GA-P55-USB3 incompatible with 2x4 GB RAM? Weird problems

[lores]

Hi!

I'm starting a new topic because a nearly two weeks' research hasn't been able to solve my problem.


I've bought the GA-P55-USB3 a few months ago together with a 2x2 GB G.Skill 1333 MHz kit. It had been working perfectly (with i5 760) until I tried to replace my old 2x2 GB set with a new 2x4 GB set from Corsair, Corsair XMS3 DIMM Kit 8GB PC3-12800U CL9-9-9-24 (DDR3-1600) (CMX8GX3M2A1600C9) (1600 MHz at 1,65 V via XMP; 1333 at 1,5 V via SPD). As soon as i removed the 2x2 GB kit and inserted the 2x4 GB one, problems appeared. I have been able to track them down to the following:

- regardless of the DRAM Voltage, Voltage Refs, Termination Voltage, Uncore, QPI, QPI Voltage, DRAM Timings, DRAM Divider and DRAM Freq settings, the computer is highly unstable whenever the 2x4 GB kit alone is used - which manifests in reboots, kernel panics in Linux and BSODs in Windows;
- memtest86+ 4.1 doesn't detect any errors whatsoever after running for 10 h multiple times at multiple settings;
- all other components are 100% stable;
- whenever only one stick of the 2x4 GB set is used, it works perfectly;
- if I revert to 2x2 GB (in dual channel), it works perfectly, also with all the settings kept from the 2x4 GB kit BIOS configuration.

After 10 days of tests and arriving at these conclusions, I was convinced that either the 8 GB RAM is not compatible with my mobo or that it is partially corrupt (e. g. interleaving in dual channel). Therefore, I ordered another kit with the very same parameters as my good old 2x2 GB from GSkill - 8 GB (2x4 GB) TeamGroup Elite DDR3-1333 DIMM CL9 Dual Kit. To my astonishment, after the first boot with the new 8 GB set, the very same problems reappeared! The symptoms are the same (see the list); the settings don't seem to make any difference; memtest is still OK.

An even more surprising result I achieved with 4x4 GB - I kept one TeamGroup 4 GB stick in channel A and the other one in channel B and then added a Corsair 4 GB stick to each channel; dual channel mode active. Result: stability! (But need to test this config further in order to confirm this).

I tried reflashing the BIOS to F10g, F9 and F7 (is there any newer BIOS version for this mobo?); I cleared the CMOS and the DMI pool data every time.


My question is: how the heck is this happening? Is the mobo kaputt? Is it lacking compatibility? Is this possible anyway?

Hoping to get some input here, thanks!

[Aussie Allan]

 Lores....Welcome buddy!

 I already know you think this not applicable but that 10 hour Memtest you mentioned,....was this with one stick or Both?.....10 passes per stick is what's required to absolutely rule out a defective memory module, ...........4 years ago I got two CPUs in a row that were shot!....it does happen, mind you it took me about 20 hours to come to that conclusion the second time!

  And at the moment I know it's frustrating as hell not knowing what the actually problem is, ..... but you said with 16Gb of memory on board filling all the slots there no problem what so ever!....... in fact, stable as hell!...........is this now the new problem you want fixed , because if it is, I'm lost where you want me to start!....and no, not even station drivers has a later BIOS

 Aussie Allan

[lores]

Thanks for your quick reply!

I had run memtest over the corsair 8 GB pair overnight. Over the single sticks, I had just run a pass each. Same with the new TeamGroup pair.

I had stated that 4x4 GB be stable. However, I fear I might have been wrong - I'm testing this config right now and have had a few reboots already; they could be attributed to wrong timings etc. (the two kits use different SPD settings) or it could be the same problem, just distributed over 16 GB, 8 of which are in the same channel (which seems to be working correctly). That is the reason why I first thought that 4x4 GB be completely stable (waaay fewer reboots)...

[Aussie Allan]

 Right , ...Memtest by it's very nature run each stick at the rating it was released at......this is why you run them individually.

 10 passes is an industry ( and home user) unspoken rule that a pass at this level mean the stick are good! and you can absolutely rule this out of the equation.

 If you don't do this first, .... it will be nagging at the back of your and my mind that we never really tested these properly, .... and with the issue you have, every thing you do will have to be by the book to find the problem by ruling out one thing at a time.......sorry but I think you should start by testing the memory properly.........(waaay fewer reboots)...could also be pointing to a duff module as it could be spreading the parity so to speak!

 the other thing I thought of was , .... PSU, what Wattage ?

