Is my 5820K this crappy or am I missing something?

[AndroidVageta]

No, because I confirmed the bad memory in another mobo, which gave me that verify. Until then I couldn't be sure.

I initially had 4x512MB in there not 2x2GB that you assumed, so that is perhaps what confused you?

And yes, I know about populating all slots putting more physical stress on the memory controller, bu that's not the issue here. I should also stress that there was no overclocking involved.

Running Memtest now before I nap

Won't be mad thorough but maybe it'll pick something up before the overnight test this evening.

 

[qubit]

Hopefully you'll see something solid. If you do, try running with less RAM and see how that goes, then memtest it again.

 

[cadaveca]

No, because I confirmed the bad memory in another mobo, which gave me that verify. Until then I couldn't be sure.

I initially had 4x512MB in there not 2x2GB that you assumed, so that is perhaps what confused you?

And yes, I know about populating all slots putting more physical stress on the memory controller, bu that's not the issue here. I should also stress that there was no overclocking involved.

did you run memtest for 9 days? because until the most recent release, that's how long proper testing took. You can not expect testing to work properly if you don't do it properly. Now it takes 4 passes...and it shows "1/4" when it starts. Then there's all these people complaining about how their ram doesn't pass the "Hammer" test, and how that's OK...actually... it's not.

 

[OneMoar]

did you run memtest for 9 days? because until the most recent release, that's how long proper testing took. You can not expect testing to work properly if you don't do it properly. Now it takes 4 passes...and it shows "1/4" when it starts. Then there's all these people complaining about how their ram doesn't pass the "Hammer" test, and how that's OK...actually... it's not.

go home dave you are drunk
memtest x86+ ftw
and yes I am in camp bit-rot/hammer test errors mean jackshit because I have yet to hear,see or readable a stability issue caused by bit-rot outside of a database server thats not running ECC

 

[cadaveca]

go home dave you are drunk
memtest x86+ ftw
and yes I am in camp bit-rot/hammer test errors mean jackshit because I have yet to hear,see or readable a stability issue caused by bit-rot outside of a database server thats not running ECC

..yet every single ram kit I have passes the hammer test... DDR3 from 1333 to 3100 MHz, and DDR4 from 2133 MHz to 3200 MHz...

Point is, the new UEFI-capable Memtestx86 takes only 4 passes (still takes well over a day), and does a much better job than past versions.

 

[OneMoar]

..yet every single ram kit I have passes the hammer test... DDR3 from 1333 to 3100 MHz, and DDR4 from 2133 MHz to 3200 MHz...

Point is, the new UEFI-capable Memtestx86 takes only 4 passes (still takes well over a day), and does a much better job than past versions.

none of my systems passed the hammer test and they aren't crashing left and right so ...
and yea a complete cycle of memtest would take untill the next ice age
but generally if the ram is suspect and is indeed the issue it will show within the first 4 to 6 hours and likely a lot sooner if the ram is indeed at fault

 

[AndroidVageta]

Well that was helpful. I go to turn on my monitor only to see that Memtest is "Out of Range" according to my monitor. The fuck is that about??? Goddamn...so now I'm running it AGAIN and will make sure to keep the monitor on...

 

[cadaveca]

Yeah, there is no "initialization" signal upon turning on the monitor that tells it what resolution to run (hence the "out of range").

 

[qubit]

did you run memtest for 9 days? because until the most recent release, that's how long proper testing took. You can not expect testing to work properly if you don't do it properly. Now it takes 4 passes...and it shows "1/4" when it starts. Then there's all these people complaining about how their ram doesn't pass the "Hammer" test, and how that's OK...actually... it's not.

9 days, seriously? Who's got that kind of time? That sounds really excessive and I've never seen that requirement stated anywhere. I've seen you talk about memtest before and I've not seen you say it before either, although I obviously don't see every comment you make lol, so I could have missed it.

