Re: difference between cuda and stream
Mike Reed <[email protected]> Tue, 8 Jan 2013 20:05:46 +0000
| Newsgroups | gmane.comp.distributed-net.rc5 |
|---|---|
| Message-ID | <CACwkm_Oaf8dF26dLq66QZfTaFCRM3Sx+aORPb8CqKxk-fXCfuw@mail.gmail.com> |
Hi, One stats unit / block is 2^28 keys in length. Not nearly something or maybe something. For graphics cards, iter 64 (i.e. 64 x 2^28) is common. Proxies give out packets up to iter 256. Kind regards, Mike On Jan 8, 2013 7:58 PM, "bert" <[email protected]> wrote: > Bruce thank you for a real post on my question and I thank you for it. > > Bert > On 1/8/2013 2:47 PM, Bruce Slade wrote: > > Bert, > Having both ATI/AMD and Nvidia cards, there is no comparison between the > two for crunching RC5 as the ATI/AMD cards are much much faster. The > fastest nvidia I have is a 560ti, and the performance of that card is > matched by an AMD 6570 which costs much less and consumes much less power > and puts out basically no heat, where the 560ti sucks power and puts out > much heat. An ATI/AMD 6770, using the stream519 client on winXP32 will > crunch a 32stats credit workunit/block/whatever in 128 seconds, which > works out to 21,600 points a day, which I think is what you are asking. > While I don't have a GTX 690, I would seriously doubt that it would match > that in points per day on RC5, and again, even if it did the heat generated > and power used would be in favor of the AMD card. > > An other item for you to keep in mind is when you start using the AMD 7770 > cards or higher, you will have to be using the rc5 openCL client as changes > in the chip design put the stream client out of business for that series. > And also you need to make sure that the motherboard PCI version will > support the upper end AMD PCI card software requirements. > > On the other hand if you also crunch folding@home, the opposite is true > as nvidia totally outperforms the amd cards. So buy the card for the > predominant distributed computing project you intend to use it mainly for. > > > On 1/7/2013 5:20 PM, bert wrote: > > Would you know how much per day a 7970 does? > > Bert > On 1/7/2013 8:02 PM, Joseph Kaye wrote: > > On 1/7/2013 6:02 PM, bert wrote: > > Since I don't own an AMD card and have been strictly NVIDIA brand,I've > been curious as to which is faster cuda based cards or stream based > cards. I have been looking at the GTX 690 and the ATI 7970(is this the > latest?) I'm thinking that a GTX 690 will do about 64k blocks a day,but > having never had a stream card I can't imagine what one would do in a > day. Thanks. > > > Bert > _______________________________________________ > rc5 mailing listrc5-Ra3b/[email protected]://lists.distributed.net/mailman/listinfo/rc5 > > Looking at the Client Speeds Database > (http://cgi.distributed.net/speed/query.php?cputype=all&arch=4&contest=rc572&multi=3) > I'd say the DAMMIT (ATI/AMD) cards are much faster. An AMD Radeon HD > 7970 is doing 3,600 Mkeys/sec. > > > > _______________________________________________ > rc5 mailing listrc5-Ra3b/[email protected]://lists.distributed.net/mailman/listinfo/rc5 > > _______________________________________________ > rc5 mailing listrc5-Ra3b/[email protected]://lists.distributed.net/mailman/listinfo/rc5 > > > > _______________________________________________ > rc5 mailing listrc5-Ra3b/[email protected]://lists.distributed.net/mailman/listinfo/rc5 > > > > _______________________________________________ > rc5 mailing list > rc5-Ra3b/QYEcJ3d140v2zMXi0fjHoOT/h/[email protected] > http://lists.distributed.net/mailman/listinfo/rc5 > > _______________________________________________ rc5 mailing list rc5-Ra3b/QYEcJ3d140v2zMXi0fjHoOT/h/[email protected] http://lists.distributed.net/mailman/listinfo/rc5