Re: Linux to support Massive Multi-Threading (or dies)!
Web Server forum
Back To The Forum Home!Search!Private Messaging System

Web Server Talk Web Server Talk > Unix and Linux reviews > Free Debian support > Linux Debian support > Re: Linux to support Massive Multi-Threading (or dies)!




  Last Thread   Next Thread Next
  Show Printable Version Email this Page Subscribe to this Thread      Post New Thread    Post A Reply      

    Re: Linux to support Massive Multi-Threading (or dies)!  
Christopher Browne


View Ip Address Report This Message To A Moderator Edit/Delete Message


 
11-16-04 07:13 PM

After takin a swig o' Arrakan spice grog, "Ranando King" <rk@magictouchcorp.com> belched out
:
> The main reason I didn't touch the IHT performance issues is because
> I'm making what might be a terrible assumption: that the Niagara
> chip will be using cores with truely "parallel thread"
> processing. IHT isn't quite that, which is the reason it suffers
> such performance hits for software not specifically designed to
> handle it. Also, the fact that you can enable HT on SMP makes it
> somewhat ok to ignore those issues for this lopsided
> comparison. Now... if we ignore the IHT features altogether as well
> as Niagara's parallel threads, we basically end up comparing a
> single processor to an 8 processor SMP system without the SMP
> bottlenecks. Under such conditions, the statements I made still
> hold. There isn't that much work to be done to support such a
> system. Current day applications will still benefit greatly from the
> decreased process/cpu core ratio.

This points to another issue, namely that of the scalability of memory
bandwidth.

Intel's implementation of HyperThreading only appears to lead to
around 20% increases in CPU throughput, and that being the best case
scenario.

If your bottleneck is memory access, then you won't get as much as a
20% improvement from hyperthreading, and might even watch performance
get WORSE, not better.

A situation where there are 16 CPU cores on one chip is entirely
likely to exacerbate the situation; if they are all fighting for
access to a single memory bus, that means memory bandwidth will be the
entirety of the system bottleneck.  I'd be really surprised if 16 CPUs
could get materially more work done than 8 if they are sharing one
memory bus.

NUMA is, of course, a good answer to that, but adds considerable
hardware complications.  Each CPU has to have its own memory bus,
which means an impressively increased pinout requirement for the CPU
chip...
--
If this was helpful, <http://svcs.affero.net/rm.php?r=cbbrowne> rate me
http://www3.sympatico.ca/cbbrowne/languages.html
"The  reality of  the  software business  today  is that  if you  find
something that  can make you ridiculously rich,  then that's something
that Microsoft is going to take away from you." -- Max Metral





[ Post a follow-up to this message ]



    Sponsored Links  




 





   All times are GMT. The time now is 09:57 PM.      Post New Thread    Post A Reply      
  Last Thread   Next Thread Next


Most Popular forums 

Forum Jump:
Rate This Thread:

Forum Rules:
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
HTML code is OFF
vB code is ON
Smilies are ON
[IMG] code is OFF
 
Medical and Health forum | Computer Games Reviews | Graphics design forum

Back To The Top
Home | Usercp | Faq | Register