Question It has to be the memory, right?

[werther595]

TL;DR: My system power-cycles on Prime95 test with XMP enabled, but seems to run the test ok without XMP enabled. Is the RAM bad, and/or what are my next troubleshooting steps?


I have been having a problem with a new build occasionally, randomly, suddenly, powering off and rebooting. It happens with random things, like with nothing open but one Firefox tab watching Youtube, or playing Minecraft. Not all the time mind you, but randomly and seemingly not under a heavy load.

I've posted about it a couple of times and the solutions I found previously (upgraded PSU, updated AMD drivers for GPU) have helped, but the problem hasn't gone away entirely.

https://forums.tomshardware.com/threads/drivers-from-asrock-or-from-amd.3644560/#post-21963499

https://forums.tomshardware.com/threads/new-build-powers-off-under-any-stress.3641056/#post-21938311

I ran a test w Prime95 (blended, 12 tests) and it power-cycled within two minutes. (CPU temps were at about 80 degrees so I don't think it is an overheating issue).

I then cleared CMOS and ran the test again. It ran fine for about 20 mins until my CPU got up to 94-degrees so I stopped it (new cooler coming in the mail today!) The only thing I had done in BIOS before was add some extra aggressive fan curves and RGB stuff and enable XMP. After this more successful test I went back in and re-enabled XMP, ran Prime95 and boom...power-cycle.

Does this mean there is definitely a problem with the RAM (OLOy Warhawk 2x8GB, 3600, CL18)? If it won't perform to XMP specs, should I RMA it and hope for better silicone? Run the test with one stick installed? to see if one or the other is bad? Unfortunately I don't have any other RAM kits around to try out.

Or could it be the mother board (no obvious signes of damage like swolen or leaking caps)? Or something else entirely that I haven't thought of yet? It has now been a couple of months and I'm getting pretty frustrated here. I had previously run the Windows memory checker overnight and found no errors in my ram (with XMP enabled). Any ideas, apart from me throwing everything out the window?

System Deets:
Ryzen 5 3600
ASRock B550M Pro4
OLOy Warhawk (2x8GB) 3600 CL18
ASRock Phantom Gaming RX 5500 XT
Super Flower Leadex III 650W
XPG Spectrix S40G 500GB NVMe
SKHynix Gold SSD 1TB
darkFlash DLM21 Mesh Micro ATX case with 6 fans
OS: Windows 10 Home (up to date)

 

[Flamebrander]

TL;DR: My system power-cycles on Prime95 test with XMP enabled, but seems to run the test ok without XMP enabled. Is the RAM bad, and/or what are my next troubleshooting steps?


I have been having a problem with a new build occasionally, randomly, suddenly, powering off and rebooting. It happens with random things, like with nothing open but one Firefox tab watching Youtube, or playing Minecraft. Not all the time mind you, but randomly and seemingly not under a heavy load.

I've posted about it a couple of times and the solutions I found previously (upgraded PSU, updated AMD drivers for GPU) have helped, but the problem hasn't gone away entirely.

https://forums.tomshardware.com/threads/drivers-from-asrock-or-from-amd.3644560/#post-21963499

https://forums.tomshardware.com/threads/new-build-powers-off-under-any-stress.3641056/#post-21938311

I ran a test w Prime95 (blended, 12 tests) and it power-cycled within two minutes. (CPU temps were at about 80 degrees so I don't think it is an overheating issue).

I then cleared CMOS and ran the test again. It ran fine for about 20 mins until my CPU got up to 94-degrees so I stopped it (new cooler coming in the mail today!) The only thing I had done in BIOS before was add some extra aggressive fan curves and RGB stuff and enable XMP. After this more successful test I went back in and re-enabled XMP, ran Prime95 and boom...power-cycle.

Does this mean there is definitely a problem with the RAM (OLOy Warhawk 2x8GB, 3600, CL18)? If it won't perform to XMP specs, should I RMA it and hope for better silicone? Run the test with one stick installed? to see if one or the other is bad? Unfortunately I don't have any other RAM kits around to try out.

Or could it be the mother board (no obvious signes of damage like swolen or leaking caps)? Or something else entirely that I haven't thought of yet? It has now been a couple of months and I'm getting pretty frustrated here. I had previously run the Windows memory checker overnight and found no errors in my ram (with XMP enabled). Any ideas, apart from me throwing everything out the window?