Anyway, I ran mine for over 24 hours and on more than one occasion, along with various quickies and it passed every time. Dodgy sticks normally fail within a few minutes to hours, so 24 hours is pretty thorough. I also swapped the sticks round and fiddled with the BIOS memory settings, wiggled the modules with the computer running and generally faffed around quite a bit with it before giving up in frustration. If you're curious, the mobo was an Abit AN8 Ultra with the latest BIOS and not overclocked while I was testing.

I bought the RAM in February 2011 and ran the tests at the time, when it didn't work properly, so quite a long time ago as you can see. After not using the RAM for all those years, I then tried it in a different mobo belonging to my friend* a few months ago and that's when I discovered that one of the sticks wouldn't even let the PC post. Now, whether that stick degraded in the meantime or just works marginally on one mobo and not at all on another one I don't know, but I'm not bothered finding out as it's academic now. btw, he just bought some cheap and nasty HP prebuilt after that, which has very little upgrade potential, even though I said I'd help him put a PC together. Oh well.

*It was an Asus A8N-VM CSM. Yes, it was my mobo too until I gave it to my friend a few years ago to keep his PC building costs down as it was only for the common internet and email usage. Oh and I've got this PC back now, bought 4GB RAM for it the other day, plugged it in and it works. It's just that it reports that it's got 4GB RAM and 3GB useable in Windows 7 64-bit and the BIOS shows 3GB. Nice, lol. The manual of course says nothing about this problem. It runs well enough as it is, though.

 

[cadaveca]

I have said it before, but yeah, test #7 alone takes like 9 days to fully test like 8 or 16 GB, can't remember. You can find the post by the software's author on his forum explaining why when he was asked why the software didn't indicate true stability. It is still a good indicator to proper stability, and the new version works fairly well, but still has some detection bugs to work out.

 

[hat]

And here I thought once it made a full pass, taking quite a small amount of time, it had tested all the RAM and it was good to go. Multiple passes just to be sure it didn't miss something. 9 days??

 

[AndroidVageta]

And here I thought once it made a full pass, taking quite a small amount of time, it had tested all the RAM and it was good to go. Multiple passes just to be sure it didn't miss something. 9 days??

If I have to test my RAM for 9 days I'm just buying new RAM!

 

[qubit]

And here I thought once it made a full pass, taking quite a small amount of time, it had tested all the RAM and it was good to go. Multiple passes just to be sure it didn't miss something. 9 days??

I don't think it misses anything as such ie the RAM is completely tested.

Flaky hardware has a tendency to work intermittently, whether it's a car, a clock or computer gear etc, so it could work 100%, then suddenly show a glitch for a moment, then go back to being good and pretending like nothing happened. Over time, these glitches get more frequent and last longer until eventual and inevitable complete hardware failure. Dry solder joints are notorious for this behaviour and also make the device temperature sensitive. I think the 9 days of testing, as crazy as it is, is designed to tease out problems like this. I'm not convinced that perfectly good hardware wouldn't actually develop faults after being stress tested so intensively over an extended period of time however. Heck, I might just try this out with my spare computers and let you all know how it goes.

@cadaveca This is my interpretation of the rationale for the extended testing. How do you see it?

 

[cadaveca]

Different areas of the ram are tested in different ways with each pass. With the older version, you can see this with just two passes..the first is a preliminary pass that takes about 15 minutes, but the second, when testing 8 GB, will take nearly 45 minutes.

So no, it's not about forcing out issues...it's that parts of the ram just weren't tested yet.

The latest version shows 4 passes as sufficient, but ultimately, proper testing devices for memory cost a significant amount of money that no free software is going to replace.

 

[RichF]

LLC helps keep you from killing your CPU at higher voltages.

Anandtech's article on it suggested the opposite: that transient spikes are higher with LLC enabled which is more risky than having vdroop cause you to set a higher voltage in BIOS for the same stability that LLC provides with a lower setting.

 

[OneMoar]

Different areas of the ram are tested in different ways with each pass. With the older version, you can see this with just two passes..the first is a preliminary pass that takes about 15 minutes, but the second, when testing 8 GB, will take nearly 45 minutes.

