Need help with memory tuning on ryzen

[Kuzoh]

I have a 5800x with a kit of 3600mhz CJR ram on a b550m mortar mobo, been using it fine for past months but it now isn't stable with the rated memory speed anymore and it's causing my PC to shut down under load and only load it doesn't happen on idle or stress test, it's very hard to pin down this I came to thinking one of two possibilities: 

1. Memory stick are no longer able to work on 3600 rated speed which I want to test and get it out of the way first
2. Bad CPU being degraded/defective in someway I already replaced the PSU but it wasn't that it still turns off during load.

I did turn off the XMP in bios and  now no more crashed under load but at 2133mhz on ryzen isn't a good experience, I want to set the frequency to 3200 at least and see if it's stable since that is considered JEDEC speed for Zen 3. I don't have another system to test this in so I can't switch out my CPU/RAM to a different mobo. 

I want to ask what timings CJR would be stable at since want to turn it down to 3200mhz from 3600mhz. Would it be okay to do XMP in bios but manually drop the frequency down to 3200mhz or should I just manually do everything with much looser timings? I also will attach a zentimings screenshot that it shows with XMP enabled if that helps. Again I just want to do 3200mhz not being sure if it will work right on the XMP timings for 3600. 

 

 

 

[WereCat]

These are very loose timings. I doubt this is causing instability. 

 

You either need to bump up the DRAM Current Capability a bit or tweak voltages. 

 

VDDP voltage is way too high and this may have caused some degradation though it's just a possibility not a sure thing. 

You shouldn't really need more than 0.9V to 0.93V at most usually at those DRAM speeds. 

 

VDDG is fine but also kinda too high. 

 

You may need to bump up ProcODT for stability but that may also then require higher DRAM voltage. 

 

Edit

 

Have you actually done some RAM stability testing? 

It's possible the crashes are caused by something else. 

[Kuzoh]

7 hours ago, WereCat said:

Have you actually done some RAM stability testing? 

It's possible the crashes are caused by something else. 

@WereCat Yes I have done TM5 before and it passed without errors that is the thing, it's hard for me to tell what's causing all this to happen, I done stress tests in 3dmark, geekbench, P95, ADIA, even tested the GPU in another system intel based and it didn't lock up the pc or crash it so I ruled out the GPU. I just don't understand why it's stable in everything else but games. Literally, don't know what else I should try other than the ram or bios update if none of them work then I guess it's CPU, I wish there was a CPU tool like intel has that could give a solid answer if the memory controller is bad, or something else.

[WereCat]

1 hour ago, Kuzoh said:

@WereCat Yes I have done TM5 before and it passed without errors that is the thing, it's hard for me to tell what's causing all this to happen, I done stress tests in 3dmark, geekbench, P95, ADIA, even tested the GPU in another system intel based and it didn't lock up the pc or crash it so I ruled out the GPU. I just don't understand why it's stable in everything else but games. Literally, don't know what else I should try other than the ram or bios update if none of them work then I guess it's CPU, I wish there was a CPU tool like intel has that could give a solid answer if the memory controller is bad, or something else.

Have you tried the DRAM Current Capability increase? (if that's even available on MSI boards). My PC will randomly shut down when playing games if I don't do this with my RAM setup even at 2133MT/s. But at 110% from default 100% I can even OC up to 3800MT/s without issues.

 

As I mentioned, try to lower that VDDP, it's possible that this is also root of the issue as it's too close to SOC voltage and VDDP and VDDG should always be lower than SOC. Try 0.930V.

 

Try setting CPU Load Line Calibration to Mode 3, it's possible that voltage droop is causing CPU instability in some cases. (I don't know how the levels are on MSI boards, on ASUS the Level 3 is the best for day to day use and on Gigabyte they have reverse order so you may need to experiment a bit just don't go to any extreme settings, I believe Mode 3 should be fine).

 

Disable Curve Optimizer if you've tried it or any negative voltage offsets.

 

[Kuzoh]

2 hours ago, WereCat said:

Have you tried the DRAM Current Capability increase? (if that's even available on MSI boards). My PC will randomly shut down when playing games if I don't do this with my RAM setup even at 2133MT/s. But at 110% from default 100% I can even OC up to 3800MT/s without issues.

