Current RAM is giving me issues, any suggestions for stable RAM for my setup? (Ryzen/ Asus Prime X370 pro) - SOLVED (Not anything to do with RAM)

[Sasbom]

Hello to all!

I'm a 3D artist by trade, and I bought the wrong Ram for my setup, thinking anything would do.

Recently, I discovered a little thing called QVL, and now I feel a little dumb.

 

This is the ram I got:

https://www.gskill.com/en/product/f4-3200c16d-16gtzb

2x16 of Trident Z 3200Mhz cas 16 (no RGB, I'm sorry  )

 

Together with:

Ryzen 1800X (overclocked to 3.8, I'm not going the full 4 because of stability issues)

Asus Prime x370 pro motherboard

Gtx 1080ti (doesn't have anything to do with this, that beast works fine and I love it)

 

The thing is, everytime I run something heavily taxing on the processor (think CPU rendering in Arnold and Vray, or generating point cloud data in Agisoft Photoscan),

the process can run for about 40 minutes max, and then the software crashes.

I've had full on BSoD's before, but thanks to the new BIOS'es they haven't occured as much as they used to.

 

I feel like I waited long enough for the BIOS to catch up, and I simply can't blame the crashes on growing pains anymore. 
I need a stable system that can handle punishingly gruesome tasks for 4, 5 hours straight without crashing, and according to what I know now, it's because of RAM limitations that I have these problems.

 

So, does anyone know of any RAM sets (preferably a set of 2 sticks of 16 Gb ram) that can run smooth and stable-like on the prime x370 pro?

I really don't want to risk being stupid this time, that's why I asked.

 

Groeten uit Nederland,

 

Sasbom

 

(Yes, my power supply handles everything very well , no problems there)
(Yes, I have tried increasing SoC to 1.2, doesn't fix anything  )
 

--

 

HELLO TO ALL FUTURE READERS:

 

If you have the same kind of crashes during long processes in a lot of heavy duty graphics related software, check this link out!
Also check your crash logs using event manager! 
 

https://support.allegorithmic.com/documentation/display/SPDOC/GPU+drivers+crash+with+long+computations

[ouroesa]

What is your temps like?

 

Have you tried running a CPU benchmark for 4-5 hours to ensure stability? Same with memory. What can happen sometimes is the VRM's getting too hot and causing issues.

 

I would start by setting the CPU back to defaults and seeing of that is stable, if not, it could be your memory (try underclocking it to 2800Mhz and see if this fixes the problem), if it is stable then it is your CPU/VRM.

[DarkStang]

I'm running 1.35v on my 1700 to keep it stable at 3.9ghz. You may want to bump your voltage up a bit and make sure you're cooling is adequate for your CPU and VRM.

 

RAM can fail during long intense tasks. But this is rare uh my experience. I think you're looking at a temp or power issue.

 

Edit: I'm running some fairly cheap EVGA 3200 ram without issue. I might be lucky though.

Edited March 21, 2018 by DarkStang
Added ram info

[Sasbom]

5 hours ago, ouroesa said:

What is your temps like?

 

Have you tried running a CPU benchmark for 4-5 hours to ensure stability? Same with memory. What can happen sometimes is the VRM's getting too hot and causing issues.

 

I would start by setting the CPU back to defaults and seeing of that is stable, if not, it could be your memory (try underclocking it to 2800Mhz and see if this fixes the problem), if it is stable then it is your CPU/VRM.

Already tried running it at stock. I got a big Dark Rock 3 CPU fan on there, and temps never go higher then maybe 80 C.

I already have my memory underclocked to 2933, and had it underclocked to 2133 before BIOS updates before, still the crashing has occured.

I haven't been able to run it at the advertised speed, that might say something as well, compabitibility-wise.

 

2 hours ago, DarkStang said:

I'm running 1.35v on my 1700 to keep it stable at 3.9ghz. You may want to bump your voltage up a bit and make sure you're cooling is adequate for your CPU and VRM.

 

RAM can fail during long intense tasks. But this is rare uh my experience. I think you're looking at a temp or power issue.

 

Edit: I'm running some fairly cheap EVGA 3200 ram without issue. I might be lucky though.

