Nehalem review up

Article: Nehalem Performance Preview
By: Vincent Diepeveen (diep.delete@this.xs4all.nl), April 11, 2009 3:50 pm
Room: Moderated Discussions
Michael S (already5chosen@yahoo.com) on 4/10/09 wrote:
---------------------------
>Vincent Diepeveen (diep@xs4all.nl) on 4/10/09 wrote:
>---------------------------
>>Jack (jumpingjack6@verizon.net) on 4/9/09 wrote:
>>---------------------------
>>>Vincent Diepeveen (diep@xs4all.nl) on 4/7/09 wrote:
>>>
>>>>
>>>>How can you draw a conclusion about Shanghai, you haven't even compared it head on with Nehalem yourself.
>>>>
>>>>Vincent
>>>>
>>>
>>>David did characterize, at the beginning of the article, that Shanghai would be
>>>fairly characterize as slightly lagging harpertown, in that it falls behind in some
>>>cases, achieves parity and others, and has some strong points.
>>>
>>>Considering that is roughly a good assessment, then it can be extrapolated that Nehalem has opened up a wide margin.
>>>
>>>Nonentheless, you can search the databases yourself, the 5570 DP Xeon can range
>>>anywhere from 1.5 to 2x faster than a Shanghai 2P (2.7 GHz). I have not found one
>>>where Shanghai even comes close. This does not make Shanghai a bad CPU, but it
>>>does make it tough for AMD to market Shanghai against Nehalem.
>>
>>Sales ballony, based upon a few cracked spec tests.
>>
>>Both are nearly identical processors in performance for the software we tried.
>>
>>Of course HT and turboboost turned off, and Shanghai a tad higher clocked than
>>E versions of Xeon, gives Shanghai a slight edge in clockrate 2.53Ghz vs 2.7 shanghai.
>>Of course with more powerbudget intel clocks higher.
>>
>
>BTW, AMD submitted SpecJbb scores for 2.9GHz Shanghai. In the past it was the
>indication that next lower clocked part, i,e, 2.8GHz, will soon be available in
>normal thermal envelop. So there is a hope for 2.8GHz 75W Shanghai coming.
>
>>If you look to the intel documents in what i7 can execute it is SSE2+ instructions
>>a cycle max. That gives 8 flops as a max, with or without HT. Is that so much higher than AMD?
>>
>>Multiplication is not faster than at AMD in throughput, in fact if you try latencies
>>of AMD are better, so a good programmer CAN be faster at AMD.
>>
>
>Let's follow you own logic. Floating-point addition is faster (=had shorter latency)
>on Intel. Should we conclude that "a good programmer CAN be faster at Intel".

There is 1 unit that is doing multiplication,
there is a lot that can do addition.

Addition has a latency of 0.5 cycle at intel so to speak and 0.33 cycle or so at AMD (i could be off by 0.17 or so as i checked the i7 handbooks quickly a while ago for all kind of stuff, not the AMD ones).

Multiplication is important for FFT and matrix calculations. Adding goes rather quick. Enough units to do it. Just 1 for multiplication.

So that matters, the rest doesn't except when it becomes a bottleneck.

