Ryzen and Threadripper

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

To post messages, you must log in.

Previous · 1 . . . 26 · 27 · 28 · 29 · 30 · Next

AuthorMessage
jsm

Send message
Joined: 1 Oct 16
Posts: 77
Credit: 34,579,085
RAC: 85,100
Isle of Man
Message 2007458 - Posted: 15 Aug 2019, 6:05:49 UTC - in response to Message 2007417.  

So the rumours that TR3 would be scrapped are put to bed. I hope that if and when we get the 64/128 top of the line that it can show similar improvements. I will keep my 2990wx as it is steadily churning out acceptable performance with its 19.04 kernel but I am really looking for 8 way memory control in any top of the range unit.
jsm
ID: 2007458 · Report as offensive     Reply Quote
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 11565
Credit: 170,669,821
RAC: 102,568
Australia
Message 2007462 - Posted: 15 Aug 2019, 6:21:43 UTC - in response to Message 2007330.  

The Intel microcode hasn't really changed in ten years.

Actually it has changed quite a bit, hence their improvements in IPC (Instructions Per Clock) with each revision.
But the changes have been tweaks to an existing architecture, not a significantly different architecture as was the case with Ryzen.
Grant
Darwin NT
ID: 2007462 · Report as offensive     Reply Quote
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 3452
Credit: 195,615,997
RAC: 537,152
United States
Message 2007478 - Posted: 15 Aug 2019, 12:52:12 UTC - in response to Message 2007458.  

So the rumours that TR3 would be scrapped are put to bed. I hope that if and when we get the 64/128 top of the line that it can show similar improvements. I will keep my 2990wx as it is steadily churning out acceptable performance with its 19.04 kernel but I am really looking for 8 way memory control in any top of the range unit.
jsm


+1
I will stop procrastinating tomorrow.
\\// Live Long & Prosper (starting tomorrow ;)
ID: 2007478 · Report as offensive     Reply Quote
Profile Ageless
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 14867
Credit: 4,086,569
RAC: 917
Netherlands
Message 2007501 - Posted: 15 Aug 2019, 14:20:12 UTC - in response to Message 2006404.  
Last modified: 15 Aug 2019, 14:21:58 UTC

Finally got a moment to address this.
Keith Myers wrote:
If you have legitimate storage needs, like a RAID array or massive database that is constantly being read and written, then certainly update to X570 for the PCIe Gen. 4 lanes. But I posted on the assumption we were speaking of appropriate SETI or BOINC crunching hardware.
My crunching hardware consists of the trio of Android devices and that's not easily going to change. I'm building my next gen gaming rig. My local computer shop owner is trying to sell me a 2600X on a B450 motherboard. I told him no. My system, my rules. :)

Keith Myers wrote:
You point out another flaw in your thinking. The X570 and PCIe Gen 4 chipsets are in reality a interim solution. Both PCIe Gen. 5 and DDR5 standards are already of out of committee and approved and will likely be seen in motherboards at the end of next year or early 2021. If your argument is that a X570 purchase "future proofs" you, yes for about a year. Since these new X570 motherboards are not cheap as the previous AMD motherboards, maybe waiting out one development cycle would be a smart choice. An X470 motherboard still has a lot of functional life left in it.
As I posted in my previous post on this subject, PCIe Gen 6 is already slated for 2021, so I wouldn't hold my breath for PCIe 5.0 either.
But as a couple of motherboard makers showed, before they were called back, is that a BIOS update can add PCIe 4.0 support to some 300/400 motherboards. So similarly I would think that a BIOS update will be able to add PCIe 5.0 (1) and 6.0 (2) support to X570, especially since the actual PCIe slots themselves don't change.

And while we don't really know if Big Navi will be PCIe 3.0 or 4.0, it is coming in 2020. So I may for the moment put in an RX580 and get the new card next year.