System Deets:
Ryzen 5 3600
ASRock B550M Pro4
OLOy Warhawk (2x8GB) 3600 CL18
ASRock Phantom Gaming RX 5500 XT
Super Flower Leadex III 650W
XPG Spectrix S40G 500GB NVMe
SKHynix Gold SSD 1TB
darkFlash DLM21 Mesh Micro ATX case with 6 fans
OS: Windows 10 Home (up to date)

If by Power Cycle you mean crash and restart, that means you have an unstable overclock. Your motherboard doesn't exactly have the best VRMs, it's a 7 phase power design. It should be your RAM, as it is running at too high a frequency for the 1 or 2 phases to it from the motherboard. Downclock it, and you should be fine. It's not the RAM, its the motherboard

 

[Darkbreeze]

I don't see, in any of those threads, OR in this one, any mention of which slots your memory modules are installed in. That's probably because people have assumed that being an mATX board it only has two, but it actually has four.

Your memory should NOT be installed in the A1 and B1 slots, which are the 1st and 3rd slots over from the CPU socket. They should be in the 2nd and 4th slots, with the 4th slot being the one closest to the edge of the motherboard. Are they in THOSE slots?

What is the EXACT model of your memory kit, because there are a LOT of kits out there that do not play well well with Ryzen parts?

OLOy is a newcomer to the memory scene, and their products are cheap budget crap. I'd be very surprised if this wasn't your problem. You would be WISE to return them and get a set of G.Skill sticks that are validated as compatible with your motherboard on the G.Skill memory configurator.

https://www.gskill.com/configurator

 

[werther595]

Thanks for your quick reply!

which slots your memory modules are installed in.

A2 and B2, per the mobo manual

What is the EXACT model of your memory kit, because there are a LOT of kits out there that do not play well well with Ryzen parts?

OLOy is a newcomer to the memory scene, and their products are cheap budget crap. I'd be very surprised if this wasn't your problem. You would be WISE to return them and get a set of G.Skill sticks that are validated as compatible with your motherboard on the G.Skill memory configurator.

https://www.gskill.com/configurator

Memory kits is: OLOy WarHawk RGB DDR4 3600 (PC4 28800) 16GB (2 x 8GB) 288-Pin Intel/AMD Ready Desktop Memory Model MD4U083618BEDA

https://www.newegg.com/oloy-16gb-288-pin-ddr4-sdram/p/N82E16820821167?Item=N82E16820821167

I had read up and heard good things, with the main knock against them being they are a new company so ?? about their "lifetime" warranty. They say Ryzen 3000 compatible, and other have had success with them. The memory uses Samsung chips. How can I test to be sure, without having alternate sticks to try? I'd spend the extra on new RAM or send these in for replacement if I was sure they were the problem. But at this point I already have an extra PSU that may or may not have been problematic, and don't want to keep buying parts just to try them out. Any good way to be sure?

 

[Flayed]

Run memtest86 maybe?

 

[werther595]

If by Power Cycle you mean crash and restart, that means you have an unstable overclock. Your motherboard doesn't exactly have the best VRMs, it's a 7 phase power design. It should be your RAM, as it is running at too high a frequency for the 1 or 2 phases to it from the motherboard. Downclock it, and you should be fine. It's not the RAM, its the motherboard

Thanks for your help!!

Is there a way to be sure, without buying a new one trying it out? I had read early reviews of the mobo that praised it for VRMs in particular and recommended it, but I certainly could have gotten a bad unit.

 

[Flamebrander]

Thanks for your quick reply!



A2 and B2, per the mobo manual



Memory kits is: OLOy WarHawk RGB DDR4 3600 (PC4 28800) 16GB (2 x 8GB) 288-Pin Intel/AMD Ready Desktop Memory Model MD4U083618BEDA

https://www.newegg.com/oloy-16gb-288-pin-ddr4-sdram/p/N82E16820821167?Item=N82E16820821167

I had read up and heard good things, with the main knock against them being they are a new company so ?? about their "lifetime" warranty. They say Ryzen 3000 compatible, and other have had success with them. The memory uses Samsung chips. How can I test to be sure, without having alternate sticks to try? I'd spend the extra on new RAM or send these in for replacement if I was sure they were the problem. But at this point I already have an extra PSU that may or may not have been problematic, and don't want to keep buying parts just to try them out. Any good way to be sure?

Just downclock them to maybe 3200 and try benchmarking it until it's unstable

 

[Flamebrander]

Thanks for your help!!

Is there a way to be sure, without buying a new one trying it out? I had read early reviews of the mobo that praised it for VRMs in particular and recommended it, but I certainly could have gotten a bad unit.

