SOLVED My shiny new processor. It crash. :(

Hi, looking for diagnosis advice. My level of knowledge is "can fit a processor, knows how (but not when) to tweak the CPU settings in the BIOS".

I have an Asus M4A79XTD EVO motherboard. For the past year, it's had a happy and loving relationship with a Phenom II X2 3.1GHz CPU and 4GB of DDR3 1600MHz of RAM.

A couple of weeks ago, I fed the system an extra 8GB of RAM on two sticks and it liked that.

Last week, I took the dual core CPU out and added in a Phenom II X6 3.3GHz CPU. Problems began.

After an initial issue where the processor was "an unrecognized AMD", which was fixed with a BIOS update, things *normally* work. But sometimes the computer just hangs or shuts down. Probably once a day on average, but a couple of times one day, not at all another. This happens regardless of O/S (I dual Win7 and Ubuntu). The only thing I can guess at is that the problem is something wrong with the CPU, unless I had a string of lucky days the week before with the new memory. But I've run a full stability test with AMD Overdrive for an hour and it was fine. In Linux, I've been keeping an eye on the temperature and it's never seemed to go over 42 C which I *think* is okay? I don't know if a new BIOS can cause problems like this.

What should I do? I don't want to return the CPU if it's okay. Do motherboards wear out? Could I have missed a spot with the thermal goop? (It slid around a bit when I was fitting the processor). Do some motherboards have problems with some CPUs? The crashes don't seem to be triggered by any particular heavy load (unless opening a blank text file counts as heavy load which is when it went down this morning).

Sorry for long-winded post: I don't know what's relevant and also go on a lot.

Thanks for any suggestions. I'd actually hoped to try overclocking with this as I heard it was good for that, but I'm not sure I should even try.

Harmony.

EDIT: Just occurred to me to add that the CPU fan on the new processor (I'm using what came with it) is a lot louder than the old one - it really sounds like it's getting a workout and the only way I got it to slow down was by turning up the fans on my case, at which point it noticeably relaxed a bit more. But the core temp in the Linux System Monitor says the CPU is hovering around the 39 C mark most of the time and in Win7, the AMD Overdrive program says it's about 30.5 C right now - (just browsing these forums).

Those temps are perfectly fine for that chip.

So you've updated the bios? To the latest version?

Also, check your RAM settings and all that, as well as the CPU. Don't leave anything in the BIOS for major settings like voltage and timings as auto, it could be causing instability. (though voltage for the CPU at stock should be fine... using cool n quiet and all that).

But the temps are perfectly fine, it's not that at all.

Mjolnir said:

Those temps are perfectly fine for that chip.

So you've updated the bios? To the latest version?

Also, check your RAM settings and all that, as well as the CPU. Don't leave anything in the BIOS for major settings like voltage and timings as auto, it could be causing instability. (though voltage for the CPU at stock should be fine... using cool n quiet and all that).

But the temps are perfectly fine, it's not that at all.

Click to expand...

Thanks for that. Everything I can cross off the list is useful. Yes, BIOS was the latest that I could find on ASUS' website.

Everything is currently set to Auto in the BIOS. I'll set things to specific values next. E.g. 800MHz, Multiplier. 4.00x, which is the lowest my BIOS seems to allow. I'm slightly confused by that actually, as I thought things were normally in the ranges of, e.g. 200MHx and multiplier of 14+. But those aren't the sort of numbers the BIOS offers. Possibly just different terminology?

Having said in my initial post that it averages a crash a day, btw. I've just had my fourth for the morning.

Thanks,

H.

Sounds like it could be a PSU issue. What powersupply are you using? It's possible the new CPU just draws more power than the old one, and your PSU isn't up to the task.

CompuTamer said:

Sounds like it could be a PSU issue. What powersupply are you using? It's possible the new CPU just draws more power than the old one, and your PSU isn't up to the task.

Click to expand...

That would never have occurred to me. I just throw hardware at it and forget about it! I have a HX520W Corsair PSU. I would have thought it would be up to the task, but I've never really thought about it. The system has:
Motherboard and CPU mentioned.
CPU fan (stock)
12GB DDR3 1600MHz Ram (2x2GB, 2x4GB)
1x 64MB SSD
4x SATA drives (2x 500GB, 2x 160GB, in two RAIDs)
1x Blu-Ray drive (assuming irrelevant as not running during crashes)
1x Radeon 4870 Graphics Card.
1x Asus Xonar D2 Sound Card
Case has two 120mm fans.

I don't draw any power off the USB (or hardly ever). I can't think of anything else in the case.

That PSU is more than fine.

