problem Computer crash when processing large amounts of data

[KrazeyKami]

When you put a DDR3-1066 module in the orange A1 socket and a DDR3-1333 in the black A2, you are actually mixing RAM types on the one channel. Howell has suggested - as a diagnostic - that you try to keep the same type on each channel.

Aaah, so u say A1-A2 = channel 1, B1-B2 = channel 2 and C1-C2 is channel 3? I always thought that A1,B1,C1 = channel 1 and A2,B2,C2 = channel 2, since this is triple channel memory... Come to think of it, that doesnt make sense calling it tripple channel, since it would mean you only have 2 channels....

This would also make sense, in the understanding that using the same type of memory DIMMS (x6) would work together on A,B and C.

But then, im facing a dilemma;
First of all, Asus says, that triple channel needs to be filled this way. A1, B1, C1 (the oranges) and then A2, B2, C2 (the blacks). Let's say that this would not be the case when using mixed types.

Then, i can imagine that what Howell suggets should work: A1,2 = 10700, B1,2 = 8500. But then, C1,2 = always a mixed channel = a problem.

This would suggest that they made memory channels with the logic of never being able to use mixed memory types that come in sets of 3?

That doesn't make much sense to me...

For sure, i will test tonight as Howell describes and see how it reacts.

p.s.,
In the manual, indeed, it says A, B, and C are the channels (so 3 of em).

I think this kinda proves, once again, that the memory OCZ provided is not of the same type ....

Kind regards,
Kami.

 

[Howell]

Thank you time.

Then, i can imagine that what Howell suggets should work: A1,2 = 10700, B1,2 = 8500. But then, C1,2 = always a mixed channel = a problem.

I envision this arrangement only as a temporary limp-along solution until you can obtain 6 sticks of HCL approved and tested memory from a company of quality. Buck has suggested some replacements.

I see this temporary because even if your system stabilizes one or more channels will be running at a speed slower than the rest. Not optimal.

FWIW, OCZ has or is leaving the memory business. The sooner you sell what you have the better price you will get.

 

[KrazeyKami]

Thank you time.



I envision this arrangement only as a temporary limp-along solution until you can obtain 6 sticks of HCL approved and tested memory from a company of quality. Buck has suggested some replacements.

I see this temporary because even if your system stabilizes one or more channels will be running at a speed slower than the rest. Not optimal.

FWIW, OCZ has or is leaving the memory business. The sooner you sell what you have the better price you will get.

Hey Howell,

I saw the modules that Buck suggested, Thanks to Buck for that. I am looking for a different set, but at the moment i'm just gonna use your method to try and prove it's a memory type conflict.

OCZ and my local reseller are still denying that it is a different type of hardware, and frankly im fed up with them: http://www.ocztechnologyforum.com/f...3-10700U-(Both-OCZ3G1333LV6GK)-on-ASUS-P6T-SE

I'm never buying their crap, that's for sure.
I'm just upset that i bought for nearly 200 euros worth of memory that appears to be incompatible with each other.

 

[Buck]

KrazeyKami,

I would be interested in seeing what happens if you were to use the HDD software from Samsung to switch your drives from SATA 2.0 to 1.0. If you look up your model at www.samsunghdd.com, under the download section, there is a software category, with two versions of the software.

buck

 

[KrazeyKami]

KrazeyKami,

I would be interested in seeing what happens if you were to use the HDD software from Samsung to switch your drives from SATA 2.0 to 1.0. If you look up your model at www.samsunghdd.com, under the download section, there is a software category, with two versions of the software.

buck

Hey Buck,

I'm willing to try that, since i'm desperate to find the true reason for this rig getting unstable.

@Howell,
I did what you said;
Atm im running 10700@A1,A2, 8500@B1,B2, and nothing@C1,C2.
The memory now shows up as dual channel, and guess what... solid as a rock.

I'm trying everything that before would crash my rig within 2 minutes, but it's running smooth and stable.

I'll wait a little longer, and then add 10700 to C1. If all goes well, i'll add the 8500 to C2, thus creating a mixed memory on channel C, and see what happens.

I'm not sure anymore what it proves, but to me it is a sign that for some reason, this memory just won't run on Triple Channel when being mixed together.

My vendor has agreed to send me some additional 10700 memory and take back the 8500. If my rig then stays stable @ Auto settings, im pretty much convinced that the types were different. If it crashes again, i'll look back in the direction of 12GB of RAM needing more voltage.