It's got good VRMS, but "good" is also based off of others. For the CPU, it's great. On the RAM side, it's decent, but still not good for high overclocks. Try putting more voltage if it's lower than 1.35, or lowering it if it's over. This would be in the RAM voltage

 

[werther595]

It's got good VRMS, but "good" is also based off of others. For the CPU, it's great. On the RAM side, it's decent, but still not good for high overclocks. Try putting more voltage if it's lower than 1.35, or lowering it if it's over. This would be in the RAM voltage

Fair enough. I'll try bumping speeds up a little from the JDEC settings and see what we get. Since DRAM calculator for Ryzen doesn't support B550 yet, would there be somewhere I could find suggestions for some of the less frequently discussed settings in there? Is the instability usually a result of the excessive clock speed itself, or the voltage?

Thanks again for getting back to me on this. I'm losing both hair and money by the fistful over this!

 

[Darkbreeze]

There is nothing wrong with that motherboard for a Ryzen 3600. It's perfectly fine. Not sure why there is conversation about overclocking. I don't see any mention of overclocking unless I missed something. Running memory at XMP or A-XMP profile speeds isn't "overclocking", not for our purposes. It certainly isn't relevant unless you are ACTUALLY overclocking beyond the XMP profile configuration.

No, there is no way to check that these are actually compatible with your board. OLOy does not have a validation list like Corsair and G.Skill have. I think you are wasting your time. Having Samsung IC's doesn't mean squat if the rest of the module is shoddy.

 

[Flamebrander]

There is nothing wrong with that motherboard for a Ryzen 3600. It's perfectly fine. Not sure why there is conversation about overclocking. I don't see any mention of overclocking unless I missed something. Running memory at XMP or A-XMP profile speeds isn't "overclocking", not for our purposes. It certainly isn't relevant unless you are ACTUALLY overclocking beyond the XMP profile configuration.

No, there is no way to check that these are actually compatible with your board. OLOy does not have a validation list like Corsair and G.Skill have. I think you are wasting your time. Having Samsung IC's doesn't mean squat if the rest of the module is shoddy.

XMP is still technically overclocking as it is running higher than DDR4 standard which is 2133. The thing is that the VRMs are responsible for converting the 12 volt rail from a power supply to a say 1.35 volts which is XMP standard. MOSFETs take care of most of it, but the capacitors are responsible for "cleaning" this power, or stabilizing it from sudden jumps or dips in the voltage. Running at higher speeds requires higher voltages and more consistent power, meaning that the capacitors may be overwhelmed if there aren't enough.

 

[Flamebrander]

Fair enough. I'll try bumping speeds up a little from the JDEC settings and see what we get. Since DRAM calculator for Ryzen doesn't support B550 yet, would there be somewhere I could find suggestions for some of the less frequently discussed settings in there? Is the instability usually a result of the excessive clock speed itself, or the voltage?

Thanks again for getting back to me on this. I'm losing both hair and money by the fistful over this!

Try running it at 3200 mhz with infinity fabric at 1600 mhz. Voltages should be 1.33 for DRAM, and CPU is really up to you, but 1.33 volts should also work. Check if this is stable. If so, then it just means that your motherboard isn't ccapable of handling higher speeds than that.

 

[Darkbreeze]

That's like saying that allowing the CPU to boost is "technically overclocking" since it's running at higher than the base clock frequency. Get real. I don't know who you're trying to "technically" to, but I assure you, it's not necessary.

I've been tweaking memory configurations, as much as you could back then anyhow, since before 99% of our members and visitors were even born. So yeah, it IS technically overclocking, but it's not, because there is no amount of testing that SHOULD be necessary in order to run any given set of sticks from a reputable manufacturer at the XMP/A-XMP/D.O.C.P profile settings. The manufacturer has already done that, EXTENSIVELY, so that you don't have to. ACTUAL overclocking, whether by tightening the timings or changing the frequency, would require extensive testing for stability on the part of the user. Running at profile settings doesn't.

We don't consider running a memory kit at it's profile configuration as "overclocking", because in the grand scheme of things, it's not. It is, but, it's not. If we call that overclocking we might as well start calling every system with it's CPU boost behavior enabled an overclocked machine too.

I can further assure you, I don't need explanations as to what VRMs do, what capacitors do, what voltage regulation does or anything else you apparently believe you are able to "teach" me. K? Thanks.

