![]() |
![]() |
#1 |
"David Kirkby"
Jan 2021
Althorne, Essex, UK
32×11 Posts |
![]()
I'm hoping to upgrade my computer within the next couple of weeks from a
* Single Xeon 8167M (2.0 GHz, 26-core) with 2 x 32 GB RDIMMs, which is sub-optimal as they should be in installed in groups of 4 to to * Pair of Xeon Platinum 8171Ms (2.6 GHz, 26 core) with 8 x 32 GB, which is an optimal configuration. I'm currently in the process of doing a PRP test to see if 2^332646233 -1 is prime or not. https://www.mersenne.org/report_expo...2646233&full=1 I'm only 10% done, and don't want to wait until early May to perform the upgrade, when the results of that test will be known. I suspect it would be a bit foolish to upgrade the computer, then immediately continue with that large PRP test. So what I am thinking is it might be beneficial to do just double-checking of what others have done. Does that seem sensible? What's the best ways to do it? I just reserved https://www.mersenne.org/report_expo...ll=1&ecmhist=1 for a LL double check, but are wondering if that's the best approach to take. I don't know, but I assume LL tests of exponents in the range 60,000,000 to 70,000,000 will note take too long with 26 cores, but maybe I am wrong. I'll set up two workers. From my point of view, it would be nice to get some feedback if there are problems. Dave Last fiddled with by drkirkby on 2021-03-01 at 21:07 |
![]() |
![]() |
![]() |
#2 |
P90 years forever!
Aug 2002
Yeehaw, FL
163578 Posts |
![]()
You'll need to do some throughput benchmarks once you upgrade. I suspect maximum throughput will come with more than 2 workers.
I don't see any reason why one of your workers cannot continue with your 100Mdigit PRP test. The rest of the workers can do double-checks or new first-time PRP tests. There is no "right" answer to your question. Do whatever brings you the greatest joy. |
![]() |
![]() |
![]() |
#3 |
"David Kirkby"
Jan 2021
Althorne, Essex, UK
9910 Posts |
![]()
The reason I did not wish to immediate continue to the 100 million digit prime test, is because it's a long test, and if hardware problems are found, I would rather not screw up a test that I've already been running a week or so. If any problems are found, I would rather swap the CPUs back - I'm buying them used, and they are not normally for sale to end users. Intel will not disclose any information about them. But a benchmark at https://www.cpubenchmark.net/cpu.php...220&cpuCount=2 for a dual socket 8171M would indicate it has good performance for the money. The RAM will be new from Kingston, so I don't see any problems with that, but the CPUs are a bit of an unknown.
|
![]() |
![]() |
![]() |
#4 |
"Viliam Furík"
Jul 2018
Martin, Slovakia
2×223 Posts |
![]()
60M or 70M tests will run very quickly on 26 cores, with your memory configuration.
Don't worry too much about hardware errors on PRP tests. If your hardware can reliably run an OS, it shouldn't be able to do much harm in the PRP test, because of the Gerbicz error check (GEC). If you are worried, run a few LL DCs after the upgrade. If nothing serious happens, you are good to continue with the big test. |
![]() |
![]() |
![]() |
Thread Tools | |
![]() |
||||
Thread | Thread Starter | Forum | Replies | Last Post |
Major iOS Upgraded Disaster | a1call | Lounge | 39 | 2019-07-19 15:58 |
No results for Recently set up GPU 2 72 computer | jschwar313 | GPU to 72 | 3 | 2016-01-31 16:41 |
Upgraded to GTX690 | Chuck | GPU Computing | 14 | 2013-04-16 00:04 |
New computer, problems with mprime | apocalypse | Hardware | 4 | 2009-10-14 19:00 |
Computer Problems | moo | Hardware | 4 | 2007-07-24 23:20 |