As for the X570 prices, I'm not going for the Godlike or Ultimate Super Duper high priced versions. I don't need a motherboard with 3 PCIe x16 slots. One is enough, as I run just one videocard. If I then check a vendor I prefer, the cheapest X570 is just 34 euros more expensive than the cheapest X470, while the amount of X470s still available is dropping fast. (checking motherboards immediately available)

So no thanks, if you want to spend a dollar less on an older motherboard that's being phased out rapidly, you can go ahead.
And if we're all to hold back because something new is just over the horizon, no one here would ever buy new hardware again, because there's always something new over the horizon the moment you bought something. Anyone waiting for X670?

(1) https://en.wikipedia.org/wiki/PCI_Express#PCI_Express_5.0
(2) https://en.wikipedia.org/wiki/PCI_Express#PCI_Express_6.0
Jord

According to Giorgo of the Ancient Astronaut Theorists I do not help with tech questions via private message. He's right: please use the forums for that.
ID: 2007501 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 9629
Credit: 885,789,490
RAC: 1,726,452
United States
Message 2007540 - Posted: 15 Aug 2019, 18:52:14 UTC - in response to Message 2007501.  

As I posted in my previous post on this subject, PCIe Gen 6 is already slated for 2021, so I wouldn't hold my breath for PCIe 5.0 either.
But as a couple of motherboard makers showed, before they were called back, is that a BIOS update can add PCIe 4.0 support to some 300/400 motherboards. So similarly I would think that a BIOS update will be able to add PCIe 5.0 (1) and 6.0 (2) support to X570, especially since the actual PCIe slots themselves don't change.

Yes, for a while some X470 motherboards could do PCIe 4.0. But then AMD laid down the law and told them flatly NO! And then took the decision out of the mobo vendors hands by pulling the ability to offer PCIE Gen. 4 on X470 out of the AGESA code which only AMD controls and has write access to. So all the newest BIOS have had PCIE Gen. 4 pulled from mobo BIOS' and there is nothing the vendors can do about it. They have to take the AGESA code that AMD offers and then build their BIOS features on top of what stock features AMD says the X470 chipset is supposed to have.

So no thanks, if you want to spend a dollar less on an older motherboard that's being phased out rapidly, you can go ahead.
And if we're all to hold back because something new is just over the horizon, no one here would ever buy new hardware again, because there's always something new over the horizon the moment you bought something. Anyone waiting for X670?


This is a standard computer axiom that I will not argue with. If you are always "waiting for the next big thing" you will never purchase any new technology as the goal line is always moving forward. About every 6 months on average over the age of the computer industry. I was simply offering the idea that X570 is not mandatory to enjoy Ryzen 3000. There are many capable X370/X470 platforms still available for a decent price that will do the job. I felt absolutely no imperative to purchase 5 new motherboards when they came out just because they are an improvement over my X470 motherboards because they have one feature mine doesn't and I will never use.
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 2007540 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 9629
Credit: 885,789,490
RAC: 1,726,452
United States
Message 2008987 - Posted: 23 Aug 2019, 21:01:50 UTC
Last modified: 23 Aug 2019, 21:25:38 UTC

New information posted in the Ryzen Ram Calculator thread about the new "made for Ryzen 3000" G.Skill Neo 3600Mhz Ram kits. Now that Samsung B-die is no longer being made and becoming very difficult to find New Old Stock, I wondered what the memory stick makers would come up with for performance RAM. Seems that the new Neo kits are being made with Hynix CJR dies which were the second best performing back when Samsung B-dies held top spot. The don't have the same latencies across the board like the B-dies, but the first primary number is pretty good. From OCN post.
So these new Gskill Neo3600mhz 14-15-15-35 are not b-die but rather Hynix CJR. You think these are topped out or anything left in the tank on normal voltage <1.5v?


So until I see actual Techno reviews of those kits. Caveat emptor. But they will probably turn out decent choices for Ryzen 3000 build for performance for what is available now.

