More than one B2 bound per kn-pair?

**jasong** · 07-21-2007, 08:46 PM

I was over at Mersenne forum asking about Prime95? Apparently, you can keep a P-1 Stage 1 file, and re-use it with different Stage 2 bounds.

I was wondering(assuming we're already keeping up with first, which might not be true) if we might find more factors, in a given amount of time, if we try more than one B2 value.

When I do things on my Pentium-D, with B1=120,000, and B2=1,700,000, Stage 1 takes 3hrs10min, and Stage 2 takes 2hrs15min.

**Joe O** · 07-21-2007, 08:57 PM

Would you do me a favour and try a run with 1800000 and report the times.
Thanks.

**glennpat** · 07-22-2007, 10:04 PM

Here are my results:

Intel dual core 6600 at 2.40Ghz with 2GB system memory.
Prime95 available memory set to 1 GB.

With B1=120,000 and B2=1,800,000, Stage 1 takes 1hr 53mins, and Stage 2 takes 1hr 32mins.
With B1=120,000 and B2=1,700,000, Stage 1 takes 1hr 53mins, and Stage 2 takes 1hr 26mins

**jasong** · 07-23-2007, 02:21 AM

testing now, when I get up in the morning and remember, I'll post.

**jasong** · 07-23-2007, 09:33 AM

A few seconds less than 2hrs42min.

That's just for Stage 2.

**Joe O** · 07-23-2007, 12:35 PM

Gentlemen,
Based on your timings, I think that we should increase B2 to at least 1,800,000. If you would try timings at 1,850,000 and 1,900,000 we could use that information.
The goal is to have Stage2 timings equal to Stage1, or at least very close.

**glennpat** · 07-24-2007, 04:28 PM

Here are my results:

Intel dual core 6600 at 2.40Ghz with 2GB system memory.
Prime95 available memory set to 1 GB.

With B1=120,000 and B2=1,850,000, Stage 1 takes 1hr 53mins, and Stage 2 takes 1hr 35mins.
With B1=120,000 and B2=1,900,000, Stage 1 takes 1hr 53mins, and Stage 2 takes 1hr 35mins

Another one with B2=1,900,000 and stage 2 was 37 seconds different then the other one with the same B2. Looks like the time for B2 at 1,850,000 and 1,900,000 are about the same.

**Joe O** · 07-24-2007, 05:05 PM

You know what I am going to ask for next<G>:
B1=120,000 and B2=2,000,000
but also
B1=130,000 and B2=2,000,000
To get a feel for the increase in B1 time, and to see if that reduces the B2 time as well.

**jasong** · 07-24-2007, 11:00 PM

I'm really glad to see my post accomplished something.

[sappy]
My life really improved a few years ago when I decided to spend more time worrying about others needs.
[/sappy]

**Joe O** · 07-25-2007, 12:14 AM

Jasong,
I could really use your timings for these cases as well. I also need to know how much memory you have installed on your machine, and how much is available to Prime95.

**jasong** · 07-25-2007, 04:46 PM

Originally Posted by Joe O

Jasong,
I could really use your timings for these cases as well. I also need to know how much memory you have installed on your machine, and how much is available to Prime95.

okay. I'll set it up now.

**glennpat** · 07-25-2007, 04:55 PM

I added in the results from my other post along with my new results.

Intel dual core 6600 at 2.40Ghz with 2GB system memory.
Prime95 available memory set to 1 GB.

I added in the results from my other post along with my new results.

With B1=120,000 and B2=1,700,000, Stage 1 takes 1hr 53mins, and Stage 2 takes 1hr 26mins
With B1=120,000 and B2=1,800,000, Stage 1 takes 1hr 53mins, and Stage 2 takes 1hr 32mins.
With B1=120,000 and B2=1,850,000, Stage 1 takes 1hr 53mins, and Stage 2 takes 1hr 35mins.
With B1=120,000 and B2=1,900,000, Stage 1 takes 1hr 53mins, and Stage 2 takes 1hr 35mins
With B1=120,000 and B2=2,000,000, Stage 1 takes 1hr 52mins, and Stage 2 takes 1hr 40mins.
With B1=130,000 and B2=2,000,000, Stage 1 takes 2hr 02mins, and Stage 2 takes 1hr 39mins

**jasong** · 07-26-2007, 09:33 PM

I tried to set up my computer to do the tests. It finished overnight, but the results were really peculiar, I must have screwed things up somehow.

Should I try again, or just forget about it? I have the feeling that, while architectures could have different timings, Stage 1 and Stage 2 are proportionately the same when you're comparing the same B1/B2 values.

**Joe O** · 07-26-2007, 11:30 PM

Originally Posted by jasong

I tried to set up my computer to do the tests. It finished overnight, but the results were really peculiar, I must have screwed things up somehow.

Should I try again, or just forget about it? I have the feeling that, while architectures could have different timings, Stage 1 and Stage 2 are proportionately the same when you're comparing the same B1/B2 values.

Stage1 might be proportional to the raw speed of the processor, but B2 is heavily dependent on memory. How much memory do you have on your machine? What is your processor speed?
Just run with B1=130,000 and B2=2,000,000.
Thanks.

**jasong** · 07-29-2007, 09:12 PM

Originally Posted by Joe O

Stage1 might be proportional to the raw speed of the processor, but B2 is heavily dependent on memory. How much memory do you have on your machine? What is your processor speed?
Just run with B1=130,000 and B2=2,000,000.
Thanks.

Apologies, I didn't see this until just now. It's running now.

Edit: I'm running it with maxmem=1200MB

**jasong** · 07-30-2007, 01:23 AM

With maxmem=1200MB Stage 1 takes 3hrs25.5min and Stage 2 takes 2hrs51min

**jasong** · 08-01-2007, 12:49 AM

Because of a message from vjs

dual-core Pentium-D 2.8GHz. 2Gig of RAM with 1200MB allocated to Prime95.

**Joe O** · 08-01-2007, 10:42 AM

Extrapolating from the timings you both have provided, it looks like:

Code:

B1=130,000 and B2=2,200,000

should be optimal.
Jasong, you should get 3 runs per day.
Glennpat, you should get 6 runs per day per core.

**jasong** · 08-02-2007, 06:05 PM

changed.

**glennpat** · 08-04-2007, 04:00 PM

I started running the new bounds a couple of days ago. The stage 2 time went to 1 hour 48 minutes.

Thread: More than one B2 bound per kn-pair?

Thread Tools

Rate This Thread

Display

More than one B2 bound per kn-pair?

Posting Permissions