mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > News

Reply
 
Thread Tools
Old 2018-12-14, 20:44   #12
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

116416 Posts
Default

Quote:
Originally Posted by chalsall View Post
Why not simply wait for the official announcement?
If I was more patient/reasonable/uninterested, I wouldn't be spending what I am on gpus and electricity, nor my own time and effort in various ways on this project. GIMPS total throughput would be some percentage less.

https://www.goodreads.com/quotes/536...e-unreasonable

Last fiddled with by kriesel on 2018-12-14 at 20:45
kriesel is online now   Reply With Quote
Old 2018-12-14, 20:56   #13
chalsall
If I May
 
chalsall's Avatar
 
"Chris Halsall"
Sep 2002
Barbados

2×4,657 Posts
Default

Quote:
Originally Posted by kriesel View Post
If I was more patient/reasonable/uninterested, I wouldn't be spending what I am on gpus and electricity, nor my own time and effort in various ways on this project. GIMPS total throughput would be some percentage less.
Yeah, but only a little less... You might be a bit surprised by just how much many of us spend on this hobby.

Have you ever heard the expression "Manage the situation.
chalsall is online now   Reply With Quote
Old 2018-12-14, 21:22   #14
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

22·3·7·53 Posts
Default

Quote:
Originally Posted by chalsall View Post
Yeah, but only a little less... You might be a bit surprised by just how much many of us spend on this hobby.
I don't see chalsall in the top producers lists. Using an alias? Is that common? Not sniping, just curious.

Last fiddled with by kriesel on 2018-12-14 at 21:43
kriesel is online now   Reply With Quote
Old 2018-12-14, 21:32   #15
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

100001101100112 Posts
Default

Quote:
Originally Posted by kriesel View Post
I don't see chalsall in the top producers lists. Using an alias?
Not only does Chris facilitate/run GPUto72 (which does great things to use GPU's to their potential and saving CPU time), he also runs Prime95 and mprime on many machines. He runs it to ensure that the machines that he services/deploys in a corporate environment stay functioning.

Yes, he uses a different name on PrimeNet. As do I.

Yes it is common. Do you use your full name as your forum ID? Do you use an alias? Is that common? I don't see your handle in my phonebook.

Last fiddled with by Uncwilly on 2018-12-14 at 22:00
Uncwilly is offline   Reply With Quote
Old 2018-12-14, 21:41   #16
masser
 
masser's Avatar
 
Jul 2003
wear a mask

5×7×41 Posts
Default

Quote:
Originally Posted by kriesel View Post
I don't see chalsall in the top producers lists. Using an alias?
Great, now we're comparing the sizes of our kits?
masser is offline   Reply With Quote
Old 2018-12-15, 01:06   #17
simon389
 
Aug 2013

22×3×7 Posts
Default

I thought CPUs were the only things that could do LL tests. But here I read you can use a simple GPU to do LL tests? I just finished building a bunch of quad channel Skylake X computers to search for LL primes. Was my work/money wasted? If GPUs work I could have just created a GPU rack like what cryptocurrency miners use.

Quote:
Originally Posted by kriesel View Post
If it still matters, you have my standing offer from some time ago. Current fastest hardware is GTX1080Ti ready with CUDALucas v2.06 May 5 2017 beta, capable of knocking out a ~82M exponent in about 3 days, backed up by a GTX1080, and a 16-core e5-2670 system with prime95 V29.4b8. The GTX1080Ti at 84+GhzD/day for 85M is ~10% faster than a 2080 Ti for LL per https://www.mersenne.ca/cudalucas.php (14th fastest model in the list, and that includes combined throughputs of some duals and a quad; 10th fastest single). And no, that 82M or 85M is not a clue to the nature of Mp51; it's just what I checked the 1080Ti's run time on recently. It would take me some time to put up mlucas on the system.
simon389 is offline   Reply With Quote
Old 2018-12-15, 01:12   #18
retina
Undefined
 
retina's Avatar
 
"The unspeakable one"
Jun 2006
My evil lair

166C16 Posts
Default

Quote:
Originally Posted by simon389 View Post
I thought CPUs were the only things that could do LL tests. But here I read you can use a simple GPU to do LL tests? I just finished building a bunch of quad channel Skylake X computers to search for LL primes. Was my work/money wasted?
Maybe. Maybe not. It depends upon your goal(s). Some GPUs can do an LL test very quickly, but the cost is very high. Other, cheaper, GPUs take longer to do one test, and might use more energy per test. Most CPUs can do multiple tests at once, and take longer per individual test, but overall can do more tests per time period.