[Edit] Came across this nice review of nine kits of memory tested on Ryzen 3000 at XMP and OC settings.
https://www.computerbase.de/2019-08/ram-oc-amd-ryzen-3000-test/
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 2008987 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 9629
Credit: 885,789,490
RAC: 1,726,452
United States
Message 2009013 - Posted: 24 Aug 2019, 0:47:06 UTC

Updated the daily driver this afternoon with the latest official BIOS that fixes the no booting problem on Linux 19.04 or later with RdRand error. Was able to boot into my Ubuntu 19.04 development partition with kernel 5.0.0.21 with no issues. Just an FYI for anyone else running a Ryzen 3000 cpu.
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 2009013 · Report as offensive     Reply Quote
Profile FloridaBear
Avatar

Send message
Joined: 28 Mar 02
Posts: 117
Credit: 6,473,972
RAC: 7,471
United States
Message 2009408 - Posted: 26 Aug 2019, 20:07:44 UTC - in response to Message 2004808.  

A 12 core, 24 thread 3900 with a TDP of 65W now sounds interesting enough for me to definitively wait with buying the hardware part of my new system.


The 3900x has a default TDP of 105W, but will go to 142W without PBO unlock, so about a 40% overhead. If the 3900 is roughly the same (e.g. 40% overhead), that puts it at about 90W max. If I set PPT at 90W on my 3900X, the clocks settle in at about 3.5 GHz. Temps run at about 54C at this speed (with a radiator, YMMV).

On another note, I really like the ability to set the maximum allowed temperature in the BIOS, which makes for pretty worry-free crunching. Currently, I set my max temp to 78C and unlock PBO, which yields the following (all values approximate):

Temp: 78C (remains constant under full load)
PPT: ~155 watts
Clocks: ~3.96 GHz

This is running an NZXT Kraken x62, so short of sub-ambient cooling, this is pretty maxed out for a 3900X.


ID: 2009408 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 9629
Credit: 885,789,490
RAC: 1,726,452
United States
Message 2009415 - Posted: 26 Aug 2019, 21:05:42 UTC

I am currently just using a fixed multiplier of 41.5 to lock all the cores of my 3900X. I have a 360mm radiator. I'm happy with the temps.

CPU Core Voltage: +1.24 V
CPU SOC Voltage: +1.08 V
DRAM Voltage: +1.42 V

CPU Temperature: +74.0°C
CPU Socket Temperature: +44.0°C
Motherboard Temperature: +35.0°C
Chipset Temperature: +51.0°C
Tsensor 1 Temperature: +33.0°C
CPU VRM Temperature: +50.0°C
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 2009415 · Report as offensive     Reply Quote
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 3452
Credit: 195,615,997
RAC: 537,152
United States
Message 2009576 - Posted: 28 Aug 2019, 12:33:39 UTC

Has anyone seen any more details on the release date on the 3950x?

Keep looking and the last announcement I find is sometime in September.

Plus I just saw an article that says the shortage of the 3900x continues.

Tom
I will stop procrastinating tomorrow.
\\// Live Long & Prosper (starting tomorrow ;)
ID: 2009576 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 9629
Credit: 885,789,490
RAC: 1,726,452
United States
Message 2009621 - Posted: 28 Aug 2019, 17:00:31 UTC - in response to Message 2009576.  
Last modified: 28 Aug 2019, 17:05:45 UTC

Has anyone seen any more details on the release date on the 3950x?

Keep looking and the last announcement I find is sometime in September.

Plus I just saw an article that says the shortage of the 3900x continues.

Tom

Haven't heard anything other than probably September, not set in stone with only 4Q 2019 as the target date.

Considering the 3900X and the 3950X will grab the top 5% of the binned chiplets, I think TSMC is having yield issues with the 7nm process still that reduces the general quality of the dies.

