Slow performance from an FX-55 ClawHammer?

[Electrolyte]

After being toyed around with BOINC yet again, losing a HDD (almost 2 today) and a RAM stick because of it, I'm switching back to Seventeen or Bust - I really hope you guys never use BOINC for this project.

To the point - Has anyone noticed slow performance from AMD CPUs? In other projects my FX-55 blows my P4 3.2GHZ Northwood away, but in Seventeen or Bust, it's vice versa. The P4 has about 1.77 million cEMs/s on both tests (which is 3.54 million total) and my FX-55 only just reaches 3 million, with every application closed mind you. Both systems are running the same operating system, just the P4 has absolutely everything closed 24/7, only SB runs on it.

Do Intel CPUs just run faster than AMDs on SB?

[Matt]

PRP testing is best suited to P4s. AMD CPUs are best utilised in the sieving sub-project, you can find more about this in the sieving sub-forum.

[Electrolyte]

Instantly looking at the sticky for sieving got me confused :(

[Joe O]

Instantly looking at the sticky for sieving got me confused :(
It's really very simple.
1) PM or email Matt to get a password for his reservation site.
2) When you get your password reserve a 10G range.
3) In the meantime create a directory on your hard drive
4) Download jjsieveSSE2.exe from the first post in this thread near the bottom. (http://www.free-dc.org/forum/showthread.php?t=3501) and put it in the directory you just created.
5) Download the Newest Dat 8k's 991<n<50M from the beginning of the same post you got the program from and put it in the same directory.
6) Create a bat file with the following lines:
jjsieveSSE2 -s <datfilename>
pause
where <datfilename> is the name of the dat file you downloaded in 5
7) When you get your 10 g range create a nextrange.txt file in the same directory with the following contents:
pmax=110130000000000
pmin=110120000000000
Done
This assumes you have reserved the range 110120-110130
8) Start your bat file
9) Submit your results

Any questions?

[Electrolyte]

Much better, I think I'll stick my FX-55 onto the sieving project then so it gets the most from it's power. Thanks!

[Matt]

In my PM to you I said to get a 1000G range, you're probably better off following Joe's instructions and starting with a smaller range initially.

[Electrolyte]

EDIT: Actually, took a 100G range. Will probably just keep doing 100G ranges since they only take approx 1 day to finish.

[Joe O]

Advanced instructions:
nextrange.txt can contain more than one range

pmax=110130000000000
pmin=110120000000000
pmax=110140000000000
pmin=110130000000000
pmax=110160000000000
pmin=110140000000000
pmax=110260000000000
pmin=110240000000000
Done

You can edit nextrange.txt at any time. It reads the file until the first Done. There is a bug in that it leaves garbage behind after the Done. This will be fixed in the next version.

You can reserve a 1000G range from Matt and then enter it as 10 100G ranges in nextrange.txt.

[Electrolyte]

So far it's found 4 factors, but they have "(duplicate)" at the end of them. Does this mean it's just found a factor that already exists?

[Joe O]

So far it's found 4 factors, but they have "(duplicate)" at the end of them. Does this mean it's just found a factor that already exists?
Yes, sort of. If you are happy with this answer then stop reading.

It doesn't have to be the exact same factor for the duplicate message to appear. It just has to be a factor for a k n pair that has already been factored.

[vjs]

Personally I'd suggest a 1000G range. If you grow tired of the range just note where you stopped and let us know.

I personally like finishing the large ranges and it makes book keeping easier less chance of errors etc.

BTW thank's for sieving we need alot of help in this area.

[Electrolyte]

Okay, the range is finished. What factors file do I use to submit the results? fact.txt?

EDIT: Never mind, found it in a thread.

[hhh]

If you like efficiency, you might want to check out the joint sieving part of SoB and PrimeSierpinskyProject. It's situated here:
http://www.mersenneforum.org/forumdisplay.php?f=48
It's basically the same thing, just that you use a larger sob.dat-file, which makes it more effective, and you sieve lower ranges, which makes you find more factors.
Your contribution would make a nice difference in that part of the project.
Though, you would not find factors for Sob that are needed soon. On the other hand, for PSP, yes.
You can check it out if you have the time. We'll be glad to help you getting strarted, if necessary.
Yours H.

[Electrolyte]

I don't think I'll be on this for long. Already made a mistake, using rar instead of zip since I didn't notice WinRAR uses RAR by default, so I'm kinda annoyed right now.

[hhh]

I think Joe will be able to handle this *problem*. Don't panic.
If you take large enough chunk, you can just let your computer crunch for one month, or longer, without even touching it. As for the manual part as limiting factor, it can be reduced to a minimal effort.
:thumbs: H.

[Electrolyte]

A problem with the long ones is I use a lot of different software on this machine, so it's more prone to crash. Even when I did ctrl+C on the siever and restarted the system, it lost it's position (no idea how) and refused to continue from where it had left off. I ended up restarting the range from scratch and lost 6 hours CPU time :(

I may try to get a smaller system (like a Pentium Pro or Pentium 2) running the siever. It'll take longer to do them, but at least I won't have to submit them as much. At the moment I've set the FX-55 on the standard SB client - my main goal in the whole of distributed computing is to find a prime on this project and to help complete it, since compared to other projects, I know this one is getting somewhere. The others, on the otherhand, I feel as if they're getting nowhere, either by not telling us or are telling us in too technical detail so we won't understand if it's good or bad.

But that's my opinion, of course.

[vjs]

Actually once the client is up and running if you stop it you don't have to restart it from the bat, you can simply click on the cmov or sse2. Should start right back up where it left off.

rar etc doesn't matter we can figure it out no problem.

Really suggest using larger ranges though.

As for helping the project find a prime, sieve is really the best bet, you might not get the personal glory but its certainly has the highest%eff.