Odd memory problem

33 replies [Last post]
Kofoed
Offline
Joined: 06/01/2010
Posts: 17

Hello 3Dgameman forum.
I was hoping one of you could give me some advice on a fairly odd memory problem my pc have.
The thing is, whenever I use the memory slot 3 or 4. It won't boot at all! When I use the first 2 memory slots, it works just fine though.
I have tried in 2 different motherboards now, and with 2 sets of memory.
So if it is not the ram, neither the motherboard?
What can it be?!
I can't use 3 or 4 slot in total either. I got 4 memory modules but can only use the 2 :(

EDIT: some system spec if needed.
Corsair XMS3 memory
Amd 965 BE cpu
GA-990FXA-UD3 motherboard
And a cooler master silent 1000W psu
Thanks in advancd.

eire1274
eire1274's picture
Offline
Joined: 09/12/2003
Posts: 1221

It does it with every memory module?

Weird... I've used that exact motherboard on a build, and used all 4 slots (4Gb x 4). Is that memory on the Gigabyte compatibility chart? We were using the Corsair Dominator which met board specs right on.

That's all I can say, with two motherboards and two memory kits, must be some obscure timing method that the motherboard just can't read. I know it's not the AMD chip (which is your memory controller).

Nick McDermott - 3dGM Admin

Kofoed
Offline
Joined: 06/01/2010
Posts: 17

I figured by turning off channel inteeleaving. It boOts, but shows only 4GB installed.
Also in windows it says 8GB installed 3,92usable
Can it be my CPU? Or anything else?

eire1274
eire1274's picture
Offline
Joined: 09/12/2003
Posts: 1221

That shows that banks 3 and 4 are available and recognized, but not addressed (because the slots aren't being interleaved).

The CPU only has two memory channels, which the board breaks apart into 1,3 and 2,4. Essentially, if you get memory on 1 and 2, then the CPU isn't at fault.

Have you tested the memory on another board?

STRIKE OUT: More and more it seems like a straight out incompatibility.

Nope, just verified that the mobo and memory are compliant. Two bad motherboards in a row? Hard to believe, as the board is Gigabyte's top dog series on the AMD side.

Tivon, when did we lose the strike out command? I'm positive I was using it on the new software...

Nick McDermott - 3dGM Admin

Kofoed
Offline
Joined: 06/01/2010
Posts: 17

Im pretty sure it should all work together, can it really be 2 bad motherboards?
Seems very very unlikely :P

sherretz
sherretz's picture
Offline
Joined: 07/10/2003
Posts: 100

Maybe a BIOS flash would help?

What was the other mobo that had the problem? Or was it another board of the same model?

E8400 Stock; CoolerMaster H50; Asus P5Q Deluxe; G. Skill 2*2GB + 2*1GB PC2 1066; Sapphire 4850; WD 6400AAKS; Win 7 Pro; PCP&C Silencer 750W; Antec 300

Kofoed
Offline
Joined: 06/01/2010
Posts: 17

I tried flashing it to the newst bios F4 it's called.
My older motherboard was an Asus M4A89GTD PRO
I decided to upgrade for the new AMD cpu's that suppose to be released soon! :)

EDIT: Just to make sure, there is No way that a PSU can cause this? Right :P???

sherretz
sherretz's picture
Offline
Joined: 07/10/2003
Posts: 100

I PSU could only cause this if it fried and managed to take part of the mobo with it. In other words, you would have other problems (like being unable to turn the computer on and a bad smell in the room) before you would see this.

Unless you had a fried PSU recently and are still using the same mobo.........

Kofoed
Offline
Joined: 06/01/2010
Posts: 17

Well, a power rail could that have taken damage dosen't mean it have to frie anything.
I still think its very unlikely though.

But really? Could I have gotten a bad mobo?
Seems odd :/

eire1274
eire1274's picture
Offline
Joined: 09/12/2003
Posts: 1221

I thought this was two of the same 990FX boards. Yes, if you only have one, you could likely have gotten one that was damaged following it's QA at Gigabyte.

First thing to do, test every memory module INDIVIDUALLY, using Ultimate Boot CD (link below for the ISO). If every memory module passes, then it is likely the board is bad and is not handling memory banking correctly. Gigabyte is excellent in handling RMAs in most areas and will have you up quick.

http://www.ultimatebootcd.com/download.html

Nick McDermott - 3dGM Admin

Kofoed
Offline
Joined: 06/01/2010
Posts: 17

I just find it rather wierd.
My old motherboard is having the exact same issue :P
I haven't had time to test them just yet. But I will let you know by the end of the day.

sherretz
sherretz's picture
Offline
Joined: 07/10/2003
Posts: 100

Any obvious damage on the memory chips? What about the DIMM connectors? Any plating scraped off or otherwise marred?

At this point I'm just shooting in the dark, but stuff like that is easy to check.

eire1274
eire1274's picture
Offline
Joined: 09/12/2003
Posts: 1221

What's got me here... have you shuffled the DIMMs around? Put the two from 3,4 into 1,2? If so, did they work?