 Aussie Allan

[lores]

You're right; I've now been running memtest on the first Corsair stick since the last post and will have both Corsairs checked completely (10x each) in some 13 hours (if I get the 2x4 GB Corsair kit to work, I'm happy and don't care about TeamGroup; additionally, I've been messing around with that kit for nearly two weeks now, as opposed to TG which I've had for just a day now).

Are you 100% sure that a positive memtest is an ultimate warranty that a given stick is healthy? I'm not sure how relevant the chip is to dual channel operation; I read somewhere that old RAMs used to have problems with rank/channel interleaving (in dual channel mode). Therefore, I could imagine that a stick could pass memtest 10 times but still be unable to operate in dual channel in spite that the motherboard be fine.

The PSU is OCZ StealthXStream2 500W, serving:

Intel Core i5-760, stock MHz
Gigabyte P55-USB3
RAM...
GeForce GTX 460 1 GB (PalitNE5X460SF1102), OCed
HD322GJ + HD103SJ

[Aussie Allan]

  I was hoping you weren't going to say that!....... did you know the video card you have , suck down 270Watts under load, leaving 230 watts for the rest of the system at 100% loading........May still not be ultimately the problem, but Bloody good chance it is!

 Goggle up a PSU configurator and load in your system, I like to have a head room of about 30% over what they say, .... power supplys drop in performance with age, and two, leave a bit for adding toys down the line!

 Persevere with the memtest though, not much graphics here and you may even kill two birds with one stone!

 Aussie Allan

[lores]

Hm, I strongly doubt that my PSU could be the root of the problem: The PC used to work for half a year without a single issue, although the i5 was OCed to 4x4 GHz 24/7. Since buying the 2x4 GB kit, I have been running the tests at stock CPU settings and I couldn't possibly imagine that the additional 4 GBs be drawing the amount of power that a 50% OC of my CPU used to draw all the time. As far as I know, ddr3's power consumption is not much compared to other PC components. But still - in the following tests, I will be running my i5 at 1,3 GHz and my 460 GTX at 550 MHz to rule out the PSU as culprit - thank you for that remark!

Do you think that CPU clock skew [ps], CPU Clock Drive [mV] or PCI-E Clock Drive [mV] are relevant? I tried adjusting the skew, but the mobo doesn't allow me to choose the direction, so I could just go from 0 ps to ca 600 ps; doing so in 100 ps steps didn't help. I also tried fiddling with the Clock Drive (possible values: 700 mV - 1000 mV in 100 mV steps), but still no go... Which values are recommended in my case?



EDIT:
http://www.antec.outervision.com/index.jsp recommends about 450 W for my config.

[Aussie Allan]

 With the PSU at least 6 months old I'm not going to rule it out just yet considering your safety margin is 50 Watts.....and even if it's a real goodie rated at 90%.....you've just blown the margin, but we'll see.

 Been out of 775s for too long now to comment really, afraid you would have a better Idea then me.... with all CPU / MB combos being so variable, ...be hard pressed to find anyone to put there money on an answer to that one......push come to shove, ....No! I don't think it's relevant at this stage

  Persevere with the memtest........one last thing, have you had the CPU out resiliently ? or anything else for that matter just prior to things going tits-up ?

 Aussie Allan

[lores]

I guess you're right about the watt margin, I should have went for a bit more. But ATM, my priority is getting one of the 2x4 GB kits to work...
i5 760 is 1156, not 775.

"one last thing, have you had the CPU out resiliently ? or anything else for that matter just prior to things going tits-up ?" - what do you mean? The CPU (as well as the rest of the system) was and is 100% ORTHOS-stable. It's worth noting that I can replace the 4 GB sticks (be it one or two pairs) with my old 2x2 GB kit and all works like a charm again, even if I keep all the settings.

[Aussie Allan]

  1156 ....sorry .....answering to many posts,they start to blend into one after a while!, front lobes going mushie after 11:00 lately, ....and the spelling

 What do I mean!....well actually it's nothing personal, ..... Just doing a little mental hunting really, I've put in hundreds of CPUs and resonantly got a speck of solder in the LGA nest from a solder sucker repairing a dry joint........felt a real Plonker when I finally found it .

  Get Memtest out of the way,  no use getting ahead of our selves.

