Random Instability 2600x / Vega

[pio]

Okay, so I have replaced everything in my kid's machine, quite literally.  And absolutely nothing has fixed it.

 

The problem:

Random freezing, sound stops, computer unresponsive and display stays on.

The rig specs:

Cooler Master TD500 Mesh v2

Asus Prime B450M-A II (upgraded from ASRock B450m Pro4)

AMD Ryzen 5 2600x (upgraded from Ryzen 2600)

Cooler Master Hyper 212 Halo Black (upgraded from Wraith Prism)

2x16GB Crucial Ballistix DDR4-3200 (have tried 2 different sets of similar 3200 sticks)

Sapphire RX Vega 64 (have used RX 580, GTX 960, GTX 970, and Vega64)

Cooler Master v 850 i Gold PSU (previous eVGA GT650)

 

I'm curious if perhaps having DDR4-3200 at 16-18-18-36-1T is maybe the cause?  I recall there being high speed DDR4 issues on earlier Ryzen systems.  It seems to pass all stability tests though.  All Unigine benchmarks, it'll run through fine.  3d things like stress tests and games seem to make it freeze faster, but its completely random.  It can be 5 minutes in playing a game to hours of playing a game.  Or it can freeze at idle when I'm not paying any attention at all.  P95 and Cinebench lead to no results either, its performing fantastically in Cinebench and P95 seems to run fine for hours.  But after swapping all this hardware out, I'm out of ideas other than maybe a RAM speed compatibility problem or something.  Thoughts?

[bonami2]

1 hour ago, pioneerisloud said:

Okay, so I have replaced everything in my kid's machine, quite literally.  And absolutely nothing has fixed it.

 

The problem:

Random freezing, sound stops, computer unresponsive and display stays on.

The rig specs:

Cooler Master TD500 Mesh v2

Asus Prime B450M-A II (upgraded from ASRock B450m Pro4)

AMD Ryzen 5 2600x (upgraded from Ryzen 2600)

Cooler Master Hyper 212 Halo Black (upgraded from Wraith Prism)

2x16GB Crucial Ballistix DDR4-3200 (have tried 2 different sets of similar 3200 sticks)

Sapphire RX Vega 64 (have used RX 580, GTX 960, GTX 970, and Vega64)

Cooler Master v 850 i Gold PSU (previous eVGA GT650)

 

I'm curious if perhaps having DDR4-3200 at 16-18-18-36-1T is maybe the cause?  I recall there being high speed DDR4 issues on earlier Ryzen systems.  It seems to pass all stability tests though.  All Unigine benchmarks, it'll run through fine.  3d things like stress tests and games seem to make it freeze faster, but its completely random.  It can be 5 minutes in playing a game to hours of playing a game.  Or it can freeze at idle when I'm not paying any attention at all.  P95 and Cinebench lead to no results either, its performing fantastically in Cinebench and P95 seems to run fine for hours.  But after swapping all this hardware out, I'm out of ideas other than maybe a RAM speed compatibility problem or something.  Thoughts?

 

 

MemTest: RAM reliability tester for Windows

HCIDESIGN.COM

 

That the program i use for ram. And seemed to be the most effective.

You need to launch multiple time the app to fill the ram. And the max amount change for some reason everytime you restart the computer.

 

Sometime i run 2.2gb per app other days 3.5gb. No idea why but at least it work great.

Edited March 19, 2023 by bonami2

[ENTERPRISE]

3 hours ago, pioneerisloud said:

Okay, so I have replaced everything in my kid's machine, quite literally.  And absolutely nothing has fixed it.

 

The problem:

Random freezing, sound stops, computer unresponsive and display stays on.

