Discussion AMD Ryzen MegaThread! FAQ and Resources

[goldstone77]

AGESA 1.0.0.6b Might Fix The Ryzen Linux Performance Marginality Problem
Written by Michael Larabel in AMD on 13 September 2017 at 08:22 PM EDT. 73 Comments

Motherboard vendors have begun pushing out BIOS updates for Ryzen motherboards using the AMD AGESA 1.0.0.6b revision and it's reported that it does resolve the "Performance Marginality Problem" affecting early Ryzen Linux customers.

While newer Ryzen CPUs are running great on Linux without the performance marginality problem as described by AMD, since yesterday I have begun receiving unconfirmed reports from Phoronix readers that the recent AGESA 1.0.0.6b revision does address the issue via this software update.

But unfortunately details on AGESA 1.0.0.6b are light. Most of the web pages / threads referencing AGESA 1.0.0.6b are actually inquiring about the changes to be found in this update, but few actual details on this update that has been surfacing in the past two weeks via BIOS updates.

After beginning to receive reader statements since yesterday that AGESA 1.0.0.6b addresses the Linux "performance marginality problem", I reached out to some AMD contacts when waking up this morning. Unfortunately, as of ending out the day, I have yet to receive any comment from any of them whether this update is expected to address the problem.

Since then as well I have received a comment from the reader of being able to reproduce the performance marginality problem under Windows. There is now the kill-ryzen-win script that also seeks to illustrate Ryzen issues when using Visual Studio C/C# compilation. But we'll see if the latest AGESA update takes care of that too or if hearing anything else from AMD.

 

[juanrga]

AGESA 1.0.0.6b Might Fix The Ryzen Linux Performance Marginality Problem
Written by Michael Larabel in AMD on 13 September 2017 at 08:22 PM EDT. 73 Comments

Motherboard vendors have begun pushing out BIOS updates for Ryzen motherboards using the AMD AGESA 1.0.0.6b revision and it's reported that it does resolve the "Performance Marginality Problem" affecting early Ryzen Linux customers.

While newer Ryzen CPUs are running great on Linux without the performance marginality problem as described by AMD, since yesterday I have begun receiving unconfirmed reports from Phoronix readers that the recent AGESA 1.0.0.6b revision does address the issue via this software update.

But unfortunately details on AGESA 1.0.0.6b are light. Most of the web pages / threads referencing AGESA 1.0.0.6b are actually inquiring about the changes to be found in this update, but few actual details on this update that has been surfacing in the past two weeks via BIOS updates.

After beginning to receive reader statements since yesterday that AGESA 1.0.0.6b addresses the Linux "performance marginality problem", I reached out to some AMD contacts when waking up this morning. Unfortunately, as of ending out the day, I have yet to receive any comment from any of them whether this update is expected to address the problem.

Since then as well I have received a comment from the reader of being able to reproduce the performance marginality problem under Windows. There is now the kill-ryzen-win script that also seeks to illustrate Ryzen issues when using Visual Studio C/C# compilation. But we'll see if the latest AGESA update takes care of that too or if hearing anything else from AMD.

"Performance Marginality Problem" is a marketing term given by AMD representative to refer to the segmentation fault. This fault wasn't a problem of performance but one of stability and integrity of data. Corruption of data generated errors and/or hard reboots.

UPDATE: I have been reading the comments of users at Phoronix. NO, the 1.0.0.6b AGESA does NOT solve the issue.

 

[imrazor]

Any progress with Ryzen and virtualization? Last I heard it was possible with less than 15 threads, but passthrough was an issue due to IOMMU groups being a mess on most motherboards.

 

[goldstone77]

Any progress with Ryzen and virtualization? Last I heard it was possible with less than 15 threads, but passthrough was an issue due to IOMMU groups being a mess on most motherboards.

GPU Passthrough for Virtualization with Ryzen: Now Working
Level1Linux
Published on Jun 25, 2017
[video="https://www.youtube.com/watch?v=aLeWg11ZBn0&ab_channel=Level1Linux"][/video]
RYZEN GPU PASSTHROUGH SETUP GUIDE: FEDORA 26 + WINDOWS GAMING ON LINUX
SUBMITTED BY WENDELL ON JUNE 25, 2017
This might answer your question.

 

[goldstone77]

I posted this information same discussion in the ThreadRipper mega thread.

The upcoming DDR5 standard which should be available on consumer platforms in 2-3 years is aiming to hit speeds as high as 6400 MHz (standard) so it will be interesting to see what G.Skill manages to do with those. It should be stated here that G.Skill has already achieved frequencies of over 5.5 GHz with their Trident Z series alongside a Kaby Lake-X processor and have said to launch DDR4-4800 MHz memory in the future.

Click here for link
It will be interesting to see the performance gains with infinity fabric if they can achieve those same higher frequencies thus reducing the interconnect latency! 3200MHz-4200MHz would be another 15% FPS increase to FPS or other latency sensitive applications. 4800MHz would offer another 6.86% FPS increase to FPS. ~21.86% is no joke! AMD needs to really be working hard on this!