 Aussie Allan

[lores]

Okay. The CPU's been sitting in the socket for six months now and it's still stable, so I don't suspect it. What I'm not sure about is the motherboard, but - as you say - let's wait for memtest first... Thanks for your assistance this far!

[lores]

Okay, I tested both Corsairs separately with 10+ passes in my rig - no errors whatsoever. On another PC with an i3, I also thoroughly tested the new TeamGroups: same results. Additionally, I've been briefly running the 2x4 GB Corsair pair on that i3 PC and haven't experienced any reboots (but, again, it was just ca half an hour of use).

I got another pair of 2x2 GB DDR3 1333 MHz sticks, from Elixir (stable). They work perfectly in my rig, but as soon as I run them together with my old 2x2 GB G.Skill pair (resulting in 4x2 GB), the instability emerges once again. This leads me to the conclusion that my RAM isn't the culprit... Do you agree?

What I'd very much like to know now is whether such a faulty behaviour could possibly be caused by a corrupt memory controller in my i5. If not, the only option I see is the GA-P55-USB3 mobo... Are there any newer BIOSes than F10g (beta)?


EDIT: (Why can't I modify my posts from yesterday?)
I wasn't able to find any settings to get the 4x4 GB Corsair + TeamGroup 16 GB set to work stably: although they seem to work way better than a single 2x2 GB pair, there still are recurring reboots/BSODs. I've never managed them to run through one standard 3DMark Vantage test under W7 64 bit, although a more GPU stressful and power hungry Furmark seems stable; I can run through a full 3DMark06 test under XP 32 bit. This behaviour doesn't depend on the settings of my GPU or CPU (3DMark Vantage also fails with my GPU at 550 MHz (instead of ca 800 MHz) and CPU at 1,3 GHz (instead of 4 GHz) undervolted at some 0.8-0.9 V (via auto)). With one 2x2 GB pair (be it G.Skill or Elixir), the test is no problem whatsoever. With 4x4 GB, the PC also sometimes reboots without any considerable load (no running foreground apps). I suspect that it had seemed rather stable because there were proportionally fewer inter-channel operations with 4x4 GB than in a straight-forward 2x4 GB config.

[Aussie Allan]

Okay, I tested both Corsairs separately with 10+ passes in my rig - no errors whatsoever. On another PC with an i3, I also thoroughly tested the new TeamGroups: same results. Additionally, I've been briefly running the 2x4 GB Corsair pair on that i3 PC and haven't experienced any reboots (but, again, it was just ca half an hour of use).

I got another pair of 2x2 GB DDR3 1333 MHz sticks, from Elixir (stable). They work perfectly in my rig, but as soon as I run them together with my old 2x2 GB G.Skill pair (resulting in 4x2 GB), the instability emerges once again. This leads me to the conclusion that my RAM isn't the culprit... Do you agree?

A) Not 100% but Yes , I tend to agree with your findings, ....why I said not 100% is even though there identical timings there still from different manufactures...........with dual and triple channel ...even a different batch from the same manufacture can get different results....hence dual and tri kits on the market.........

What I'd very much like to know now is whether such a faulty behaviour could possibly be caused by a corrupt memory controller in my i5. If not, the only option I see is the GA-P55-USB3 mobo...

A) where down too CPU and M/B (channel/memory controller)........with the PSU as a long shot outside........good to get the Memory out of the equation though..... next one to rule out would be the CPU unfortunately..........Does your mates friendship stretch far enough for some Mobo time with your processor ?  Failing this another CPU in the problem board could give the answer too.

 Are there any newer BIOSes than F10g (beta)?

.....A) not that I'm aware of...station drivers is the cutting edge of new releases and they show F10g as well!

  Conclusion so far....you need to test the other bits and bobs but CPU , possible but not likely, starting to smell a motherboard hunt coming shortly.

 Gotta call it how I see it.......see how the CPU stands up on another board  or the other option I mentianed.



 Aussie Allan

[lores]

Okay, thanks. So I guess I'll have to look around for some CPU/mobo to rule out my i5...
What troubles me is that memtest doesn't show any errors whatso. Are there any other mem testing progs designed for dual channel? I heard that memtest's tests 5 and 8 utilise dual channel, but that doesn't seem to be enough...

[lores]

Since I've replaced my i5 with an i3, all has been working like a charm (seemingly, at least). I will do further testing and will also try to run my i5 on another motherboard.