The rig specs:

Cooler Master TD500 Mesh v2

Asus Prime B450M-A II (upgraded from ASRock B450m Pro4)

AMD Ryzen 5 2600x (upgraded from Ryzen 2600)

Cooler Master Hyper 212 Halo Black (upgraded from Wraith Prism)

2x16GB Crucial Ballistix DDR4-3200 (have tried 2 different sets of similar 3200 sticks)

Sapphire RX Vega 64 (have used RX 580, GTX 960, GTX 970, and Vega64)

Cooler Master v 850 i Gold PSU (previous eVGA GT650)

 

I'm curious if perhaps having DDR4-3200 at 16-18-18-36-1T is maybe the cause?  I recall there being high speed DDR4 issues on earlier Ryzen systems.  It seems to pass all stability tests though.  All Unigine benchmarks, it'll run through fine.  3d things like stress tests and games seem to make it freeze faster, but its completely random.  It can be 5 minutes in playing a game to hours of playing a game.  Or it can freeze at idle when I'm not paying any attention at all.  P95 and Cinebench lead to no results either, its performing fantastically in Cinebench and P95 seems to run fine for hours.  But after swapping all this hardware out, I'm out of ideas other than maybe a RAM speed compatibility problem or something.  Thoughts?

I can't see those RAM specs being an issue, even with the older Gen Ryzen. However, have you tried the system running non XMP profile to see if the issue stops ? 

 

I assume everything is running stock when testing? 

 

Lastly, have you tried the latest BIOS or even perhaps flashing back to an older BIOS closer to the later end of the Ryzen 2600X life ? I have seen newer BIOS versions with newer AGESA focus on newer CPU's at the detriment of older CPU's. I have found that to be rare but not impossible.

[Cerberus]

IMO, my ryzen 1x and 2x cpus would not boot stable without anything from the motherboard QVL list, xmp only worked on one set of two that I tried. I'm leaning towards memory.

[pio]

10 hours ago, ENTERPRISE said:

I can't see those RAM specs being an issue, even with the older Gen Ryzen. However, have you tried the system running non XMP profile to see if the issue stops ? 

 

I assume everything is running stock when testing? 

 

Lastly, have you tried the latest BIOS or even perhaps flashing back to an older BIOS closer to the later end of the Ryzen 2600X life ? I have seen newer BIOS versions with newer AGESA focus on newer CPU's at the detriment of older CPU's. I have found that to be rare but not impossible.

I had tried a few different BIOS's on the ASRock board.  I also had tried a variety of settings, and figured it was the board, hence replacing the board.  On the current Asus B450M-AII board its on, I haven't touched the BIOS yet.  It's all defaults / PBO / XMP.

I DID, however, try changing RAM down to 3000 from 3200 last night after making this post.  The machine hasn't crashed as of yet.  So, its made it a full sleep cycle for me without freezing.  That's actually progress!   

[schuck6566]

1 hour ago, pioneerisloud said:

I had tried a few different BIOS's on the ASRock board.  I also had tried a variety of settings, and figured it was the board, hence replacing the board.  On the current Asus B450M-AII board its on, I haven't touched the BIOS yet.  It's all defaults / PBO / XMP.

I DID, however, try changing RAM down to 3000 from 3200 last night after making this post.  The machine hasn't crashed as of yet.  So, its made it a full sleep cycle for me without freezing.  That's actually progress!   

I'm running a 2700x on a x470 board. the bios updates are kinda whack with these including some that have certain extra's needed for 3rd gen ryzen cpu's to be able to use all 4 mem slots right, maybe they also effect ealier cpu's if running later bios without the "added" stuff being run? Including a shot of the relevant bios updates & info for my gigabyte board so U can get an idea of what I mean.

[pio]

1 hour ago, schuck6566 said:

I'm running a 2700x on a x470 board. the bios updates are kinda whack with these including some that have certain extra's needed for 3rd gen ryzen cpu's to be able to use all 4 mem slots right, maybe they also effect ealier cpu's if running later bios without the "added" stuff being run? Including a shot of the relevant bios updates & info for my gigabyte board so U can get an idea of what I mean.

Yeah I haven't checked the Asus BIOS update page as of yet, but the ASRock B450 that I replaced.....finding the right version BIOS for a 2600x was a pain.  I even called ASRock support, and they sent me the "right" BIOS for the chip.  It didn't matter though.

[Sir Beregond]

Might be a dumb question but what is the FCLK set to? I had random freezes and slowdowns when I accidentally decoupled my FCLK from memory frequency. My RAM was at 3733 (1866) but my FCLK was still at 1800. FCLK on auto sure didn't see my OC settings. It should for normal XMP, but just thought I'd ask.