And by the way, your "method" in the post above, for the memory configuration, is a bunch of nonsense. Complete BS. 99% of DDR4 memory kits should be run at the profile voltage of 1.35v, not 1.33v. And the CPU voltage shouldn't even BE TOUCHED AT ALL unless you are manually overclocking, which I'm pretty sure ISN'T a part of this thread. You need to be accurate with the information and advice you post and right now, you aren't. It's pretty clear you know just enough to be dangerous with somebody else's system, and that's about it.

 

[Darkbreeze]

@werther595 , are you overclocking the CPU, at all? And by that, I don't mean running the PBO/PBO2 profiles, or XFR2, or any other kind of automatic overclock. I mean MANUALLY OVERCLOCKING the CPU?

 

[werther595]

Definitely not manually over clocking the CPU. I have Ryzen Master installed, and even that is on Standard mode (as opposed to OC mode) via the basic view. The mobo specs for memory list 3200 as the highest regular clock capable, but cetainly lists that it is capable of 3600 (OC) and higher.

If it is a memory failure, I assume I would get the same crash with one faulty stick as I did with two, right? But if it is a motherboard issue, would running 1 stick at xmp be less taxing? I have no idea, I'm just spitballing here since I don't have a "control" set of ram or mobo to try out.

 

[Flayed]

I would run memtest86 with both the sticks using the XMP profiles. If you get any errors then RMA the memory or get a refund and buy memory from a well-known and trusted brand.

 

[Darkbreeze]

I would run memtest86 with both the sticks using the XMP profiles. If you get any errors then RMA the memory or get a refund and buy memory from a well-known and trusted brand.

This. But, make sure they are installed in the A2 and B2 slots. Second and fourth slots over from the CPU going towards the edge of the motherboard. And honestly, even Memtest86 doesn't catch all problems in every case, so if it passes all four passes, and you NEED to do all four passes which WILL take several hours usually, then run the extended Windows memory diagnostic tool as well if you want further confirmation. Really though, if it passes four passes of Memtest86 it becomes a LOT less likely to be anything related to memory.

 

[werther595]

I'll try memtest86 and see what happens. Thanks again for everyone trying to help. I know you all don't get paid for this so I appreciate the time and know-how. I hadn't built a PC in over 20 years. Some things have changed, and some things are painfully the same. Haha

 

[Darkbreeze]

A lot HAS changed, but when you get down to the square root, it's the same number it always was.

 

[Flamebrander]

That's like saying that allowing the CPU to boost is "technically overclocking" since it's running at higher than the base clock frequency. Get real. I don't know who you're trying to "technically" to, but I assure you, it's not necessary.

I've been tweaking memory configurations, as much as you could back then anyhow, since before 99% of our members and visitors were even born. So yeah, it IS technically overclocking, but it's not, because there is no amount of testing that SHOULD be necessary in order to run any given set of sticks from a reputable manufacturer at the XMP/A-XMP/D.O.C.P profile settings. The manufacturer has already done that, EXTENSIVELY, so that you don't have to. ACTUAL overclocking, whether by tightening the timings or changing the frequency, would require extensive testing for stability on the part of the user. Running at profile settings doesn't.

We don't consider running a memory kit at it's profile configuration as "overclocking", because in the grand scheme of things, it's not. It is, but, it's not. If we call that overclocking we might as well start calling every system with it's CPU boost behavior enabled an overclocked machine too.

I can further assure you, I don't need explanations as to what VRMs do, what capacitors do, what voltage regulation does or anything else you apparently believe you are able to "teach" me. K? Thanks.

And by the way, your "method" in the post above, for the memory configuration, is a bunch of nonsense. Complete BS. 99% of DDR4 memory kits should be run at the profile voltage of 1.35v, not 1.33v. And the CPU voltage shouldn't even BE TOUCHED AT ALL unless you are manually overclocking, which I'm pretty sure ISN'T a part of this thread. You need to be accurate with the information and advice you post and right now, you aren't. It's pretty clear you know just enough to be dangerous with somebody else's system, and that's about it.

Undervolting a RAM stick can't and won't be harder to run and in fact it is easier on the MOSFETs and Chokes. Also, I was only "teaching" you because I wasn't sure what you knew and what you didn't. And yes, a CPU boost algorithm is overclocking as it is taking it higher than what it is binned for. Oh yeah and btw if this guy is overclocking his RAM, he's probably taking his CPU higher and I believe he said that already.

