Client Development?

[XtremeG]

Hello,

I've been sobbing for the past 3 years now and i've been using the 2.5.0 client for the past 2 years. It's not that it's slow (or so) and it's more then stable not for that.

But in the past 2 years there have been many new developments (x64 etc, core 2 duo etc etc) and the client has never been updated.

Thus i'm getting the feeling that if the client gets a new version maybe we can get alot more cem/s out of our pc's, which would only benefit the overall project.

Anyone might have an answer for me?

[engracio]

Hello,

I've been sobbing for the past 3 years now and i've been using the 2.5.0 client for the past 2 years. It's not that it's slow (or so) and it's more then stable not for that.

But in the past 2 years there have been many new developments (x64 etc, core 2 duo etc etc) and the client has never been updated.

Thus i'm getting the feeling that if the client gets a new version maybe we can get alot more cem/s out of our pc's, which would only benefit the overall project.

Anyone might have an answer for me?

I know it has been hectic here and on the PSP side with all of the updates and upgrades and faster sieving.:thumbs: Not to mention the newest prime.:|party|:

Best thing to do is actually ask Mr Prime95 and see if he has any upgrade worth mentioning like maybe a faster client?:) But here is the thing, why fix something when it is really not broke?:D This client without stating the obvious, found three (3) SOB primes, two (2) of them are secondpass/double check.:Pokes:



e:)

[Keroberts1]

is that suposed ot mean the client is broken or isn't? It missed 2 primes during the first pass!!! Although i would blame that more on the CPU's taht run these tests but is it possible to maybe even slow the client down alittle to decrease the error rate? We should definatly be doing some checking to see waht the actual error rates are. Would anyone be able to post stats on exactly how many of the double check tests produced different results?

[engracio]

No I do not think the client is broken, It found the 2 missed primes since the last client upgrade. I agree the missed primes more than likely were caused by overclocking which were prevalent and extreme when first pass came around. I myself was doing it. Now the cpu's are much more efficient and cooler running.

The guru's which track this things might be working on it already just to satisfy their personal curiousity. Hopefully they share it with us.

[Matt]

While there has been little in the way of client updates on the PRP side of things, there has been much activity in sieving client development. First there was proth_sieve upgrades, then JJSieve, SR2Sieve and now BOINC sieving. If you don't like the PRP client then maybe the BOINC sieve client is the one for you?

[umccullough]

I thought I read somewhere that LLR was a faster primality testing method than PRP. PSP and Riesel have gone this route already, why not SoB?

[riptide]

An interesting observation I'm read a few times recently.... one notably from Sturle

That overclocking can cause errors on Prime hunting is almost universally accepted? I test my overclocked systems with PRIME95.exe. AS most 'clockers do with a math stress+error checking program. If its stable with Prime95 without errors would it be resonable to suggest that it should be stable enough for SOB and other sieving apps? Or is the margins of error and tolerances different in both?

[umccullough]

If you look on some of the hardware testing sites, you'll definitely see mention of Prime95...

Was just reading an article on Toms Hardware and noted several references to "stability testing" using Prime95. They overclock memory and/or processors and then check stability using it. (go there and search for "prime95" to see some examples of this usage)

This seems like a reasonably good method of testing for errors in processor and memory - you just compare the residue after a test to a known-good one to see if any bits got twiddled during the test.

[philmoore]

In response to Riptide's post: I suspect that many overclocked machines were originally stress tested successfully with Prime95, which should also be an indication of being reliable for this project. However, with time and possible degradation of the heatsink, fans, etc., an overclocked machine which originally passed the test could later develop problems. I don't overclock, and I redo stress testing on active machines about once a year. I would probably do it more often if I overclocked.

[engracio]

In response to Riptide's post: I suspect that many overclocked machines were originally stress tested successfully with Prime95, which should also be an indication of being reliable for this project. However, with time and possible degradation of the heatsink, fans, etc., an overclocked machine which originally passed the test could later develop problems. I don't overclock, and I redo stress testing on active machines about once a year. I would probably do it more often if I overclocked.

I was going to response in the same vein. From my experience as the oc got older, I would at times get sudden reboots. When it happens a few times, I would check the temp and memory/fsb settings. Then I would lower either both or one of the setting down a notch. Most of the time that would fix the problem and cleaning out the dust balls off the box. Seasonal temps would also affect the oc. Most of my oc are medium level set for the box to last as long as possible. Not the super duper max out speeds. I oc just to get enough oomhp to get as much juice I can get out of the box.

[Keroberts1]

try sieving for OC'd boxes. Its a little different but all of the results are verified when you submit then so there is no chance of missing a prime. Just a chance that an extra test will be assigned that din't have to be. the potential waste is minescule compared to passing a prime and waiting for DC t
to catch it.

[engracio]

try sieving for OC'd boxes. Its a little different but all of the results are verified when you submit then so there is no chance of missing a prime. Just a chance that an extra test will be assigned that din't have to be. the potential waste is minescule compared to passing a prime and waiting for DC t
to catch it.

If the post is a suggestion for me, most if not all of my dc for SOB have been mainly thru sieving, check the stats. Seems like they were all good. For others which have doubt about their oc, that would be another route that they can take.:thumbs:

[Tallbill]

Heh, they should post who had a certain test and missed finding one of the primes only to be discovered later by another user.