 

Holy moly, 1.35 for SoC sounds like a lot of power. I'll test it out tho, maybe I'll get some stability in return. I already have a couple of projects lying around to

benchmark with, which right now produce consistent errors on my workbench while doing OK on similar systems.
I already fed my ram more than enough power to keep it happy, theoretically (sometimes even went as far as 1.5v, which isn't recommended)

 

In my case, my RAM seems to be failing consistently, and when blue screens occur, it's a different message each time. I have acces to workstations that have

the same specs as mine, exept that they run intel. They never crash, they just take a long time to finish these arduous tasks (which is totally normal for rendering and the like).

I really think it's a RAM issue, as Ryzen has been known for some time to be quite picky, and it doesn't help that the Asus prime x370 pro also has a bit of a difficult time.

[DarkStang]

9 hours ago, Sasbom said:

Holy moly, 1.35 for SoC sounds like a lot of power. I'll test it out tho, maybe I'll get some stability in return. I already have a couple of projects lying around to

benchmark with, which right now produce consistent errors on my workbench while doing OK on similar systems.
I already fed my ram more than enough power to keep it happy, theoretically (sometimes even went as far as 1.5v, which isn't recommended)

I think the max safe voltage is 1.45, so you've got some room to play with. Just make sure your heat doesn't run away.

If you really think RAM is the problem you can run memtest overnight and see if it reports any issues. (https://www.memtest86.com/download.htm)

[JR88]

Some corsair set is probably on the QVL. 

[blake86]

21 hours ago, Sasbom said:

Hello to all!

I'm a 3D artist by trade, and I bought the wrong Ram for my setup, thinking anything would do.

Recently, I discovered a little thing called QVL, and now I feel a little dumb.

 

This is the ram I got:

https://www.gskill.com/en/product/f4-3200c16d-16gtzb

2x16 of Trident Z 3200Mhz cas 16 (no RGB, I'm sorry  )

 

Together with:

Ryzen 1800X (overclocked to 3.8, I'm not going the full 4 because of stability issues)

Asus Prime x370 pro motherboard

Gtx 1080ti (doesn't have anything to do with this, that beast works fine and I love it)

 

The thing is, everytime I run something heavily taxing on the processor (think CPU rendering in Arnold and Vray, or generating point cloud data in Agisoft Photoscan),

the process can run for about 40 minutes max, and then the software crashes.

I've had full on BSoD's before, but thanks to the new BIOS'es they haven't occured as much as they used to.

 

I feel like I waited long enough for the BIOS to catch up, and I simply can't blame the crashes on growing pains anymore. 
I need a stable system that can handle punishingly gruesome tasks for 4, 5 hours straight without crashing, and according to what I know now, it's because of RAM limitations that I have these problems.

 

So, does anyone know of any RAM sets (preferably a set of 2 sticks of 16 Gb ram) that can run smooth and stable-like on the prime x370 pro?

I really don't want to risk being stupid this time, that's why I asked.

 

Groeten uit Nederland,

 

Sasbom

 

(Yes, my power supply handles everything very well , no problems there)
(Yes, I have tried increasing SoC to 1.2, doesn't fix anything  )

ryzen doesen't like very much dual rank memory stick, what your kit 99.9% probably is. it is almost impossible to go higher than 2900 mhz. What 's the speed you set in bios?

Edit: Ok you already have downclock. So you can do just one thing. Take off one of your stik and try if the problem persist. If yes try the second stick alone. If you still have problem, most probably it s not a ram failure, but the problem is somewhere else

[Sasbom]

I've currently got it on 2933Mhz, 

 

Here's what I'll do this Friday, currently away from my main workstation:

only use 1 RAM stick, up the SoC to 1.2 again, and run memtest on 2133 Mhz, which is what the BIOS defaults it to.

If that goes well, I'll go and run some actual benchmarks with CPU rendering and mesh processing.

 

This is my current setup:

 

PSU: dark power pro 650W

mobo: Asus Prime x370 pro (on the 5 - 15 bios - supposedly the most stable one yet)

RAM: G.Skill trident z 3200Mhz (2 sticks totalling 32 Gb) @2933 Mhz

CPU: Ryzen 1800x @3,8GHz

GPU: Aorus GTX 1080ti Extreme edition

 

Could it be a power issue? I reckoned 650W would be quite up to the task of handeling my system but now I'm beginning to doubt that again.

 

[ouroesa]

7 minutes ago, Sasbom said:

I've currently got it on 2933Mhz, 

 

Here's what I'll do this Friday, currently away from my main workstation:

only use 1 RAM stick, up the SoC to 1.2 again, and run memtest on 2133 Mhz, which is what the BIOS defaults it to.

