| Author |
Message |
|---|
|
|
|
What times are you guys getting for fermi cards like 470, 480, 570 etc?
It's taking me 15 minutes for one unit, which is barely different than the other card for this, an oc'd 6970.
Maybe I need to overclock the shader; it's at the default of 1150MHz. Thing is I usually only overclock ati cards, all the nvidia cards I've had were quadros and stuff so I have no experience in this wrt what tools to use etc..
An equivilent to sapphire trixx would be nice.
Thanks
____________
 |
|
|
|
|
|
I'm not sure that a link to another BOINC project is prohibited by forum rules, but I'll risk to direct you for example here:
http://wuprop.boinc-af.org/results/projet.py?projet=PrimeGrid&application=Proth+Prime+Search+%28Sieve%29+%28cuda23%29
Look beneath at GPU section, you can see what is average time for every GPU which host participates in PrimeGrid and WUProp projects.
____________
|
|
|
|
|
|
Hi, its my lucky day.
I was looking recently for Tesla performance on primgrid without result. Then hYdrociTy join us with a nice Tesla video card and x3mEn shared a good web page.
But i was a little surprise when I saw Tesla performance. Its take me 1000 sec for a work unit (It can take less, cause I am running another boinc project on both core of my cpu) and I got a GTX 560 TI o'c Normaly I like EVGA, but for this time, I go for MSI cause of Twin Frozer. When running 24/24, it keep gpu at maximum 56 Celcius.
Your card have good stats : Number of processor cores: 448
Processor core clock: 1.15 GHz but I just got 384 processor cores and I do some good speed.
Well, have fun and lets continue crunching numbers!
|
|
|
|
|
|
Thanks for the link x3mEn.
Yea so this Tesla C2070 is pretty stable now, ~980 seconds for a unit.
The default fan tends to keep the card at 80-85 degrees Celsius, but this computer room ambient temperature is always 22 degrees. The card seems to get warmer than the 6970 when both are at 50% fan, but It seems the 6970 fan is somewhat larger.
I guess if I run the card in TCC mode without a monitor plugged in the card may be a bit faster so I'll report back when I get to it.
@ Pogo yea I actually don't feel scared with crunching on the Tesla since it was intended to be run 24/7 anyway. The 6970 I give a break now and then with a few long gaming sessions :)
____________
|
|
|
|
|
|
My MSI-brand GTX 570 runs the current PPS Sieve WUs in 740-750 seconds (~12.5 minutes). I haven't touched the clocks which came from the factory at 786/2100/1572. Card temps are reported around 80c; in my room the ambient is unfortunately about 25c-30c this time of year. Fan is usually at about 55%.
--Gary
p.s. ubuntu 10.10, driver 270.41.06 |
|
|
|
|
|
The C2070 runs 85c at overclocked of 700 core, 1400 shader, 1536 memory at 85c and finishes a wu in 790 seconds now :)
____________
|
|
|
|
|
|
My NVS 3100M takes around 9 hours to complete one of these. No, I'm not kidding! :)
____________
|
|
|
|
|
|
I've looked but can't seem to find the thread so I'll ask.. is the current primegrid pps cuda app limited to windows 64 bit and dp gpu cards? I have a 460 but not 64 bit windows. |
|
|
Scott Brown Volunteer moderator
Project administrator
Volunteer tester
Project scientist
Send message
Joined: 17 Oct 05
Posts: 2165
ID: 1178
Credit: 8,777,295,508
RAC: 0
                                     
|
I've looked but can't seem to find the thread so I'll ask.. is the current primegrid pps cuda app limited to windows 64 bit and dp gpu cards? I have a 460 but not 64 bit windows.
PPS Sieve and CW sieve cuda apps run on any cuda capable card on either 32-bit or 64-bit OS.
LLR cuda application (in testing) will require double precision cards and (unless something changes) a 64-bit OS.
on PRPnet, geneferCUDA (for the higher ranges only) requires double precision cards, but runs on either 32-bit or 64-bit OS.
____________
141941*2^4299438-1 is prime!
 |
|
|
|
|
LLR cuda application (in testing) will require double precision cards and (unless something changes) a 64-bit OS.
thanks Scott that's the part I was interested in. |
|
|