Nick McDermott - 3dGM Admin

Kofoed
Offline
Joined: 06/01/2010
Posts: 17

Hello again, I talked with Gigabyte support and they said the memory is not "fully" compitable.
But at corsair site,
http://www.corsair.com/memory-by-product-family/xms-classic/xms3-ddr3-memory/cmx4gx3m2a1600c9.html
It. Says "Guaranteed to work with all intel and AMD platforms"
Its this motherboard/CPU not dual channel? :O
Also, I am going to check the memory slot for damage and set the voltage and timing manual.
No luck this far
..
And to Eire1274 - Yes, but i figured, the slots that I though was 1+2, is most likely 4+2.. Atleast that is what the motherboard manualn says :P
It does not work in slot 3 and 1 though, its all dead :(

EDIT: Alittle update, when I boot windows with Channel Interleaving off, and 6GB installed. CPU-z detects all slots is being used. But windows won't use more than 3.92GB :/

Something wierd happend, after rebooting the pc, the bios says only 4gb installed memory, but if i uinstall the ram, and put then back in. It detects them again.
But only untill next reboot. However, windows keep saying 6gb installed even when the bios says 4gb.
Bad mobo?

Edit: I had my friend try the ram, he is using amd phenom 965, as myself.
M4A89GTD PRO USB3 motherboard - And all 4 memory slots, also windows said he was able to use 8gb ... :) my old M4A89GTD motherboard does not work with more than 2 :(.. Just like this one from

suspicious!!

Kofoed
Offline
Joined: 06/01/2010
Posts: 17

I might aswell RMA it before its too late, although im still not sure if it will work even, after that.
Also, since the memory worked in my friends machine. Can it still be bad RAM, not supported or something?
Im out of ideas..

sherretz
sherretz's picture
Offline
Joined: 07/10/2003
Posts: 100

Does his RAM work in your machine? That's the only way to boil it down to the chips themselves. You could also have him run a memtest86 on your RAM to see if it finds anything wrong.

Just out of curiosity, do you see any obvious signs of damage to the RAM slots, or do any of them look like they are not fully connected to the motherboard in any way?

Kofoed
Offline
Joined: 06/01/2010
Posts: 17

No visuel damage to the ram, or slots.
And his RAM didn't work in my machine either :/ So I guess I got another bad mobo!
Since this exact issue happend twice now. Can any other part perhaps cause this?
Like, could the PSU give too much power and frie each and every mobo I try with :P?
Sadly we didn't do a memtest, but I did one on my own and it reported no issue on any of the 2 working slots, with any memory.

eire1274
eire1274's picture
Offline
Joined: 09/12/2003
Posts: 1221

I tend to get excellent boards from Gigabyte. My issue is that two boards in a row (different chipset AND different manufacturer) have shown similar behavior, and the chances of that are slim to nil. I'm really thinking you've got some toasted memory, DIMMs that don't want to pair up.

A PSU frying a mobo... typically won't just kill the memory bus interleaving system and allow the remainder of the chipset to survive. Also, two in a row, doesn't make sense, especially as the memory system is controlled by the ATX power connector's connection to the mobo chipset, not looking at voltage fluctuation due to 12-V input at the CPU (which can cause strange behavior due to the CPU being powered by an independent voltage control unit on the mobo).

What I need is for you to do is get somebody to test each DIMM INDIVIDUALLY (only one put into the board at a time) using MEMTEST86+ or a similar bootable memory tester. The Ultimate Boot CD I listed above includes several memory testers including MEMTEST86+. One DIMM showing issues can be the root of this whole problem.

Nick McDermott - 3dGM Admin

Kofoed
Offline
Joined: 06/01/2010
Posts: 17

Alright, I see if I can lend my friends system for a little again tomorrow :P
Still think its odd how the memory worked in his, and his working memory dosen't work in my system though. But what do I know :)

Kofoed
Offline
Joined: 06/01/2010
Posts: 17

Alright, I see if I can lend my friends system for a little again tomorrow :P
Still think its odd how the memory worked in his, and his working memory dosen't work in my system though. But what do I know :)
EDIT:
Sorry for double post

eire1274
eire1274's picture
Offline
Joined: 09/12/2003
Posts: 1221

No no no, test it on YOUR SYSTEM! We are looking at incompatibility between the two parts, memory and mobo.

What memory does he have in his system? Same kind? Or different kind? If it's the same, then it's an obvious incompatibility.