So no, it's not about forcing out issues...it's that parts of the ram just weren't tested yet.

The latest version shows 4 passes as sufficient, but ultimately, proper testing devices for memory cost a significant amount of money that no free software is going to replace.

so nothing we didn't already know about memtest or any software stability testing in general
I fail to understand your point also I would think that running a mem-test for days on end would have the potential to induce errors that would normally never occur under normal operating conditions e.g how the hammer test checks for bitrot/leaking

 

[MxPhenom 216]

Eh what did I do wrong?

Feeding false information

 

[AndroidVageta]

So I'm chalking this up to bad RAM. My friend and I are having the same issues basically and we're both running the same RAM (he has 1 and a half kit and I have the other half) and if I lower the RAM speed stability increases. Not perfectly, mind you, but going from 2400 to even 2133 nets me MUCH fewer crashes. Things that before were nearly guaranteed lock ups are now not a problem.

Even someone on Newegg said they had to run the RAM at 2000 to achieve stability. Makes sense.

So yeah I know think the motherboard and CPU are fine, it's just some bonk ass RAM.

 

[OneMoar]

I don't think its the ram I think its the IMC trying feeding the system-agent some more voltage

 

[R-T-B]

Is this an X99-A with Crucial Ballistix Sport? Yeah I have the EXACT same issue. I run at DDR4-2000 with lower latency and it works fine. Go to 2400 and I have issues. I think it's a combination of flakey ram and a sensitive mobo.

EDIT: Nvm, this is an Asrock I see. But I do have a similar issue with the same ram for whatever it's worth.

 

[AndroidVageta]

Is this an X99-A with Crucial Ballistix Sport? Yeah I have the EXACT same issue. I run at DDR4-2000 with lower latency and it works fine. Go to 2400 and I have issues. I think it's a combination of flakey ram and a sensitive mobo.

EDIT: Nvm, this is an Asrock I see. But I do have a similar issue with the same ram for whatever it's worth.

Yeah, think the RAM is either shitty or the motherboard is possibly picky. Either way I doubt it's a mobo issue. Friend of mine went through four motherboards before we started looking into RAM...there's no way he had that many bad motherboards. Even exchanged his 5960X for good measure. So that's 4 motherboards, 2 CPU's, and a new 1300 watt PSU that he went through while still having the same issues as I am.

It's the RAM, no doubt. What else could it be that explains why my system is running perfectly with lower RAM speeds and nothing else changed?

 

[OneMoar]

Yeah, think the RAM is either shitty or the motherboard is possibly picky. Either way I doubt it's a mobo issue. Friend of mine went through four motherboards before we started looking into RAM...there's no way he had that many bad motherboards. Even exchanged his 5960X for good measure. So that's 4 motherboards, 2 CPU's, and a new 1300 watt PSU that he went through while still having the same issues as I am.

It's the RAM, no doubt. What else could it be that explains why my system is running perfectly with lower RAM speeds and nothing else changed?

there is a reason intel specs the 5960x to run at a maximum speed of for the ram 2133 anything above that and you are _expected_ to need to manually tweak the voltages and or timings
XMP is by no means perfect id bump the system-agent by +0.050 to +0.100 and see if helps I would also do as dave suggested and put everything on manual and set things by hand

 

[cadaveca]

I am actually more inclined to think it's the board's BIOS tuning myself.

But what model of ram exactly are we talking about... I just might have a set...

 

[OneMoar]

I am actually more inclined to think it's the board's BIOS tuning myself.

But what model of ram exactly are we talking about... I just might have a set...

you have so many kits around if there was ever a dram shortage you could be a very rich man :lol:

 

[AndroidVageta]

I am actually more inclined to think it's the board's BIOS tuning myself.

But what model of ram exactly are we talking about... I just might have a set...

The Crucial Ballistix Sport DDR4-2400 2x4GB to be exact.

And it might be settings, etc...all I know at this point is that it is a RAM issue. As long as it's not the mother board or CPU I'm OK haha!