So define to yourself the goal you want to achieve, really low latency / really high throughput, and purchase gear based upon that.
retina is offline   Reply With Quote
Old 2018-12-15, 01:27   #19
GP2
 
GP2's Avatar
 
Sep 2003

22×3×5×43 Posts
Default

Quote:
Originally Posted by simon389 View Post
I thought CPUs were the only things that could do LL tests. But here I read you can use a simple GPU to do LL tests? I just finished building a bunch of quad channel Skylake X computers to search for LL primes. Was my work/money wasted? If GPUs work I could have just created a GPU rack like what cryptocurrency miners use.
CPUs are still very competitive for doing LL tests, and Skylake X is a good choice. Make sure you get the latest beta version 29.5 to get the most out of AVX-512 on a Skylake. Most of the users here use CPUs rather than GPUs.

GPUs can also do LL tests, and the modern ones do them pretty quickly, but they do tend to cost more and especially, they use a lot more electricity. How much does electricity cost in your area? That is a major consideration. In the end what matters isn't raw speed so much as total cost (amortized cost of hardware plus electricity) per test.
GP2 is offline   Reply With Quote
Old 2018-12-15, 02:21   #20
simon389
 
Aug 2013

22×3×7 Posts
Default

Quote:
Originally Posted by GP2 View Post
How much does electricity cost in your area? That is a major consideration. In the end what matters isn't raw speed so much as total cost (amortized cost of hardware plus electricity) per test.
15 cents / kW hr

Definitely rocking the 29.5 with AVX 512 optimizations. i7 9800X with DDR4 3600 can do a 88M LL @ 1.95ms/iter @ 300watt total consumption. Competes the test in about 48 hours.

Glad to hear my work was not in vain :D
simon389 is offline   Reply With Quote
Old 2018-12-15, 02:44   #21
preda
 
preda's Avatar
 
"Mihai Preda"
Apr 2015

1,217 Posts
Default

Quote:
Originally Posted by GP2 View Post
CPUs are still very competitive for doing LL tests, and Skylake X is a good choice. Make sure you get the latest beta version 29.5 to get the most out of AVX-512 on a Skylake. Most of the users here use CPUs rather than GPUs.

GPUs can also do LL tests, and the modern ones do them pretty quickly, but they do tend to cost more and especially, they use a lot more electricity. How much does electricity cost in your area? That is a major consideration. In the end what matters isn't raw speed so much as total cost (amortized cost of hardware plus electricity) per test.
IMO, the only dimension where CPUs win vs. GPUs is in ease of use / ease of deployment. That means that's it's easier to just install the OS, and prime95/mprime, and you're good to go, no headache.

On every other dimension, (in my measure), the GPUs win:
- price: i7-7820x + RAM + Cooler is more expensive than Vega64 (8GB GPU RAM included, cooler included)
- power: i7-7820x (w AVX) uses > 200W power (excluding the motherboard), Vega64 in my normal setup (--setsclk 5) uses 150W.
- perf: at the wavefront, i7-7820x does 2.7 ms/it (depends on RAM speed, this is with "3000" RAM), and the Vega64 does 2.0ms/it.

OTOH installing the GPU drivers (OpenCL or CUDA), software, etc. is a PITA unfortunately.

Last fiddled with by preda on 2018-12-15 at 02:45
preda is online now   Reply With Quote
Old 2018-12-15, 02:59   #22
simon389
 
Aug 2013

22·3·7 Posts
Default

Quote:
Originally Posted by preda View Post
IMO, the only dimension where CPUs win vs. GPUs is in ease of use / ease of deployment. That means that's it's easier to just install the OS, and prime95/mprime, and you're good to go, no headache.

On every other dimension, (in my measure), the GPUs win:
- price: i7-7820x + RAM + Cooler is more expensive than Vega64 (8GB GPU RAM included, cooler included)
- power: i7-7820x (w AVX) uses > 200W power (excluding the motherboard), Vega64 in my normal setup (--setsclk 5) uses 150W.
- perf: at the wavefront, i7-7820x does 2.7 ms/it (depends on RAM speed, this is with "3000" RAM), and the Vega64 does 2.0ms/it.

OTOH installing the GPU drivers (OpenCL or CUDA), software, etc. is a PITA unfortunately.
Makes sense to me! I wish I had learned this a month ago. Oh well.

Last fiddled with by simon389 on 2018-12-15 at 03:11
simon389 is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Stockfish game: "Move 8 poll", not "move 3.14159 discussion" MooMoo2 Other Chess Games 5 2016-10-22 01:55
Aouessare-El Haddouchi-Essaaidi "test": "if Mp has no factor, it is prime!" wildrabbitt Miscellaneous Math 11 2015-03-06 08:17
Would Minimizing "iterations between results file" may reveal "is not prime" earlier? nitai1999 Software 7 2004-08-26 18:12

All times are UTC. The time now is 03:40.

Tue Sep 29 03:40:39 UTC 2020 up 19 days, 51 mins, 0 users, load averages: 1.25, 1.44, 1.56

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.