MemTest64 crashes even though other memory tests work

[Segundoh]

I’m using a G.Skill Flare X5 6000 CL36 with EXPO enabled. Build: https://pcpartpicker.com/list/fpfDLs
I ran MemTest64 with Task Manager and HWiNFO. Both TestMem5 and Task Manager crashed. I also got an error message later saying "MemTest64.exe - System Error: Exception Processing Message 0xc0000005 - Unexpected parameters." I had run it for 8 hours at this point, and there weren't actually any memory errors reported in MemTest64 even with this message. There was one time my entire screen went black for some reason. I'm not sure if there were any memory errors here since I couldn't see anything. Here's a pic of that:

https://imgur.com/a/uVxXqMx

I'm also getting error popups when I try to open other programs, but I guess that's expected if I have 100% CPU usage and almost 100% RAM usage.

I troubleshooted by running other tests. The other tests I ran include 24 hours of HCI MemTest, 4 passes of MemTest86 (I don't have the paid version), 3 cycles each of TestMem5 Absolut, Extreme, and 1usmus configurations (I'll run more cycles though), 1 hour of OCCT memory stress tests, AIDA64 memory benchmarks, and 24 hours of Y-Cruncher stress tests which cycle between all of the available tests. I tried to do 30 GB in Linpack Xtreme, but that made my SSD usage jump up to 100%; from what I understand from running HCI MemTest, that's not a very accurate memory test then. How should I run that then? I'll try to install Linux to use GSAT, and maybe I'll pay for Karhu. Additionally, are there any other memory stress tests or benchmarks I should be running?

I've only run into issues with MemTest64; is something just wrong with the program itself then, or is my RAM unstable? Could this be a sign of a bad motherboard, CPU, or other bad parts?

 

[EarthDog]

I can't see anything in that image... it's blurry. Also, host it here, please, so it doesn't go away in time... we're drag and drop.

So, I see a bunch of testing with one/most applications testing successfully... but, have you actually tried to use your PC like you normally would? Is it stable FOR YOUR USES (the important bar to cross)?

 

[Segundoh]

I can't see anything in that image... it's blurry. Also, host it here, please, so it doesn't go away in time... we're drag and drop.

So, I see a bunch of testing with one/most applications testing successfully... but, have you actually tried to use your PC like you normally would? Is it stable FOR YOUR USES (the important bar to cross)?

I’ll take another picture if that happens again and I’ll drag and drop that. It seems to stable for my uses. It works fine for general browsing purposes like using Edge and Firefox. It seems to work in games too as I haven’t noticed any issues with those. I’m thinking of doing some game development, streaming, and video editing at some point, so I’ll try that out.

To repeat what I said in the other thread, it also doesn’t boot up when I disable Power Down mode for some reason. Maybe it’s different for DDR5 RAM, but I thought it was generally a good idea to disable that. If I disable that, my PC doesn’t boot up. Is that a RAM, mobo, or BIOS issue? It also doesn’t work I enable memory context restore and EXPO, though that’s more understandable.

 

[EarthDog]

I don't even know what that is....never heard about disabling that (what is it supposed to benefit?). Leave it enabled if it's making your system unstable.

 

[Segundoh]

I don't even know what that is....never heard about disabling that (what is it supposed to benefit?). Leave it enabled if it's making your system unstable.

That’s fair. I’ll stick to this thread then. I forgot to say earlier that I’m not sure why EXPO through AMD Advanced Options is making my PC fail to boot up then. It seems to be changing more values, so that might be part of it. I’ll just stick to the Ai Tweaker EXPO then.

As for Power Down mode, it’s supposed to prevent my RAM from powering down so it’s always on. I can see how that would cause issues though it did work on my AM4 system with DDR4 RAM. I’m worried that means my memory or mobo is bad, but I suppose it’s not if it’s working stably otherwise; it’s likelier that it’s a BIOS issue since the BIOS versions have been terrible for my mobo. It is stable as long as I don’t change these, so it all seems fine.

 

[EarthDog]

Seems like it's advice similar to the other three (or more) forums you've already posed this question to.

 

[Segundoh]

Seems like it's advice similar to the other three (or more) forums you've already posed this question to.

That’s fair lol. I’m trying to get as much advice as I can before the return period for my RAM expires. I found out too late that AM5 ASUS boards are bad, and it’s outside of the return period for that. Oh well, I guess that happens sometimes though. Thanks for the help!

 

[Woomack]

The error in the first post suggest it can be also a memtest86 error, so it's hard to say if it's RAM. The best is to run the latest bootable memtest86 version. Windows software sometimes shows problems that don't exist.

