Any ASUS Prime X370 Pro owners out there ... besides me??

[mackerel]

I'm not 100% sure, but I think there might have been a way to tell from the box info if you still have that. Both my Ryzen CPUs are early ones, and in theory affected by this. I couldn't say I have knowingly encountered it though, also doing distributed computing when it is worth it. Maybe there is a specific combination of factors that need to be present but it has been so long since I looked at the issue I can't recall.

 

[Keith Myers]

No retail box of mine has ever had the batch code

I'm not 100% sure, but I think there might have been a way to tell from the box info if you still have that. Both my Ryzen CPUs are early ones, and in theory affected by this. I couldn't say I have knowingly encountered it though, also doing distributed computing when it is worth it. Maybe there is a specific combination of factors that need to be present but it has been so long since I looked at the issue I can't recall.

I have never seen the cpu batch code of production printed on any retail box of mine. It lists the cpu serial number and the product SKU, but that's it. The cpu is visible in the retail box in its clamshell pack on the side. But once you install it and cover it up with a heat sink, you cannot tell the date of production.

That is why the replacement cpu got its picture taken in its clamshell pack as the first thing after breaking the security seal on the box.

 

[DaveB]

You have to look at the batch code number printed on the IHS. Like my first 1700X was Batch 1707 or week 7 of 2017. If I remember right, week 25 was second week of July. Only way to be sure is look which means pulling the heat sink off and scrubbing off the heat sink compound. The replacement cpu got its picture taken with the cell phone so I know what week it was made without having to disassemble.

The consensus I have read is that the segmentation violation is NOT affecting ONLY Linux users. They are just the likely ones to trip up the cpu because they would be the ones heavily loading the cpu with multi-thread workloads like compiling kernels. If you are similarly heavily loading the cpu with multi-threaded work in the Windows environment, you could have the seg fault too. I am that case in fact since I run the Windows system just as hard as the Linux system with distributed computing on all cores 24/7. I had segfaults on my workload tasks. The tasks write out the error dump down to the registers values and it is pretty easy to see the condition that threw the upset. So I took advantage of the RMA offered.

I had lots of niggling issues with all my early Ryzens and ended up either returning or selling them. Then late last fall I picked up a week 40 something 1700X and it had no issues of any kind. Plus it easily ran at 4 GHz with the memory at 3200 CL14. So yeah, I believe the issue affected many things but they got it fixed so all the new Ryzens should be good to go.

 

[Keith Myers]

My week 38 1800X has been without issue since the day I put the system together. Now the 1700X is finally dialed in. I just this week adjusted voltages gleaned from the OCN memory thread post attributed to BIOS 4008. Many changes in the new AGESA code. Biggest difference is that you have to dial back the memory voltage to stock 1.35V and dial the SoC voltage down considerably from where it was running in earlier BIOS releases. I have found that 0.95V for the SoC made a huge difference in stability at the higher clocks. Exactly opposite of what was canon back in the early days requiring ever increasing SoC voltage for higher clocks. Side benefit is the cpu runs cooler along with higher memory clocks and tighter timings.

 

[Thentilian]

You have to look at the batch code number printed on the IHS. Like my first 1700X was Batch 1707 or week 7 of 2017. If I remember right, week 25 was second week of July. Only way to be sure is look which means pulling the heat sink off and scrubbing off the heat sink compound. The replacement cpu got its picture taken with the cell phone so I know what week it was made without having to disassemble.

The consensus I have read is that the segmentation violation is NOT affecting ONLY Linux users. They are just the likely ones to trip up the cpu because they would be the ones heavily loading the cpu with multi-thread workloads like compiling kernels. If you are similarly heavily loading the cpu with multi-threaded work in the Windows environment, you could have the seg fault too. I am that case in fact since I run the Windows system just as hard as the Linux system with distributed computing on all cores 24/7. I had segfaults on my workload tasks. The tasks write out the error dump down to the registers values and it is pretty easy to see the condition that threw the upset. So I took advantage of the RMA offered.

My week 38 1800X has been without issue since the day I put the system together. Now the 1700X is finally dialed in. I just this week adjusted voltages gleaned from the OCN memory thread post attributed to BIOS 4008. Many changes in the new AGESA code. Biggest difference is that you have to dial back the memory voltage to stock 1.35V and dial the SoC voltage down considerably from where it was running in earlier BIOS releases. I have found that 0.95V for the SoC made a huge difference in stability at the higher clocks. Exactly opposite of what was canon back in the early days requiring ever increasing SoC voltage for higher clocks. Side benefit is the cpu runs cooler along with higher memory clocks and tighter timings.

