Trying to work out if intel's new hyperthreading gizmo is just marketing speil or will actually give tangible benefits. The programs that push PC's the most at the moment are games, and as yet HT is unproven. Games designers won't support anything that's not going to be 'standard' so it seems like a chicken and egg situation.

Thoughts?

 

i was in the simlar boat tbh, i really wantd HT but for £540 its outrageous. I bought my P4 2.8 friday, and i mentioned it to the bloke.

he said with games it can make them very juddery and slow at some points. But you can also turn off HT and just run the raw clock speed.

Si

 

Well, our initial testing doesn't really show much in the way of benefits. We have written code (number crunching stuff) to test HT and are seeing maybe 5% decrease in processing time.

OK, early days and as we fathom our way through optimising the code to take advantage of HT we might see improvements, but as it stands I would vote HT to be hype.

Cheers

Ian

 

Help. Much better now than the first processor steppings.

EDIT: Have a gander at this Anandtech page. The numbers speak for themselves. Of course the apps in question are written with a good understanding of thread pooling and locality of reference, and may lend themselves particularly well to the sort of optimisations that HT enables. YMMV.

[Edited by MrDeference - 1/12/2003 9:33:53 PM]

 

Neil,

Did some benchmarking the other week on DL580's with HT processors and there was minimal (ie 1%) performance improvement under EX2K. It's probably not worth it at the moment, you'd be better spending the cash on an array controller ot more memory.

If you really want to know the detail, give me a call and I'll send you some slides from ESG.

Phil

 

On toms hw i saw a 900% increase with HT enabled.

Whether to take this with a pinch of salt or not.

But it was running a flash application.

See if i can dig it out.

Si

 

Si, how many times have a told you to take Tom's Bribeware with a pinch of salt?