About that last point, something that was allready said by alot of people, i did up the voltage when having 12GB inserted. It didn't make it stable.

I tried 1.64-1.66 DRAM Voltage, and 1.3-1.35 QPI voltage. I tried different timings, up to 9-9-9-20, and tried DBCLK 1066-1333. Nothing helped to make the rig stable, with 12GB inserted. Funny point is, when only 1 set of 6GB inserted (doesn't matter which one), the rig was stable, on all those settings mentioned.

I always thought that on purpose mixing of 2 types of memory, atleast @ my experience with dual channel, was never a problem: the rig would simply run on the lowest memory speed. Still it is strange that the rig will only crash when copying large amounts of data, while having run smoothly for over a year, playing games, VMWare, running various programs, doing stuff on internet, running 24x7, etc.

i'll wait for my new DIMMS and see what happens with 6x 10700 @ Auto settings.

 

[Stereodude]

Wait, so it still runs dual channel even though the two channels are running at different speeds? I didn't think that was possible.

 

[KrazeyKami]

Wait, so it still runs dual channel even though the two channels are running at different speeds? I didn't think that was possible.

Well, according to OCZ... its the same memory, running at the same speed.
That's according to OCZ.

According to the facts, yes, my computer was running rock solid with 10700 in A1,A2, and 8500 in B1,B2. CPU-Z confirmed the memory was running @ Dual Channel, and it ran on the lowest speed (533 MHz).

After that, i added 10700 to C1... CPU-Z confirmed the memory was running @ Triple Channel. Guess what happened? It crashed, almost instant. After that, i removed the DIMM from C1, and added it to C2. It crashed. I removed the DIMM from C2, added the 8500 DIMM to C1, it crashed. I removed it and placed it in C2, it crashed.

After confirming it crashes in Triple Channel *Mixed*, i went on ahead and tested it on Dual Channel *Mixed*. The rig remained stable.
I even took it a step further, and inserted the memory in A1,A2 and C1,C2, just to rule out that the DIMM slots aren't defect. The rig remained as solid as ever.


So all in all, what does this mean?
This proves that for some reason, the memory won't run stable when mixed, and @ Triple Channel. The DIMM slots are not the culprit, nor is the Triple Channel mode by itself. It's only crashing when mixing the 2 types of memory @ Triple Channel. @ Dual Channel, the 2 types are running solid.

It freaks me out. Why would this memory run;
- stable when not mixed @ Double Channel;
- stable when not mixed @ Triple Channel;
- stable when mixed @ Dual Channel;
-Unstable when mixed @ Triple Channel?

Afaiac, this proves that the memory is not stable when mixed in Triple Channel.

OR: The rig just cant handle more then 4 DIMMS @ the standard voltages. Now, i tried to up voltages etc. as eleborately described in my previous posts, so don't think that is the problem.
We will know for sure once i have my 6 DIMMS of 10700 memory.

 

[Howell]

Well, I'm glad you were able to get some answers even though in the end you may never know exactly what was happening.

Ideally, we pay someone to figure out what will work and in return they give us an HCL. If a vendor is inept in developing their HCL then I don't give them my money. You and I have better things to do than free QA for a vendor.

 

[time]

Kami, in your first post you said, "No blue screens (although that option is checked ON)".

Now that you're a bit further along the path, I'd like to point out that the correct option is "Automatically restart" and it should be cleared, NOT checked.

Hope this helps.

Incidentally, the two different RAM types you have possibly DO have the same memory chips (you may not be able to see the numbers on the chips due to the heat spreader), but they clearly have different SPD tables.

Normally, a manufacturer would use faster chips before uprating their RAM modules ...

 

[KrazeyKami]

Kami, in your first post you said, "No blue screens (although that option is checked ON)".

Now that you're a bit further along the path, I'd like to point out that the correct option is "Automatically restart" and it should be cleared, NOT checked.

Hope this helps.

Incidentally, the two different RAM types you have possibly DO have the same memory chips (you may not be able to see the numbers on the chips due to the heat spreader), but they clearly have different SPD tables.

Normally, a manufacturer would use faster chips before uprating their RAM modules ...

Hey Time,

I know about the Blue Screen option I may have worded it wrong, but yea, the Automatically Restart thingie is cleared. Still there are no Blue Screens, and no usefull remarks in Event Viewer. There is no Memorydump made either.
This probably simply means that the computer is indeed suprised by the sudden restart. Thats why the 2 event messages saying "sudden power loss" and "unplanned restart" are written after i booted into Windows again.

