which program?

drakkar67 · 2005-11-26, 14:20

for LLRing which program is best for...
Pentium M (2 MB cache)
Pentium 4 prescott
Pentium 4 D (64 bit)

and for for sieving which mode is best for...
Pentium M
Pentium 3

are the residuals same for LLR3.6.2 and PRP24.14?

thanks
drakkar67

ltd · 2005-11-26, 14:39

For PRPing i use PRP24.14 but with the new llr3.6.2 version both applications should be at the same speed again. The residues are not identical. That is one of the reasons why at PSP there will be no automated second pass testing. Tests should be assigned to machines running the same application type.

Idtentical residues are produced by
Group 1:
winpfgw ( not used for PSP anymore)
Group 2:
prp (all versions)
llr version <=3.1 ( version not used anymore as it is much slower)
Group 3:
llr >=3.5
llrnet

I have no idea which programm is fastest on which platform at the moment.

For sieving my best experiences have been with proth_sieve_cmov on
PIII and all AMD starting with the "TB" design
For the Pentium M i made no tests but due to the fact that it is relative slow on PRP i think the cmov version should be best.

Lars

drakkar67 · 2005-11-26, 16:55

thanks Lars

Citrix · 2005-11-26, 23:37

Pentium M and P4's should be used with PRP and not for sieving. Sieving does not use their full potential.

As for the fastest program, the latest PRP and LLR work at the same speed, so it is your choice to make. (Winpfgw, proth.exe are much slower and should be avoided)

Citrix

drakkar67 · 2005-11-27, 15:58

My notebook with Pentium M 1.6 is very slow on PRP tests. IMHO at this sieve depth it is more useful at sieving.
It's now PRPing at about 7 hours per test (3,5 PRP a day). I tried it at 11200 sieve depth and found about 100 factors a day (but I don't know exactly).
0-1 M...2 factors
1-2 M...5 factors
2-20 M...no factors, I donno why but maybe it was sieved w/2-20 M
20-50 M...about 90 factors
So my notebook factors 5 pairs a day effectively. It's better than PRPing 3,5 pairs.
It's now PRP'ing the 1703000 range. After finishing that range I may give it a sieving range, I haven't decided yet.

drakkar67

Citrix · 2005-11-28, 07:00

But if you PRP, you might find a prime and eliminate a few 100,000 candidates.

Anyway, the timming for your Pentium M sounds right, that is how long a PRP test takes.

Citrix

OmbooHankvald · 2005-11-28, 08:43

Quote:

Originally Posted by drakkar67

My notebook with Pentium M 1.6...

It's funny.. I too have a Pentium M notebook 1.6.
I've currently put it on sieving 50K/G/T/whatever letter is used-chunks.

Should I rather put it at PRP'ing?

Citrix · 2005-11-28, 08:54

I think a Pentium M has a Pentium 4 design and should be equally slow for sieving. So PRP should be the best.

But for right now putting the PC on sieving is not a bad idea since getting to 15T is slightly more important than finding a prime.

Hope this makes sense.
Citrix

axn · 2005-11-28, 09:54

Quote:

Originally Posted by Citrix

I think a Pentium M has a Pentium 4 design and should be equally slow for sieving. So PRP should be the best.

Pentium M is more like PIII than P4. It has a much higher IPC than P4 in regular code. So in that respect, P-M's performance is more like Athlon than P4. (Note:- talking about P-M, and not P4-M).

EDIT:- Can someone post some timings for P-M and P4 in both sieving and PRP? That should conclusively tell us which is good for what!

Mystwalker · 2005-11-28, 11:22

I have a Pentium-M 1.7 GHz, which does 320 kp/s.
My Pentium4 2.4 GHz does 200 kp/s.

But when running PRP concurrently, the sieving rate only drops to 156 kp/s - showing that there is still a lot of room for improvement. PRP speed drops to 50%, compared to the not-sieving situation.

I have no comparisons for PRP, but from what I've heard of GIMPS, P-M and P4 basically have the same performance per clock, which sounds sensible, because the SSE2 engine is used most of the time.

IMHO, a P-M should definitely sieve.

drakkar67 · 2005-11-28, 13:09

My P-M 1.6 (2 MB L2),
sieving 330-335 kp/s (in SSE2 mode, with 0-50 M SoB.dat, in 11200-11300 range)
PRP test about 16 ms/iteration (192 K FFT)
(it may be better at high FFT ranges, but we have much time to reach them)

My P4 1.6 (512 K L2),
PRP test about 9-10 ms/iteration (192 K FFT)

My P4 celeron 2.4 (256 K L2),
PRP test about 7.5-8 ms/iteration (192 K FFT)

(I think it may slow down greatly for PRP>256 K FFT)

None of them overclocked. Upto 15000 range P-M may be better at sieving but mine will PRP at least a week until it finishes its range.

