Recommend me a video card for number crunching!
8 Jul 2011 16:37:17 UTC
Topic 195852
(moderation:
Hello all!
I'm about to buy a PCIe16x graphics card in order to improve the (einstein@home) crunching performance of my computer.
Specs:
Dualcore Athlon64 (Socket AM2)
4GB DDR2 RAM
Windows XP SP3
Board is ABIT AN52 with one PCIe16x bus.
I'm looking for a graphics card which will be used exclusively for number crunching since I don't play any games. This means that I don't need insane amounts of VRAM on the card but a very fast GPU, and another thing - I don't want the card to be too noisy since I'm running my main rig in the bedroom 24/7 and I need to be able to sleep right next to it without resorting to ear plugs :)
I'm going to buy the card from this store:
http://www.links.hr/index.php?option=katList&id_kategorija=0518
The prices are issued in Croatian Kuna (HRK); the budget is not an issue but I'd like to keep it below $200 if possible (1000 HRK). Do you see any cards you'd recommend? I'd really appreciate any help with this one, as I'm stuck in stone age with an ancient entry-level Radeon card and I know little about video cards to begin with.
Recommend me a video card for number crunching!
)
better link:
http://www.links.hr/index.php?option=katList&deftab=yes&page=sviartikli&id_kategorija=0518
First of all remember that
)
First of all remember that Einstein doesn't support OpenCL so you can't use ATI cards here at this time. You're limited to NVIDIA cards if you want to crunch Einstein with it. The top performing NVIDIA GPU today is the GTX580 which is listed at 4169 at the online store you mentioned. The next card below that is the GTX570 for which there are 2 models, 1 at 3059, the other at 2899. The only advantage of the more expensive model is that it has DP, whatever that is.
To get below the 1000 mark you're going to have to settle for an older model. Is that wise? Well, GPUs are evolving rapidly and already some of the older cards are not recommended at GPUgrid. I have no idea how long older cards will be suitable for Einstein but it's only a matter of time before they are not suitable here either, due to insufficient RAM, inability to meet the deadline or obsolete compute capability. I recommend investing in something that is likely to be useful for a long time so I recommend GTX570. The GTX580 costs a lot more but gives only a small increase in performance IMHO.
My 570 is an Asus brand and it's very quiet at 85% fan speed. The online store you linked to is also selling Asus brand so I think you would be happy with the noise level of the 570 based card.
BOINC FAQ Service
Official BOINC wiki
Installing BOINC on Linux
The extra VRAM is useful for
)
The extra VRAM is useful for Einstein@home with newer Fermi based cards including the 4xx and 5xx cards. With this setup you can run multiple work units at once. You will need around 350MB for each work unit that you run. A decent mid-range card would be the GTX 560 which is available in 1 or 2GB versions but those cards are quite a bit higher than 1000. With the link you provided, I see a GTX 550 1GB card priced at 1099 which should be a decent card as well. With a 1GB card you should be able to run at least two work units at once. 3 might be cutting it close but you can try. The GTX 570 or greater would be the best choice for running at least 3 work units at once.
Regarding noise, the 5xx series cards seem to be a bit quieter than the previous generation. However, you will want to turn up the fan speed enough to where the GPU does not overheat under load. Einstein@home generally will not load the GPU 100% even with multiple work units running at once which helps keep the temperature down a bit.
Please note that as of this
)
Please note that as of this post there are 18.3 days worth of work left in the BRP3 search, this is the only work that can be run on a GPU here at Einstein right now.
There has been some talk about other GPU-apps in development but no timeline for when they will be released has been given.
RE: There has been some
)
Err ?!? You ought check out this page then. :-)
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
RE: RE: There has been
)
That points to this thread which contains this message. "I hope there will be one later" isn't much of a timeline IMHO. I would call that "some talk".
BOINC FAQ Service
Official BOINC wiki
Installing BOINC on Linux
Alright, I'm buying one of
)
Alright, I'm buying one of these two video cards - the price difference is small but I have no idea which one is better for crunching. Here's what I'm eyeing:
Card 1:
http://www.links.hr/index.php?naziv=Grafi%E8ka+kartica+PCI-E+GAINWARD+GeForce+GTX+550+Ti+GS%2C+1GB+DDR5%2C+HDMI%2C+DVI%2C+D-SUB&option=artikl&id_kategorija=05180306&id_artikl=051.803.778
Card 2:
http://www.links.hr/index.php?naziv=Grafi%E8ka+kartica+PCI-E+EVGA+GeForce+GTX+550+Ti+FPB%2C+1GB+DDR5%2C+HDMI%2C+DualDVI&option=artikl&id_kategorija=05180306&id_artikl=051.803.816
Even if there is no work for GPUs right now, a new video card will certainly come in handy down the road. =)
The only difference I see
)
The only difference I see between those 2 cards is the price and the video outputs. Make sure the one you get has an output connector that matches your monitor's input connector.
Regarding availability of work, if Einstein runs out of GPU tasks you can always use it at GPUgrid.
BOINC FAQ Service
Official BOINC wiki
Installing BOINC on Linux
RE: That points to this
)
Yes, which indicates that it's in the testing phase. If anyone has accurate predictions of the results of that testing - prior to the tests being done that is, so that we can then perhaps speed up development even more by obviating their need - then the developers would love to hear from you. Don't be shy, step forward ! :-)
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
I ended up purchasing GeForce
)
I ended up purchasing GeForce GTX 550 Ti and the computation speed has increased twelvefold. Thanks to all who offered assistance. :)