What Core 2 To Buy Now?

Somewhere on teh interweb Fishface typed:
~misfit~ wrote:

Ummm, 334 x 9 you say? OK, no problem. SuperPi 1M takes 19.031
seconds.

So, perhaps an extra second and a half penalty for the lesser cache,
around the same difference between the 4MB and 2MB cache processors,
as I recall. I just checked the CPU usage while this was running,
and it appears only partially multithreaded. The total CPU usage was
50 percent, with a higher and lower spiky curve oneach core...

So should I re-think my overclock setting with this info? I'm
running 300 x 11. Would I be better off with, say 330 x 10? 367 x 9? 412
x 8? 471
x 7 or even 550 x 6? My FSB is adjustable from 200 to 800 and I
already know the CPU'll do 3.3GHz easilly at a reasonable
temp/vcore. Would I be better off with a faster FSB and lower multi?

Well, the higher FSB and lower multiplier should result in more
memory bandwidth. I don't think the processors are as bandwidth
starved as they once were, so it may or may not help, depending upon
the application. I was very disappointed when I had to lower my
speed with the Q6600 and raise the multiplier.

Yes, very little difference that I can ascertain.

SuperPi 1M

300 x 12 17.750
413 x 8 17.641

However, the CPU temps have increased a few degrees (same clockspeed, same
vcore) so that's got to mean that the CPU is doing more work right? (I'm
running SETI@home)

Cheers,
--
Shaun.

Fishface wrote:

~misfit~ wrote:

Ummm, 334 x 9 you say? OK, no problem. SuperPi 1M takes 19.031 seconds.

So, perhaps an extra second and a half penalty for the lesser cache, around the
same difference between the 4MB and 2MB cache processors, as I recall. I just
checked the CPU usage while this was running, and it appears only partially
multithreaded. The total CPU usage was 50 percent, with a higher and lower
spiky curve on each core...

I tried setting the "Core Affinity" of SuperPi mod to a single core and it took .008
seconds longer and it pegged the one core to 100%. I'm such a scientist...

Somewhere on teh interweb ~misfit~ typed:
Somewhere on teh interweb Fishface typed:
~misfit~ wrote:

Ummm, 334 x 9 you say? OK, no problem. SuperPi 1M takes 19.031
seconds.

So, perhaps an extra second and a half penalty for the lesser cache,
around the same difference between the 4MB and 2MB cache processors,
as I recall. I just checked the CPU usage while this was running,
and it appears only partially multithreaded. The total CPU usage was
50 percent, with a higher and lower spiky curve oneach core...

So should I re-think my overclock setting with this info? I'm
running 300 x 11. Would I be better off with, say 330 x 10? 367 x
9? 412 x 8? 471
x 7 or even 550 x 6? My FSB is adjustable from 200 to 800 and I
already know the CPU'll do 3.3GHz easilly at a reasonable
temp/vcore. Would I be better off with a faster FSB and lower multi?

Well, the higher FSB and lower multiplier should result in more
memory bandwidth. I don't think the processors are as bandwidth
starved as they once were, so it may or may not help, depending upon
the application. I was very disappointed when I had to lower my
speed with the Q6600 and raise the multiplier.

Yes, very little difference that I can ascertain.

SuperPi 1M

300 x 11 17.750
413 x 8 17.641

However, the CPU temps have increased a few degrees (same clockspeed,
same vcore) so that's got to mean that the CPU is doing more work
right? (I'm running SETI@home)

Corrected above. I had 300 x 12, it should have been 300 x 11. CPU speed of
3.3GHz in both cases.
--
TTFN,

Shaun.

Somewhere on teh interweb Fishface typed:
Fishface wrote:

~misfit~ wrote:

Ummm, 334 x 9 you say? OK, no problem. SuperPi 1M takes 19.031
seconds.

So, perhaps an extra second and a half penalty for the lesser cache,
around the same difference between the 4MB and 2MB cache processors,
as I recall. I just checked the CPU usage while this was running,
and it appears only partially multithreaded. The total CPU usage was
50 percent, with a higher and lower spiky curve on each core...

I tried setting the "Core Affinity" of SuperPi mod to a single core
and it took .008 seconds longer and it pegged the one core to 100%. I'm
such a scientist...

ROFL! Yeah, I didn't think it was using both cores. I'd say that the .008
extra is possibly due to the CPU power that the OS is using. (I'm such a
speculator.)
--
TTFN,

Shaun.

What MB are you using and what ram?


bob



The benchmarks look pretty good to me, despite what people are saying
about
the lacking L2. The degree of overclock is definitely not a given. My
particular
E4500 will not overclock as Shaun's.

Bob Johnson wrote:

What MB are you using and what ram?

I tried it in a Gigabyte P35-DS3L and a Biostar 965PT, both proven to
overclock my E6400 to at least 3.2 GHz. RAM is 2 GB SuperTalent
T667UB1G/M (DDR2-667 with Micron D9 chips), proven to overclock
to at least 440 MHz with 2.1 volts at 4-4-4-10 timings. I tried the BSEL
mod to make it appear to be a 1066 MHz part. I only tried up to 1.375
volts vCore, though.