X99 lost a memory channel

[porina]

Asrock X99 Extreme 3 was working fine with a 5820k and 4x4GB of ram (16GB total). Tonight I dropped in my E5 2683 v3 ES which worked fine in the past. Everything seemed fine until I saw I only had 12GB once booted into windows.

 

What I've tried:

• Reseat the ram in same slots
• Move ram around different slots
• Reseat CPU
• Reseat cooler (CoolerMaster 240mm AIO)
• Reseat CPU again, after cleaning the pads on the CPU with isopropyl alcohol
• Inspect the CPU socket for bent pins. Everything looked fine.
• Tried looser mount tension on waterblock

 

I'm suspecting the problem is something to do with the CPU and the socket but I think I've done what I can unless there are any other ideas? Currently I'm running two sticks and getting 8GB as expected. I haven't tried only running the other pair... on that note, I could swap the existing pair with the removed pair just to eliminate one of the modules spontaneously dying. I'll do that shortly.

[WhitetailAni]

You didn't specify - but did the 5820K work with all 16GB of RAM?

[porina]

Just now, ragnarok0273 said:

You didn't specify - but did the 5820K work with all 16GB of RAM?

It's in the first line I wrote. It was running boinc immediately before the CPU swap and I even double checked the reported ram to project server was showing 16GB and the old CPU, so it wasn't yet updated with the new CPU and ram.

[WhitetailAni]

2 minutes ago, porina said:

It's in the first line I wrote. It was running boinc immediately before the CPU swap and I even double checked the reported ram to project server was showing 16GB and the old CPU, so it wasn't yet updated with the new CPU and ram.

Did you check that 16GB works with the old CPU?

[porina]

1 minute ago, ragnarok0273 said:

Did you check that 16GB works with the old CPU?

Yes as far as 16GB was detected and the system ran 24/7 on compute for some months now. Dropping in my Xeon as I'm joining a compute challenge in an hour and could use the extra horsepower. The 5820k was put in as I was going to sell it but never did get around to it.

 

Just swapped the pair in the system with the other pair, both reporting 8GB so modules seem ok. I'll double check the other pair of ram slots on mobo next and expect one of these to not be working currently.

[porina]

Ok, this doesn't make sense... I moved the two modules into the other two slots and... still both are detected. Nani the f...

[WhitetailAni]

Just now, porina said:

Ok, this doesn't make sense... I moved the two modules into the other two slots and... still both are detected. Nani the f...

Test your 5820K again.
It could be that the Xeon has a dead channel.

[porina]

Just now, ragnarok0273 said:

It could be that the Xeon has a dead channel.

Well the above just disproved that.

 

I have seen:

All 4 modules fitted = 12GB.

2 modules only in slots 3+4 = 8GB.

Swap other 2 modules in slots 3+4 = 8GB.

2 modules only in slots 1+2 = 8GB

 

I'm just going to put all 4 modules in again and see what happens.

[porina]

Now it detects all 16GB. Maybe something was loose and all that moving around sorted it, but it wasn't fun.

[porina]

Having had a closer look at the system, I wonder if ram settings might have played a part in it.

 

The ram is Ripjaws 4 3333. I was running it at 2133 with the 5820k as I've had stability problems with running this ram above 3000 in the past. It is very fussy about the mobo/CPU. When I swapped the CPU it booted up without intervention. No problem? There were some irregularities in Windows. Both the mobo bios and Windows reported the ram was 2666, even though the Xeon wont run ram above 2133. I also noted from the reported timings it was using the XMP profile timings, not JEDEC ones for 2133. So it is actually slower timings than it should be. I can't find an XMP toggle in the bios. Might do a complete reset just in case of some weird setting in there from previous. For the Xeon it is better to run JEDEC 2133. I would swap modules from another system but I know I'll end up breaking another system if I were to try that now.

 

So the above may be a contributing factor, otherwise just call this haunted hardware.

[porina]

Loaded bios defaults. Everything seems sensible now, with auto ram settings picking JEDEC 2133 timings. Maybe some things were mis-configured even though it booted 1st time without complaint after CPU swap.

[freeagent]

X58 has been known to exhibit similar behaviour. Can be a little annoying..

[porina]

3 minutes ago, freeagent said:

X58 has been known to exhibit similar behaviour. Can be a little annoying..

I've heard of similar for X99 too, but don't recall encountering anything as weird as this until now. Still, if I had to blame anything, it'll be the mobo bios. I can't find an XMP toggle in it, and it is on the latest bios. Loading defaults made it work fully as expected, so my working assumption is that some settings weren't properly reset when I just swapped the CPU over. It was close enough to appear to work, but took the reset to actually load correct values. I've heard of people recommending a "load defaults" after a bios update for a similar reason, and have seen some weird behaviour after doing a bios update if it isn't reset. Depending on the mobo, I think many bios updates will force a settings reset anyway, but doesn't hurt to check. Anyway, it is working and my OCD isn't being triggered with 12GB of ram.