[pio]

8 hours ago, Sir Beregond said:

Might be a dumb question but what is the FCLK set to? I had random freezes and slowdowns when I accidentally decoupled my FCLK from memory frequency. My RAM was at 3733 (1866) but my FCLK was still at 1800. FCLK on auto sure didn't see my OC settings. It should for normal XMP, but just thought I'd ask.

Right now its at 1500 / 1500.  What's weird, the thing has the appearance of stability even at DDR4-3800 / 1800 FCLK, but I'm sure that can't be right on a 2600x.

[ENTERPRISE]

1 hour ago, pioneerisloud said:

Right now its at 1500 / 1500.  What's weird, the thing has the appearance of stability even at DDR4-3800 / 1800 FCLK, but I'm sure that can't be right on a 2600x.

1800 I didn't think was possible on Ryzen 2000. I would be surprised if it was stable. I assume your base clock is also stock ?

[Sir Beregond]

9 hours ago, pioneerisloud said:

Right now its at 1500 / 1500.  What's weird, the thing has the appearance of stability even at DDR4-3800 / 1800 FCLK, but I'm sure that can't be right on a 2600x.

Yeah, Zen+ right? I imagine that would be lower than what Zen 2 or 3 FCLK can do which is why you are probably seeing it more stable at 3000 for the memory.

[StrunkenCap]

Have you tried stock jedec just to see If It's related to memory ? While Zen+ can do +3600 some chips are dud and starts having problems above +3000

[pio]

8 hours ago, ENTERPRISE said:

1800 I didn't think was possible on Ryzen 2000. I would be surprised if it was stable. I assume your base clock is also stock ?

I didn't think 1800 was possible on 2000 series either, but it does it lol.  I was just curious its limitations.  Definitely probably not stable.  1500 FCLK / DDR4-3000 is where its sitting at right now still, I haven't seen it freeze since changing it down to there, but I also haven't been running it as heavily now either.

 

Yes, Base Clock is at stock 100.  Literally was a "PBO and Go" setup, and was originally using XMP for the RAM at 1600 / DDR4-3200 (hence thinking maybe memory was too high).

[pio]

2 minutes ago, StrunkenCap said:

Have you tried stock jedec just to see If It's related to memory ? While Zen+ can do +3600 some chips are dud and starts having problems above +3000

Stock JEDEC would be what, 2933 right?  Yeah, I can plop it down another notch.  That might finish getting it more stable, it does seem knocking it down to 3000 helped.

[Sir Beregond]

2 hours ago, pioneerisloud said:

Stock JEDEC would be what, 2933 right?  Yeah, I can plop it down another notch.  That might finish getting it more stable, it does seem knocking it down to 3000 helped.

Stock JEDEC for DDR4 is 2133 I think.

[StrunkenCap]

3 hours ago, pioneerisloud said:

Stock JEDEC would be what, 2933 right?  Yeah, I can plop it down another notch.  That might finish getting it more stable, it does seem knocking it down to 3000 helped.

Basically what your RAM defaults to without XMP

[pio]

Okay, so far its passed 10 hours of blended p95 (with AVX).  Hasn't frozen yet *knocks on wood.

 

Stock PBO, stock vcore, stock multiplier (so like 3.95GHz all core boost).  RAM at DDR4-3000 / 1500 FCLK, 1.125v SOC.

 

I was trying to get a good overclock, but 4.1GHz was taking near 1.40v and temps were getting a little too high with the Hyper 212 Halo Black on there, so yeah, not overclocking with that cooler.  Figured that was going to be the case, but tried anyway.

 

I'll let this continue running for a little bit longer, probably a nice round 12 hours to be safe.  I guess I'll call it good if it makes it that far, everything at basically bone stock.  Still better than 3.8GHz for a 4c / 8t part, like it was set to before (and still crashing).  Definitely looks like RAM speed and FCLK were the problem though.  XMP profile puts it at 3200 / 1600, and clearly this chip doesn't like that.

Edited March 22, 2023 by pioneerisloud

[ENTERPRISE]

1 hour ago, pioneerisloud said:

Okay, so far its passed 10 hours of blended p95 (with AVX).  Hasn't frozen yet *knocks on wood.

