mersenneforum.org  

Go Back   mersenneforum.org > Fun Stuff > Lounge

Reply
 
Thread Tools
Old 2020-10-28, 17:19   #1
mathematizer
 
Oct 2020

1000002 Posts
Default Hello, from another newbie

Hope this is the right place for this - looks like the introductions post was very outdated.

I was a math major for my undergrad and have been interested in prime numbers and number theory for a while (though my expertise is not nearly at the level of most users here.)

I'm a middle school teacher who finally got around to building a home PC that can participate in the search.

I had been using my CPU (i5 9600k), but received a suspicious result. I tried troubleshooting a little bit, but haven't had much luck correcting it - my "free time" is somewhat limited, especially helping my young child navigate partially remote education.

I switched to gpuowl with my GPU (Radeon 5700xt), and have been working on some manual tests, which seem to be running more reliably.

Just wanted to introduce myself - seems like a great community to be a part of.
mathematizer is offline   Reply With Quote
Old 2020-10-28, 17:35   #2
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

2×3×1,489 Posts
Default

Welcome. That suspect result will mark that CPU for a while. You may get larger exponents for about 3 months. If you haven't installed Prime95 v30 yet on the machine, do it now. The code for running PRP's has robust error checking and correcting built in. And the PRP's will generate a file that allows us to check the result without having to do a full recheck.
Uncwilly is online now   Reply With Quote
Old 2020-10-28, 18:43   #3
mathematizer
 
Oct 2020

25 Posts
Default

Thanks - I had recently upgraded to v30, so hopefully that file was generated, though the CPU did an LL test. I then ran the PRP on my GPU and submitted that result.

Wasn’t sure whether it was worth continuing to PRP/LL test on the CPU since I didn’t want to produce more suspicious results.

Last fiddled with by mathematizer on 2020-10-28 at 18:46
mathematizer is offline   Reply With Quote
Old 2020-10-28, 18:55   #4
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

10010100010112 Posts
Default

Welcome. That cpu has an igp. https://ark.intel.com/content/www/us/en/ark/products/134896/intel-core-i5-9600k-processor-9m-cache-up-to-4-60-ghz.html
You could consider also running mfakto, or even experiment with gpuowl, on the igp.

A PRP proof file is only generated by software that supports it at the completion of a PRP test begun with it. Next time...
PRP with GEC (Gerbicz error check) will prevent computation errors from sneaking through (almost 100% detection and retry). The LL Jacobi symbol check has 50% detection probability. Run PRP on your cpu and it should be fine. Or prime95 will show you the cpu error rate when the GEC detects errors.

There's lots of GIMPS reference info here
(Who better to appreciate the learning curve than a teacher!)

Is your 5700XT gpu reliable in gpuowl? I have one that in Windows 10 on a multigpu rig causes driver instability and can unpredictably take down the driver or the whole system.

Last fiddled with by kriesel on 2020-10-28 at 19:01
kriesel is online now   Reply With Quote
Old 2020-10-28, 19:16   #5
mathematizer
 
Oct 2020

2016 Posts
Default

I built the computer as a quarantine project when the schools went remote at the beginning of the pandemic. Had a received a variety of parts from friends and needed to troubleshoot some bad parts. My original CPU had a bad memory controller, which took forever for me to track down to replace it.

I’m aware the current CPU has an integrated graphics chip, but am confused why that’s relevant. It seems you are suggesting that because the CPU provided a suspicious result, the IGC might be an alternative for putting it to use, using one of the two programs that you mentioned?

Regarding the proof file, my CPU did complete the test, however it was an LL test that produced the suspicious result, not a PRP - I suppose that test doesn’t produce the file?

Regarding the 5700xt, it’s been working well. I originally had purchased a 5600xt, but a friend was upgrading and passed this one along to me. I’ve been able to run it overnight and through my workday without issues. Seems to zip through assignments. I only had one test produce a couple errors, and confidence was still high on it. I haven’t had issues with crashes.

It was a bit of a process figuring out how to get gpuowl to work (I’ve been away from the PC world awhile as my school is OSX based). I eventually found a post (Thanks!) with a windows build. After failing a few times, I got around to reading the instructions and seem to be using it effectively now. :)
mathematizer is offline   Reply With Quote
Old 2020-10-28, 19:21   #6
Runtime Error
 
Sep 2017
USA

18310 Posts
Default

I have taken 100989029 as a LL double check to see if I can clear your name. It will be a few days. Your GPU should also generate proof files if you are using gpuowl v7+. here for windows


Edit:
Quote:
Originally Posted by mathematizer View Post
I’m aware the current CPU has an integrated graphics chip, but am confused why that’s relevant. It seems you are suggesting that because the CPU provided a suspicious result, the IGC might be an alternative for putting it to use, using one of the two programs that you mentioned?
You can run prime stuff on all three simultaneously: the CPU, the ICG, and the GPU!!!

Last fiddled with by Runtime Error on 2020-10-28 at 19:25
Runtime Error is offline   Reply With Quote
Old 2020-10-28, 19:34   #7
mathematizer
 
Oct 2020

408 Posts
Default

Thanks, Runtime Error!