As for the memory modules, i'm pretty sure that OCZ knows to a certain level what they are talking about. It's their product anyway. But, even tho if they used the same modules, chips and what not, it is clear that the SPD's are different. That's all i wanted to get confirmed from OCZ, so that maybe they can explain why this memory will behave as such when mixed. Instead all i get is 2 senteces: "The memory is the same" and "http://www.ocztechnology.com/support". Anyway, i guess the prove has to be in the pudding and i'll just have to test it myself with new hardware.

 

[KrazeyKami]

Hi everyone,

Yesterday i got a call from my reseller, telling me he had the new memory DIMMS. I took em home, plugged em in and confirmed it was PC3-10700 memory. I returned the PC-3 8500 memory to the reseller. I now have 6 DIMMS of PC3-10700 memory in my computer.

The system is running rock solid. I copied many TB's, while downloading, playing games, doing VMWare, after that i ran Memtest and Prime95 overnight, and all is stable.

Now, to me this proves that the problem was the PC3-8500 memory being mixed with the PC3-10700 memory. Ofcourse OCZ still claims that this memory is the same, but the fact is that it, the memory showing up as PC3-8500, is NOT COMPATIBLE with the memory showing up as PC3-10700. All these tests and real-life findings prove it.

So, no manual settings, no voltage tweaks, no faulty memory (as the PC3-8500 was tested for hours, 0 errors, AND ran fine on its own).

In the end the problem was exactly like i suspected:
It are 2 different types of memory, running on different SPD's and NOT being able to run stable together, while performing high loads on the RAM.

Hopefully this will help other people who are having trouble with 2 types of OCZ memory that 'should be the same'.

This thread can now be closed, as the problem is resolved by replacing the 'older revision of the same type of memory'.

Thank you all for reading my gibberish and coming up with good ideas
We'll meet again at the next upgrade / failure of my rig.

Kind regards,
Kami.

 

[MaxBurn]

Stuff like this is certainly frustrating and puts you through the ringer. Us old farts have learned the hard way like you have to put in a significant amount of research on critical components like memory and then only get exactly what you need. It's noble and all to work with a local vendor but much of the time I just find they don't have what I need.

 

[Howell]

I'm glad you were able to get to the root of the problem.

The remaining issue here is that there are real world problems that memory specific tests will not reveal. As a diagnostician that is difficult to stomach.

 

[Howell]

I wonder if testing with iometer would have revealed this problem rather than risking real data.

 

[KrazeyKami]

I wonder if testing with iometer would have revealed this problem rather than risking real data.

Iometer... never knew this existed, but it looks like just the right tool to test these things... Good to know it exists!
As for real data, yea, but i did it with a copy of my 'main' drive; the original data was never in any harm

 

[LunarMist]

Stuff like this is certainly frustrating and puts you through the ringer. Us old farts have learned the hard way like you have to put in a significant amount of research on critical components like memory and then only get exactly what you need. It's noble and all to work with a local vendor but much of the time I just find they don't have what I need.

Or not use OCZ.

 

[Mercutio]

The remaining issue here is that there are real world problems that memory specific tests will not reveal.

For what it's worth, I've never found memory testing to be all that helpful. Most of the time bad RAM just doesn't reveal itself until it's in a system. Even hardware testers are inadequate to that task.

 

[Buck]

For what it's worth, I've never found memory testing to be all that helpful. Most of the time bad RAM just doesn't reveal itself until it's in a system. Even hardware testers are inadequate to that task.

Quite true Merc. Hence, I've become single-minded as to which RAM I will sell or use.

 

[Mercutio]

Exactly. Everything I buy needs to be Crucial, Mushkin or Corsair and it all needs to conform to JEDEC spec before I'll buy it. I've been bitten a few times by RAM that wouldn't even work unless it was overvolted. I'm not touching that stuff again.

 

[Howell]

I especially like that Crucial memory has a life time transferable warranty which makes used sticks valuable. Does Mushkin, Kingston or Corsair offer the same?

 

[Buck]

I especially like that Crucial memory has a life time transferable warranty which makes used sticks valuable. Does Mushkin, Kingston or Corsair offer the same?

I don't know. I only use Kingston, and I have never had a RAM failure from them in the last 10 years. So, I haven't had to transfer warranty.