Memory context restore option causes the motherboard to skip additional training when it's not required. In theory, it shouldn't be required at all after the motherboard train RAM settings and saves them. For some reason, AMD motherboards do that anyway, each time you start/restart the PC. I saw some mixed comments and looks like on some motherboards this specific memory context option not always works as expected. On the other hand, how many times you start/restart your PC so it has to be enabled?

ASUS mobos have the best BIOS support in the last few years. However, they focus much more on Strix and Maximus/Crosshair series. TUF is their "budget gaming" series and sometimes are problems. If you picked any other brand then you could have the same or worse problems. For me ASUS Strix/Crosshair mobos are the best for AMD. ASRock is sometimes acting weird. Gigabyte was still fine but not better than ASUS. I had no chance to test MSI. All other brands, like Biostar, were meh.

I'm typically testing RAM in the easy way:
1. run bootable memtest86 - last ISO, so it supports new chipsets / one full pass is enough
2. run AIDA64 or anything else that loads RAM to ~95%, 7-Zip benchmark or Prime95 is good too
3. run mixed load tests with 3D, for that 3DMark stability test is surprisingly good

There are many other programs, but the above works for me.

If you test RAM at EXPO/XMP, then the best is to clear CMOS, enable the profile and nothing else. Enabling/disabling PBO sometimes gives different results too. On ASRock, enabled PBO forces some voltages to run at different values than you set manually. On ASUS it shouldn't act this way, but who knows what's on TUF mobos.

 

[Segundoh]

The error in the first post suggest it can be also a memtest86 error, so it's hard to say if it's RAM. The best is to run the latest bootable memtest86 version. Windows software sometimes shows problems that don't exist.

Memory context restore option causes the motherboard to skip additional training when it's not required. In theory, it shouldn't be required at all after the motherboard train RAM settings and saves them. For some reason, AMD motherboards do that anyway, each time you start/restart the PC. I saw some mixed comments and looks like on some motherboards this specific memory context option not always works as expected. On the other hand, how many times you start/restart your PC so it has to be enabled?

ASUS mobos have the best BIOS support in the last few years. However, they focus much more on Strix and Maximus/Crosshair series. TUF is their "budget gaming" series and sometimes are problems. If you picked any other brand then you could have the same or worse problems. For me ASUS Strix/Crosshair mobos are the best for AMD. ASRock is sometimes acting weird. Gigabyte was still fine but not better than ASUS. I had no chance to test MSI. All other brands, like Biostar, were meh.

I'm typically testing RAM in the easy way:
1. run bootable memtest86 - last ISO, so it supports new chipsets / one full pass is enough
2. run AIDA64 or anything else that loads RAM to ~95%, 7-Zip benchmark or Prime95 is good too
3. run mixed load tests with 3D, for that 3DMark stability test is surprisingly good

There are many other programs, but the above works for me.

If you test RAM at EXPO/XMP, then the best is to clear CMOS, enable the profile and nothing else. Enabling/disabling PBO sometimes gives different results too. On ASRock, enabled PBO forces some voltages to run at different values than you set manually. On ASUS it shouldn't act this way, but who knows what's on TUF mobos.

Sorry, what error in the first post are you discussing? I might have misspoken in my first post; MemTest86 passed with no errors after the default number of passes which I believe is 4 passes. And, I see, so Windows programs aren't the most reliable. I won't worry about MemTest64 then.

The memory context restore thing makes sense too. Maybe I'm just not giving it enough time to train and save my RAM settings before I enable Memory Context Restore? I think I gave it like a month the first time I tried MCT, so I'm not sure if that's correct. It seems that Memory Context Restore is not enabled for the first reboot since my boot time is still 30-40 seconds. Then, if I reboot it again, my PC fails to boot up.

I see, so every mobo manufacturer has issues to an extent. I looked up this issue though, and it seems people are having issues enabling Memory Context Restore even with Crosshair and Strix mobos though I haven't heard anything about Maximus mobos. I hope they'll eventually bring their BIOS fixes to TUF GAMING mobos as well.

1. I ran the free version of MemTest86 for 4 cycles, and that passed. I'm not sure if I ran all 11 tests, but I believe I did if that was the default settings; I didn't change the default settings at all.
2. Which AIDA64 benchmark should I be running? I ran the CPU and memory ones. There's apparently a good one called a cache and memory benchmark, but I can't find that in the newest version of AIDA64 Extreme. I ran the 7-Zip benchmark and Prime95 Blend alongside FurMark for about 2-3 hours, and that seemed fine.
3. What is the 3DMark stability test? Do I have to pay for that? I only have the demo with Time Spy, Fire Strike, and Night Raid.