 

As I mentioned, try to lower that VDDP, it's possible that this is also root of the issue as it's too close to SOC voltage and VDDP and VDDG should always be lower than SOC. Try 0.930V.

 

Try setting CPU Load Line Calibration to Mode 3, it's possible that voltage droop is causing CPU instability in some cases. (I don't know how the levels are on MSI boards, on ASUS the Level 3 is the best for day to day use and on Gigabyte they have reverse order so you may need to experiment a bit just don't go to any extreme settings, I believe Mode 3 should be fine).

 

Disable Curve Optimizer if you've tried it or any negative voltage offsets.

 

 

I haven't seen that but I'll need to see if I have that option. MSI bios has been a little buggy imo since I have to manually select the XMP profile on the sub menu, the big XMP red button doesn't actually change anything lol (thanks msi!) 

I'll try out your advice, I did raise DRAM to 1.4v just to see if that has any effect on it. Hoping it doesn't shut down on load and this can be all good again. I disabled all PBO in bios completely as I didn't want auto to set it to anything dumb.

This is what XMP set the ram to under 3200mhz, please tell me if anything is unsafe for 24/7. 

[WereCat]

21 minutes ago, Kuzoh said:

 

I haven't seen that but I'll need to see if I have that option. MSI bios has been a little buggy imo since I have to manually select the XMP profile on the sub menu, the big XMP red button doesn't actually change anything lol (thanks msi!) 

I'll try out your advice, I did raise DRAM to 1.4v just to see if that has any effect on it. Hoping it doesn't shut down on load and this can be all good again. I disabled all PBO in bios completely as I didn't want auto to set it to anything dumb.

This is what XMP set the ram to under 3200mhz, please tell me if anything is unsafe for 24/7. 

This looks fine. Still quite loose timings but more than reasonable voltages at least. 

 

I suggest you to stick with this for few days and see if you run into issues. If not then we can try to start tweaking the DRAM speed and timings. 

[Kuzoh]

Just wanted to ask also does this mean anything bad? I keep seeing them pop up in event viewer kernel power 55 and 172

 

Processor 11 in group 0 exposes the following power management capabilities:

Idle state type: ACPI Idle (C) States (2 state(s))

Performance state type: ACPI Collaborative Processor Performance Control
Nominal Frequency (MHz): 3801
Maximum performance percentage: 143
Minimum performance percentage: 46
Minimum throttle percentage: 14

 

[Kuzoh]

@WereCat I crashed after testing it today in a game it doesn't look good as much as I would've hoped for took 11 mins in game this time before it was 3. 

[WereCat]

32 minutes ago, Kuzoh said:

@WereCat I crashed after testing it today in a game it doesn't look good as much as I would've hoped for took 11 mins in game this time before it was 3. 

Damn, it's possible there is a HW being faulty. 

Can you try each stick in each slot separately? 

[Kuzoh]

1 minute ago, WereCat said:

Damn, it's possible there is a HW being faulty. 

Can you try each stick in each slot separately? 

 

Testmem5 or TM5 already even on this frequency and 3600mhz both pass even AIDA memory stress testing too. It doesn't seem to be the ram I seriously don't know what else to try at this point.

[WereCat]

32 minutes ago, Kuzoh said:

 

Testmem5 or TM5 already even on this frequency and 3600mhz both pass even AIDA memory stress testing too. It doesn't seem to be the ram I seriously don't know what else to try at this point.

Try Furmark to stress your GPU then. 

[Kuzoh]

2 hours ago, WereCat said:

Try Furmark to stress your GPU then. 

How long should I let it run for? 

I used it for almost half and hour nothing bad is happening, what am I looking for?

[WereCat]

3 hours ago, Kuzoh said:

How long should I let it run for? 

I used it for almost half and hour nothing bad is happening, what am I looking for?

Just a test if its GPU at fault. Furmark is a power hog so if GPU temp or VRM is bad it would crash the system. 30min should be enough

[Kuzoh]

@WereCat It's definitely not the GPU, I tested it without xmp last night 2133mhz is all this chip can do without rebooting in games  

[WereCat]

38 minutes ago, Kuzoh said:

@WereCat It's definitely not the GPU, I tested it without xmp last night 2133mhz is all this chip can do without rebooting in games  

Then it's possible the memory controller on CPU is busted. 

 

Try OCCT extreme with large FFT. 

Try y-cruncher FFT for 2h.