[freeagent]

I've read all sorts of things about it, people thinking it was the mounting pressure..

 

I suffered from that for years on my X58.. very fussy. But! The board has two bios, bios 1 had the newest available, bios 2 had a very early one.. switched to the oldest and that solved every single one of my memory issues that I ever had. Took me 7 years.. forgot it was even there.. lol. I went from running 6GB, to being able to fully populate all dimms with 3 different kits lol. And being able to run more than 1 type. For instance, my main ram was STT Chrome Hypers, but if I wanted to run my G.Skill BBSE it would throw fits for a bit. Even more fits when trying to run my OCZ PSC. But with the old bios could run all of those kits as well as 8GB of G.Skill Hynix sticks. Crazy stuff man. ANd by fits I mean dropping channels. They would be there then poof gone.

[Guest]

AsRock board?

It's the board, not the cpu or memory. 

I've seen this with numerous models of AsRock boards even had one that did the same thing to me. Was an extreme 4 actually.

 

[porina]

8 hours ago, freeagent said:

I've read all sorts of things about it, people thinking it was the mounting pressure..

That was one thing I kinda tried, by loosening the screws holding the waterblock in place. I've only had problems with that once with a different AIO and 115x, where temps went right up at higher tightness. Didn't make a difference in this case.

 

8 hours ago, ShrimpBrime said:

AsRock board?

It's the board, not the cpu or memory. 

I've seen this with numerous models of AsRock boards even had one that did the same thing to me. Was an extreme 4 actually.

I've gone off Asrock, although it is more due to their power delivery than anything else. I also have one each of Z370 and B450 boards where I suspect the power delivery is not adequate leading to system unstableness at stock when running sustained workloads. The Z370 may randomly crash on average once every few days unless I turn off turbo, then it'll run forever (4.3 GHz vs 4.0 GHz all core so not a great loss). The B450 crashed with different frequency depending on the CPU, getting worse with newer models. In that case I found limiting power didn't help and actually made it worse. Only thing that helped was positive voltage offset.

 

Is there a mobo maker that does decent affordable OC boards which don't suck?

MSI - horrible bios layout, slow bios update release speed, particularly bad ram compatibility with their Z170 boards compared to Asus at the time

Asus - relatively expensive compared to other brands, slower bios update release speed on anything except top models

Gigabyte - meh bios layout, don't have enough experience with OC on them yet

Asrock - sus power delivery, but pretty fast on bios updates

 

Right now I'd be extremely hesitant to buy MSI or Asrock. Asus are still in the running even if the last board I got from them was X299, and I've recently decided to give Gigabyte a chance on Z490. In case you wonder about my focus on bios update speed, I have tended to buy early on in product cycle so that's when most of the development is done.

[Analog]

9 hours ago, porina said:

I've heard of similar for X99 too, but don't recall encountering anything as weird as this until now. Still, if I had to blame anything, it'll be the mobo bios. I can't find an XMP toggle in it, and it is on the latest bios. Loading defaults made it work fully as expected, so my working assumption is that some settings weren't properly reset when I just swapped the CPU over. It was close enough to appear to work, but took the reset to actually load correct values. I've heard of people recommending a "load defaults" after a bios update for a similar reason, and have seen some weird behaviour after doing a bios update if it isn't reset. Depending on the mobo, I think many bios updates will force a settings reset anyway, but doesn't hurt to check. Anyway, it is working and my OCD isn't being triggered with 12GB of ram.

 

Just fyi, I had a very similar issue to yours. I had an ASRock X99m Fatality with a 5960X before and one of the memory channels just stopped working on that board. After trying pretty much everything and giving up and continued using it with 12GB (had 4x4GB at the time). Funnily enough it started working again on its own. Then later on, after a month or two it stopped working again and just died. CPU was fine, as I tested it on an EVGA board so I am pretty sure it has something to do with the socket or the traces. Considering that the build was 5-6y old already I just disassembled everything and sold the components. Moral of the story, no more ASRock boards for me

[Guest]

3 hours ago, porina said:

That was one thing I kinda tried, by loosening the screws holding the waterblock in place. I've only had problems with that once with a different AIO and 115x, where temps went right up at higher tightness. Didn't make a difference in this case.

