Ryzen and Threadripper

Message boards : Number crunching : Ryzen and Threadripper
Message board moderation

To post messages, you must log in.

Previous · 1 . . . 6 · 7 · 8 · 9 · 10 · 11 · 12 . . . 69 · Next

AuthorMessage
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13164
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1991975 - Posted: 30 Apr 2019, 8:13:01 UTC

I push my 2920X Threadripper right up the point where it is just below the 68°C. package temp limit. Just shy of throttling the clock speed due to temps. Still able to hold all cores north of 4000Mhz with some boosting to around 4100Mhz. I also have high end liquid cooling for the cpu.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1991975 · Report as offensive     Reply Quote
Profile Wiggo
Avatar

Send message
Joined: 24 Jan 00
Posts: 34932
Credit: 261,360,520
RAC: 489
Australia
Message 1991981 - Posted: 30 Apr 2019, 9:23:51 UTC

But then again Keith the 2920X with only 24 threads isn't going to come into the memory contention problems that a 2990WX capable of 64 threads will face, but then Tom has already found that out.

Cheers.
ID: 1991981 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13164
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1992011 - Posted: 30 Apr 2019, 17:24:56 UTC - in response to Message 1991981.  

True. The double length path to memory for two of the dies is an issue with the 4 die Threadrippers.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1992011 · Report as offensive     Reply Quote
Profile bloodrain
Volunteer tester
Avatar

Send message
Joined: 8 Dec 08
Posts: 231
Credit: 28,112,547
RAC: 1
Antarctica
Message 1992046 - Posted: 1 May 2019, 0:05:48 UTC - in response to Message 1992011.  

true and the way gen 1 and gen 1.5. did it are different that the 7nm version of it coming out
ID: 1992046 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13164
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1992055 - Posted: 1 May 2019, 0:55:18 UTC - in response to Message 1992046.  

We still don't know the I/O architecture detail of Zen 2 with respect how it is going to be implemented on Threadripper.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1992055 · Report as offensive     Reply Quote
jsm

Send message
Joined: 1 Oct 16
Posts: 124
Credit: 51,135,572
RAC: 298
Isle of Man
Message 1992394 - Posted: 3 May 2019, 5:48:14 UTC - in response to Message 1992055.  

and what do we do with 128 threads?
jsm
ID: 1992394 · Report as offensive     Reply Quote
Profile Wiggo
Avatar

Send message
Joined: 24 Jan 00
Posts: 34932
Credit: 261,360,520
RAC: 489
Australia
Message 1992397 - Posted: 3 May 2019, 6:32:24 UTC

Currently?

1: Other than save power and reduce heat by reducing the number of threads used until you get a happy medium on your CPU times, nothing (sorry).

2: Or wait until they can get better/more memory bandwidth and sharing architecture happening between cores.

3: But in the meantime you could sell it for something more appropriate for use here and maybe get a few $'s back from the deal. ;-)

Cheers.
ID: 1992397 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13164
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1992399 - Posted: 3 May 2019, 6:37:32 UTC - in response to Message 1992397.  

I think he was referring to Zen 2 Threadrippers. They have a complete change in cpu architecture from Zen1 and the earlier Threadrippers. Shouldn't have memory issues since the memory access is completely different from what I understand.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1992399 · Report as offensive     Reply Quote
Profile Wiggo
Avatar

Send message
Joined: 24 Jan 00
Posts: 34932
Credit: 261,360,520
RAC: 489
Australia
Message 1992401 - Posted: 3 May 2019, 6:45:02 UTC

It can only be hoped that they got it right this time. ;-)

Cheers.
ID: 1992401 · Report as offensive     Reply Quote
Profile bloodrain
Volunteer tester
Avatar

Send message
Joined: 8 Dec 08
Posts: 231
Credit: 28,112,547
RAC: 1
Antarctica
Message 1992415 - Posted: 3 May 2019, 10:19:41 UTC - in response to Message 1992394.  

and what do we do with 128 threads?
jsm

one thing finale.. happening on the consumer side. is multi core/thread support in software.
ID: 1992415 · Report as offensive     Reply Quote
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 5124
Credit: 276,046,078
RAC: 462
Message 1992443 - Posted: 3 May 2019, 16:05:38 UTC - in response to Message 1992415.  

and what do we do with 128 threads?
jsm

one thing finale.. happening on the consumer side. is multi core/thread support in software.


Run multiple copies of BOINC in parallel?
:)

Tom
A proud member of the OFA (Old Farts Association).
ID: 1992443 · Report as offensive     Reply Quote
jsm

Send message
Joined: 1 Oct 16
Posts: 124
Credit: 51,135,572
RAC: 298
Isle of Man
Message 1992482 - Posted: 3 May 2019, 20:11:11 UTC - in response to Message 1992443.  

Oh I like that!
Could that be done now? With WIN10 or Ubuntu? Perhaps two complete downloads and installation on separate disks so that preference files did not clash.
jsm
ID: 1992482 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13164
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1992485 - Posted: 3 May 2019, 20:37:17 UTC - in response to Message 1992482.  

