You may find wierder things in your cereal..but...

[[OC]Lucifer]

I was playing around with the few overclocked options I have with the little voltage Im allowed by this 8KHA, and those are: 133X7 (933mhz), and 100X9.5 (950mhz). And despite the bus being 33mhz slower, and being 17mhz slower, it performs faster on Sandra CPU bench. Wierd eh? 133X7=2607/1282, 100X9.5=2637/1301.

[Bakeman]

"133X7 (933mhz), and 100X9.5 (950mhz). And despite the bus being 33mhz slower, and being 17mhz slower,"

the bus is 33mhz slower, but the 17mhz is faster not slower

[[OC]Lucifer]

Thats what I meant. Sorry. :)

[Bakeman]

hey no problem!

I'm sure everyone still loves ya :)

[oc jason]

even if the mhz were 17 slower it still benc betterm the multiplier makes for a better all around bench, as the FSB is a food speed performance, the multiplier makes a bigger difference. AN that is how much the work is actually doing. I mena if you had a FSB of 1000 and a multiplier of 1, that is still 1 gig but about as fast as a 1960 calculator. The higher the xmulti the more work is done, the higher the FSB the faster it gets done

[Sonny]

That's something to think about

[[OC]Lucifer]

I understand now :)

[JigPu]

Originally posted by [OC] Jason
even if the mhz were 17 slower it still benc betterm the multiplier makes for a better all around bench, as the FSB is a food speed performance, the multiplier makes a bigger difference. AN that is how much the work is actually doing. I mena if you had a FSB of 1000 and a multiplier of 1, that is still 1 gig but about as fast as a 1960 calculator. The higher the xmulti the more work is done, the higher the FSB the faster it gets done
Not to start being insubordnite to a seinor member....
But even with a high multiplier, you are forgetting that it is less efficent at transferring data to memory. It may be able to do more calculations every second, but if it can't get those instructions fast enough from memory, its performance starts to degrade.

Besides, by my calculations, 1000x1 would be faster than 100x10. You can do 10 times as many calculations in the core, but since you are only getting data from the memory 1/10 as much as with 1Kx1 it slows down (compartivly). If you had a large enough cache to hold all the instructions to be processed while the memory isn't being accessed though, then you could get serious speed increases.

Just from my speriacne...
JigPu