Re: HyperThreading Good or Bad...

Tech-Archive recommends: Repair Windows Errors & Optimize Windows Performance

From: Alvin Bruney [MVP] (vapor)
Date: 08/07/04 

Date: Sat, 7 Aug 2004 10:00:04 -0500

It's not a hack - at least by conventional standards.

I've been reading the white paper. What jumps out at me is that HT can offer
signficant gains depending on how it is implemented. The paper indicates
that contention for shared resources among threads is a critical factor.
Based on that, I wouldn't expect an 'across the board' improvement with
hyperthreading since the application needs to be written appropriately.

The paper is a good read by the way. 

-- 
Regards,
Alvin Bruney
[ASP.NET MVP http://mvp.support.microsoft.com/default.aspx]
Got tidbits? Get it here... http://tinyurl.com/27cok
"Miha Markic [MVP C#]" <miha at rthand com> wrote in message 
news:u$KyvyIfEHA.3928@TK2MSFTNGP11.phx.gbl...
> Hi Yogesh,
>
> My (only) theoretical knowledge says that HT may gain you something or 
> loose you something.
> Since it is a sort of hack into the procesor I guess it really depends on 
> how your application behaves.
> I have no better suggestion than to try it yourself and compare the 
> results.
> Nothing that you didn't already know, eh?
>
> -- 
> Miha Markic [MVP C#] - RightHand .NET consulting & development
> miha at rthand com
> www.rthand.com
>
> "Yogesh S" <YogeshS@discussions.microsoft.com> wrote in message 
> news:EF993D1E-9370-4B15-9633-AF7A29F49BC9@microsoft.com...
>> Hi,
>>
>> We have a mission critical trading application running on 8 CPU ( P-IV 
>> 2.8 Ghz) with Hyperthreading enabled that gives us 16 logical CPU... most 
>> of our applications needs to be behave in a timely manner even a 2 or 3 
>> second lag sometimes hurts organization P&L sheet...
>>
>> based on my understanding after reading below articles... even intel 
>> itself says applicatins may gain 10-30% perf improvement but depends on 
>> application and situation
>>
>> http://www.2cpu.com/articles/43_3.html
>> http://www.2cpu.com/articles/ht_explored/ht_2.html
>> http://arstechnica.com/paedia/h/hyperthreading/hyperthreading-1.html
>>
>> Can anyone shed a light on their personal views ? given a choice would 
>> you run your application in HT or Non HT mode..?  i know its an open 
>> question as we ourselves are now planning to switch off HT and run 
>> application in Non HT mode ..
>> and closely observe application behaviour
>>
>> but would be glad if anyone could share their experiences or any specific 
>> tips applicable to heavily multi-threaded and time critical application 
>> running in HT mode.
>>
>> Thx...
>>
>> Regards
>> Yogesh S
>
>

Relevant Pages

  • Re: Newsreader: Best of the bunch?
    ... smaller specialized applications because of a common set of keys and that the ... You've changed the message you're reading? ... > same commands as you would for any other editing job. ... have the problem of what happens when the application has windows? ...
    (Debian-User)
  • Re: .NET for serious commercial applications?
    ... > applications or you want to goto mars! ... Why would you want to hack memory if you're designing nuclear power station ... no other .NET language can do that. ...
    (microsoft.public.dotnet.framework)
  • Re: creating live virtual files by concatenation
    ... several files and then start manipulating text in it across "actual ... file" boundaries that could blow up easily. ... for reading. ... The quick hack might be useful in certain situations. ...
    (Linux-Kernel)
  • Re: Input/output errors on file accesses with Fedora Core 2
    ... > I have two applications that simultaneously analyse a large file. ... I am not sure if NFS is an issue. ... cat is failing since the file is actually open for reading by tail. ...
    (Fedora)
  • Re: sysv_ipc.c broken in v1.30 (was Re: sysvshm appearse broken in -current)
    ... from shmget_existingand restore the sysv_ipc.c change and see how that works for you? ... Applications do not expect it to occur, ... in the specific rights being requested, it passes in a mode mask. ... Both of these suggest removing the check, but I want to do a bit more reading before I do that, and run some tests on other platforms to confirm my reading. ...
    (freebsd-current)