I certainly have my own suspicions over whether my name should be clear - it's a new build and I'm also new (haven't familiarized myself with steps I can take to troubleshoot hardware errors). Knowing that there's a definite issue would be a good first step before I dive into reading how to troubleshoot.

I haven't tried running the CPU (prime 95 30.3 b6) and GPU (gpuowl) software simultaneously yet, perhaps I will try a PRP test out on the CPU and see what I get.


Edit: Just read your edit - I'll definitely start with running it on the CPU and GPU for now! May dive into the IGC afterwards, as I'll need to figure out which program to run on it and how to configure it.

Last fiddled with by mathematizer on 2020-10-28 at 19:35
mathematizer is offline   Reply With Quote
Old 2020-10-28, 19:41   #8
mathematizer
 
Oct 2020

25 Posts
Default

Quote:
Originally Posted by kriesel View Post
Next time...
PRP with GEC (Gerbicz error check) will prevent computation errors from sneaking through
I don't see that specific option listed under work preference:

PRP
PRP-WR
PRP-100M
PRP-D
PRP-CF

Is GEC an additional setting in v30 that I'd select somewhere? Looking through my current exponent's active work, I see it alerting me that hardware errors have occured and there was 1 Gerbicz/double check error, but confidence is still excellent, so I imagine it's on automatically.

Last fiddled with by mathematizer on 2020-10-28 at 19:45
mathematizer is offline   Reply With Quote
Old 2020-10-28, 19:46   #9
Runtime Error
 
Sep 2017
USA

3×61 Posts
Default

Quote:
Originally Posted by mathematizer View Post
I don't see that specific option listed under work preference:

PRP
PRP-WR
PRP-100M
PRP-D
PRP-CF

Is GEC an additional setting in v30 that I'd select somewhere?

Just do PRP. Once it starts, you should notice a big ".residues" file appear in the folder and you know it will make a proof file.

The PRP error check is so insanely good that it will produce reliable results on even the worst hardware. The PRP proof files eliminate the need for lengthy double checks, which is a huge benefit to the project.

You should have no issues running PRP simultaneously on your GPU and CPU (on different exponents, of course). The IGC is probably pretty weak, but it could still potentially do some trial factoring work. Full disclosure, I have a machine that has an IGC but I don't bother with it.
Runtime Error is offline   Reply With Quote
Old 2020-10-28, 19:46   #10
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

47·101 Posts
Default

Quote:
Originally Posted by mathematizer View Post
Thanks, Runtime Error!

I certainly have my own suspicions over whether my name should be clear - it's a new build and I'm also new (haven't familiarized myself with steps I can take to troubleshoot hardware errors). Knowing that there's a definite issue would be a good first step before I dive into reading how to troubleshoot.

I haven't tried running the CPU (prime 95 30.3 b6) and GPU (gpuowl) software simultaneously yet, perhaps I will try a PRP test out on the CPU and see what I get.

Edit: Just read your edit - I'll definitely start with running it on the CPU and GPU for now! May dive into the IGC afterwards, as I'll need to figure out which program to run on it and how to configure it.
I sometimes run CPU(s), GPU(s), and IGP simultaneously on a system. (But not on certain cpu/motherboard combos; an i7-4790 blew out two motherboards that way; prime95 & igp mfakto was just too much for an old Asrock motherboard model.)
FYI, If you must choose one, for power budget reasons or otherwise, gpu runs, for high throughput at good efficiency. Two, Gpu and cpu. No constraint, GPU, cpu, IGP. IGPs tend to be slow compared to the others. Sometimes running on the IGP helps total throughput, sometimes it doesn't, depending on models and work type/program. (IGP competes with CPU cores for TDP and memory bandwidth in the same chip package.) Look into power reduction for efficiency increase on the GPU after things are working at a basic level.
Proper tuning of a gpu in mfakto.ini gpu sieving settings etc can add more performance relative to default settings, than the total throughput of an IGP.

Something else to look at later is using something like Mfactor to take advantage of hyperthreading on the cpu. Prime95 performance in primality testing is usually best without using hyperthreading there. Which leaves hyperthreads capacity for other activity; interactive use, and low-memory-bandwidth number theory stuff.

Last fiddled with by kriesel on 2020-10-28 at 19:55
kriesel is online now   Reply With Quote
Old 2020-10-28, 19:54   #11
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

2·3·1,489 Posts
Default

Quote:
Originally Posted by mathematizer View Post
Is GEC an additional setting in v30 that I'd select somewhere? Looking through my current exponent's active work, I see it alerting me that hardware errors have occured and there was 1 Gerbicz/double check error, but confidence is still excellent, so I imagine it's on automatically.
GEC = Gerbicz Error Checking. It is built into the PRP test. You are good.
Uncwilly is online now   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Greetings from a newbie piforbreakfast Lounge 53 2020-10-28 16:41
Newbie :) ednaemars Information & Answers 2 2018-04-24 07:24
I'm a Newbie davieddy Information & Answers 17 2010-11-22 00:47
newbie needs help 111393 Hardware 9 2004-03-28 05:01
Newbie ThePotato Lounge 7 2003-03-02 15:24

All times are UTC. The time now is 07:10.

Wed Dec 2 07:10:14 UTC 2020 up 83 days, 4:21, 1 user, load averages: 1.58, 1.66, 1.55

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.