Sure, you can do that now. Just start another client up. Each client has to be installed into its own unique directory. You could start with the wiki. And then follow the instructions for setting up multiple BOINC instances.

https://boinc.berkeley.edu/wiki/Controlling_BOINC_remotely

https://www.overclock.net/forum/18056-boinc-guides-tutorials/1628924-guide-setting-up-multiple-boinc-instances.html
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1992485 · Report as offensive     Reply Quote
Profile bloodrain
Volunteer tester
Avatar

Send message
Joined: 8 Dec 08
Posts: 231
Credit: 28,112,547
RAC: 1
Antarctica
Message 1992853 - Posted: 6 May 2019, 19:15:14 UTC - in response to Message 1992443.  

lol true or have half go to 1 project and the other half go to another project.
ID: 1992853 · Report as offensive     Reply Quote
jsm

Send message
Joined: 1 Oct 16
Posts: 124
Credit: 51,135,572
RAC: 298
Isle of Man
Message 1992931 - Posted: 7 May 2019, 7:21:46 UTC - in response to Message 1992853.  

At 55% ratio has dropped slightly to 1.16 so we may be past the optimum. Bearing in mind that the start of the test at45% had less than 1 I have dropped the pref to 50% for this next week.
I might well have a go at two BOINCS with the second set at lowest possible priority (19?) to see if it then used the less optimum threads but in a way that did not reduce the work on the main threads 'first instance' more than that gained on the second.
jsm
ID: 1992931 · Report as offensive     Reply Quote
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 5124
Credit: 276,046,078
RAC: 462
Message 1993099 - Posted: 8 May 2019, 12:53:35 UTC - in response to Message 1992931.  

At 55% ratio has dropped slightly to 1.16 so we may be past the optimum. Bearing in mind that the start of the test at45% had less than 1 I have dropped the pref to 50% for this next week.
I might well have a go at two BOINCS with the second set at lowest possible priority (19?) to see if it then used the less optimum threads but in a way that did not reduce the work on the main threads 'first instance' more than that gained on the second.
jsm


I took a look at your tasks and made some guesstimates and came out with as much as 10,000 more RAC than you are currently getting. So it has got me wondering what did I get wrong? :)

So now all we need to do is graph it....

Tom
A proud member of the OFA (Old Farts Association).
ID: 1993099 · Report as offensive     Reply Quote
Profile ML1
Volunteer moderator
Volunteer tester

Send message
Joined: 25 Nov 01
Posts: 20372
Credit: 7,508,002
RAC: 20
United Kingdom
Message 1993253 - Posted: 9 May 2019, 11:16:29 UTC

I wonder what RAC this will give?!


Beyond Rome: AMD's EPYC and Radeon to Power World's Fastest Exascale Supercomputer

AMD announced today that it had been selected to power Frontier, which is set to be the world's fastest exascale-class supercomputer when it comes online in 2021. The new supercomputer, which will be built with Cray's Shasta supercomputer blades, is being developed by Cray under a $600 million contract for the U.S. Department of Energy (DOE) for the Oak Ridge National Laboratory. All told, the supercomputer is projected to be faster than the top 160 supercomputers in the world, combined...


Can we persuade them to do a soak test with s@h? ;-)

Happy fast crunchin'!
Martin
See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)
ID: 1993253 · Report as offensive     Reply Quote
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 5124
Credit: 276,046,078
RAC: 462
Message 1993254 - Posted: 9 May 2019, 11:58:10 UTC - in response to Message 1993253.  

I wonder what RAC this will give?!


Beyond Rome: AMD's EPYC and Radeon to Power World's Fastest Exascale Supercomputer


I suspect it could absorb/process and return the entire output of the Seti servers and barely notice it....

Tom
A proud member of the OFA (Old Farts Association).
ID: 1993254 · Report as offensive     Reply Quote
Profile Mike Special Project $75 donor
Volunteer tester
Avatar

Send message
Joined: 17 Feb 01
Posts: 34259
Credit: 79,922,639
RAC: 80
Germany
Message 1993261 - Posted: 9 May 2019, 12:32:58 UTC - in response to Message 1993253.  
Last modified: 9 May 2019, 12:57:55 UTC

I wonder what RAC this will give?!


Beyond Rome: AMD's EPYC and Radeon to Power World's Fastest Exascale Supercomputer

AMD announced today that it had been selected to power Frontier, which is set to be the world's fastest exascale-class supercomputer when it comes online in 2021. The new supercomputer, which will be built with Cray's Shasta supercomputer blades, is being developed by Cray under a $600 million contract for the U.S. Department of Energy (DOE) for the Oak Ridge National Laboratory. All told, the supercomputer is projected to be faster than the top 160 supercomputers in the world, combined...


Can we persuade them to do a soak test with s@h? ;-)

Happy fast crunchin'!
Martin




1.5 Exaflops

https://www.planet3dnow.de/cms/46402-amd-epyc-cpus-amd-radeon-instinct-gpus-and-rocm-open-source-software-to-power-worlds-fastest-supercomputer-at-oak-ridge-national-laboratory/


With each crime and every kindness we birth our future.
ID: 1993261 · Report as offensive     Reply Quote
Profile Jord
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 15184
Credit: 4,362,181
RAC: 3
Netherlands
Message 1993381 - Posted: 10 May 2019, 22:40:39 UTC

The new AMD Ryzen Gen 3 look interesting enough that I plan to buy one for a system I plan on building later this year. I'll wait for some good motherboards with PCIe 4.0 and perhaps that AMD will be friendly enough to release a newer Radeon VII that can utilize that bus. The CPU I'm eyeing is the Ryzen 7 3700X.

https://www.pcgamesn.com/amd/ryzen-3000-release-date-ryzen-3rd-gen-specs-performance-price
https://www.tomshardware.com/news/amd-ryzen-3000-everything-we-know,38233.html
ID: 1993381 · Report as offensive     Reply Quote
Previous · 1 . . . 6 · 7 · 8 · 9 · 10 · 11 · 12 . . . 69 · Next

Message boards : Number crunching : Ryzen and Threadripper


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.