Hyperthreading and overclocking

Has anyone else noted that one can overclock much farther with HT set off?

My 2.4C goes to about 2.7 with HT on, and 3.5 without

-
JH stood up at show-n-tell, in ,
and said:

Has anyone else noted that one can overclock much farther with HT set
off?

My 2.4C goes to about 2.7 with HT on, and 3.5 without

I've noticed that one main problem, for me, when pushing too high...is that
the resultant crashes are due to both processors getting stuck fighting over
a thread.


--
Strontium

"It's no surprise, to me. I am my own worst enemy. `Cause every
now, and then, I kick the livin' **** `outta me." - Lit

Strontium wrote:

I've noticed that one main problem, for me, when pushing too high...is that
the resultant crashes are due to both processors getting stuck fighting over
a thread.

Interesting. How can you tell?

I've noticed that one main problem, for me, when pushing too high...is
that
the resultant crashes are due to both processors getting stuck fighting
over
a thread.


Agreed, that's what it seems like... it always hangs, especially when I am
decoding RealVideo (but with other stuff too) and running Prime95
Always hangs, prime95 does not spit out an error, so I get the impression it
is the Hyperthreads that are the issue.

Note the large difference... the 2.4C goes 2.7 with HT enabled, 3.5 without!

My 2.4 is running fine at 3.2 with HT on (hangs at 3.3). I'll need to check
with HT off...
What's better, 2.7 with HT on or 3.5 with HT off?

LiviHam

"Noozer" wrote in message
news:2QtVb.435062$X%5.90658@pd7tw2no...

"JH" wrote in message
...
I've noticed that one main problem, for me, when pushing too high...is
that
the resultant crashes are due to both processors getting stuck
fighting
over
a thread.


Agreed, that's what it seems like... it always hangs, especially when I
am
decoding RealVideo (but with other stuff too) and running Prime95
Always hangs, prime95 does not spit out an error, so I get the
impression
it
is the Hyperthreads that are the issue.

Note the large difference... the 2.4C goes 2.7 with HT enabled, 3.5
without!

Is it a noticable difference?

My 2.6 goes to 3.24 with HT on.

Is it a noticable difference?

My 2.6 goes to 3.24 with HT on.




I have not benchmarked both, but seeing the rendering process with Nero
Recode when re-compressing a DVD the frames per second is much faster with
the 3.5 setting. This and rendering with Premiere are the two times and only
times I stress the CPU.

Note the large difference... the 2.4C goes 2.7 with HT enabled, 3.5
without!

Is it a noticable difference?

My 2.6 goes to 3.24 with HT on.



Ok, checked it...

recode a movie 9to5 with NERO took 13minutes 28 seconds at 3.5 without HT
and 15.22 at 2.8 with HT on.

25 per cent difference in clock, 14.5 per cent difference i time. Looks like
HT account for about 10 per cent..... looks like I need to have a 10 per
cent speed advantage without HT! Less than that.. turn it off

-
Anton Gysen stood up at show-n-tell, in ,
and said:

Strontium wrote:

I've noticed that one main problem, for me, when pushing too
high...is that the resultant crashes are due to both processors
getting stuck fighting over a thread.

Interesting. How can you tell?

By watching the process that is stuck, in task manager.


--
Strontium

"It's no surprise, to me. I am my own worst enemy. `Cause every
now, and then, I kick the livin' **** `outta me." - Lit

Strontium wrote:
-
Anton Gysen stood up at show-n-tell, in
, and said:

Strontium wrote:

I've noticed that one main problem, for me, when pushing too
high...is that the resultant crashes are due to both processors
getting stuck fighting over a thread.

Interesting. How can you tell?

By watching the process that is stuck, in task manager.

Before elaborating a little (ie: what are you watching and looking for),
what do you actually mean by "fighting" over a thread? A thread is allocated
to a CPU for a certain amount of time. There's nothing to fight over, since
you only have one CPU (SMP machines can fight over who has a certain cache
line in its cache, but good OS design reduces this).

--
Michael Brown
www.emboss.co.nz : OOS/RSI software and more
Add michael@ to emboss.co.nz - My inbox is always open

-
Michael Brown stood up at show-n-tell, in
, and said:

Strontium wrote:
-
Anton Gysen stood up at show-n-tell, in
, and said:

Strontium wrote:

I've noticed that one main problem, for me, when pushing too
high...is that the resultant crashes are due to both processors
getting stuck fighting over a thread.

Interesting. How can you tell?

By watching the process that is stuck, in task manager.

Before elaborating a little (ie: what are you watching and looking
for), what do you actually mean by "fighting" over a thread? A thread
is allocated to a CPU for a certain amount of time. There's nothing
to fight over, since you only have one CPU (SMP machines can fight
over who has a certain cache line in its cache, but good OS design
reduces this).

I'm not sure what you are asking. But, I've seen certain processes
(variable) get stuck....and the only explanation (from my puny knowledge of
CPU and virtual CPU knowledge), it appears to me that both processors are
fighting over the time for that thread. Either that, or one processor will
not let go of it. Either that, or the thing is just whacked. Come to think
of it, I like that description better.


--
Strontium

"It's no surprise, to me. I am my own worst enemy. `Cause every
now, and then, I kick the livin' **** `outta me." - Lit