 

[werther595]

So last night I (a) installed a new CPU cooler, (b) enabled XMP, and (c) ran a few sessions of memtest86. The first session lasted about 4 hours and generated 1 error on Pass1 (I'll try to attach a photo so you can see the details). I then had to stop that test and move the computer to another room. I restarted memtest86 and let it run all night. It went through 5 passes and didn't generate any errors.

https://flic.kr/p/2jPJiib

View: https://flic.kr/p/2jPJiib


- Does any error in memtest86 mean the RAM is faulty, or could it be for some other reason?

- Does memtest86 correct errors it encounters? I thought it was weird that I saw an error on Pass1 the first time, then no errors in 5 passes the second time, unless memtest86 corrected the error it found.

After all this, I ran prime95 again, with XMP still enabled and the new cooler, and everything ran fine with no crashes for about 15-20 mins. I know that isn't long enough for a full stress test, but it was more than I could do before.

This all makes me wonder...does RAM overheat, and if so would it cause crashes like I've been having? The OLOy Warhawk has a HUGE heatspreader or whatever you want to call it, obviously designed more for the RGB than dissipating heat. My new cooler is also HUGE (Scythe Fuma2) and my case is fairly small. This means my front case fans are about 6 inches from the RAM blowing air in, the cooler is right behind the RAM pulling air to the back...could this be cooling the RAM enough to solve an overheating problem? I have no idea, just throwing random ideas out there.

 

[Flayed]

So last night I (a) installed a new CPU cooler, (b) enabled XMP, and (c) ran a few sessions of memtest86. The first session lasted about 4 hours and generated 1 error on Pass1 (I'll try to attach a photo so you can see the details). I then had to stop that test and move the computer to another room. I restarted memtest86 and let it run all night. It went through 5 passes and didn't generate any errors.

https://flic.kr/p/2jPJiib

View: https://flic.kr/p/2jPJiib


- Does any error in memtest86 mean the RAM is faulty, or could it be for some other reason?

Yes the ram is faulty.

- Does memtest86 correct errors it encounters? I thought it was weird that I saw an error on Pass1 the first time, then no errors in 5 passes the second time, unless memtest86 corrected the error it found.

No it is just a test. I don't think there is a way of fixing faulty ram, it needs to be replaced.

After all this, I ran prime95 again, with XMP still enabled and the new cooler, and everything ran fine with no crashes for about 15-20 mins. I know that isn't long enough for a full stress test, but it was more than I could do before.

This all makes me wonder...does RAM overheat, and if so would it cause crashes like I've been having? The OLOy Warhawk has a HUGE heatspreader or whatever you want to call it, obviously designed more for the RGB than dissipating heat. My new cooler is also HUGE (Scythe Fuma2) and my case is fairly small. This means my front case fans are about 6 inches from the RAM blowing air in, the cooler is right behind the RAM pulling air to the back...could this be cooling the RAM enough to solve an overheating problem? I have no idea, just throwing random ideas out there.

 

[Flayed]

This all makes me wonder...does RAM overheat, and if so would it cause crashes like I've been having? The OLOy Warhawk has a HUGE heatspreader or whatever you want to call it, obviously designed more for the RGB than dissipating heat. My new cooler is also HUGE (Scythe Fuma2) and my case is fairly small. This means my front case fans are about 6 inches from the RAM blowing air in, the cooler is right behind the RAM pulling air to the back...could this be cooling the RAM enough to solve an overheating problem? I have no idea, just throwing random ideas out there.

While I suppose it is possible for memory to overheat it is not an issue I have heard of actually happening. I don't think DDR4 really gets that hot. My motherboard doesn't seem to have a temperature sensor for the memory which leads me to believe it is not an issue.

Memory being faulty however is an issue that actually happens. I have had DOA ram in the past a set of Corsair value select DDR (or was it DDR2?) a long time ago.

 

[werther595]

Alright then, rock and roll. I'm off to contact customer service...Thanks for your help

 

[Darkbreeze]

Undervolting a RAM stick can't and won't be harder to run and in fact it is easier on the MOSFETs and Chokes. Also, I was only "teaching" you because I wasn't sure what you knew and what you didn't. And yes, a CPU boost algorithm is overclocking as it is taking it higher than what it is binned for. Oh yeah and btw if this guy is overclocking his RAM, he's probably taking his CPU higher and I believe he said that already.

Just stop dude. This is wrong on so many levels I don't even want to have to break it down and clarify the points where you fly off the tracks, but will IF I need to. Believe whatever you want to believe, just don't spread it around to our members. We actually care about giving them accurate information, not senseless opinions based on, I don't know what. That's the nice version. I do bite when poked, so let's just leave it at that.