If that goes well, I'll go and run some actual benchmarks with CPU rendering and mesh processing.

 

This is my current setup:

 

PSU: dark power pro 650W

mobo: Asus Prime x370 pro (on the 5 - 15 bios - supposedly the most stable one yet)

RAM: G.Skill trident z 3200Mhz (2 sticks totalling 32 Gb) @2933 Mhz

CPU: Ryzen 1800x @3,8GHz

GPU: Aorus GTX 1080ti Extreme edition

 

Could it be a power issue? I reckoned 650W would be quite up to the task of handeling my system but now I'm beginning to doubt that again.

 

That PSU should me more than sufficient, you can calculate your power usage here http://www.coolermaster.com/power-supply-calculator/

 

That said, the PSU might be defective. You can test this by running a full system stability test for maximum power draw. If it crashes (which is should do almost instantly) then the PSU might be at fault. 

 

Do you have access to any other machine where you can test some of the hardware? Easiest route as you can eliminate hardware issues one at a time. 

 

Good luck! 

[Sasbom]

Just now, ouroesa said:

That PSU should me more than sufficient, you can calculate your power usage here http://www.coolermaster.com/power-supply-calculator/

 

That said, the PSU might be defective. You can test this by running a full system stability test for maximum power draw. If it crashes (which is should do almost instantly) then the PSU might be at fault. 

 

Do you have access to any other machine where you can test some of the hardware? Easiest route as you can eliminate hardware issues one at a time. 

 

Good luck! 

Currently not in a situation where I can swap hardware unfortunatly. Max power draw has never crashed it immediatly, 

some VERY demanding tasks always last for about 40 minutes until giving up (using real world applications, not benchmarks).

 

Thanks for the great advice. I'll go and torture my PC this Friday/Saturday to see what's wrong.

[Sasbom]

BUMP

 

holy guacamoly, turns out the SoC definetly could use a bit more juice, bumped it up to 1.2, and the DRAM to 1.5.
I thought I tried turning up the SoC, but that was definitely not the case.

I've been running the dense cloud generation process on ultra high in Agisoft photoscan with 60 camera's (a lot for the stuff I typically do),
and the PC has been running at max capacity for about 50 minutes now.

We'll see if it holds up, and if it does, it looks like the problem might have been solved. Who knew I was depriving my CPU of precious voltage?
I'll update when the process has finished. It looks like it might take a little while, so I'll be off nervously playing the piano or something.

 

UPDATE:
More than an hour in, looks like SoC was definitely underfed. I don't get why the whole Ryzen experience has proven so difficult for me but
it sure is educational.

UPDATE 2:
It survived the depth calculating part of the operation, but then, after 2 hours, 15 minutes into the actual generation of the dense cloud, it soft-crashed again.
This time however, I think its my fault, I was running chrome next to Photoscan.

I'll run memtest tonight!

[Sasbom]

BUMP 2:

Allright new insights have emerged. I ruled out that memory was the issue. Turns out, that intensive tasks that involve at least some form of graphical manipulation tend to suffer

crashes because of a low TDR time. The software can sometimes "hang" during a computation, which might confuse the gpu driver. 

I also suffered the same problems when rendering with my CPU (think Vray and Arnold), which I now know are due to the same problem.

 

If you suffer the same problems, I implore you to check out this guide, made by the substance company. I actually use substance painter too, which at start up

will show you a warning if TDR and TDRDDRI is set too low. I always ignored the message, but I shouldn't have.

https://support.allegorithmic.com/documentation/display/SPDOC/GPU+drivers+crash+with+long+computations

 

I found out by using Event manager, and looking up the error codes. Apparantly, the issues I'd been having aren't to do with my RAM memory or CPU at all!
I always found it weird that my system would stress test OK, that's why I always use my own projects that fail due to crashes (that render fine on other PC's might I add).

 

The bump in performance I got out of learning about and tweaking my system even further has been quite neat. The reason I'd been getting further into the processes I tested with was probably due to the hanging time decreasing.

EDIT: I just tested it out, and the same scene on the same settings that has given me so much errors and crashes now finally processes completely.

Here is a piece of wall I scanned to use for texturework, which I had tremendous trouble with. This is a 17 million dot Point cloud, processed from a set of 60 photo's. I'm glad everything works now.