ECC memory

[Exidor]

So I did some research on testing ECC memory.

It seems that in non server environments and even in server environments there really is no simple reliable way to know if ECC is working.

I have an AMD 5900x running with some ECC memory. Now both memtest and BIOS tell me that the ram is running in ECC mode. But many people point out that this only means the memory reports itself as ECC not that it actually works as ECC.

So you are supposed to provoke memory errors and check the logs for and actually look for ECC corrected errors.

 

Now this sounds esoteric and  bullshitty at best.

I mean people tell you to underwolt or overclock your RAM until you get errors or even better use a heat gun?

Is that seriously the only way to test this?

 

Is there no test for this?

I mean surely in critical server infrastructure where people pay for 100% uptime someone somewhere must have figured out a way to test this?

[Guest]

It's registered or unregistered ECC? Sounds like unregistered.

 

You have to remember that ECC memory is meant for redundancy. 

So left at defaults, no OC = stability and redundancy.

 

You do nothing to check the memory. It does the error correcting by it's self. 

 

A heat gun..... People are stupid. Were did you read that at? It's not even amusing.... melting IC's off the memory board sounds like a fun way to test for errors. lol.

[DieselWeasel]

I can't answer you becuase I don't know, but why even bother with this if your computer is working? I just asumme my ECC memory is working it's magic. I have no reason to doubt it becouse there haven't been any problems.

[Exidor]

3 minutes ago, ShrimpBrime said:

It's registered or unregistered ECC? Sounds like unregistered.

 

You have to remember that ECC memory is meant for redundancy. 

So left at defaults, no OC = stability and redundancy.

 

You do nothing to check the memory. It does the error correcting by it's self. 

 

A heat gun..... People are stupid. Were did you read that at? It's not even amusing.... melting IC's off the memory board sounds like a fun way to test for errors. lol.

 

Yea I know....

You kind of start googling ECC memory testing and you end up falling down a rabbit hole.

 

It is unregistered memory Kingston KSM32ED8/16ME I run it stock.

 

The whole point of this exercise is to find out if the memory is running in ECC or not.3

As AMD does not officially support ECC on the 5000x platform many people state point blank that even if your BIOS and memtest report that you are running ECC the memory actually does not run in ECC mode. And that only very specific memory and MOBO combinations are verified to work.

[Guest]

2 minutes ago, Exidor said:

 

Yea I know....

You kind of start googling ECC memory testing and you end up falling down a rabbit hole.

 

It is unregistered memory Kingston KSM32ED8/16ME I run it stock.

 

The whole point of this exercise is to find out if the memory is running in ECC or not.3

As AMD does not officially support ECC on the 5000x platform many people state point blank that even if your BIOS and memtest report that you are running ECC the memory actually does not run in ECC mode. And that only very specific memory and MOBO combinations are verified to work.

That's why I think it's unregistered ECC memory. 

 

ECC memory either runs or it doesn't. I have registered memory and it will not post on desktop platforms, the motherboard/Cpu does not have the required resources to run it. Just like Some Tesla cards, which I happen to have, will not post on a desktop motherboard. 

 

So if it IS registered ECC memory, it's running in ECC mode.

 

In the bios, in the memory section, you should have an ECC setting. If enabled, the board is running in ECC mode and that's how you would know.

[Exidor]

Yea it is enabled in BIOS.

 

As I said the only reason I even started looking into this is because a lot of people state that as the support is unofficial most of the RAM MOBO combinations even when they state that they run ECC do not actually run in ECC.....

And the only way to know is when you encounter an ECC corrected error and check the logs.

 

 

And as I am not that knowledgeable about such things it got me confused and I started questioning the set up.

Googling does not help. Even places like lvl1 that usually avoid the worst of the BS have very confusing statements about this in the forums.