>BTW, what sort of multiplication has shorter latency on AMD? I can think only about
>integer 64x64=>128b that is very rare in hand-coded asm and never generated by compilers.
>
>>Yet these differences are that tiny, that any claim there of 50%+ is total ballony.
>
>It is absolutely correct that on single-threaded dense computational kernels running
>out of L1/L2 cache and at the same clock frequency Intel's Merom, Penryn, Nehalem
>aas well as AMD's Greyhound (Barcelona) and Shanghai cores are all within few per
>cents +- of each other. If the kernels did not include packed SSE instructions then
>you could add Intel' Dothan, Yonah and AMD Hammer to the same list. Actually, in
>my own scalar kernels I found Nehalem rather consistently lagging behind Penryn.
>although that was not the case for SIMD kernels.
>However that absolutely correct observation has nothing to do with the industry standard benchmarks on real CPUs that
>A. Multithreaded
>B. Run at different clock frequencies
>C. Not dense
>D. Could fit in onchip cache on one CPU but not on another (that what hurts Bracelona most)
>
>Points A+C and D give big edge to Nehalem because it has both SMT and much faster
>external memory access than any of competitors.
>
>>
>>It's just compiler and L3 based all these claims for just a few software programs.
>>
>>If you see clearly how moving from intel c++ 10.0 to 11.0 is a huge improvement
>>at core2 already, obviously the compiler team did do a great job.
>>
>
>Are you sure that Oracle and IBM JVMs compiled with Intel compiler?
>
>>I see i7 as a very logical step from core2, yet for performance a tiny step, if
>>you look single socket. Of course this allows intel now to scale to 2 and maybe soon 4 sockets.
>>
>>>However, if you are truly interested in how Nehalem stacks up against Shanghai,
>>>you can search the following, scores for both processors are now in the database:
>>
>>A $100 billion company, well it used to be, that just gets benchmarked for a few
>>programs, that's not a rather good idea.
>>
>>Turboboost, HT, more power budget, it sure helps to look better.
>>
>>It's not a help in practice; the X series that eat so much more power, that these
>>can increase their clockfrequency a lot more with turboboost than the E series,
>>whereas all HPC centers will be buying nearly always the E series and some, also
>>datacenters, already announced turboboost will get turned off. If i look at ebay
>>now, the L series 54xx still are a lot more expensive than the E series, how comes?
>>
>>>http://www.spec.org
>>>http://www.tpc.org
>>>http://www.vmware.com/products/vmmark/results.html
>>>
>>>
>>>Also, using a QX9770 in the comparision is not a bad idea, but it is also not a
>>>server branded CPU. It is irrelevant anyway, a few hundred MHz won't change the result.
>>>
>>
>
< Previous Post in ThreadNext Post in Thread >
TopicPosted ByDate
Nehalem review upDavid Kanter2009/04/07 02:43 AM
  Nehalem review upnoone2009/04/07 05:48 AM
  Strange jbb on HarpertownHenrik S2009/04/07 07:29 AM
    Strange jbb on HarpertownDavid Kanter2009/04/07 10:19 AM
      Strange jbb on HarpertownHenrik S2009/04/07 08:33 PM
        Strange jbb on HarpertownChris2009/04/07 11:54 PM
          Strange jbb on HarpertownHenrik S2009/04/08 01:40 AM
  Nehalem review upVincent Diepeveen2009/04/07 07:34 AM
    Nehalem review upJack2009/04/09 03:51 PM
      Nehalem review upVincent Diepeveen2009/04/10 12:58 AM
        Nehalem review upMichael S2009/04/10 02:45 AM
          Nehalem review upEduardoS2009/04/10 06:01 AM
            Nehalem review upMichael S2009/04/10 06:56 AM
          Nehalem review upEugene Nalimov2009/04/10 08:12 AM
          Choice of C compiler doesn't matter much for Java...Henrik S2009/04/10 09:10 AM
            Choice of C compiler doesn't matter much for Java...EduardoS2009/04/10 01:49 PM
              Choice of C compiler doesn't matter much for Java...Henrik S2009/04/11 06:13 AM
                Choice of C compiler doesn't matter much for Java...EduardoS2009/04/11 10:30 AM
                  Large pagesDavid Kanter2009/04/11 01:02 PM
                  Choice of C compiler doesn't matter much for Java...Henrik S2009/04/11 10:06 PM
                    Choice of C compiler doesn't matter much for Java...Paul2009/04/12 12:53 AM
                      Choice of C compiler doesn't matter much for Java...iz2009/04/12 01:59 AM
                      Choice of C compiler doesn't matter much for Java...Henrik S2009/04/12 06:37 AM
                    Choice of C compiler doesn't matter much for Java...EduardoS2009/04/12 07:08 AM
                      Choice of C compiler doesn't matter much for Java...Henrik S2009/04/12 08:25 AM
                        Choice of C compiler doesn't matter much for Java...EduardoS2009/04/12 04:24 PM
                          Choice of C compiler doesn't matter much for Java...Henrik S2009/04/12 09:18 PM
                            Thread costsDavid Kanter2009/04/12 11:12 PM
                              Thread costsHenrik S2009/04/14 01:08 PM
            Choice of C compiler doesn't matter much for Java...Michael S2009/04/11 07:53 AM
              Choice of C compiler doesn't matter much for Java...Henrik S2009/04/11 10:08 PM
          Nehalem review upVincent Diepeveen2009/04/11 03:50 PM
            Nehalem review upMichael S2009/04/11 04:12 PM
              Nehalem review upVincent Diepeveen2009/04/12 02:01 AM
                Nehalem review upMichael S2009/04/12 04:07 AM
  Nehalem review uprwessel2009/04/07 01:01 PM
    Nehalem review upslacker2009/04/08 08:11 AM
      Energy vs. powerDavid Kanter2009/04/08 09:11 AM
        Energy vs. powerVincent Diepeveen2009/04/10 01:08 AM
          Energy vs. powerslacker2009/04/10 08:26 AM
            Energy vs. powerRagingDragon2009/04/10 09:19 AM
              Energy vs. powerDavid Kanter2009/04/10 10:47 AM
              Energy vs. powerJack2009/04/10 03:44 PM
                Energy vs. powerslacker2009/04/10 06:00 PM
                  Energy vs. powerJack2009/04/10 06:31 PM
                  Energy vs. powerDavid Kanter2009/04/10 11:16 PM
      Nehalem review uprwessel2009/04/08 01:32 PM
  Minor font issuegpriatko2009/04/07 03:35 PM
    Minor HTML issueDavid Kanter2009/04/07 08:38 PM
      Minor HTML issueDavid Kanter2009/04/07 08:39 PM
  Good work, i look forward to linux and SP2 numbers (NT)PiedPiper2009/04/08 12:52 AM
  Nehalem review upJoe Chang2009/04/10 02:59 AM
Reply to this Topic
Name:
Email:
Topic:
Body: No Text
How do you spell avocado?