Re: Hyperthreading & 8.0

James Ralston wrote:

>On 23 Sep 2002, Ryan wrote:
>  
>
>>What does hyperthreading provide?
>>    
>>
>
>Hype.
>
>To elaborate: their were multiple Linux kernel hackers present at the
>USENIX 2002 Technical Conference (back in 2002-06).  I won't name
>names, but one of the kernel hackers who was present described it as
>"a hack Intel created for marketing purposes".
>
>On 23 Sep 2002, Ryan wrote:
>  
>
>>from the benchmarks I have seen, they have reported a ~20-30%
>>performance increase.
>>    
>>
>
>Non-independent benchmarks are simply marketing hype (no pun
>intended); they're completely worthless for any objective comparative
>purposes.
>
>My $0.02: wait until independent benchmarks of hyperthreaded P4 chips
>v. non-hyperthreaded P4 chips come out before making any purchasing
>decisions based on the technology.
>
>  
>

  Actually I've seen real world increases with hyper threading.  On the 
other hand I've seen hyperthread completely tank certain benchmarks. 
 From what I've seen it goes something like this:

-Non cache bound, cpu bound activities  = +10-20%   (example compiling a 
kernel with make -j 4)
-Cache bound activities =  -20-30%  
-Single threaded single app =  little to no effecr
-Multipurpose server not cpu bound = little to no effect


-- 
There is no such thing as obsolete hardware.
Merely hardware that other people don't want.
(The Second Rule of Hardware Acquisition)
Sam Flory  <[email protected]>