I realized yesterday that setting PBO to Enabled alongside disabling Power Down mode makes my PC fail to boot up. I read that Auto and Enabled are basically the same thing for PBO, but it seems they're not exactly the same thing; still, sicne theyre the same and my CPU is indeed boosting to 5.13GHz, I guess I'll just go with Auto for now. I'll have to test these two alongside memory context restore. I've tested EXPO on its own for about a month, and I haven't noticed any issues with that.

 

[EarthDog]

An aside.........

May I ask why you want these things enabled? I've never heard of most of these items, yet, you're pushing to enable them as if they are the next best thing since sliced bread, lol. Do you know what they do or are we just reading something from somewhere (where, link?!) and wanting to copy it (that's OK too, just trying to figure out the point in all of this and if it helps other users or is it just a one off)?

Worth noting, I have the latest version of AIDA64 Engineer, and in the stress testing section, it has the usual suspects which have been there for years (CPU, FPU, Cache, System Memory, local disks, GPUs). But if you're testing memory, you should only check off system memory and cache.

 

[Woomack]

Sorry, what error in the first post are you discussing? I might have misspoken in my first post; MemTest86 passed with no errors after the default number of passes which I believe is 4 passes. And, I see, so Windows programs aren't the most reliable. I won't worry about MemTest64 then.

You said you had an error like below:

I ran MemTest64 with Task Manager and HWiNFO. Both TestMem5 and Task Manager crashed. I also got an error message later saying "MemTest64.exe - System Error: Exception Processing Message 0xc0000005 - Unexpected parameters."

Windows programs run with everything else in the background. They're not bad, but sometimes errors can be caused by some other software or a driver. When you run memtest86 directly from a bootable storage then it's not using anything else. In the end, you want to test it in Windows. However, to be sure it's RAM, it's easier to check it outside of the OS.

The memory context restore thing makes sense too. Maybe I'm just not giving it enough time to train and save my RAM settings before I enable Memory Context Restore? I think I gave it like a month the first time I tried MCT, so I'm not sure if that's correct. It seems that Memory Context Restore is not enabled for the first reboot since my boot time is still 30-40 seconds. Then, if I reboot it again, my PC fails to boot up.

You can't change the way how the motherboard is training RAM, unless in BIOS are additional options. Typically, you don't have to and even shouldn't change anything. MCR is a new option and if I'm right it wasn't even tested on all motherboards as vendors started to work on adding it to a standard boot procedure some months after motherboards' release. I don't remember where, but I saw an article or larger post about it.

I see, so every mobo manufacturer has issues to an extent. I looked up this issue though, and it seems people are having issues enabling Memory Context Restore even with Crosshair and Strix mobos though I haven't heard anything about Maximus mobos. I hope they'll eventually bring their BIOS fixes to TUF GAMING mobos as well.

Maximus is for Intel. I just said it because higher ROG series motherboards have much better BIOS support and significant updates are for much longer. Most motherboards don't have any BIOS updates after about 2-3 months after release. The only exception is new CPU support or something similar. Top gaming/OC series motherboards have optimizations/OC and RAM tuning up to 8-12 months and sometimes beta versions with some unique options are even later.

1. I ran the free version of MemTest86 for 4 cycles, and that passed. I'm not sure if I ran all 11 tests, but I believe I did if that was the default settings; I didn't change the default settings at all.
2. Which AIDA64 benchmark should I be running? I ran the CPU and memory ones. There's apparently a good one called a cache and memory benchmark, but I can't find that in the newest version of AIDA64 Extreme. I ran the 7-Zip benchmark and Prime95 Blend alongside FurMark for about 2-3 hours, and that seemed fine.
3. What is the 3DMark stability test? Do I have to pay for that? I only have the demo with Time Spy, Fire Strike, and Night Raid.

Memtest86 has to make one full pass, but I'm almost sure that some tests just take time and are not required.
AIDA64 has a stability test. You pick memory and cache tests only so it loads RAM to ~95% and also loads memory controller. Don't pick all tests as it will cause the CPU to heat up more, but the RAM/IMC load won't be as high as with these 2 mentioned tests only.
3DMark stability tests required a paid version. I just said it as an example, and you don't have to use it and pay for it. It covers stability in games as often RAM passes stability tests, but still crashes in some specific games. You can use many other things.