drakkar

2005-11-26, 14:20	#1
drakkar67 Feb 2003 Istanbul 2²·13 Posts	which program? for LLRing which program is best for... Pentium M (2 MB cache) Pentium 4 prescott Pentium 4 D (64 bit) and for for sieving which mode is best for... Pentium M Pentium 3 are the residuals same for LLR3.6.2 and PRP24.14? thanks drakkar67

2005-11-26, 23:37	#4
Citrix Jun 2003 2·811 Posts	Pentium M and P4's should be used with PRP and not for sieving. Sieving does not use their full potential. As for the fastest program, the latest PRP and LLR work at the same speed, so it is your choice to make. (Winpfgw, proth.exe are much slower and should be avoided) Citrix Last fiddled with by Citrix on 2005-11-26 at 23:39

2005-11-28, 11:22	#10
Mystwalker Jul 2004 Potsdam, Germany 3·277 Posts	I have a Pentium-M 1.7 GHz, which does 320 kp/s. My Pentium4 2.4 GHz does 200 kp/s. But when running PRP concurrently, the sieving rate only drops to 156 kp/s - showing that there is still a lot of room for improvement. PRP speed drops to 50%, compared to the not-sieving situation. I have no comparisons for PRP, but from what I've heard of GIMPS, P-M and P4 basically have the same performance per clock, which sounds sensible, because the SSE2 engine is used most of the time. IMHO, a P-M should definitely sieve. Last fiddled with by Mystwalker on 2005-11-28 at 11:24

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
GPU LLR program	jasong	GPU Computing	19	2011-08-23 03:32
So you think you can program	rogue	Lounge	5	2009-10-02 15:02
Program	Primeinator	Information & Answers	5	2009-07-16 21:42
Program for GPU	tribal	Information & Answers	5	2009-03-19 20:54
Old Program	moo	Software	0	2006-06-27 00:19

2005-11-26, 14:39	#2
ltd Apr 2003 304₁₆ Posts	For PRPing i use PRP24.14 but with the new llr3.6.2 version both applications should be at the same speed again. The residues are not identical. That is one of the reasons why at PSP there will be no automated second pass testing. Tests should be assigned to machines running the same application type. Idtentical residues are produced by Group 1: winpfgw ( not used for PSP anymore) Group 2: prp (all versions) llr version <=3.1 ( version not used anymore as it is much slower) Group 3: llr >=3.5 llrnet I have no idea which programm is fastest on which platform at the moment. For sieving my best experiences have been with proth_sieve_cmov on PIII and all AMD starting with the "TB" design For the Pentium M i made no tests but due to the fact that it is relative slow on PRP i think the cmov version should be best. Lars

2005-11-26, 16:55	#3
drakkar67 Feb 2003 Istanbul 64₈ Posts	thanks Lars

2005-11-27, 15:58	#5
drakkar67 Feb 2003 Istanbul 110100₂ Posts	My notebook with Pentium M 1.6 is very slow on PRP tests. IMHO at this sieve depth it is more useful at sieving. It's now PRPing at about 7 hours per test (3,5 PRP a day). I tried it at 11200 sieve depth and found about 100 factors a day (but I don't know exactly). 0-1 M...2 factors 1-2 M...5 factors 2-20 M...no factors, I donno why but maybe it was sieved w/2-20 M 20-50 M...about 90 factors So my notebook factors 5 pairs a day effectively. It's better than PRPing 3,5 pairs. It's now PRP'ing the 1703000 range. After finishing that range I may give it a sieving range, I haven't decided yet. drakkar67

2005-11-28, 07:00	#6
Citrix Jun 2003 2×811 Posts	But if you PRP, you might find a prime and eliminate a few 100,000 candidates. Anyway, the timming for your Pentium M sounds right, that is how long a PRP test takes. Citrix

2005-11-28, 08:54	#8
Citrix Jun 2003 2×811 Posts	I think a Pentium M has a Pentium 4 design and should be equally slow for sieving. So PRP should be the best. But for right now putting the PC on sieving is not a bad idea since getting to 15T is slightly more important than finding a prime. Hope this makes sense. Citrix

2005-11-28, 13:09	#11
drakkar67 Feb 2003 Istanbul 2²·13 Posts	My P-M 1.6 (2 MB L2), sieving 330-335 kp/s (in SSE2 mode, with 0-50 M SoB.dat, in 11200-11300 range) PRP test about 16 ms/iteration (192 K FFT) (it may be better at high FFT ranges, but we have much time to reach them) My P4 1.6 (512 K L2), PRP test about 9-10 ms/iteration (192 K FFT) My P4 celeron 2.4 (256 K L2), PRP test about 7.5-8 ms/iteration (192 K FFT) (I think it may slow down greatly for PRP>256 K FFT) None of them overclocked. Upto 15000 range P-M may be better at sieving but mine will PRP at least a week until it finishes its range. drakkar