Then there are the Epyc processors which are getting the top 1% of the binned chiplets.
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 2009621 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 9629
Credit: 885,789,490
RAC: 1,726,452
United States
Message 2009672 - Posted: 28 Aug 2019, 21:37:45 UTC

Another leaked Geekbench4 test of the upcoming Ryzen Threadripper 32-core cpu. Article over at Tom's Hardware.
https://www.tomshardware.com/news/amd-ryzen-threadripper-3000-32-core-castle-peak-4.3ghz,40261.html
Geekbench4 result.
https://browser.geekbench.com/v4/cpu/14448604
Individual core clocks.
https://browser.geekbench.com/v4/cpu/14448604.gb4
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 2009672 · Report as offensive     Reply Quote
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 3452
Credit: 195,615,997
RAC: 537,152
United States
Message 2009802 - Posted: 29 Aug 2019, 13:47:12 UTC - in response to Message 2009672.  

Another leaked Geekbench4 test of the upcoming Ryzen Threadripper 32-core cpu. Article over at Tom's Hardware.
https://www.tomshardware.com/news/amd-ryzen-threadripper-3000-32-core-castle-peak-4.3ghz,40261.html
Geekbench4 result.
https://browser.geekbench.com/v4/cpu/14448604
Individual core clocks.
https://browser.geekbench.com/v4/cpu/14448604.gb4


+1

Sometimes I wish I still had my 2990wx rig. Other times I am glad it is gone. It was very irritating to be reduced to 25-27 threads to get decent wall clock processing time.
The new Threadrippers still look like they will be a better buy for straight ahead fast Seti cpu crunching than the 64 core Rome Epyc(sp?) crunching.

Tom
I will stop procrastinating tomorrow.
\\// Live Long & Prosper (starting tomorrow ;)
ID: 2009802 · Report as offensive     Reply Quote
Profile Ageless
Volunteer tester
Avatar

Send message
Joined: 9 Jun 99
Posts: 14867
Credit: 4,086,569
RAC: 917
Netherlands
Message 2009891 - Posted: 29 Aug 2019, 20:57:43 UTC

Piggybacking this news to this thread, although AMD does not seem to be affected: GlobalFoundries' lawsuit against TSMC seeks ban on Apple, Nvidia products.

US SEMICONDUCTOR GIANT GlobalFoundries has filed lawsuits against Taiwan Semiconductor Manufacturing Company (TSMC) in the US and Germany, alleging the firm of infringing 16 of its patents.

GlobalFoundries, which last year bowed out of developing 7nm chips due to a "lack of demand", alleges that TSMC infringes its intellectual property in its 7nm, 10nm, 12nm, 16nm, 28nm manufacturing processes. Given these processes are used to make more than half of TSMC's chips, the firm is seeking "significant damages" which reports claim could reach billions of dollars.

Among the big names accused of infringing upon GlobalFoundries' intellectual property are Apple, Asus, Broadcom, Cisco, Google, Nvidia, Qualcomm, Lenovo, and Motorola.

This means, if the courts were to back GlobalFoundries' corner and issue an injunction, Apple's iPhones and iPads, Nvidia GPUs and any smartphone running a Qualcomm SoC made by TSMC could be banned in the US.

AMD, despite being one of TSMC's clients - specifically for its 7nm Ryzen 3000 CPUs - is not named in the lawsuit and would likely be unaffected by any outcome in favour of GlobalFoundries.


So best get your hardware while you still can...
Jord

According to Giorgo of the Ancient Astronaut Theorists I do not help with tech questions via private message. He's right: please use the forums for that.
ID: 2009891 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 9629
Credit: 885,789,490
RAC: 1,726,452
United States
Message 2009909 - Posted: 29 Aug 2019, 22:18:02 UTC - in response to Message 2009891.  

You know this will never go to trial. It will languish in the court docket for years until both parties reach a settlement agreement before it goes to trial. Likely some cross-licensing agreement and a bit of cash thrown in for good measure to make the whole thing disappear. This type of lawsuit happens regularly in the tech industry with the outcome I outlined above as the end result.
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 2009909 · Report as offensive     Reply Quote
Profile ML1
Volunteer moderator
Volunteer tester

Send message
Joined: 25 Nov 01
Posts: 9891
Credit: 7,396,093
RAC: 83
United Kingdom
Message 2010533 - Posted: 3 Sep 2019, 12:20:04 UTC

AMD completely trounces Intel on this one on both performance and price!