Edit: Another 18.28% for 6400MHz as well for a total of 41.14%.
Additional information:
The FPS increase was also calculated by truegenius in this post made by Juanrga click here for link

if we plot graph of % difference between 3200 & 2133 then we can see that ryzen is providing 1.5x to 2x performance compared to intel ( eg if intel's performance is improving by 8% then ryzen's by 15% ), which means ryzen benifits significantly much more from faster ram than intel does.

In the table below you can see the internal latency deficit Ryzen suffers vs Intel.

The gap is 5x higher. 98-69 = 29ns gap. 29/6 ≈ 5.

Click here for link
Looking closer as truegenius's calculations the benefit over Intel is actually half of what I reported, and I will change my statements. 3200MHz-4200MHz would be another 7.5% FPS increase over Intel processors or other latency sensitive applications. 4800MHz would offer another 3.43% FPS increase over Intel. ~10.93% is no joke! AMD needs to really be working hard on this!

Another 9.14% for 6400MHz as well for a total of 20.07%.
Also, here is a link to where I started the discussion on AMD forums.
https://community.amd.com/message/2790538#comment-2790538

 

[juanrga]

DELETED

 

[Rogue Leader]

Recent BIOS update on Gigabyte mobos is killing RyZen CPUs

http://wccftech.com/amd-ryzen-cpu-gigabyte-x370-damage-burning-bios-update/

That article is a MONTH old, a new BIOS was released the same day fixing this issue.

You may have noticed a lot of your posts and other posts have been deleted here recently. If you all are incapable of having a discussion without treating eachother with respect and not mudslinging both eachother and products you will be posting elsewhere.

This therad is about 5 posts away from being closed if you all want to continue on this path. You also won't be posting on any other threads either if you keep this up.

 

[goldstone77]

Recent BIOS update on Gigabyte mobos is killing RyZen CPUs

http://wccftech.com/amd-ryzen-cpu-gigabyte-x370-damage-burning-bios-update/

That article is a MONTH old, a new BIOS was released the same day fixing this issue.

You may have noticed a lot of your posts and other posts have been deleted here recently. If you all are incapable of having a discussion without treating eachother with respect and not mudslinging both eachother and products you will be posting elsewhere.

This therad is about 5 posts away from being closed if you all want to continue on this path. You also won't be posting on any other threads either if you keep this up.

Yes sir, acknowledged! We all need to do better. I know differences of opinions can cause tensions to rise, but let's try too focus on facts and do away with personal attacks while staying on topic. Arguing facts with evidence, and not wallowing in a squabble match. I'm sorry for my part, and would sincerely like to keep this thread alive!

 

[juanrga]

Recent BIOS update on Gigabyte mobos is killing RyZen CPUs

http://wccftech.com/amd-ryzen-cpu-gigabyte-x370-damage-burning-bios-update/

That article is a MONTH old, a new BIOS was released the same day fixing this issue.

You may have noticed a lot of your posts and other posts have been deleted here recently. If you all are incapable of having a discussion without treating eachother with respect and not mudslinging both eachother and products you will be posting elsewhere.

This therad is about 5 posts away from being closed if you all want to continue on this path. You also won't be posting on any other threads either if you keep this up.

I deleted the post. I copied it from another part and I didn't notice the datetime. Sorry for the stupid mistake.

 

[8350rocks]

http://www.guru3d.com/news-story/eight-nvme-m2-ssds-in-raid-on-x399-threadripper-reach-28-gbs.html

Der8auer manages to run 8 NVMe SSDs in RAID 0 on X399, get 28 GB/s bandwidth...far surpassing any current HEDT platform available.

 

[aldaia]

AMD to launch 12nm Ryzen in February 2018, says mobo makers

AMD has informed its partners that it plans to launch in February 2018 an upgrade version of its Ryzen series processors built using a 12nm low-power (12LP) process at Globalfoundries, according to sources at motherboard makers.

The company will initially release the CPUs codenamed Pinnacle 7, followed by mid-range Pinnacle 5 and entry-level Pinnacle 3 processors in March 2018, the sources disclosed. AMD is also expected to see its share of the desktop CPU market return to 30% in the first half of 2018.

AMD will launch the low-power version of Pinnacle processors in April 2018 and the enterprise version Pinnacle Pro in May 2018.

Their corresponding chipsets, the 400 series, will also become available in March 2018 with X470- or B450-based motherboards to be the first to hit the store shelves. The chipsets are still designed by ASMedia and its orders for the chipsets are expected to grow dramatically starting January 2018.

Thanks to stable chip orders for Microsoft's and Sony's game consoles, increased demand for graphics cards, growing sales for its Ryzen 7/5 processors, new Ryzen Pro product line for the enterprise sector and the top-end Ryzen Treadripper processors, AMD managed to achieve 19% sequential growth in second-quarter 2017 revenues and expects the amount to grow further by 23% in the third quarter.

AMD said it does not comment on products that have not been announced.

 

[juanrga]

So it requires new 400-series chipsets and new mobos.

 

[-Fran-]

If they require new MoBos I'll flip a table D:!

Cheers!

