• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

WPrime 2.00 Benchmark Thread

Rroff said:
Is it me or does this benchmark seem quite sensitive - possibly even more sensitive to memory bandwidth/latency than just pure CPU? I haven't had time to test yet just as a casual observation.
Click to expand...

Very much so. From my days of trying to get the best score I could on hwbot I found out that memory timings were more important than just pure cpu speed.

There are several hundred q9650's cpu's in the world faster than my time despite being 100mhz or more slower on cpu speed. If you need to back off your cpu speed by 100 or 200Mhz to get faster memory bandwidth/latency then you will find that more than makes up the drop in cpu speed. I presume the i7 will be no different to the results with a q9650 with regards to memory latency/bandwidth?????

There are also lots and lots of tweaks which can be done to reduce your times as well but I could never be bothered with them.
 
We'll have to see. On Deneb and Bloomfield it looks like a very close correlation between pure clockspeed and the results (as long as the number of cores / threads remains the same and CnQ etc is off).

I'm interested to see the effect of cache. If we get some Athlon II results that should make it clearer.

NB speed and memory timings don't seem to make a huge difference for Deneb at least. There's an effect but it's not huge.
 
Last edited:
And a clean install of xp 32 bit does wonders for your times versus windows 7 as well ;)

Interesting link here on latency:

http://www.clunk.org.uk/reviews/cor...c8d-6gb-triple-channel-kit-review/Page-4.html

Difference with i7 at 4Ghz between 1600Mhz memory at 8-8-8-24 and 7-7-5-18 is 6.33 secs versus 6.318 secs.

Much bigger difference of 3 seconds on the 1024M run.

Of course ram speed makes and hence bandwidth makes an even bigger difference.
 
Also if you check reviews between the i7 860 and 920's on two different platforms, the 920's are about about 10% quicker with prime32 clock per clock persumably solely due to the superior memory bandwidth on the lga1366 platform?
 
HeX said:
What are your memory timings? I am running the same CPU as you at the same speed, but your scores seem to be considerably faster than mine...
Click to expand...

Cpuz 1.53 reads my timings wrongly, my RAM setup is:
V = 1.65v, frequency = 800, Clocks 9-9-9-24

But this was run in XP 32-bit. - For me superPI is quicker in XP 32-bit than W7 64-bit.

Just tried in W7 and get 12.697, 14.286, 17.052, 32.509s - Slower on 4, 3 and 1, faster on 2 than XP:confused:
 
Last edited:
Untitled.png


Run on 8,6,4,2 Threads.
 
HeX said:
Ah those are much more like what I'm getting. Surprising how much slower it is in W7 than XP.
Click to expand...

So am I however in the hwbot world this has been known for some time. xp 32 bit makes a big difference to your timings.

I can't even get close to all 6 of my best hwbot timings now I am on windows 7.:(

Which probably means for this thread people ought to say which op they are using as it can make 10% difference. Even more with a lite version fresh install of windows xp 32 bit. Trust me, I spent 3 months trying to maximise my scores ;)
 
hi,

nice thread. My results:

27wu1w0.jpg


EDIT: The cores are the other way round :rolleyes: I have not discovered the magic phenom that is faster on single thread than triple.
 
You must log in or register to reply here.
Back
Top Bottom