confirmed CUDA and nvidia are the card of choice. I did folding @ home. my two GTX 285 cards did 100% load for 2 years 24/7.
PROTIP: install after market cooler and make sure cooling your boxen is a priority, the reference card is LOUD at 100%. I put artic cooling acellero extremes on my cards.. three 92 mm fans on each card and they are near silent @ 100% 24/7.
Topic: Explain CUDA? (Read 1327 times)
u can start by openning forum mining miners sections,
there is cuda compiled software there
there is cuda compiled software there
When I used openCL on my GTX460, it mined fine at ~60 MH/s, but when ever I stopped mining, it would crash my computer.. I switched to rpcminer-cuda with no performance loss and everything works fine.
On my GT420M CUDA works like a charm
Take a look at my mod:
http://forum.bitcoin.org/index.php?topic=27907.0
When I used openCL on my GTX460, it mined fine at ~60 MH/s, but when ever I stopped mining, it would crash my computer.. I switched to rpcminer-cuda with no performance loss and everything works fine.
I tried mining on Nvidia 9800GT + rpcminer-cuda on ubuntu 10.04 (25 Mhashes, tops).
There was a whole bunch of problems. Download cuda 3.0 SDK only (newer won't work at all).
After building, copy *.cubin from windows rpcminer distribution.
After all that I may say CUDA doesn't worth it. rpcminer-cuda is much slower (probably 15-20%) than poclbm + OpenCL.
(In case if you didn't know, Nvidia can run OpenCL too.)
There was a whole bunch of problems. Download cuda 3.0 SDK only (newer won't work at all).
After building, copy *.cubin from windows rpcminer distribution.
After all that I may say CUDA doesn't worth it. rpcminer-cuda is much slower (probably 15-20%) than poclbm + OpenCL.
(In case if you didn't know, Nvidia can run OpenCL too.)
Can anyone explain CUDA to me? From what I gather its just a seperate way of computing data for nvidia cards...but do I need to "activate" it everytime I use it or is it an "install and be lazy" solution?
CUDA is nvidias language for talking to its cards when you want the card to crunch data (as opposed to displaying a scene in a video game). basically its a set of rules on how to get data to the card, what to do with it, and how to get it out.
its a "install and forget" deal.
Can anyone explain CUDA to me? From what I gather its just a seperate way of computing data for nvidia cards...but do I need to "activate" it everytime I use it or is it an "install and be lazy" solution?
I'm just a little confused
I'm just a little confused
Jump to: