How Many Processor Cores Are Enough?
|
From: Jon Forrest on 27 Sep 2006 19:48 Dennis M. O'Connor wrote: > "Jon Forrest" <forrest(a)ce.berkeley.edu> wrote ... >> Today I read that we're going to get quad-core processors >> in 2007, and 80-core processors in 5 years. This has >> got me to wondering where the point of diminishing returns >> is for processor cores. > > It depends on your application. Examples: A web server > like Apache can effectively use a lot more cores than > an old DOS game. But future games on the PC may be > able to effectively exploit every core you can provide them. No doubt, but I'm talking about general purpose computing. I bet there is diminishing return curve for web servers too. The curve would look different at each site, but it would exist. Jon
From: Rick Jones on 27 Sep 2006 20:24 Casper H.S. Dik <Casper.Dik(a)sun.com> wrote: > It all depends on the bandwidth. (Which means it ain't a pretty > picture for Intel as long as they keep the FSB) Is it really just a question of bandwidth? I would have thought that application (I'm assuming the system vendors deal with the OSes) behaviour would be equally important. How different is having an FSB for a single socket with N cores on the chip than having a "link" for a single-socket with N cores on the chip? I would think that as the cores per chip increase, the issues that the folks selling large SMP's deal with will become known to the single-socket crowd. rick jones -- portable adj, code that compiles under more than one compiler these opinions are mine, all mine; HP might not want them anyway... :) feel free to post, OR email to rick.jones2 in hp.com but NOT BOTH...
From: Rick Jones on 27 Sep 2006 20:26 Jon Forrest <forrest(a)ce.berkeley.edu> wrote: > I bet there is diminishing return curve for web servers too. The > curve would look different at each site, but it would exist. Yes - particularly if the web server was constrained to a single NIC - albeit some NICs out there now can spread their interrupt load across cores. Increasing core counts won't do all that much for individual TCP connections - getting very much parallelism in a single connection isn't really possible. rick jones -- a wide gulf separates "what if" from "if only" these opinions are mine, all mine; HP might not want them anyway... :) feel free to post, OR email to rick.jones2 in hp.com but NOT BOTH...
From: Terje Mathisen on 28 Sep 2006 00:34 Casper H.S. Dik wrote: > Jon Forrest <forrest(a)ce.berkeley.edu> writes: > >> Today I read that we're going to get quad-core processors >> in 2007, and 80-core processors in 5 years. This has >> got me to wondering where the point of diminishing returns >> is for processor cores. > > Sun has been shipping 8 core CPUs since, I think, late last year. > > It all depends on the bandwidth. (Which means it ain't a pretty > picture for Intel as long as they keep the FSB) That 80-core Intel demo chip has a vertically mounted SRAM chip as well, providing 20 MB (afair) directly to each code. For any problem where those 20 MB * 80 = 1.6 GB of SRAM can hold everything in a nicely distributed manner, you're going to see _very_ impressive performance indeed, particularly since they also have a (presumably very fast) mesh network connecting the individual cores. Intel's press releases talks about aggregate bandwidth in the TB/s range for this 80-core chip, from which we can calculate that each core must have at least 12.5 GB/s. Since the SRAM is directly attached, a 256-bit interface seems very reasonable, in which case the SRAM can idle along at about 400 Mhz. Alternatively, a somewhat narrower interface running at higher frequency would give the same result. Seems reasonable to me! And yes, I'd like to have one and see what I could do with it. :-) Terje -- - <Terje.Mathisen(a)hda.hydro.com> "almost all programming can be viewed as an exercise in caching"
From: Dennis M. O'Connor on 28 Sep 2006 01:50
"Jon Forrest" <forrest(a)ce.berkeley.edu> wrote i... > Dennis M. O'Connor wrote: >> "Jon Forrest" <forrest(a)ce.berkeley.edu> wrote ... >>> Today I read that we're going to get quad-core processors >>> in 2007, and 80-core processors in 5 years. This has >>> got me to wondering where the point of diminishing returns >>> is for processor cores. >> >> It depends on your application. Examples: A web server >> like Apache can effectively use a lot more cores than >> an old DOS game. But future games on the PC may be >> able to effectively exploit every core you can provide them. > > No doubt, but I'm talking about general purpose computing. What's this "general purpose computing" you speak of ? Is it Spider Solitaire ? MS Office ? Gnu CC ? Web browsing ? Or is it protein folding ? -- Dennis M. O'Connor dmoc(a)primenet.com |