Is AMD still doing this RMA? It's been quite a while, but I have noticed funky issues with my CPU when loading up handbrake or streaming while gaming. Stability seems to take a dump sometimes even though it's rock solid for everything else.

If they are not offering the RMA, no point in taking the CPU cooler off to find out what week my chip was manufactured in.

Although, maybe if I decide to get a 2000 series Ryzen 8 core, then I'll pull it off to find out. RMA, if available, then sell the unit brand new in the box. Normally I wouldn't do something like that, but if my CPU does have this defect, it could make it hard to sell.

I still need to see if it's worth even getting an X470 chipset for the new 2000 series, but I don't think it is. Most of what I read on it so far is just better BIOS and some memory support. But it's all brand dependent on what is and isn't good for memory, and that's about the only thing I really want is to be able to run fast memory for gaming and streaming.

 

[Keith Myers]

Is AMD still doing this RMA? It's been quite a while, but I have noticed funky issues with my CPU when loading up handbrake or streaming while gaming. Stability seems to take a dump sometimes even though it's rock solid for everything else.

If they are not offering the RMA, no point in taking the CPU cooler off to find out what week my chip was manufactured in.

Although, maybe if I decide to get a 2000 series Ryzen 8 core, then I'll pull it off to find out. RMA, if available, then sell the unit brand new in the box. Normally I wouldn't do something like that, but if my CPU does have this defect, it could make it hard to sell.

I still need to see if it's worth even getting an X470 chipset for the new 2000 series, but I don't think it is. Most of what I read on it so far is just better BIOS and some memory support. But it's all brand dependent on what is and isn't good for memory, and that's about the only thing I really want is to be able to run fast memory for gaming and streaming.

Yes, AMD has to warranty the flawed parts. Just contact AMD Support to start the process. You have to download a PDF document to fill out and return to start the process.

I have been sitting on the fence whether the newer parts were justified in cost compared to 1st Gen and X370. Based on the reviews I didn't think so. But then new information has come to light that all the reviews missed. The presence of PE or Performance Overdrive now has tipped the scales in the new parts favor. None of the reviews tested the new capability because the review samples were loaded with pre-production BIOS'. Only the latest AGESA BIOS' have the PE feature. Think of if as AMD's version of Intel's Mult-Core Enhancement. The feature once enabled in the BIOS boosts all cores up to a fixed multiplier that is 100 Mhz typically on top of what XFR2 and Precision Boost 2.0 enable. It also allows you to remove the normal power envelope limits that AMD normally sets. That feature is one I would find appealing since I run all cores loaded at all times. And I want the highest achievable cpu clocks because I do lots of cpu work.

The memory overclocking is more consistent too. Stilt has said the IMC is the exact same from Gen 1 and Gen 2 because the firmware from PHY in the chip is the same. So no miracle memory overclocks matching Intel. But you should be able to more consistently meet the manufacturers XMP spec up to about 3533Mhz supposedly. Some 3600Mhz have been reached. As with Gen.1, Samsung B-dies and single-rank are best choice for achieving rated ratings.

 

[Woomack]

The memory overclocking is more consistent too. Stilt has said the IMC is the exact same from Gen 1 and Gen 2 because the firmware from PHY in the chip is the same. So no miracle memory overclocks matching Intel. But you should be able to more consistently meet the manufacturers XMP spec up to about 3533Mhz supposedly. Some 3600Mhz have been reached. As with Gen.1, Samsung B-dies and single-rank are best choice for achieving rated ratings.

If BIOS/UEFI is good then on new AGESA you can make 3733-3866 on every IC, it doesn't have to be Samsung. I was able to set 3733+ on Hynix, Micron and Samsung in dual and quad channel on 1st gen Ryzen/TR. Well, I was testing it only on ASRock motherboards but I don't think there will be issues on other boards too ... as long as BIOS is good.

16GB modules in dual/quad channel I could make 3466 stable
8GB Samsung B modules dual/quad channel on X370/X399 - 3733 stable, 3866 max
8GB Hynix modules dual channel on X370 - 3733, max boot 3866 and could pass some tests
8GB Micron modules dual/quad channel on X370/X399 - 3733 stable, 3866 could barely boot

Johan45's results on X470 using Samsung sticks are not much better. Stable results are about the same and the only difference is that he could boot at 4000 while I can't. Maybe it's just BIOS and highest multi is not working, hard to say. Still regardless if it's new or old motherboard, I wouldn't count on stable results much above 3600 or at tight timings and high frequency at the same time.

Here is Crucial/Micron D9 in quad channel @3733 CL20 ... relaxed timings but it works fine at 1.35V.