 

[goldstone77]

Succeeding it will be Picasso, AMD’s second generation APU based on the Zen microarchitecture. These processors, much like Pinnacle Ridge, will be refreshed parts built on GlobalFoundries’ new 12nm LP process technology. Both generations will also maintain compatibility with AMD’s AM4 socket on the desktop.

http://wccftech.com/amd-pinnacle-ridge-matisse-cpus-leaked-launching-2018-2019-respectively-am4-picasso-apu-succeeding-raven-ridge/

 

[juanrga]

socket != chipset

 

[gamerk316]

socket != chipset

See LGA775 for reference 😛

 

[jdwii]

socket != chipset

See LGA775 for reference 😛

Basically how long did we see Phenom products come out and be supported on AM2+/AM3

Let me leave this here haha

http://www.gigabyte.us/Motherboard/GA-MA790FX-DQ6-rev-10#support-cpu

Darn AM2+ boards saw Athlon 64 FX-62 support all the way to the Phenom II x6 1100T

FX-62 2006 release
Phenom II x6 1100T 2010

Zen 2017
Zen3 2020

Main difference now more is integrated into the CPU(SOC) so one would think support would even be easier.

 

[goldstone77]

Socket AM4 has 1331 pin slots. Plenty of pins to work with!

 

[-Fran-]

socket != chipset

See LGA775 for reference 😛

Basically how long did we see Phenom products come out and be supported on AM2+/AM3

Let me leave this here haha

http://www.gigabyte.us/Motherboard/GA-MA790FX-DQ6-rev-10#support-cpu

Darn AM2+ boards saw Athlon 64 FX-62 support all the way to the Phenom II x6 1100T

FX-62 2006 release
Phenom II x6 1100T 2010

Zen 2017
Zen3 2020

Main difference now more is integrated into the CPU(SOC) so one would think support would even be easier.

Even more, I do know for a fact that Asus gave un-official support to Pilediver into 890FX chipset MoBos, even when AMD said they wouldn't have support for it. I helped a friend investigate on the subject and we found some Beta BIOS'es for his Asus Crosshair IV after he got an FX8350 migrating from an AthlonII X3. That is a *huge* generational jump that according to "official slides" was not meant to happen.

That is why I don't usually believe "technical reasons" when Intel or AMD say "oh, guys, sorry, you gotta have a shiny new mobo for this".

Cheers!

 

[TMTOWTSAC]

They may be a little gunshy about that kind of thing now. Half the threads about fx 9590's here come down to mobo issues. Either boards that don't officially support it but sorta work anyway, or boards that are supposed to support it and burn out their power delivery systems. It shouldn't be the same situation given the power usage disparity will be much less, or even negative, but there isn't a lot of information about the 12nm LP node as of yet.

 

[goldstone77]

They may be a little gunshy about that kind of thing now. Half the threads about fx 9590's here come down to mobo issues. Either boards that don't officially support it but sorta work anyway, or boards that are supposed to support it and burn out their power delivery systems. It shouldn't be the same situation given the power usage disparity will be much less, or even negative, but there isn't a lot of information about the 12nm LP node as of yet.

Juanrga and Jdwii said they seen on other forums that the talk is lower operating voltage, which they assume ~3-400MHz boost to frequency. But they are still seeing limitation with a frequency wall 4.1GHz.

 

[juanrga]

They may be a little gunshy about that kind of thing now. Half the threads about fx 9590's here come down to mobo issues. Either boards that don't officially support it but sorta work anyway, or boards that are supposed to support it and burn out their power delivery systems. It shouldn't be the same situation given the power usage disparity will be much less, or even negative, but there isn't a lot of information about the 12nm LP node as of yet.

Juanrga and Jdwii said they seen on other forums that the talk is lower operating voltage, which they assume ~3-400MHz boost to frequency. But they are still seeing limitation with a frequency wall 4.1GHz.

Those aren't my words.

 

[goldstone77]

They may be a little gunshy about that kind of thing now. Half the threads about fx 9590's here come down to mobo issues. Either boards that don't officially support it but sorta work anyway, or boards that are supposed to support it and burn out their power delivery systems. It shouldn't be the same situation given the power usage disparity will be much less, or even negative, but there isn't a lot of information about the 12nm LP node as of yet.

Juanrga and Jdwii said they seen on other forums that the talk is lower operating voltage, which they assume ~3-400MHz boost to frequency. But they are still seeing limitation with a frequency wall 4.1GHz.

Those aren't my words.

I take it back than. What were you saying?

 

[juanrga]

I said I expect 100--300MHz improvement for Pinnacle Ridge, and this also applies to the frequency wall, because the process is improved.

 

[jdwii]

They may be a little gunshy about that kind of thing now. Half the threads about fx 9590's here come down to mobo issues. Either boards that don't officially support it but sorta work anyway, or boards that are supposed to support it and burn out their power delivery systems. It shouldn't be the same situation given the power usage disparity will be much less, or even negative, but there isn't a lot of information about the 12nm LP node as of yet.

that's over VRM those cpus used 220 watts of course that wouldn't work on 4+1 phase boards lol hell the fx 8350 barely did