Intel's Hyperthreading

Is this hyperthreading all its cracked up to be? Does AMD have a anything
similar to compete?

Paul

Hyperthreading is way overrated. Early tests I have read show running processor intensive programs at the same time instead of in sequence will increase performance by up to 10% in some cases and decrease performance by 5% in other cases. The performance hit is blamed on the relatively small size of the level 2 cache. The newer chips with 1 MB or 2 MB L2 cache should help.

Paul wrote:
>
> Is this hyperthreading all its cracked up to be? Does AMD have a anything
> similar to compete?
>
> Paul

--
Mike Walsh
West Palm Beach, Florida, U.S.A.

On Sun, 02 Nov 2003 17:15:41 GMT
Will Dormann <> wrote:

> Mike Walsh wrote:
>
> > Hyperthreading is way overrated. Early tests I have read show
> > running processor intensive programs at the same time instead of in
> > sequence will increase performance by up to 10% in some cases and
> > decrease performance by 5% in other cases. The performance hit is
> > blamed on the relatively small size of the level 2 cache. The newer
> > chips with 1 MB or 2 MB L2 cache should help.
>
>
> I could be wrong, but I think the main advantage of Hyperthreading is
> that it makes the computer *feel* faster by still retaining
> responsiveness under heavy CPU load. (rather than cutting down the
> time it takes to encode a video, for example)

In principle a hyperthreaded CPU can handle two interrupts
simultaneously, just as a dual processor machine can. In practice, as
with all things hyperthreading, the efficiency with which it does so
depends on whether the particular code blocks on one or more of the
nonredundant elements in the pipelines.

IMO this is potentially its greatest benefit, but it's mostly of use in
I/O intensive environments, not memory- or calculation-intensive ones,
which means that it's not going to be terribly useful for most "home"
applications. It might, however, help analog video capture, which is
essentially taking a bitstream from PCI device, possibly doing some
processing on it, and moving it to another. But there it's only going to
help if the PCI bus is not already saturated--a dual-bus machine might
help there, but again you're getting into "serious server" territory and
duals would probably be a viable option in such a machine.

> -WD

--
--
--John
Reply to jclarke at ae tee tee global dot net
(was jclarke at eye bee em dot net)

Mike Walsh wrote:

> Hyperthreading is way overrated. Early tests I have read show running processor intensive programs at the same time instead of in sequence will increase performance by up to 10% in some cases and decrease performance by 5% in other cases. The performance hit is blamed on the relatively small size of the level 2 cache. The newer chips with 1 MB or 2 MB L2 cache should help.


I could be wrong, but I think the main advantage of Hyperthreading is
that it makes the computer *feel* faster by still retaining
responsiveness under heavy CPU load. (rather than cutting down the
time it takes to encode a video, for example)


-WD

Probably true. A processor intensive application running in the backgroup should not affect the response of another program as much. This probably would have made a big difference a few years ago, but with a 2 Ghz or faster processor it should not make much of a difference. Intel's advertisements are pushing the idea of using a hyperthreading processor using a gigabit network, but most computers with heavy traffic on gigabit networks are servers with more than one processor, so this point may be moot.

Will Dormann wrote:
>
> I could be wrong, but I think the main advantage of Hyperthreading is
> that it makes the computer *feel* faster by still retaining
> responsiveness under heavy CPU load. (rather than cutting down the
> time it takes to encode a video, for example)
>
> -WD

--
Mike Walsh
West Palm Beach, Florida, U.S.A.

On Sun, 2 Nov 2003 15:24:35 -0000, "Paul" <> wrote:

>Is this hyperthreading all its cracked up to be? Does AMD have a anything
>similar to compete?
>
>Paul
>

The short answer is No, not yet (2006?)

"Mike Walsh" <> wrote in message
news:...
>
> Hyperthreading is way overrated. Early tests I have read show running
processor intensive programs at the same time instead of in sequence will
increase performance by up to 10% in some cases and decrease performance by
5% in other cases. The performance hit is blamed on the relatively small
size of the level 2 cache. The newer chips with 1 MB or 2 MB L2 cache should
help.
>
> Paul wrote:
> >
> > Is this hyperthreading all its cracked up to be? Does AMD have a
anything
> > similar to compete?
> >
> > Paul
>
> --
Ther is no statement-like answer to this question. First thing is that the
programs you're running must support dual processing in order to see a
valuable performance gain. Second thing is that the gain depends of the
programs. There are some sites where the why and how are explained in
detail, but the logic behind the process is complex.
Finally, on top of pure performance gain, there is also the benefit of
parallel processing.
With my P4 / 800 - 2.4 GHz and HT enabled, I can consult my Email while
rendering video in Premiere. No way to do this (without dropped frames)
without hyperthreading!

ElJerid wrote:
> With my P4 / 800 - 2.4 GHz and HT enabled, I can consult my Email while
> rendering video in Premiere. No way to do this (without dropped frames)
> without hyperthreading!


Don't you mean *capturing* video? Rendering should not be affected by
CPU load. The output file will be the same, but it'll just take longer
to do.


-WD

"Will Dormann" <> wrote in message
news:lXcpb.7587$...
> ElJerid wrote:
> > With my P4 / 800 - 2.4 GHz and HT enabled, I can consult my Email while
> > rendering video in Premiere. No way to do this (without dropped frames)
> > without hyperthreading!
>
>
> Don't you mean *capturing* video? Rendering should not be affected by
> CPU load. The output file will be the same, but it'll just take longer
> to do.
>


Right!
It' s capturing as well as exporting.
Thanks for correcting.

AMD does not have anything similar yet.

--
DaveW



"Paul" <> wrote in message
news:3fa5219f$0$113$.. .
> Is this hyperthreading all its cracked up to be? Does AMD have a anything
> similar to compete?
>
> Paul
>
>