Intel's Groveport Platform

By: Michael S (, April 17, 2017 4:18 pm
Room: Moderated Discussions
Brendan ( on April 17, 2017 3:13 pm wrote:
> Hi,
> Michael S ( on April 17, 2017 1:43 pm wrote:
> > Brendan ( on April 14, 2017 12:23 pm wrote:
> > > Michael S ( on April 14, 2017 8:09 am wrote:
> > > For single threaded they are slow (compared to "equally modern" alternatives). However, when you have
> > > more threads than CPUs (e.g. many thin clients running many completely unrelated processes, many separate
> > > compiler/linker processes, etc) a normal server at the same price (e.g. a pair of 8-core Xeon E5s)
> > > is even slower (due to the time threads have to wait doing nothing between timeslices).
> > >
> >
> > Doing nothing between timeslices?
> > May be, I am taking you more seriously than you deserve?
> > Or your way of thinking and expressing thoughts is just to different from mine?
> This is extremely basic stuff. If there's 10 threads that want to use a CPU and only
> one CPU, at any point in time 9 threads can't be using that CPU and would be waiting
> ("doing nothing"); and 10 CPUs that are half as fast would be a huge improvement.

O.k. You expressed yourself in most unusual (for me) way, but atleast we are on the same tune.

> > IMHO, it's bloody obvious than if KNL has any chance at all to be competitive against "normal" Xeon
> > on non-HPC loads then it's *only* when there is a lot of parallel tasks ready all the time. *Much* more
> > tasks than mere 32 that are needed for full utilization of a pair of hyperthreaded 8-core Xeons.
> Yes; and it would be extremely foolish to assume that HPC is the only case where there is a lot of parallel tasks.

But HPC, at least some classes of it, is the one of the few tasks, and likely the most important among them by far, where KNL's dual 512-bit SIMD units can be advantageous. Very-high-bandwidth, but not very low latency, HMC-alike memory is also advantageous only for relatively small class of non-HPC workloads.
And before you say "deep learning", pleae look at KNL LINPACK scores and compare them with NVidia P100 at similar price and power consumption. And then consider that P100 itself is not competitive for for deep learning relatively to its cheap and not so cheap brethren.

Also, large-scale HPC is where the strongest part of KNL, its OmniPath links, really shine.

> Unfortunately, for some bizarre reason, some "HPC people" think that they're special little snowflakes.

What's the story about snowflakes? The word suddenly became popular... Even in Oz, where snow is not the most common, or so I heard.

> It would also be extremely foolish to assume "parallel tasks ready all the time". For fluctuating
> loads you want to cope with the peak demand. You don't want a system that crumbles during the "9
> am office worker rush" because someone decided it'd be overkill for the "overnight lull".
> > And no, according to Intel's own estimates, even in the best case scenario
> > (SPECInt2006_base) KNL does not quite match performance of dual Xeon E5 2620
> > v4, which street price is likely 2.5-3 times lower than Growerport.
> Are you suggesting that HPC are the only special little snowflakes that use floating
> point; or that HPC is beaten by dual Xeon E5 2620 for both HPC and non-HPC?
> - Brendan

No, i didn't say anything like that. KNL is definitely much faster than dual-2620v4 on vectorizable FP, and even somewhat faster (not a lot, 15% or so) at SPECFp2006_rate, probably due to great memory bandwidth.
But you were talking about workloads that resemble SPECInt_rate, don't you?
< Previous Post in ThreadNext Post in Thread >
TopicPosted ByDate
Intel's Groveport PlatformMS2017/04/11 01:41 PM
  Intel's Groveport PlatformDaveC2017/04/11 05:47 PM
    Intel's Groveport PlatformMichael S2017/04/12 07:14 AM
  Intel's Groveport PlatformBrendan2017/04/12 05:49 AM
    Intel's Groveport PlatformMS2017/04/12 03:23 PM
      Intel's Groveport PlatformBrendan2017/04/12 07:50 PM
        Intel's Groveport PlatformMichael S2017/04/13 10:51 AM
          Intel's Groveport PlatformBrendan2017/04/14 07:40 AM
            Intel's Groveport PlatformMichael S2017/04/14 08:09 AM
              Intel's Groveport PlatformBrendan2017/04/14 12:23 PM
                Intel's Groveport PlatformMichael S2017/04/17 01:43 PM
                  Intel's Groveport PlatformBrendan2017/04/17 03:13 PM
                    Intel's Groveport PlatformMichael S2017/04/17 04:18 PM
                      Intel's Groveport PlatformBrendan2017/04/17 05:35 PM
                        Intel's Groveport Platformanon2017/04/17 06:32 PM
                          Intel's Groveport PlatformBrendan2017/04/17 08:31 PM
                            Intel's Groveport Platformanon2017/04/17 10:39 PM
                              Intel's Groveport PlatformBrendan2017/04/17 11:48 PM
                                Intel's Groveport PlatformMichael S2017/04/18 01:40 AM
                                Intel's Groveport Platformanon2017/04/18 03:32 AM
                                  Intel's Groveport PlatformBrendan2017/04/19 04:43 PM
                                    Intel's Groveport Platformanon2017/04/21 05:10 PM
                                      Intel's Groveport PlatformBrendan2017/04/24 06:21 AM
                                    Intel's Groveport PlatformJukka Larja2017/04/21 11:09 PM
                            Intel's Groveport PlatformMichael S2017/04/18 01:17 AM
                            Intel's Groveport PlatformMichael S2017/04/18 01:29 AM
                        Intel's Groveport PlatformMaynard Handley2017/04/17 09:40 PM
                          Intel's Groveport PlatformBrendan2017/04/17 10:13 PM
                          snowflakesMichael S2017/04/18 01:06 AM
                            snowflakesAaron Spink2017/04/18 07:21 AM
                              strawberries Daniel B2017/04/20 04:31 AM
Reply to this Topic
Body: No Text
How do you spell green?