[Jurrunio]

There is this, seems like they know what they're doing but I dont really understand it either

https://serverfault.com/questions/643542/how-do-i-get-notified-of-ecc-errors-in-linux

[Guest]

1 minute ago, Exidor said:

Yea it is enabled in BIOS.

 

As I said the only reason I even started looking into this is because a lot of people state that as the support is unofficial most of the RAM MOBO combinations even when they state that they run ECC do not actually run in ECC.....

And the only way to know is when you encounter an ECC corrected error and check the logs.

Well I suppose you could "TEST" the memory by using A stress test program like OCCT Linpack.

If there are errors, does NOT always mean blue screen. Even on standard DDR4 memory modules.

 

Also an unstable CPU will cause errors the memory may not be able to correct. I have seen Registered ECC memory blue screen on many occasions. 

[svmlegacy]

10 minutes ago, Exidor said:

As AMD does not officially support ECC on the 5000x platform many people state point blank that even if your BIOS and memtest report that you are running ECC the memory actually does not run in ECC mode.

While AMD does not support ECC for Ryzen 5000, they have stated that it isn't explicitly disabled. It is safe to assume it's enabled, as long as the BIOS itself shows that it is.

Run the system. If you have trouble, then look into the MCE logs.

[Exidor]

2 minutes ago, Jurrunio said:

There is this, seems like they know what they're doing but I dont really understand it either

https://serverfault.com/questions/643542/how-do-i-get-notified-of-ecc-errors-in-linux

 

Yea I think I already read that.

Confusing as hell....

 

This year was really annoying I also switched to HDMI 2.1 and that is a dumpsterfire....

 

 

You would think that you can get a binary answer on stuff working or not working. Without resolving to fortunetelling.

[Guest]

Here's a useful article.

 

https://www.pugetsystems.com/labs/articles/How-to-Check-ECC-RAM-Functionality-462/

[Guest]

My favorite part of the article.

 

 

Quote

 

Conclusion

For something that is critical for servers and some workstations, you would think that checking to see if ECC is working would be a simple matter. Unfortunately, we have found that there is no consistent, conclusive way to determine if ECC RAM is working properly.

 

 

Looks like it's fortunetelling for you. Sorry for that.

[Exidor]

2 minutes ago, ShrimpBrime said:

Here's a useful article.

 

https://www.pugetsystems.com/labs/articles/How-to-Check-ECC-RAM-Functionality-462/

 

Thanks.....

I am loving the conclusion....

For something that is critical for servers and some workstations, you would think that checking to see if ECC is working would be a simple matter. Unfortunately, we have found that there is no consistent, conclusive way to determine if ECC RAM is working properly. Even if none of the three methods we showed indicate that ECC is working, all that means is that none of them could detect that ECC is working. On the positive side, on every system we have needed to verify that ECC is working we have been able to confirm it by using one of the methods we showed in this article.

[Guest]

Just now, Exidor said:

 

Thanks.....

I am loving the conclusion....

For something that is critical for servers and some workstations, you would think that checking to see if ECC is working would be a simple matter. Unfortunately, we have found that there is no consistent, conclusive way to determine if ECC RAM is working properly. Even if none of the three methods we showed indicate that ECC is working, all that means is that none of them could detect that ECC is working. On the positive side, on every system we have needed to verify that ECC is working we have been able to confirm it by using one of the methods we showed in this article.

 

My Crystal Ball says your system is running fine error free for the unforeseeable future so long as you don't fuck with it.......... MMMMMmmmm

 

[Exidor]

I mean.....

This is why we cant have nice things.

 

You would have thought that for something that is supposed to be redundant security there would be a reliable way to check it..... FFS.

 

 

Reminds me of the whole HDMI and USB power delivery cluster fuck.

[Exidor]

Oh well maybe I spent an extra 100 euros on ram that is just slower

[Guest]

Just now, Exidor said:

I mean.....

This is why we cant have nice things.

 

You would have thought that for something that is supposed to be redundant security there would be a reliable way to check it..... FFS.

