[RC5] clocks / key differences

dan the person motion at es.co.nz
Fri Nov 12 19:22:36 EST 1999


For the millionth time (don't people ever read the FAQ?)....

intel p5 and p6 generation chips are faster per MHz than AMD K6es at RC5.  It
doesn't mean AMDs suck in general, it is just that one particular small
instruction that is used allot in the RC5 algorithm happens to be faster on
intel chips. (RC5 is a very bad general purpose CPU benchmark).

As for the K6-III being a newer chip; the K6-2 is a K6 with 3dnow, the k6-3 is
a k6-2 with onchip fullspeed L2 cache.  Neither of these are used by rc5 so
the 'rc5 adjusted age' of the k6-3 is the same as a k6 is the same as a
pentiumMMX.

Also, the p5 chips get a boost as they are able to get a faster rate from
utilising the MMX unit. (For some reason k6s are slower using the MMX core).

If it makes you feel any better, something that really is newer technology is
doing better.  Jake Knust spoke in this list and reckoned an AMD Athlon 600
gets "3.033 kkeys per mhz" that works out at 330 (rounded) clocks/key.
(However the older p6 based celerons seem to get close 346 c/k).

However Bruce Ford reckons an Athlon optimised core could get "248 clocks per
key."  Now all we need is someone with time on their hand to develop such a
core.


Dennis Lubert wrote:

> Does anyone know how many clocks per key the AMD-K6-III needs ? My 200MHz
> MMX needs ~ 463 c/k (keyrate 432000 k/s --> 200 000 000 Hz / 432 000 ~=463)
> but my K6-III @ 450 MHz has a keyrate of 764000 k/s that would result in
> (450 000 000 / 763 000 ~=)590 c/k !!!!Shouldn't the (newer) K6-III have a
> better rate than this ? Or is there something wrong with my PC ? DO I
> normally need to have a keyrate of ~900 000 k/s ?
>

--
To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest



More information about the rc5 mailing list