I've just finished crunching a short S5 WU with the S5T0301 patch on my overclocked AMD 64 4000+. The result (here) came back as valid and it was ~11% faster than the standard app.
yes, the new 'patched'-client seems to be stable and is producing valid WU´s.
i think akos can - "step-by-step and time-by-time" - push up more power of the einstein client.
i hope :)
regards and big thx for your work !!!
rock, rock, rock ;)
I have a dual Athlon MP 2000+ (Palomino core, 3DNow! and SSE capable) box that I've just put the original S5T0000 patch on. It's no speed demon by today's standards (~12 hours for long wu's), but I'll be stepping it through the various patches and report how a low-midrange PC responds to the optimizations.
Yes, your CPU has SSE2 so you should use the patch file S5S0007.dat or an app already patched with this file. Akos has moved into SSE3 versions of his patch process so please be careful not to use one of those.
Yes, your CPU has SSE2 so you should use the patch file S5S0007.dat or an app already patched with this file. Akos has moved into SSE3 versions of his patch process so please be careful not to use one of those.
The newest is S5T0711.dat - for SSE and also for SSE3 CPUs - use it. On my SSE/SSE2 AMD A64 works fine and its faster as S5S0007.
Can anybody do a time measurment with the same wu ( official / S5Txxxx ) ?
( I have only 10+ hours wus... )
I have been running S5S0007 patched apps on about 15 boxes with SSE and SSE2 - PIIIs, P4s (Williamette), and Athlon XP. Mostly short results and some longs. I haven't got enough data for a proper analysis yet but the speedup is not far short of 20%. The formula I'm using for calculating speedup percentage is:-
[Stock Secs - Patched Secs] / Patched Secs X 100
On one Celeron 900 running at 1062MHz, with data file h1_0122.5 Stock time was 9642 secs (average of 4) and Patched time was 8037 secs (average of 8). The formula gives a speedup of 20%.
I do it this way because:-
9642 secs/WU represents 8.96 results per day
8037 secs/WU represents 10.75 results per day
The increase is 1.79 results per day so I'm producing 1.79 on 8.96 which is 20% more work per day.
A calculation for an Athlon XP also overclocked and running longs (h1_1150.5) gives a 16.6% speedup.
To date I have seen more than 60 results with the S5S0007 patch and there have been zero invalids last time I looked. More than 80% have already validated and those still pending are in the "initial" state rather than "No consensus yet" state so there are no "possibly invalids" yet.
A 20% speedup is going to save a couple of months on the expected time for S5 (not to mention the saving on my electricity bill :).) So where's the next 20% going to come from?? (Only kidding) :).
I give you my warmest appreciation for your work - and a donation when you set up the paypal account.
I was keen to try to take things carefully for the sake of the project, fearing that a "feeding frenzy" might cause problems. However your technique of "throwing the pearls off the back of a speeding truck" seems to be working out OK.
So we should consider the proposal for a small test group to have been run over by that speeding truck and unfortunately it died in the process before the ambulance could be called :).
I'm not sure what I'm doing wrong but I'm getting this ... *sigh*
These new optimized apps seems to get more and more complicated to set up,
it's starting to go over my head, and I'm still quite clever for beinga a girl.
RE: P.S. S5T0004 looks like
)
Yes, 4 is giving bad results, move back to 3, or to the 3 standard.
I've just finished crunching
)
I've just finished crunching a short S5 WU with the S5T0301 patch on my overclocked AMD 64 4000+. The result (here) came back as valid and it was ~11% faster than the standard app.
Great work akosf!
yes, the new 'patched'-client
)
yes, the new 'patched'-client seems to be stable and is producing valid WU´s.
i think akos can - "step-by-step and time-by-time" - push up more power of the einstein client.
i hope :)
regards and big thx for your work !!!
rock, rock, rock ;)
Norman
Hi all, I have a dual
)
Hi all,
I have a dual Athlon MP 2000+ (Palomino core, 3DNow! and SSE capable) box that I've just put the original S5T0000 patch on. It's no speed demon by today's standards (~12 hours for long wu's), but I'll be stepping it through the various patches and report how a low-midrange PC responds to the optimizations.
TTFN,
-- Tony
Hello! Athlon XP 2600+
)
Hello!
Athlon XP 2600+ 24/7. Than I can help?
RE: Hello! Athlon XP 2600+
)
Yes, your CPU has SSE2 so you should use the patch file S5S0007.dat or an app already patched with this file. Akos has moved into SSE3 versions of his patch process so please be careful not to use one of those.
Cheers,
Gary.
RE: RE: Hello! Athlon XP
)
The newest is S5T0711.dat - for SSE and also for SSE3 CPUs - use it. On my SSE/SSE2 AMD A64 works fine and its faster as S5S0007.
RE: Can anybody do a time
)
I have been running S5S0007 patched apps on about 15 boxes with SSE and SSE2 - PIIIs, P4s (Williamette), and Athlon XP. Mostly short results and some longs. I haven't got enough data for a proper analysis yet but the speedup is not far short of 20%. The formula I'm using for calculating speedup percentage is:-
[Stock Secs - Patched Secs] / Patched Secs X 100
On one Celeron 900 running at 1062MHz, with data file h1_0122.5 Stock time was 9642 secs (average of 4) and Patched time was 8037 secs (average of 8). The formula gives a speedup of 20%.
I do it this way because:-
9642 secs/WU represents 8.96 results per day
8037 secs/WU represents 10.75 results per day
The increase is 1.79 results per day so I'm producing 1.79 on 8.96 which is 20% more work per day.
A calculation for an Athlon XP also overclocked and running longs (h1_1150.5) gives a 16.6% speedup.
To date I have seen more than 60 results with the S5S0007 patch and there have been zero invalids last time I looked. More than 80% have already validated and those still pending are in the "initial" state rather than "No consensus yet" state so there are no "possibly invalids" yet.
A 20% speedup is going to save a couple of months on the expected time for S5 (not to mention the saving on my electricity bill :).) So where's the next 20% going to come from?? (Only kidding) :).
I give you my warmest appreciation for your work - and a donation when you set up the paypal account.
Cheers,
Gary.
Oh and another thing. I
)
Oh and another thing.
I was keen to try to take things carefully for the sake of the project, fearing that a "feeding frenzy" might cause problems. However your technique of "throwing the pearls off the back of a speeding truck" seems to be working out OK.
So we should consider the proposal for a small test group to have been run over by that speeding truck and unfortunately it died in the process before the ambulance could be called :).
Now... Where the hell has that truck gone??? ....
Is that the dust trail over that hill??? ....
Cheers,
Gary.
I'm not sure what I'm doing
)
I'm not sure what I'm doing wrong but I'm getting this ... *sigh*
These new optimized apps seems to get more and more complicated to set up,
it's starting to go over my head, and I'm still quite clever for beinga a girl.