I realized yesterday that setting PBO to Enabled alongside disabling Power Down mode makes my PC fail to boot up. I read that Auto and Enabled are basically the same thing for PBO, but it seems they're not exactly the same thing; still, sicne theyre the same and my CPU is indeed boosting to 5.13GHz, I guess I'll just go with Auto for now. I'll have to test these two alongside memory context restore. I've tested EXPO on its own for about a month, and I haven't noticed any issues with that.

On ASUS, auto was acting like disabled (unless I changed some more advanced options or switched PBO to advanced and set curves). On ASRock auto PBO acts like enabled and overrides some manual settings (or maybe it's a problem with BIOS).

On my Ryzen 7600/ASUS Strix, I set PBO max clock +200MHz, CPU voltage manually at 1.20V and curve -30. I think that everything else is at auto. The CPU boosts up to 5.35GHz at ~1.18-1.20V. RAM is simply with enabled EXPO profile.
I had more complicated settings with 101.97GHz bclk / 5.4GHz and 6320 RAM, but I forgot to save the profile and I had to clear CMOS. Later I just didn't want to waste time and left it like above.

Btw. like EarthDog already said, you may search for problems that don't exist (or are not possible to solve, at least in the easy way), so maybe just keep it simple and when the PC is stable then just ignore everything else. People post many things around the web. Not everything is correct or the best option for everyone. Not everything affects your experience in the way it's worth to waste a couple of days on tests.

 

[Segundoh]

An aside.........

I heard from a few links that Memory Context Restore greatly improves boot times like this one:

https://www.reddit.com/r/Amd/comments/11m10fz

Most of the links do say that peoples systems have become unstable after enabling this setting, so I suppose I shouldn’t worry too much about. I haven’t heard anything about disabling Power Down mode causing issues though. This is supposed to prevent RAM from reducing their voltages so they always run at their max voltages. It’s supposed to improve performance by at least a little bit.
This link says that:

https://www.reddit.com/r/overclocking/comments/nu1jsv

It’s set to Auto by default, so I suppose it doesn’t really make a difference in most intensive situations. Not really sure what Auto means though. I would like better RAM during general browsing situations and all though. That same link says it should be fine to disable Power Down Mode, but maybe it’s more finicky for DDR5 RAM since most of the links I found were for DDR4 RAM.

As for AIDA64, I downloaded the Extreme edition free trial. I think I still have a bit of time left in that trial, so I’ll try to download the Engineer edition free trial. That makes sense though; I’ll only enable those two for my memory stress test.

 

[EarthDog]

So, in reading these passages, I don't see any reason why I would want to enable them (or try to get them working) in the first place. Apologies for being so blunt, but, for me at least, these things look like waste of time with no real benefits. Just someone flipping switches and reporting anecdotes.

If you want better RAM, buy better or overclock what you have (though overclocking RAM doesn't yield much anyway)... but these methods just don't seem to have much merit as described. AMD is quite limited compared to Intel so you're only looking around DDR5 6200-6400 at most and that depends on the BIOS and memory kit/if it's on the QVL.

Boot times, once memory is trained, is longer on AMD platforms, but not terrible on the boards I've tested using the latest BIOS. I mean, how often are you turning your PC on? Maybe let it sleep instead if the boot times are bothering you?

 

[Segundoh]

You said you had an error like below:

Oh yeah, I misspoke with that first one. I only ran MemTest64 and Task Manager. I ran TestMem5 separately. Regardless, I see what you’re saying. So Windows programs are less reliable than bootable storage ones like MemTest86. It seems fine based on four passes of that and I used all the default tests.

And I see, so I can’t change how it trains RAM and MCR is unreliable. That’s good to know.

It makes sense why there weren’t any issues with Maximus then lol. And I see, so I suppose the BIOS updates for my B650 will stop soon. There is a new beta BIOS, but I was reluctant to test a beta build in case it damages anything; I’m sure it’s fine though.

I’ll try that AIDA64 stress test then. Thanks for the guidance with that! Maybe I’ll even buy 3DMark too.

And I see, so does enabling PBO set other values in addition to the CPU clock speed? Maybe Enabled sets different things than Auto since the boost speed, at least, is the same with both. What other values should I check? As of now, I didn’t manually change any PBO settings other than that, but I may later.

And that does make sense too. Just because a few people said it doesn’t mean that it’ll work for me and improve anything for me. That makes sense. I’m worried because I’m thinking these are hardware issues with my mobo or RAM or something else, but I guess it’s likelier that they’re BIOS issues.

Post magically merged: Mar 22, 2023

So, in reading these passages, I don't see any reason why I would want to enable them (or try to get them working) in the first place. Apologies for being so blunt, but, for me at least, these things look like waste of time with no real benefits. Just someone flipping switches and reporting anecdotes.

