Help with stress testing Ryzen 2200G and general RAM OC questions.

[bengalih]

System Specs:

- AMD Ryzen 3 2200G
- ASRock AB350 Pro4 ATX mobo
- G.SKILL F4-2666C15D-8GVR (DDR4 2666\Timing 15-15-15-35\CAS Latency 15\Voltage 1.2V)

Background:
Was unable to properly initialize Ryzen video drivers using anything above the 2400 XMP profile.
Based on some advice here I disabled Gear Down Mode and used even timings on the RAM.

Currently I'm running with the following RAM config:
- 2933 Mhz
- 14-14-16-32
- Gear Down disabled
- Command Rate 1T
- DRAM voltage 1.35

As stated I came to these values just based on some info here and a few hours learning about tuning yesterday and some trial and error.
The Ryzen DRAM calculator gives me 14-14-14-46 and 1.365 for *FAST* settings, but I seem not to be able to get tRP any tighter than 16.

With my current settings I ran memtest86 for a few passes and all seemed ok.
I then ran the blend test in Prime95 and it seemingly rebooted my computer after about 30-40 minutes.
I then sat there while it ran again and saw it BSOD the system after about 15-20.

I installed the Ryzen Master application so I could check temps.
I then ran OCCT Linpack test for just under an hour and nothing happened.
CPU temperatures seemed to never get higher than about 75.

I then started up Prime95 again and it has been running for 2.5 hours without issue.
I am seeing CPU temps as high as 83 (and not getting much lower than 79).

Questions: (quite a few, please forgive my newbness...)

1) If my RAM is rated as 2666@15-15-15-35 is running at 2933@14-14-16-32 too ambitious?

2) What is the effective result/consequences of me putting my tRAS so low? The RAM is rated for 35 but I am at 32 and DRAM calc says I should do 46 (even with lower timings). I know the textbook definitions of these terms, but I don't really know how numbers between 32-46 will actually affect the system and how much tRAS itself matters toward performance and stability.

3) Is it strange that Prime95 crashed 2-3 times for me under 45 minutes and now it is running stable for 2.5 hours? If I let it run 12-24 hours and everything is stable how do I explain the 3 prior crashes when nothing else has changed?

4) If I do get more crashes, what additional things might I try to change other than lowering the frequency or loosening the timings? Can raising the voltage a bit more really help here?

5) Is my CPU temp too high at ~83 ?

6) What other tests can I run to ensure stability? Right now I'm just working on RAM. I don't know if I plan to OC the CPU/GPU so now I'm just looking for the best results on the RAM.

7) I have some Patriot PV48G300C6K RAM coming in tomorrow. It is DDR4 3000 16-16-16-36. Basically the same cost as the G.SKILL I've got. Since this RAM is rated up to 3000 instead of the 2666 G.SKILL, would this seem like the better option? How likely is it that the G.SKILL will actually OC better than the superior rated Patriot? I realize there may not be a definitive answer, but it seems like if I'm spending all this time getting the G.SKILL to 2933 and the Patriot is already rated at 3000, then its a no-brainer (?).

Please feel free to give me any other general advice here of things I may be overlooking. I feel I've learned a lot in the past day just reading about these methods, but I know its still just a drop in the bucket.

 

[wingman99]

A good memory test is running 4 instances of MemTest HCI With the total amount of available memory you see in task manger divided into the 4 instances of MemTest HCI.

When you restart the PC the secondary auto memory training will change sometimes giving a failed test and then with a reboot a good test. You can try lowering your memory speed when this happens.

 

[bengalih]

I think I disabled the memory training. Is it recommended to keep it on?

I'll try the HCI test. I'm still concerned about the Prime95 testing. I had 3 failures within 20 to an hour my first few times running the blend test.
After that however I ran the blend test for like 12 hours without issue and now a custom RAM test on Prime95 has been going on 8 hours with no issue.

Is it possible that the components go through some sort of burn-in period where they adjust? I just don't have enough experience to explain why I got 3 failures and now it seems smooth sailing.

They only things that changed between the bad and good tests were:
- I ran OCCT for about an hour
- I installed Ryzen Master (but didn't use it to adjust any settings - just installed to monitor temps).


?

 

[wingman99]

When you boot up the PC the motherboard will train the secondary timings that are on AUTO. Don't disable AUTO secondary memory timings training or the PC might not POST. You can set secondary memory timings manually. Folks and I have had failed testing with one boot up of the AUTO secondary memory training, then other boot ups no problem testing memory because of good AUTO memory training. Sometimes it takes a lot of boot ups to have a failed memory test from AUTO secondary memory timing failure. Memory or PC does no burn in, only burn out of specifications. I usually reboot 1-20 times running memtest HCI 30% each time to see if my PC will support the memory speed I selected with AUTO secondary memory timings.

 

[bengalih]

When you boot up the PC the motherboard will train the secondary timings that are on AUTO. Don't disable AUTO secondary memory timings training or the PC might not POST. You can set secondary memory timings manually. Folks and I have had failed testing with one boot up of the AUTO secondary memory training, then other boot ups no problem testing memory because of good AUTO memory training. Sometimes it takes a lot of boot ups to have a failed memory test from AUTO secondary memory timing failure. Memory or PC does no burn in, only burn out of specifications. I usually reboot 1-20 times running memtest HCI 30% each time to see if my PC will support the memory speed I selected with AUTO secondary memory timings.

Ok...but how exactly does it know which secondary settings to keep and where do you see these settings?

For instance - if on your 12th boot-up HCI passes the test which it failed the other 11 times (due to the secondary memory timings being different) - what is to prevent the computer from changing those timings on the next boot-up (and the next, and the next, ...) ??

 

[wingman99]

Ryzen Timing Checker LINK: https://www.techpowerup.com/download/ryzen-timing-checker/ The auto secondary timings the motherboard sets from what I have seen an experienced does not effect stability after reboot when setting memory speed lower than the speed that has the stability problems with rebooting. There is less tolerance when running memory faster. Also better motherboards have less rebooting stability problems when running the memory speed faster than 2666+ speed.