[Beowulf] The Walmart Compute Node?

[Peter St. John]

Vincent,
Ah, thanks; I was neglecting ops per cycle variations among cores.
Peter

On Nov 9, 2007 12:21 PM, Vincent Diepeveen <diep at xs4all.nl> wrote:
> Ok easy theoretic calculation, and it's still very rude of course:
>
> 1 core 2.4Ghz * 3 instructions a cycle * (sse)2 = 7.2 * 2 = 14.4 Gflop
> 4 cores of a quad core ==> 57.6 gflop
> 3 nodes ==> 3 * 57.6 =  172.8 gflop
>
> Now of course your software won't be able to get that out of the
> hardware at core2,
> at new K8 cores perhaps it goes a tad better (though they aren't
> there yet).
>
> More careful calculation for core2 you can do using 2 instructions a
> cycle:
>
> 172.8 * 2 / 3 = 115.2 gflop
>
> If you really want to build this cheapo, checkout pricewatch. I'm
> sure one of you can manage it for $1000.
>
> On other hand at sycortex.com under 'news' i see a 72 cpu solution,
> which in their case is 72Gflop (hope i'm wrong)
> offered for under $15k.
>
> Let's say roughly a factor 5-10 difference in price compared to pc's
> even if we add power for the coming 3 years?
>
> Thanks,
> Vincent
>
>
> On Nov 9, 2007, at 5:58 PM, Peter St. John wrote:
>
> > Vincent,
> > I'm missing something in the arithmetic. "3 nodes of quadcore" is 12
> > cores? delivering 100 "GFlops" would require something like 8 GHz? So
> > perhaps you mean, 3 nodes of dual socket, quadcore CPU  (24 cores) at
> > 4GHz? And you can get that for $1500?
> > Thanks,
> > Peter
> >
> > On Nov 9, 2007 11:44 AM, Vincent Diepeveen <diep at xs4all.nl> wrote:
> >> Larry, all what you write is very interesting and of course i hope
> >> for you your product line gets a big succes.
> >> Just like IBM's blue gene, the major expertise of your product line
> >> is that it is only interesting to governments who need major
> >> amounts of
> >> crunching power (the other conditions left aside such as no big RAM
> >> requirements as that usually means you need good branch prediction
> >> and so on),
> >> and who have million dollar budgets, and probably have a program
> >> lying around where this hardware can get used for.
> >>
> >> The price of a box with say 100 "1 gflop" cpu's, delivering in total
> >> 100 gflop isn't gonna be $1500 i guess, whereas for 1500$ one can
> >> build hands down
> >> 3 nodes with a quadcore, delivering not only *more* than 100 gflop,
> >> but also capable of doing other software than just crunching; it's
> >> also possible to put
> >> a lot of RAM inside and it's also possible to run software that's
> >> making a lot of use from the branch predictor.
> >>
> >> For sure you're not qualifying for a $2500 setup, and with those
> >> freak qualifications you qualify bigtime for this mailing list of
> >> course :)
> >>
> >>
> >> On Nov 9, 2007, at 3:42 PM, Larry Stewart wrote:
> >>
> >>> Robert G. Brown wrote:
> >>>
> >>>> On Thu, 8 Nov 2007, Jim Lux wrote:
> >>>>
> >>>>> In general, a N GHz processor will be poorer in a flops/Watt
> >>>>> sense than a 2N GHz processor.
> >>>>
> >>> Well that just isn't so.  It seems pretty clear from IBMs BlueGene/
> >>> L, as well as the SiCortex processors, that the
> >>> opposite is true.  The new Green 500 list is brand new, and there's
> >>> not much on it yet, but the BG/L is delivering 190MF/Watt
> >>> on HPL, whereas the machines made out of Intel and AMD chips are
> >>> half that at best.
> >>>
> >>>>>
> >>>>> The power draw is a combination of a fixed load plus a frequency
> >>>>> dependent load, so for the SAME processor, running it at N/2 GHz
> >>>>> consumes more than 50% of the power of running it at N GHz.
> >>>>
> >>> This probably IS true, but high performance cores have a lot more
> >>> logic in them to try to achieve performance: out of order
> >>> execution, complex branch prediction, register renaming, etc. etc.
> >>> A slower core can be a lot simpler with the same silicon process,
> >>> so a decent lower-clock design will be more power efficient than a
> >>> fast clock design.
> >>>
> >>>>>
> >>>>> If you go to a faster processor design, the frequency dependent
> >>>>> load gets smaller (smaller feature sizes= smaller capacitance to
> >>>>> charge and discharge on each transition).  The core voltage is
> >>>>> also usually smaller on higher speed processors, which also
> >>>>> reduces the power dissipation (smaller number of joules to change
> >>>>> the voltage from zero to one or vice versa).  So, in general, a
> >>>>> 2N GHz processor consumes less than twice the power of a N GHz
> >>>>> processor.
> >>>>
> >>> The flaw in this argument is that a slower clock design can use the
> >>> same small transistors and the same current state of the art
> >>> processes and it will use many fewer transistors to get its work
> >>> done, thus using very much less power.  Our 1 GF core is 600
> >>> milliwatts, for example.
> >>> Even after adding all the non-core stuff - caches, memory
> >>> controllers, interconnect, main memory, and all overhead, it is
> >>> still around 3 watts per GF.
> >>>
> >>>>
> >>>> In ADDITION to this is the fact that the processor has to live in a
> >>>> house of some sort, and the house itself adds per processor
> >>>> overhead.
> >>>> This overhead is significant -- typically a minimum of 10-20 W,
> >>>> sometimes as much as 30-40 (depending on how many disks you
> >>>> have, how
> >>>
> >>> This factor does not scale this way!  With low power processors,
> >>> you can pack them together, without the endless support chips, you
> >>> can use low power inter-chip signalling, you can use high
> >>> efficiency power supplies with their economies of scale.  If you
> >>> look inside
> >>> a PC there are two blocks doing useful work - memory and CPUs, and
> >>> a whole board full of useless crap.  Look inside a machine designed
> >>> to be a cluster and there should be nothing there but cpus and
> >>> memory.
> >>>
> >>>
> >>> --
> >>> -Larry / Sector IX
> >>>
> >>> _______________________________________________
> >>> Beowulf mailing list, Beowulf at beowulf.org
> >>> To change your subscription (digest mode or unsubscribe) visit
> >>> http://www.beowulf.org/mailman/listinfo/beowulf
> >>>
> >>
> >> _______________________________________________
> >> Beowulf mailing list, Beowulf at beowulf.org
> >> To change your subscription (digest mode or unsubscribe) visit
> >> http://www.beowulf.org/mailman/listinfo/beowulf
> >>
> >
>
>