Remove the 2x2GB RAM sticks and see if you have the same issue. Don't mix ram modules like that if you can avoid it.. 8GB should be more than enough anyway..

Mjolnir said:

That PSU is more than fine.

Remove the 2x2GB RAM sticks and see if you have the same issue. Don't mix ram modules like that if you can avoid it.. 8GB should be more than enough anyway..

Click to expand...

That would be a shame. Though a memory problem would explain how the system gets through stability tests fine. Maybe I did just have a lucky week with the memory before I put the new processor in.

I do need the 12GB however - hard though that is to believe. Aside from video encoding (not rips, art projects) the reason for all the cores and memory is for virtual machines. I've been setting up a RAM disk in Windows, which gets loaded up with VM(s). That amount of RAM enables me to (a) give the VM a decent amount of memory itself and simultaneously (b) run the whole thing in RAM without writing anything to disk. Even with an SSD, that's a substantial improvement. Any data that the VM needs to preserve is written off to a network drive elsewhere as and when updated.

The memory sticks all have the same speed and latency and are the same brand (Corsair), albeit not the same line. The 2GB sticks are from their Dominator line, the 4GB ones from their Vengeance line. I don't know why they can't just have a line called "Plays Well With Others" or "Happy To See You" instead of all this aggression, but there you go.

The modules are paired with each other, but not with the other pair. I thought it was okay to have different ones so long as they were done in matched pairs and had the same spec. Is there any way that I can test the memory? I know Linux has that memtest thing on start up?

We have the same board in use, with a phenomII 965BE.
The fan noise is correct, pretty loud and getting louder when gaming. Fortunately we fitted a 6870 underneath, compensating for the cpu fan noise

Sadly though, from day one we are struggling with the memory. About the same issues you describe, we loosened the timings, that seemed to help, but not for long, changed voltage on the ram, helped just as long, eventually we came across some official and tweaked memory settings which did better. The pc has a Antec EarthWatts EA-650 psu, sufficient for all tasks.
Funny though, ram voltage should be 1.65v, but works just as well at 1.5v. We downclocked the ram, we tried everything.
Now it's more or less stable with the occasional hickup.

I know from the past memory sticks should have the same amount of banks for full compatibility, 32x32 or 64x64. Don't know if this is still the case today. Don't know the numbers either. Mixing memory sticks should be possible, but maybe it's the cause of your problems.

Bisquit said:

We have the same board in use, with a phenomII 965BE.
The fan noise is correct, pretty loud and getting louder when gaming. Fortunately we fitted a 6870 underneath, compensating for the cpu fan noise

Sadly though, from day one we are struggling with the memory. About the same issues you describe, we loosened the timings, that seemed to help, but not for long, changed voltage on the ram, helped just as long, eventually we came across some official and tweaked memory settings which did better. The pc has a Antec EarthWatts EA-650 psu, sufficient for all tasks.
Funny though, ram voltage should be 1.65v, but works just as well at 1.5v. We downclocked the ram, we tried everything.
Now it's more or less stable with the occasional hickup.

I know from the past memory sticks should have the same amount of banks for full compatibility, 32x32 or 64x64. Don't know if this is still the case today. Don't know the numbers either. Mixing memory sticks should be possible, but maybe it's the cause of your problems.

Click to expand...

Interesting. Thanks. I'm starting to think that it might be the combination of the motherboard with the CPU. I haven't got round to taking the memory out yet and I haven't had a crash in three days whilst the morning I started this thread, it went down four times iirc. The sole difference so far has been following Mjollnir's advice and setting everything I can to a specific value, rather than leaving things on Auto. I also turned off the AMD Turbo whatever that is. Rock steady so far... And yes, I know that means it will probably crash as soon as I post this.

If it goes wrong again, I'll take the memory out and try it like that.

The fan whines like a spoilt child though. I have to get a replacement for that. I don't know whether to just get a different fan or a whole new cooler. The one that came with the chip looks pretty beefy and copperish to my untrained eyes.

Thanks for all comments.

If you need a new cooler, and need to stay cheap-ish, http://www.svc.com/rr-b10-212p-gp.html is decent. If you have a bunch of burning cash to throw around, and want silence, try http://www.svc.com/nh-d141.html.

sounds more like a memory issue to me, are your 2x2 and 4x2 sticks the same speed, brand? If not, and if you got to have 12gigs, I would sugest taking out the 2x2 sticks and replace them with matching 4x2 sticks. you'll get 16gigs but might solve your issue.