 

[Keith Myers]

Thanks for the report Woomack. I have never been able to get either my BOINC Ryzen's stable at anything other than 3200 CL14. I had 3333 CL14 working for a while on 3203 BIOS but then I ruined that by updating to later BIOS'

I have the Ryzen 1700X on Win 10 stable on 4008 BIOS. What finally clicked for me on that BIOS is down-volting both the SoC and the Vdimm. It is finally stable at 0.95V SoC and 1.36V memory at 3200 CL14. It is the most stable since the system was built in March 2017. I am loathe to change anything. I probably should update the Ryzen 1800X Linux cruncher to 4008 BIOS too. But I am in the "if it aint broke . . . don't fix it" camp so far. It too would probably be able to drop its SoC and memory voltage too based on the 1700X experience.

If the crunchers weren't full production machines at all times I probably would be more inclined to experiment. But testing for stability means taking them offline for 4-6 hours trying to find a likely stable candidate and that reduces production considerably. Then only running BOINC for a week actually proves the system is stable. Any failure results in going back to stability testing again resulting in lost production again.

 

[Aballarde]

Founded solution to High Definition Audio Bus

Thanks for the report Woomack. I have never been able to get either my BOINC Ryzen's stable at anything other than 3200 CL14. I had 3333 CL14 working for a while on 3203 BIOS but then I ruined that by updating to later BIOS'

I have the Ryzen 1700X on Win 10 stable on 4008 BIOS. What finally clicked for me on that BIOS is down-volting both the SoC and the Vdimm. It is finally stable at 0.95V SoC and 1.36V memory at 3200 CL14. It is the most stable since the system was built in March 2017. I am loathe to change anything. I probably should update the Ryzen 1800X Linux cruncher to 4008 BIOS too. But I am in the "if it aint broke . . . don't fix it" camp so far. It too would probably be able to drop its SoC and memory voltage too based on the 1700X experience.

If the crunchers weren't full production machines at all times I probably would be more inclined to experiment. But testing for stability means taking them offline for 4-6 hours trying to find a likely stable candidate and that reduces production considerably. Then only running BOINC for a week actually proves the system is stable. Any failure results in going back to stability testing again resulting in lost production again.

 

[Aballarde]

First of all I would like to thank you for trying to help me to solve the issue in Device Manager HIGH DEFINITION AUDIO BUS code 10.

After searching and trying different ways to fix the problem, I stumble with a way to make CODE 10 for High Definition Audio Bus, go away for good, and in appreciation I would like to share my findings and solution in behalf of those who may have the same issue specially with ASUS PRIME X370- PRO motherboard.

It happens to be that: Graphics driver software for RADEON graphics cards and here is what is all about:

" after i have disabled Enable Ulps, the High Definition Audio Controller Code 10 error in the device manager was gone (because stupid windows 10 could "see" again the gpu sound device again, the device (fury x gpu's) wasn't anymore in energy saving sleep state) and the start menu and taskbar freezes too".

" Anyone that has troubleshooted problems with AMD/ATI graphics card crossfire configurations will probably know about Ultra Low Power State (ULPS). ULPS is a sleep state that lowers the frequencies and voltages of non-primary cards in an attempt to save power. This holds true for single card users as well. The downside of ULPS is that is can cause performance loss and some crossfire instability. This also holds true for laptops with or without crossfire and can increase performance, but will use more battery power.
You can disable this feature manually by entering regedit , select EDIT at the top of the page , enter ULPS in the 'find what' box , then click 'find next' to search for all instances of 'Enable ULPS' (there will be several with crossfire). Each time a instance is found , change the default value of "1" to a "0" then at EDIT again choose 'find next' until there are no more."

So I follow instructions and disabled ULPS for both of my RADEON graphics cards and the problem is gone. Here you can find more detailed information in how to solve the issue: https://community.amd.com/thread/224008

I hope that this information is helpful to any other PC user with ASUS PRIME X370-PRO motherboard

 

[Keith Myers]

There has been precious little comments or posts over on the Prime X370 Pro motherboard thread at OCN ever since the X470 boards came out. I have mostly moved to the Crosshair VII Hero and 2700X platform now. I only have my 2nd Prime Pro and a 1800X cruncher working currently. I managed to partially kill my first Prime Pro/1700X system in a case move and replaced it with the C7H/2700X duo which has become my staple in my crunching farm.

The C7H/2700X is a dream to work with. Just shows what ASUS learned from the first generation boards and processors. I have 3 C7H/2700X systems running at 4.1 or 4.0 Ghz all cores loaded 100% of the time at 1.29 or 1.33V Vcore. All systems are running on The Stilt's B-die single-sided 1.4V DDR4 preset memory profile which clocks them at 3466 Mhz at CL14 latency and fast timings. All C7H/2700X have been virtually flawless and completely stable in my crunching farm.

