Solved! first gen Ryzen RAM speed

[Zepp]

I originally had a first gen Ryzen(1800X) that would not boot with my 3200 speed RAM kit on XMP profile, but worked fine if it was set to 2933 auto timings.

I no longer have that PC but recently I felt like building another first gen Ryzen to play with.
I got a 2nd hand 1800X and a new Asrock X370 Fatality. The RAM is 3 year old Patriot Vipers 3200 2x8GB from a previous PC.

not surprisingly XMP 3200 wouldn't boot but 2933 wouldn't even boot either.
2800 auto timings will boot but I eventually got a BSOD kmode_exception_not_handled

I ran memtest at 2800 and got a flood of errors after about 5min.
stepped down to 2733 and had no errors.

Is it safe to assume this is just what this particular 1800X mem controller can handle or is it possible that the RAM is at fault?

 

[DAPUNISHER]

I had the Gaming X model of the board. You have the latest bios? Memory support improved quite a bit from the bios it shipped with. I think I had a OG Ryzen 1600 doing XMP CL16 3200. It could be your used 1800X has some degradation too.

 

[Zepp]

I had the Gaming X model of the board. You have the latest bios? Memory support improved quite a bit from the bios it shipped with. I think I had a OG Ryzen 1600 doing XMP CL16 3200. It could be your used 1800X has some degradation too.

I updated BIOS when I first got windows installed to the latest non beta release which was from 2022.
It's strange it goes from v7.3 to v10.08b

ASRock > Fatal1ty X370 Gaming K4

ATX<br />ASRock Super Alloy<br />Supports AMD Socket AM4 A-Series APUs (Bristol Ridge) and Ryzen Series CPUs (Summit Ridge & Raven Ridge)<br />Supports DDR4 2933+ (OC) (Ryzen CPU) / 2400 (A-series APU)<br />2 PCIe 3.0 x16, 4 PCIe 2.0 x1, 1 M.2 (Key E)<br />NVIDIA Quad SLI™, AMD Quad...

www.asrock.com

I'm not sure if the latest beta is likely to improve anything but I'm inclined to believe it's not the RAM's fault. RAM worked fine at 3200 XMP on a zen2 system for years.

Seems I just have bad luck with gen1 Ryzen and RAM speeds.

 

[DAPUNISHER]

I have a Pro 2400G running its rated 2933MTs because it's in an ASRock deskmini. I have no idea if the IMC will handle more or not. My worst luck was with Ryzen 3600s. I had at least 4 I used for builds and none of them would pass stress testing with CL 16 or 18 3600. My quick and dirty was set XMP then manually set the speed to 3466MTs.

 

[Shmee]

At this point, you may as well run a Zen 3 chip on that board. They aren't that much money, especially used, and have much stronger IMCs. You may benefit from running a beta BIOS however, for better Zen 3 support/memory tweaking.

 

[Soulkeeper]

Your best bet with X370 and zen1 is to shoot for 2933 mem speed. It is very hard to get above that.

 

[Insert_Nickname]

Your best bet with X370 and zen1 is to shoot for 2933 mem speed. It is very hard to get above that.

2933 seems to be about the limit. My 1700 would do that with Samsung memory, but wouldn't post with anything higher no matter what I did. Using Micron dual rank DIMMs, the limit was 2400MHz. It just wouldn't go higher.

 

[kschendel]

I had a bunch of 1600's and a 1400 for a while, all running 8GBx2. About half of them could do 3000 MT/s no problem, a few had to step down to 2933, one had problems with anything over 2800. I think your CPU that tops out at 2733 is probably a weak sample, or maybe the original owner overstressed it which can weaken the memory controller.

 

[Zepp]

2933 seems to be about the limit. My 1700 would do that with Samsung memory, but wouldn't post with anything higher no matter what I did. Using Micron dual rank DIMMs, the limit was 2400MHz. It just wouldn't go higher.

ya it's strange because I remember some of the big reviewers back in 2017 getting zen1 systems up past 3000 showing how infinity fabric scaled with RAM speed. I guess they probably just had golden samples.

Your best bet with X370 and zen1 is to shoot for 2933 mem speed. It is very hard to get above that.

I was close to getting a B450 but decided to just keep it 2017. I was surprised to find a still new X370 and that it's BIOS updates enabled up to Ryzen 5000.

 

[DAPUNISHER]

I ran every gen but the 4000 series APUs, and Bristol Ridge (who would?) in that X370 Gaming X. I think you are spot on about the reviewers getting the best binned CPUs to review. That was SOP for a long time.

 

[thecoolnessrune]

I originally had a first gen Ryzen(1800X) that would not boot with my 3200 speed RAM kit on XMP profile, but worked fine if it was set to 2933 auto timings.

I no longer have that PC but recently I felt like building another first gen Ryzen to play with.
I got a 2nd hand 1800X and a new Asrock X370 Fatality. The RAM is 3 year old Patriot Vipers 3200 2x8GB from a previous PC.

not surprisingly XMP 3200 wouldn't boot but 2933 wouldn't even boot either.
2800 auto timings will boot but I eventually got a BSOD kmode_exception_not_handled

I ran memtest at 2800 and got a flood of errors after about 5min.
stepped down to 2733 and had no errors.

Is it safe to assume this is just what this particular 1800X mem controller can handle or is it possible that the RAM is at fault?

