Responding to my own post, I recently tested the i7 haswell variant of our Limulus and was 1.7 GFLOPS short of 0.5 TFLOPS (498.3 GFLOPS) running HPL. I suppose with a little more fussing I could find 2 extra GFLOPS, but I'm impressed with what AVX2 with FMA can do, at least for HPL, I have not checked anything else yet.
Latest results are here: http://limulus.basement-supercomputing.com/wiki/LimulusBenchmarks Has anyone else seen any similar big jumps in performance? (of course compiler support is needed) -- Doug > Thanks Joe, > > BTW, after getting our software stack to run on Haswell, > I managed to get 385 GFLOPS (HPL, double precision, CPU only) > on the entry level Limulus. That is approximately $15.6/GFLOP > More information: > http://limulus.basement-supercomputing.com/wiki/LimulusBenchmarks > > I'll be publishing more HPC benchmarks as I run them. > And, I'll have more Hadoop benchmarks as well. > > -- > Doug > > > >> [Disclosure: we do have a business relationship with Basement >> Supercomputing. ] >> >> I have to say, that one of the most popular aspects of the show for us, >> was the Limulus system that Doug Eadline had set up in our booth. This >> is a terrific system. >> >> I really like the concept of a personal supercomputer, one where, as >> Doug puts it, you own the reset switch. We tried, and did not succeed, >> in building interest in "muscular desktops" with huge amounts of >> processors, IO, and graphics. It cost way to much to build these. >> >> Doug comes along and in very beowulf-ish fashion, says "hey, lets build >> a very low electrical power many core distributed system from lower cost >> parts". It took a few years from project inception, but the concept is >> sound. Far more sound than the "muscular desktop" strategy. >> >> He ran hadoop on it, and was showing off running the overall system. I >> think the configs for hadoop might be slightly different than whats on >> the product page: >> http://www.basement-supercomputing.com/index.php/products/hikashop-menu-for-products-listing >> , but its similar enough that you can get the concept. >> >> I was simply blown away by this. The response came anywhere from "thats >> exactly what we need" to "this is very cool" with everything in between. >> They are awesome boxes, and it looks like it could fill a very nice >> niche for a number of folks. >> >> If you've not checked out the Limulus project in the past >> (http://limulus.basement-supercomputing.com/ and twtr @LimulusProject) , >> definitely have a look at it. Returning to our core value proposition, >> of power at lower cost, with creative designs, is deeply gratifying to >> me. Have a gander at the benchmark page: >> http://limulus.basement-supercomputing.com/wiki/LimulusBenchmarks >> >> It needs to be emphasized that this is not a toy system. You can do >> real work on it, and its trivial to setup and get going. >> >> Bravo to Doug! >> >> >> -- >> Joseph Landman, Ph.D >> Founder and CEO >> Scalable Informatics, Inc. >> email: [email protected] >> web : http://scalableinformatics.com >> twtr : @scalableinfo >> phone: +1 734 786 8423 x121 >> cell : +1 734 612 4615 >> _______________________________________________ >> Beowulf mailing list, [email protected] sponsored by Penguin Computing >> To change your subscription (digest mode or unsubscribe) visit >> http://www.beowulf.org/mailman/listinfo/beowulf >> >> -- >> Mailscanner: Clean >> > > > -- > Doug > > -- > Mailscanner: Clean > > _______________________________________________ > Beowulf mailing list, [email protected] sponsored by Penguin Computing > To change your subscription (digest mode or unsubscribe) visit > http://www.beowulf.org/mailman/listinfo/beowulf > -- Doug -- Mailscanner: Clean _______________________________________________ Beowulf mailing list, [email protected] sponsored by Penguin Computing To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