 

I've gone off Asrock, although it is more due to their power delivery than anything else. I also have one each of Z370 and B450 boards where I suspect the power delivery is not adequate leading to system unstableness at stock when running sustained workloads. The Z370 may randomly crash on average once every few days unless I turn off turbo, then it'll run forever (4.3 GHz vs 4.0 GHz all core so not a great loss). The B450 crashed with different frequency depending on the CPU, getting worse with newer models. In that case I found limiting power didn't help and actually made it worse. Only thing that helped was positive voltage offset.

 

Is there a mobo maker that does decent affordable OC boards which don't suck?

MSI - horrible bios layout, slow bios update release speed, particularly bad ram compatibility with their Z170 boards compared to Asus at the time

Asus - relatively expensive compared to other brands, slower bios update release speed on anything except top models

Gigabyte - meh bios layout, don't have enough experience with OC on them yet

Asrock - sus power delivery, but pretty fast on bios updates

 

Right now I'd be extremely hesitant to buy MSI or Asrock. Asus are still in the running even if the last board I got from them was X299, and I've recently decided to give Gigabyte a chance on Z490. In case you wonder about my focus on bios update speed, I have tended to buy early on in product cycle so that's when most of the development is done.

X299, we didnt mention EVGA, which they didnt make any X570 am4 boards.

 

What do think of EVGA

[porina]

1 hour ago, ShrimpBrime said:

What do think of EVGA

No experience. They're too small in the mobo space to be a serious consideration for me.

[Guest]

2 minutes ago, porina said:

No experience. They're too small in the mobo space to be a serious consideration for me.

From what I understand, they make a decent board. But you're right, small segment.

 

Kinda like BioStop boards.

[CommanderAlex]

5 hours ago, porina said:

Right now I'd be extremely hesitant to buy MSI or Asrock. Asus are still in the running even if the last board I got from them was X299, and I've recently decided to give Gigabyte a chance on Z490. In case you wonder about my focus on bios update speed, I have tended to buy early on in product cycle so that's when most of the development is done.

Yeah MSI is pretty slow with BIOS updates that I've found after owning a Z97 board and my X299 board I am using right now. I just saw yesterday they finally introduced a Beta BIOS (Which I will never download a Beta BIOS to update on any board, period) for the resizable bar support. I saw the other day EVGA has finally done a slew of BIOS updates for Intel's chipsets with resizable bar support on Facebook. 

 

BIOS updating with MSI is not that important for me, unless upgrading for security patches/new processor unsupported on older BIOS, but with X299 right now, this may be the end of the road for BIOS updates considering Intel has not released any information on new HEDT processors, perhaps they will be on a newer chipset (X499 maybe since AMD has X399 for their Threadripper lineup). 

[porina]

8 minutes ago, ShrimpBrime said:

From what I understand, they make a decent board. But you're right, small segment.

 

Kinda like BioStop boards.

Biostar?  I think I'm more likely to buy a Biostar mobo than EVGA. At least they're in the mobo space for longer, even if their products don't usually make it to the western markets, and their rep isn't the best. The concern with small players and/or weird products is the question just how long are they going to stick with supporting the product?

 

Just now, SpiderMan said:

BIOS updating with MSI is not that important for me, unless upgrading for security patches/new processor unsupported on older BIOS, but with X299 right now, this may be the end of the road for BIOS updates considering Intel has not released any information on new HEDT processors, perhaps they will be on a newer chipset (X499 maybe since AMD has X399 for their Threadripper lineup). 

I don't expect anything new on X299. It is based off whatever they're doing in server, and there hasn't been anything relevant there since Cascade Lake. Cooper Lake was very targeted in market and wouldn't have added anything to HEDT. In my view the next chance at Intel HEDT is a variation of Ice Lake server, so it will be new socket, new platform, and 10nm. Given how late Ice Lake server is, they might skip it and instead use Sapphire Rapids instead.

[Guest]

35 minutes ago, porina said:

Biostar?  I think I'm more likely to buy a Biostar mobo than EVGA. At least they're in the mobo space for longer, even if their products don't usually make it to the western markets, and their rep isn't the best. The concern with small players and/or weird products is the question just how long are they going to stick with supporting the product?

 

I don't expect anything new on X299. It is based off whatever they're doing in server, and there hasn't been anything relevant there since Cascade Lake. Cooper Lake was very targeted in market and wouldn't have added anything to HEDT. In my view the next chance at Intel HEDT is a variation of Ice Lake server, so it will be new socket, new platform, and 10nm. Given how late Ice Lake server is, they might skip it and instead use Sapphire Rapids instead.

No, I meant BioStop. I've killed every single Biostar board ever touched. All of them lol. 

 

The just... stopped working. Probably because of my abuse, they seemed fine at stock. (Am2 mostly). Lol