Dragonfly and Hyperthreading....

Freddie Cash fcash-ml at sd73.bc.ca
Mon Feb 21 16:13:34 PST 2005


> From the point of view of cpu power per watt AMD wins hands down. As
>  I said, this is becoming the #1 concern for anyone running more
> then a few machines.  Insofar as raw performance goes... well, I
> think AMD is edging-out Intel in most areas these days.  The story
> the benchmarks tell is that the cpus are typically within 15% of
> each other anyway, and Intel can only really win benchmarks that are
> optimized specifically for its cpus or which are FP intensive.
> Intel still edges AMD on FP performance, but that's about it.  The
> real problem for Intel is that producing on-par performance is
> costing them 20% more power (or even more!), and a lot of additional
> beefing up of caches, memory subsystems, and so forth.  And
> consumers have started complaining about the fan noise.

The *really* interesting point of the Opteron vs. Xeon, or Athlon/64
vs. P4 benchmarks are that the AMD CPUs are within 10% of the Intel
CPUs, yet they are clocked up to 1000 MHz *slower*.  That right there
is the biggest indication that Intel has screwed up somewhere in the
core of the CPU.

Sure, the P4 can be cranked up to 4 GHz, but what's the point if the
Athlon64 at 2.8 GHz gives you just as much performance, for less cost,
less heat waste, and less energy??

Intel's implementation of SMT (known as HyperThreading) is nothing
more than a hack to try and keep the P4's overly-long pipeline full of
instructions.  It's a hack because it requires an HT-aware scheduler
to take full advantage of it (trying to run an SMP kernel on an HT
system will usually slow things down as the SMP scheduler tries to
balance the workload across what it believes is two real CPUs, with
separate caches, separate ALUs, and separate FPUs, which doesn't work
on an HT CPU).  And those systems that do have an HT-aware scheduler
don't always do a good job of keeping things running smoothly.  You
get much more reliable performance when HT is disabled.

Other CPUs that use SMT (like IBM's POWER4 or POWER5) built into the
design from the get-go (rather than kludged on afterward like with the
P4) work much better.  Don't have much info on this beyond what's on
Ars Technica, though.

-- 
Freddie Cash, CCNT CCLP        Helpdesk / Network Support Tech.
School District 73             (250) 377-HELP [377-4357]
fcash at xxxxxxxxxx               helpdesk at xxxxxxxxxx





More information about the Users mailing list