Dual AMD EPYC 7742 Crushes Quad Intel Xeon 8180M's In Geekbench 4

Patrick Kennedy, Editor-in-Chief at ServeTheHome, recently set a new world record on Geekbench 4 with pair of AMD EPYC 7742 processors. The publication also compared the pair of EPYC 7742 chips against four Intel Xeon Platinum 8180M processors, with the AMD system being the clear winner.

In one corner, we have the AMD EPYC 7742, which comes out punching with 64 cores and 128 threads, and the Intel Xeon Platinum 8180M with its 28 cores and 56 threads in the opposing corner. The AMD system consists of two EPYC 7742 and tallies up to 128 cores and 256 threads while the Intel system has four Xeon Platinum 8180M for a total of 112 cores and 224 threads...



Spectacular.

Way to go!

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: 2010533 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 9629
Credit: 885,789,490
RAC: 1,726,452
United States
Message 2010547 - Posted: 3 Sep 2019, 18:01:42 UTC

I wonder if we might get an announcement from AMD next Tuesday on the arrival of the expected Threadripper 3000 cpus.
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 2010547 · Report as offensive     Reply Quote
Profile ML1
Volunteer moderator
Volunteer tester

Send message
Joined: 25 Nov 01
Posts: 9891
Credit: 7,396,093
RAC: 83
United Kingdom
Message 2010577 - Posted: 4 Sep 2019, 0:03:10 UTC

Here's an interesting note for scheduling changes that comes with an even more interesting comment:


Linux 5.4 Kernel To Bring Improved Load Balancing On AMD EPYC Servers

... AMD EPYC/Zen processors now overrides the node reclaim distance to better account for the CPU's architecture. From one of the code comments, "AMD EPYC machines use this because even though the 2-hop distance is 32 (3.2x slower than a local memory access) performance actually *improves* if allowed to reclaim memory and load balance tasks between NUMA nodes 2-hops apart."...



Curious counter-intuitive stuff...

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: 2010577 · Report as offensive     Reply Quote
Profile George Project Donor
Volunteer tester
Avatar

Send message
Joined: 23 Oct 17
Posts: 120
Credit: 1,332,291
RAC: 3,544
United States
Message 2010869 - Posted: 6 Sep 2019, 5:41:34 UTC - in response to Message 2009621.  

Has anyone seen any more details on the release date on the 3950x?

Keep looking and the last announcement I find is sometime in September.

Plus I just saw an article that says the shortage of the 3900x continues.

Tom


Haven't heard anything other than probably September, not set in stone with only 4Q 2019 as the target date.

Considering the 3900X and the 3950X will grab the top 5% of the binned chiplets, I think TSMC is having yield issues with the 7nm process still that reduces the general quality of the dies.

Then there are the Epyc processors which are getting the top 1% of the binned chiplets.

This is what I get for not paying attention for awhile. I was unaware of the court case of TSMC and the status of the chiplets for the 3900X & 3950X. I'm only waiting on AMD's release of the 3950X to finish my build and get going with Linux. I sure hope I don't run into problems with memory allocation and threads doing things like down-clocking.
George

ID: 2010869 · Report as offensive     Reply Quote
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 9629
Credit: 885,789,490
RAC: 1,726,452
United States
Message 2010873 - Posted: 6 Sep 2019, 6:44:03 UTC - in response to Message 2010547.  

Distressing news today from an insider. The expected new Threadripper cpus made from Zen 2 chiplets will not be pin compatible with existing X399 motherboards. They will require new sockets and motherboard designs and chipsets.

So the idea I had of simply dropping a Threadripper 3920X into my X399 motherboard is now squashed.
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 2010873 · Report as offensive     Reply Quote
Previous · 1 . . . 26 · 27 · 28 · 29 · 30 · Next

Message boards : Number crunching : Ryzen and Threadripper


 
©2019 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.