 

Stock PBO, stock vcore, stock multiplier (so like 3.95GHz all core boost).  RAM at DDR4-3000 / 1500 FCLK, 1.125v SOC.

 

I was trying to get a good overclock, but 4.1GHz was taking near 1.40v and temps were getting a little too high with the Hyper 212 Halo Black on there, so yeah, not overclocking with that cooler.  Figured that was going to be the case, but tried anyway.

 

I'll let this continue running for a little bit longer, probably a nice round 12 hours to be safe.  I guess I'll call it good if it makes it that far, everything at basically bone stock.  Still better than 3.8GHz for a 4c / 8t part, like it was set to before (and still crashing).  Definitely looks like RAM speed and FCLK were the problem though.  XMP profile puts it at 3200 / 1600, and clearly this chip doesn't like that.

Glad you are making progress. I thought it might be XMP with respects to pushing speeds that the CPU wasn't going to cope with. Fingers crossed you get to the 12 hour mark and you can call it a day. 

[pio]

3 hours ago, ENTERPRISE said:

Glad you are making progress. I thought it might be XMP with respects to pushing speeds that the CPU wasn't going to cope with. Fingers crossed you get to the 12 hour mark and you can call it a day. 

Yessir, and you were right.  It passed 12 hours earlier today at bone stock PBO / DDR4-3000 / 1500 FCLK.  Have it going now again for another 12 hours with a slight undervolt on the CPU, to bring the all core boost back up over 4GHz.  It "should" pass here (just a measly -0.100v).  It hasn't frozen once since changing these settings last night (before the undervolt).  I think the memory speeds / FCLK were exactly the problem, even though I was just using the XMP / DOCP profile.

[Sir Beregond]

21 hours ago, pioneerisloud said:

Yessir, and you were right.  It passed 12 hours earlier today at bone stock PBO / DDR4-3000 / 1500 FCLK.  Have it going now again for another 12 hours with a slight undervolt on the CPU, to bring the all core boost back up over 4GHz.  It "should" pass here (just a measly -0.100v).  It hasn't frozen once since changing these settings last night (before the undervolt).  I think the memory speeds / FCLK were exactly the problem, even though I was just using the XMP / DOCP profile.

Glad to hear it. Yeah while I am not as familiar with pre-Zen 2/3 Ryzen, I had a feeling that was going to be the case. Each gen had its limits on FCLK and FCLK directly tied in some form to the memory speed. Had a feeling E was on to something with even 3200 being too high for it.

 

Glad that fixed it.

[neurotix]

Look into running stressapptest in Windows. It is a Linux program, and what Google uses to test stability on their servers. An hour of it should uncover any issues. See the memory thread I started in the memory section. Though tbh I run Mint almost 24/7, you need to use Windows Subsystem for Linux to be able to run it in a command line on Windows.

 

Also check into AMD chipset drivers and make sure you have the latest version for your platform. Additionally, especially if the problem just started, look into Windows updates and possibly remove some of the latest updates and then pause updates for over a month and see if that helps.

[pio]

So after some tweaking, I think, THINK its stable.  It's passed over 12 hours P95 blended (with AVX).  Didn't freeze.  4.1GHz all core, and tightened timings a lot tighter than stock.  Downclocked RAM to 3166 instead of 3200 since 3200 results in freezing.

 

Going to let it sit and fold for a few days as is, see how it does.  If it does fine, I'll remove the CPU worker on F@H, and let it run a few more days.  If it passes that....then I'll OC the GPU some.   I think its stable so far though FINALLY!

 

[neurotix]

1 hour ago, pioneerisloud said:

So after some tweaking, I think, THINK its stable.  It's passed over 12 hours P95 blended (with AVX).  Didn't freeze.  4.1GHz all core, and tightened timings a lot tighter than stock.  Downclocked RAM to 3166 instead of 3200 since 3200 results in freezing.

 

Going to let it sit and fold for a few days as is, see how it does.  If it does fine, I'll remove the CPU worker on F@H, and let it run a few more days.  If it passes that....then I'll OC the GPU some.   I think its stable so far though FINALLY!

 

Sounds great, especially if it's for your son.

 

Hope it remains stable when he's using it.