What timines do you have the mem at? what multuplier for the ram. may want to down clock the mem, if your putting 1066 and 1600 mem in you will would get stability issues and would want to set the mem at 1066.

when i do ram upgrades, I always buy kits for the amout I want, not a kit to add to my old mem to get there..

haha... that sucks... im currently considering getting a xeon processor over an i7

You need to determine the slowest memory(timings) modules in there. If you can tell us the part numbers or the links we could check. When you find the slowest memory, you need to set memory settings manually to those. Thubans have different memory controller so different size or brand memories wouldn't work as well as older cpus.

It could have to do with which slots the RAM are in. Certain motherboards would have you place identical sticks (size specific) in the 0/2, while the other are in the 1/3 slots. Some are 0/1, 2/3. This should be in your manual.

Again, thanks for all the replies, you're all great. First off, I've followed petteyg359's recommendation and ordered a Noctua NH-D14 because since I put the new HSF in, my box has started sounding like a hair-dryer factory, whereas before it was as loud as two gnats shagging. Bit of a tangent, but wanted to let you all know that your recommendation has been taken up.


Since manually setting the processor speed rather than the motherboard's Auto, as well as turning off AMD Turbo, I have had perhaps three crashes which is a lot less than where it was. So there's something screwy going on.

Information on the memory timings:

I have
2x 2GB sticks of DDR3 Corsair 8-8-8-24 Dual Channel (Dominator)
2x 4GB sticks of DDR3 Corsair 8-8-8-24 Dual Channel (Vengeance)

Weirdly though, looking at AMD Overdrive, it says that the CAS Latency is "9 clock". Should it not say 8? The frequency according to AMD Overdrive is 667MHz.

The manual has this to say on memory modules:

YOu may install varying memory sizes in Channel A and Channel B. THe system maps the total size of the lower-sized channel for the dual-channel configuration. Ane excess memory from the higher-sized channel is then mapped for single-channel operation.

We recommend that you install the memmory modules from the blue slots for better overclocking capability.

Always install DIMMS with the same CAS latency. For optimum compatability, it is recommended that you obtain memory modules from the same vendor.

Click to expand...

Slots A2 and B2 are the blue ones in this board and they have the 2 4GB modules in them. The part numbers in Slots 0 and 1 are CMD4GX3M2B1600C8, and 2 and 3 have CMZ8GX3M2A1600C8.

I'm going to follow catscratch's advice now and try altering the memory timings. Although now that the system is crashing much more rarely, it will take a while to identify any improvement. But I intend to get to the bottom of why my latency says 9 instead of 8, anyway.

Again, thanks all.

H.

Memory set to Auto will give very loose timings (9(CAS)-9-9-34) or worse. Please post screenshots of CPU-Z on the CPU, Memory and SPD tabs so we can get a better understanding of where things are currently.

Also, what power supply are you using? That's a lot of drives and a power hungry video card to be running with a 125W processor.

Thubans in 790 chipset boards does function but not as well as in an 890 chipset board; the chipset designed for the Thubans.

When looking for tips towared better use of Thuban on 790 chipset boards the Vcore and Bridge voltages being raised has seemed to bring a greater air of stability to the system.

I'll post screenshots later when I'm booted into the Windows partition.

The power supply is the one from earlier, a HX520W Corsair PSU.

I would consider buying a new motherboard just to get this working, but at this point, I'm becoming quite disappointed. If the motherboard has an AM3 socket, I should be able to put in any AM3 processor in my humble opinion. I've taken out the old 4GB RAM sticks mainly due to the stock cooler that came with the six-core screaming like a banshee in use so I replaced it with the NH-D14 that another poster recommended. This cooler looks like it ate several other coolers on the way to my house and is quiet as a mouse in church, but wont fit unless I remove two of the memory sticks. And that's with one of the fans removed! If I had both fans attached then I'd have to remove all of the memory sticks.

So the result so far, is that I've spent money on extra cores and RAM which I do need for my work, only to end up buying a replacement cooler (okay, I bought a quality one, instead of any old one, but it would have been a waste to buy something which wasn't much better than the stock for noise) and had to ditch 4GB RAM to use it. And now I'm looking at the prospect of buying a new motherboard. I'm philosophical, but I would like this situation to be better than it is.

Okay. I've done CPU-Z screenshots of the CPU, Memory and SPD tabs. For the memory tab, Slots #3 and #4 are identical and Slots #1 and #2 are empty since I took the memory out.

I get a crash anywhere from 0 per day (if I'm lucky) to three, maybe four. The system never did this before the new processor.

If I do end up having to buy a new board, any recommendations? But I would really prefer not to right now unless it's the only way.

Thanks for all the help, you people.

H.

I also forgot to ask -- did you do a fresh Windows installation after replacing the chip? Shouldn't be absolutely required, but certainly couldn't hurt.