 

 

Reminds me of the whole HDMI and USB power delivery cluster fuck.

I think on a desktop environment with Winblows, security is always a risk. Stability is always a risk.

 

Wish on that Crystal Ball that the next windows update doesn't F my entire OS from the way I liked it before. Which seems to be every other update, they F my install somehow lol.

 

There's no ECC memory in the world that can fix windows errors. It's just natural to it. XP might be the most stable and most secure these days. Hackers wouldn't even know what to do with the basics of the OS on todays software lol. Which most of it probably won't run these days XD

[Exidor]

39 minutes ago, ShrimpBrime said:

I think on a desktop environment with Winblows, security is always a risk. Stability is always a risk.

 

Wish on that Crystal Ball that the next windows update doesn't F my entire OS from the way I liked it before. Which seems to be every other update, they F my install somehow lol.

 

There's no ECC memory in the world that can fix windows errors. It's just natural to it. XP might be the most stable and most secure these days. Hackers wouldn't even know what to do with the basics of the OS on todays software lol. Which most of it probably won't run these days XD

That is true sad and funny especially considering that most of XP source-code got leaked...

 

And I hate windows 10 with a passion. I first used it on my laptop as it came preinstalled Surface Book 2 and even on a Microsoft designed high end device it was crap. So I stuck to win 7 for a very long while.
If all the sound VRR raytracing HDR Atmos Direct X and other crap that is Nvidia drivers and HDMI standards worked on linux I would have jumped ship long ago.

My hope is that with Windows becoming more like apple and not in a good way and companies pushing video game streaming at some point in the future there will be real linux support......

 

But I am mostly worried about RAM corrupting file transfares as I am a bit of a data hoarder....

[Guest]

14 minutes ago, Exidor said:

That is true sad and funny especially considering that most of XP source-code got leaked...

 

And I hate windows 10 with a passion. I first used it on my laptop as it came preinstalled Surface Book 2 and even on a Microsoft designed high end device it was crap. So I stuck to win 7 for a very long while.
If all the sound VRR raytracing HDR Atmos Direct X and other crap that is Nvidia drivers and HDMI standards worked on linux I would have jumped ship long ago.

My hope is that with Windows becoming more like apple and not in a good way and companies pushing video game streaming at some point in the future there will be real linux support......

 

But for I am mostly worried about RAM corrupting file transfares as I am a bit of a data hoarder....

I think you'll be just fine. Just use it. 

 

Use W7 quite often when benchmarking. I gave up the OS because I wanted to play DX12 games like everyone else lol. (for a daily)

 

Last night on W7, 1600+ Palomino AThlon XP 1.9ghz OC from 1.4ghz stock benching. The blue screens at 1.9850ghz was a result of CPU being topped out. Not a stitch more. Just wasn't cold enough. So this is a cpu hardware error, the memory was running below it's rated speeds. 

 

That's why I was saying earlier, any piece of hardware can cause a blue screen or a non blue screen error. Gpu, Cpu, add in cards, Thumb drive, SSD... chipset. It's all driver run. Sometimes data writes to memory in the wrong location, ECC can correct that. It cannot correct my Overclocks though! lol.

[Exidor]

32 minutes ago, Exidor said:

That is true sad and funny especially considering that most of XP source-code got leaked...

 

And I hate windows 10 with a passion. I first used it on my laptop as it came preinstalled Surface Book 2 and even on a Microsoft designed high end device it was crap. So I stuck to win 7 for a very long while.
If all the sound VRR raytracing HDR Atmos Direct X and other crap that is Nvidia drivers and HDMI standards worked on linux I would have jumped ship long ago.

My hope is that with Windows becoming more like apple and not in a good way and companies pushing video game streaming at some point in the future there will be real linux support......

 

But I am mostly worried about RAM corrupting file transfares as I am a bit of a data hoarder....

Double post I wanted to edit.....

Sorry