The first-gen Ryzen Memory Controller just wasn't as capable as later Ryzens. I have an ASUS Crosshair VI Hero board with 32GB of G.Skill (2x16GB Samsung B-Die) RAM, and another identical build but with 16GB (2x8GB Samsung B-Die)

Started with an 1800X. The 16GB system could do 2933. The 32GB system couldn't do over 2733. Both kits at 1.4V. In the exact same platforms I installed 3800X's. Immediately both could do 3200 out of the box with absolutely no fiddling at CL14 timings. The difference was very clear. I've now put a 5800X3D in the 32GB one and can do 3600 CL16 in that same rig. Nothing's changed in them over the time except the CPU installed.

EDIT: I said the above but think that in the name of transparency I'll mention that the BIOS did change over those times. These rigs have been in use since 2017 after all. I regularly update the BIOS on both systems. I've never found any of the BIOS's I upgraded ever allowed me to do more with the RAM in any particular CPU generation I had. I went through a number of BIOS revisions on the 1800X, including, and through the BIOS version that gave me 2nd gen Ryzen, and Windows 11 compatibility. When on those matching BIOS versions between the 1800X and the 3800X I was never able to do better speeds on the 1800X than 2733/2933. Likewise I've now had the 3800X's for a while and continued updating the BIOS through the version that gives compatibility with 3rd gen Ryzen. Was never able to get to 3600 CL16 on my 32GB system until I installed the 5800X3D, then it was as simple as setting the profile.

 

[kschendel]

Yep. The first gen Ryzen had some pretty shaky memory controller circuitry. Even in my small cluster, the good 1600's vs the bad ones were clear, with a MT/s difference of up to 300 MT/s.

It's not the motherboard. It's the CPU (and possibly memory.)

 

[Zepp]

well there is a new layer to consider now
I thought everything was in order with my system until the last few days I was having some strange power related issues.
The PC randomly restarted itself and the debug code was 'faulty CPU or no power to CPU'
then a few days later I had the system just power down completely. Troubleshooting is pointing towards my 5 year old PSU. It's a gold rated seasonic 550w. I put in an RMA request since the warranty is 10 years.

Now I am curious if the RAM instabilities could have been partially the fault of the failing PSU. 🤔

 

[kschendel]

My first gen Ryzen's liked having "Power Supply Idle Control" set to "Typical Current Idle." I wonder if your issues are related. If the PSU really is going bad, it's not out of the question that the PSU was helping muddy the waters.

The idle control thing seemed to result in weird sudden shutdowns when the machine wasn't doing anything, so maybe not the same as what you are seeing.

 

[Zepp]

final update
got the replacement PSU installed and retested RAM with memtest86
2800 speed and higher still erroring out
2734 speed remains stable
so in my case the dying PSU was not the cause of RAM instability

 

[Rigg]

I owned a bunch of first gen Ryzens. I probably bought over 10 1600's but also had a few 1700x's, an 1800x, and a 1200. Every one that I paired with 2x8 3200 C16 XMP kits were stable at rated speed/timings. These systems were thoroughly stress tested including a full round of memtest86 before installing windows. 3200 C16 kits were cheap and widely available at the time and they were my go to for most of the Zen 1 systems I built. I bought all of the CPU's new from Micro Center after Zen+ was released (most of my Zen 1's were bought in 2019) and I also bought a few after Zen 2 was out. The later manufacture dates might explain why I didn't have the memory issues everybody else seems to have had early on. I always kind of assumed the later BIOS versions were the main factor. A handful of them were using B350/X370 boards but most were using B450 boards with the latest BIOS at the time.

I was even able to daily drive a 2x8 3600 C16 XMP B-die kit at rated speed on an Asus B450 ITX. I'm sure that the ITX board and a really good CPU helped. This was a lottery winning 1600 that did 4.2 Ghz at stock vcore voltage. I did a bunch of gaming and testing on that rig and it was always super stable.

Something that may be a factor in some of the memory issues people had/have is the SOC voltage. Some boards automatically increase the SOC voltage when enabling an XMP profile with higher speeds. This is also usually the first voltage people think to manually increase when memory is unstable. The problem with that is the increased SOC voltage can actually cause memory instability instead of the other way around. This can be true on the later CPU's as well. Decreasing the SOC voltage is how I got the 3600 b-die kit fully stable with XMP on the PC I mentioned above. This may be worth a shot @Zepp.

 

[Dave3000]

final update
got the replacement PSU installed and retested RAM with memtest86
2800 speed and higher still erroring out
2734 speed remains stable
so in my case the dying PSU was not the cause of RAM instability

That does not surprise me since your CPU only supports up to DDR4-2667 according to AMD, https://www.amd.com/en/support/cpu/...-ryzen-7-desktop-processors/amd-ryzen-7-1800x. As long as it runs stable with the memory speed it supports then it's not an issue.

 

[PingSpike]

I have a 1700X that runs 2933 with 2x16GB+2x8GB crucial ballistix. The 2x16GB kit would run the XMP 3200/CL16 fine IIRC but I had to step down to 3000 to get memtest stable with the mixed kits. I then backed it off to 2933 for a little breathing room I guess.

A b-die kit was also fine but hynix stuff was absolutely terrible and could run XMP in 2 stick config so I returned it because crucial e-die was barely more expensive anyway.

I have a 3600 that I could get a ballistix 2x16 3000 kit to run stable with 1T however! Never figured out if it was the kit or the cpu or the board.

Edit: I also have to run idle current: typical on the 1700x or it will get random system freezes.

 

[Zepp]

surprise update: turns out there was a stability problem with my system related to a TPM-WMI error and also possibly a BIOS setting that I corrected and since then I have been able to run 2933 speed memtest stable like my old gen1 ryzen system.