New siever question

[maddog1]

Well, I got kinda bored with regular tests so I decided to try out sieving for a while. :)
I did my reading beforehand and I think I basically got the hang of this. The only question I have is this: What is a good starting range to reserve?
What confuses me, is that I see widely varying ranges being reserved.
Some ppl just reserve small ranges (eg 433000-433200 or 490000-490020) and some others much larger (eg 435000-450000 or 460000-480000)
I am intending to use an Athlon XP 3000+ on this and I went on and reserved the 490100-491000 range (seemed to be the first free one...) but now I'm worried whether I reserved a good initial range (would n't like it to last for 1 day or 1 month)
Any insights on this subject? What relation do the reported kp/s have with the speed through a range? My machine reports numbers from 470-610 during it's first half hour on the job (using latest 0.42 proth_sieve for Windows)
Can the ETA be deduced from this somehow?
Don't :spank: me if I asked something stupid or reserved the wrong range, I am still trying to understand the concept of sieving properly :D

[ceselb]

General rule, expect 'rate' number of G's per 10 days. Your range is then 15-20 days (probably 15, client starts with a low estimate)

[Keroberts1]

if yo uhave an athalon then you'll want to be using the cmov version too i believe

[royanee]

Originally posted by Keroberts1
if yo uhave an athalon then you'll want to be using the cmov version too i believe

Very true, if you are using an Athlon or P3, proth_sieve_cmov is the way to go. 900 G is a great range especially for your system. Also, the people who make the larger reservations typically split it up amongst multiple computers. I usually reserve 1000 G because round numbers are easy to deal with. Always glad to have a new siever. :)

[Mystwalker]

Originally posted by maddog1
Well, I got kinda bored with regular tests so I decided to try out sieving for a while. :)

Welcome to sieving! :cheers:


reserved the 490100-491000 range
[...]
My machine reports numbers from 470-610 during it's first half hour on the job (using latest 0.42 proth_sieve for Windows)
Can the ETA be deduced from this somehow?

It's pretty easy, as the 490100 is basically measured in "Giga-p".
So, you have a range of 900 Gp (the p is left out typically), which is 900,000,000 kp and do ~ 550 kp/sec.

So ETA is 900,000,000 / 550 kp/s = 1,636,364 seconds = 455 hours = 19 days
You see, it's a simple division - but if you like, I can include integrals and stuff. :D

[Death]

site not updated since Last Update: Tue 09-Nov-2004 03:04 (GMT)

Seems that we already got a 100,000 score, but don't know who is a winnner....

[maddog1]

Thank you guys for the quick replies and the warm welcome! :)
For those that asked, I am indeed using cmov version (as I said, I tried to do my reading before starting...)
So far, it has only found 2 dupes but ,hey, it's my first day on the job :D

[vjs]

maddog1,

Welcome to the sieve, I think if you can keep that machine in the High 500 kp/s your doing well. A 900G range like you reserved is perfect. Couple pointers/suggestions.

Read the benchmark thread in this portion for a speed comparison.

Download and install sobistrator it will make life easy, you can open and close that program when you want it doesn't automatically start/stop the client nor consume much resources while running etc.

You may also want to check out mikes pages when he comes back from vacation. Generally there is some good stuff there.

Inside the same directory as the client there will be three text files, fact, factexcel, fact range, I suggest submitting all of them.

Use cmov for sure.

If your having problems login into the server to submit

http://www.seventeenorbust.com/sieve

1. Copy the above link, Close all internet windows
2. Open a new window and paste the above into it.
3. Login and save login info
4. Bookmark and close the window after login
5. Re-open a new window and then use the bookmark.

I may be the only person having problems with this but it's how I solved it. You only have to do this once per computer.

[Keroberts1]

the speed is kinda excessive i believe I have an athalon 2400 and I'm only getting 540kps did you say athalon 1 ghz? anybody else getting much better performance than me with these?

MadDog what is the size of your dat file?

[vjs]

I thought he said an xp3000... This should get around at best 600 kp/s.

But yes good point make sure the sobdat file is about 3,0XXmb.

[maddog1]

It is indeed a 3000+ Barton, standard clock speed (13x166) w/512 MB 333 RAM.
Reported speed seems to be mostly in the range 550-630 after doing half a day's work, with the occassional spike higher or lower.
My SoB.dat is 3078 K, dated 9 Nov 04.
These values seem to be close to VJS's specs, so I guess all is going well over here :)
Thanks again for the help guys!

[Mystwalker]

Originally posted by Keroberts1
the speed is kinda excessive i believe I have an athalon 2400 and I'm only getting 540kps did you say athalon 1 ghz?

That was Silverfish in this (http://www.free-dc.org/forum/showthread.php?s=&threadid=7924) thread. ;)

[vjs]

You probably have a dual channel correct?

That might account for the speed increase.

It would be neat to see the difference between dual channel and non-dual channel.

Ie if someone has one pull one stick and see how much sieve speed changes.

I know it's effort but I don't now much a second stick would help my machine out. It would bring it upto 1G of memory first but is it worth the $$$.

[maddog1]

Sadly it's single channel (one 512M module only)
IMHO it's not worth the risk upgrading to dual at this time, since this system runs smooth as silk as it is now.
It is well known that different brands/makes/batches of memory chips (even with theoretically same specs) may not work well together for reasons not always clear. Since I don't want to get rid of the current memory alltogether and it probably is very much of a hit-and-miss affair to find the exact same memory almost one year after originally buying the system parts, I will pretend it can't be upgraded ;)
Unless you have the ability to try-before-you-buy, I'd recommend the same, or you *might* risk your sanity, which should be valued more that a few kp/s gain :)