cpu to work from boinc, but should editing from the files.
Rac is still climbing=)

---

This is an example of one of my working app_config.xml file

<app_config>
<app>
<name>setiathome_v7</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>0.1</cpu_usage>
</gpu_versions>
</app>
<app>
<name>astropulse_v6</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>1.0</cpu_usage>
</gpu_versions>
</app>
</app_config>

Hope that helps.

Actualy i use a newer version of AP crunching that uses a lot less CPU time but that is for another time.

---