That makes sense. That thing about the lower latency is just anecdotal, so there’s probably no reason to even disable Power Down mode. I can try to overclock my RAM a bit. I’ll try 6200-6400 and then I’ll try to reduce my timings a bit.

I’m not sure if my RAM is on the QVL. 6000J3636F16GX2-TZ5NR is on the QVL, but I have a 6000J3636F16GX2-FX5. It seems like they’re the same exact thing except the first one just has RGB. My mobo is at least on the FX5’s QVL list, but I’d feel better about it if my RAM was also on the mobo’s QVL list.

As for boot times, that makes sense. So it’ll get better over time. My boot times went down to about 30 seconds from 38-40 seconds before I enabled MCR. Thas still longer than my AM4 build, but, well, DDR5 just makes boot times longer from what I heard. And that’s true, I barely turn it on. I’m turning it on and off a lot now since I’m changing BIOS settings, but I normally turn it off once before I go to sleep and once after I wake up.

 

[EarthDog]

So it’ll get better over time.

First boot when changing RAM or settings is the longest, after that it should be notably shorter and be the same in subsequent boots. That is correct, DDR5 tends to take longer to boot, especially on AMD.

I turn my PC on once /day too.. I couldn't care less about my boot times, really.

 

[Segundoh]

First boot when changing RAM or settings is the longest, after that it should be notably shorter and be the same in subsequent boots. That is correct, DDR5 tends to take longer to boot, especially on AMD.

I turn my PC on once /day too.. I couldn't care less about my boot times, really.

That makes sense. Is a 30-35 second boot time normal? It seems that’s what my boot time settled at for a month before I started changing memory context restore. And, yeah, it’d be nicer to have shorter boot times, but it doesn’t really matter if it’s once a day. I can live with that.

 

[EarthDog]

I don't have an AMD system up at the moment to test... but boot times vary by board and what's on it... I'd say 20-30 seconds seems about right from pressing the power button to the Windows logon screen.

 

[Zerileous]

I agree with most everything said above, I just wanted to snip a few quotes to comment on a couple of things

.... I would like better RAM during general browsing situations and all though....

Is something going on in "general browsing situations" that makes you feel that better RAM performance is needed. I would assume this system is snappy AF and additionally that any memory tuning will be 100% inconsequential for the application of general browsing. If something is slowing down your browsing I would look elsewhere.

...

And I see, so does enabling PBO set other values in addition to the CPU clock speed? Maybe Enabled sets different things than Auto since the boost speed, at least, is the same with both. What other values should I check? As of now, I didn’t manually change any PBO settings other than that, but I may later.
....

So PBO is all about power delivery. The CPU uses an algorithm based on 3 general parameters to determine how fast it can boost: clock speed, power draw, and temperature. You can increase the clock ceiling by 200MHz in the BIOS, this is a setting that was mentioned by Woomack above. PBO itself increases the power limits using three parameters: EDC, PPT, TDC. What each of these limits does is explained here, although a lot has changed since that article was written the groundwork is fundamentally the same. Finally the "-30" that Woomack mentioned is the Curve Optimizer. I don't have a strong understanding of, or first hand experience with it, but the article kind of explains.

While it is possible to tweak these settings in windows using Ryzen Master, it is generally preferred to save a BIOS profile once ideal settings are noted. These settings are actually beneficial to tune if you do want to play around with getting more out of the system, but I would recommend a lot of research into what each setting does and how to manipulate it before you start plugging in different settings.

That makes sense. Is a 30-35 second boot time normal? It seems that’s what my boot time settled at for a month before I started changing memory context restore. And, yeah, it’d be nicer to have shorter boot times, but it doesn’t really matter if it’s once a day. I can live with that.

I get that the memory training process can be harrowing if one doesn't know what to expect, but overall having a boot time of 30-35 seconds is still awesome fast in the general scheme of PC history.

 

[Segundoh]

I don't have an AMD system up at the moment to test... but boot times vary by board and what's on it... I'd say 20-30 seconds seems about right from pressing the power button to the Windows logon screen.

It seems it’s about 35 seconds right now though I have seen it dip down to 29-30 seconds before. I’ll keep testing though. It does make sense if boot times are longer though since I have a relatively budget based B650 board.

 

[EarthDog]

I’ll keep testing though.

It does make sense if boot times are longer though since I have a relatively budget based B650 board.

Ironically, I'd think it would be LESS time considering there are less things on a B650 board compared to X670. This generally applies to the high-end... but I have no idea what motheboard you have (create a signature with your hardware... stick around - post at ONE forum! LOL).