View Full Version : Soon new standarts? (PiFast)
M.Beier
02-13-2007, 04:14 PM
I just saw this topic of current PiFast WR, and realised I got a faster time, same size calculation, just not Hexus PiFast... Why on gods earth do we use a sucky version instead of the fastest???
Im pretty sure we'd hit 15 or 16 secs at 4.5ghz, not even tweaked.. 500mhz 1:1 4-4-4-8
But I wont call this WR, since I know alot of people can do alot faster, but it kinda' sucks that the current WR is slower, due Hexus is the standart!? - Why the heck dont we just use thelab.gr's edition?
possibly, the longer it takes, the more errors bound to show up
i guess some form of stability needs to be shown at such high speeds
ozzimark
02-13-2007, 05:20 PM
Why on gods earth do we use a sucky version instead of the fastest???
because we all use the same version, results are comparable ;)
M.Beier
02-14-2007, 02:02 AM
yeah, but why not chose the fastest version to be used?
Errors apear at same clocks in the two versions, atleast thats my experience..
SP sse3 patched is also faster but we don't use it. Don't ask why, I don't set standards, who does? :D
M.Beier
02-14-2007, 07:54 AM
kiwi, well, hehe, would be awesome starting a trent :D
But I think I'll get spammed with hate-PMs if I apply "PiFast WR" in signature, when using another version, but might help in time in order to get a new version.. ;)
massman
02-14-2007, 08:14 AM
yeah, but why not chose the fastest version to be used?
Errors apear at same clocks in the two versions, atleast thats my experience..
If we change the version of PiFast now, we would have to block our entire database ...
It's not really about the fastest time to calculate PI, it's about being faster compared to another system. :)
M.Beier
02-14-2007, 09:45 AM
If we change the version of PiFast now, we would have to block our entire database ...
It's not really about the fastest time to calculate PI, it's about being faster compared to another system. :)
Disagree..
SuperPI is perfect for that purpose..
PiFast should be fastest of the fast, which it was originally ment to be, lateron an even faster edition appeared, sadly it wasnt very well promoted..
And TheLab.gr (Hipro5's homeland), should be exactly as good comparing..?
Well, funny thing. Lets say future cpus will have some kind of instruction set that dramatically improves SP/Pifast or any other bench performance but only if the bench itself has been modded to use that instruction set. Would people switch to new version or not? :D