Nick McDermott - 3dGM Admin

Kofoed
Offline
Joined: 06/01/2010
Posts: 17

Its some kingston memory.
Kingston hyper something :P 3 memory modules with 2Gb at 1333MHz
I tried to downclock my own memory to the same timing and MHz. Didn't work though.
His memory did not work in mine either :/
EDIT: Forgot to say, I alredy did test it with memtest earlier, no errors.
Also, incompatibility :/ Switching CPU didn't fix it, memory didn't either.
Only PSU/mobo and our case is different now :P
Must be the Mobo then, if you still belive the PSU can't cause it.
EDIT again: Im tired of trying the ultimate boot thing.
I just cannot get it to work. I hope memtest is alright as its the only thing I can get it to boot on :P
And by just testing 2 of the memory without any errors, they still don't work in slot 1 + 3 or 2+4 ;/
Anyhow.'
tested them on different slots, with channel interleaving on. Only the slot 3 and 4 worked. the 2 closest to the CPU.
No errors on the tested memory. They still works in my friend AMD build :P
Guess it down to the board.. Or PSU xP

oops, Looks what I found!
http://www.tomshardware.co.uk/forum/296275-12-cannot-memory-slots-dual-channel

Looks like im not the only one having this kind of a problem..! God damnit, the returntime for my motherboard(15days) have run out :/ Im stuck with this mobo now..

eire1274
eire1274's picture
Offline
Joined: 09/12/2003
Posts: 1221

Contact Gigabyte for an RMA, you have 3 years of support from them.

Yes, it is looking like this is an effect of the Phenom II / Athlon II on the 990FX platform that MAY be correctable by BIOS, something I haven't run across. The 1090T had no issue on this exact same board with all 4 slots loaded.

Check your BIOS revision: http://www.gigabyte.us/products/product-page.aspx?pid=3996#bios

Nick McDermott - 3dGM Admin

Kofoed
Offline
Joined: 06/01/2010
Posts: 17

Its bios rev 1.0, already got the latest bios.
What can a updated rev do? A possibly fix for this?

eire1274
eire1274's picture
Offline
Joined: 09/12/2003
Posts: 1221

Rev. 4 is the current release, posted August 31st.

Yes, the BIOS can change handling for CPU sets, as well as direction on multichannel handling, as the CPU is obligated to operate on system rules presented by the chipset, and chipset behavior is set by the BIOS. Definitely worth looking at.

Rev. 2 and 3 both show updates to the CPU AGESA (Central Processor Unit AMD Generic Encapsulated Software Architecture) and will likely change CPU operations from detected values to specific, including CPU timing, memory timing and detection, etc., and could likely fix your issue.

Give Rev. 4 a shot, as it also updates the TMP (Trusted Module Platform).

Added: I loaded Rev. 2 on the last build before installing Windows (that was the latest at the time), but it booted with all 4 sockets to do the BIOS update with the 1090T. Having looked in further, it's a problem seen with the x2, x3, and x4 chips that seems to have not had any effect on the x6. Perhaps changes to the memory bus within the CPU module design itself?

Nick McDermott - 3dGM Admin

Kofoed
Offline
Joined: 06/01/2010
Posts: 17

Should i ask Gigabyte to replace it with Rev4 then?
Or how do I get rev4 without having to buy it ..

eire1274
eire1274's picture
Offline
Joined: 09/12/2003
Posts: 1221

BIOS updates are free. Just download it. http://download.gigabyte.us/FileList/BIOS/mb_bios_ga-990fxa-ud3_f4.exe

Boot into Windows using just the two memory DIMMs, download and run that file. Alternatively, you can use another computer to burn a bootable (DOS) image to a disc or USB flash with the files inside that EXE (it is an executable ZIP with an AUTOEXEC.BAT, a ROM loader, and the BIOS ROM image) and boot from that, and you can update the BIOS without loading Windows.

Have you ever done a BIOS update before? If not, you may want to watch someone else do it first so you understand the process.

Nick McDermott - 3dGM Admin

Kofoed
Offline
Joined: 06/01/2010
Posts: 17

I though you ment the Motherboard model Rev version.
Such as rev 1.0, 1.1 and 1.2.
Yes, the motherboard is already using bios F4.

eire1274
eire1274's picture
Offline
Joined: 09/12/2003
Posts: 1221

That board currently only has one print revision, I believe.

Well, yeah, if you have the fourth BIOS release, then you are stuck. It is, after all, an issue with your CPU, not the motherboard.

I have verified that the X6 Phenom IIs are totally stable with high-speed memory in multi-slot dual channel on this board. X6 Athlon IIs seem OK, too. It seems that the X2-X4 chips can't tolerate high memory bus rates, and if you can slow it down to 1066 or lower you may be able to make it work, but it's iffy.

Sorry. Expect the Bulldozer Zambezi to overcome this issue entirely, as they have a whole totally revamped memory controller.

Nick McDermott - 3dGM Admin

Kofoed
Offline
Joined: 06/01/2010
Posts: 17

How can it be a issue with my CPU :P? It worked for my friend, same CPU as me.
Also, on gigabyte there is rev 1.0 rev 1.1 and rev 1.2(NEW!)
Are those different "print" of the board?

eire1274
eire1274's picture
Offline
Joined: 09/12/2003
Posts: 1221

Yes, those are different prints.

The issue is your CPU with that chipset, causing an incompatibility. Not a broken board, or broken memory, but that the board can't make that CPU talk to 4 channels of memory due to an incompatibility with that core.

Nick McDermott - 3dGM Admin