Sure the C7H was purchased at a premium over the Prime Pro product but I think the features and capabilities warrant the extra cost. No regrets at all in my upgrades.

 

[Whyzen]

Ive just gotten one and a 2700x. Made a bad RAM decision and had to return a kit. Currently awaiting the arrival of some horridly expensive ddr4 so I can resume setting my system up.

 

[Keith Myers]

No excuse not to pair Ryzen with Samsung B-die of CL16 or better latency. Just accept the cost for good stable overclocks and performance.

 

[DaveB]

No excuse not to pair Ryzen with Samsung B-die of CL16 or better latency. Just accept the cost for good stable overclocks and performance.

Yep all my Ryzen memory issues disappeared when I picked up some G.Skill TridentZ DDR4-3200 CL14 with Samsung B-Die RAM late last summer. Got on sale too - $175 for a 16GB set. Worked great at 3200 on every B350 and X370 motherboard I played with.

 

[Whyzen]

Yep all my Ryzen memory issues disappeared when I picked up some G.Skill TridentZ DDR4-3200 CL14 with Samsung B-Die RAM late last summer. Got on sale too - $175 for a 16GB set. Worked great at 3200 on every B350 and X370 motherboard I played with.

Since I wanted 32GB, I chose 2 double-sided 16GB sticks of Gskill b-die CL14. As the Conventional Wisdom predicted, I seem to be topping out at 3000. No slouch, though and a huge jump in bandwidth from my old components.

 

[Bluefalcon13]

So been pretty absent around here lately, but figured I'd post an update. Picked up a 500gb m.2 drive the other weekend finally (Samsung 970 Evo) and reinstalled Windows and stuff. Was pulling down drivers and saw there was a new bios, so tossed it on. Currently running bone stock on my OC, but low and behold, the DOCP-3000 settings now work with my poo sticker clad ram. One of these days I'll ramp the CPU OC back up to 3.7ghz, but seriously considering picking up a water block first. Probably could hit much higher then, with minimal effort. If I had some cash, I would just say **** it and buy a used 1800x and run that stock so I don't have to mess with it. Gonna skip 2xxx series to finalize my divorce and get situated in my new life, and maybe see what Zen2 has to offer.

Between the m.2 and all my old data SSDs, I have over 1TB of SSD storage, which is hilariously larger than my platter based storage [emoji23]

 

[Keith Myers]

Haven't dropped in here for a long time myself. Updated both of my FX upgraded systems to 970 EVO M.2 myself. Helped a lot in boot times over their original SATA SSD installations. Been biding my time waiting for updated TR2 motherboards and am disappointed the only real viable product so far is the MSI MEG Creation board, Watched buildzoid's commentary about what a properly designed TR2 board should have with regard to VRM phase count for the 24/32 core cpus. I have no need for that many cores. The 16 on my Ryzen 2700X are fine for supporting my gpu distributed computing. The only reason I am interested in TR is that that platform is the only AMD based platform that can support 4 double wide gpus. If the board manufacturers would design a proper HEDT X470 based board with true 4 slot cpu spacing with a PLX chip to support the PCIe lanes, I would jump on that. I would be perfectly happy with a 2900X or 2920X cpu on the TR2 motherboard platform, as long as the board has proper 4 double wide gpu slot spacing. A lot of them don't. So currently waffling on just getting a ASUS Zenith Extreme and a 2920X when it releases in October, or continue to wait for any updated TR2 boards that have properly designed VRM phase count enough to handle the XFR2/PBO'd autoboosting of the 24/32 core chips.

 

[DaveB]

Was pulling down drivers and saw there was a new bios, so tossed it on. Currently running bone stock on my OC, but low and behold, the DOCP-3000 settings now work with my poo sticker clad ram.

Glad Asus finally fixed that issue for you. I was running that same cheap G.Skill RAM at 2933 a year and a half ago on an $80 ASRock B350 motherboard.

 

[Bluefalcon13]

Glad Asus finally fixed that issue for you. I was running that same cheap G.Skill RAM at 2933 a year and a half ago on an $80 ASRock B350 motherboard.

They fixed it back when the first big bios update came. I upgraded bios a little while ago and it regressed.

It's good now. [emoji16]

 

[DaveB]

They fixed it back when the first big bios update came. I upgraded bios a little while ago and it regressed. ��‍

It's good now. [emoji16]

Yep, I went though multiple BIOSes from ASRock B350 boards that sometimes broke things while fixing others.