TR Forums

I have recently bought a 2 GB memory kit to upgrade my memory from 2 to 4 GB. Since then, I had problems of the odd kind: computer behaving slugglishly, slow, long waits, but not always. Sometimes problems to shut down, overall sluggish and unstable feeling.

I did some systematic tests and found that the memory bandwith is erratically changing from 4880Mb/s to 1220Mb/s ... randomly in about 4 out of 10 tests.

I then started to swap the memory: there are 4 slots, labeled DIMM0/DIMM1 and channel A/channel B.
DIMM0 is marked blue and DIMM1 is marked black. Originally I had, as required, the two 1GB RAMs in the two blue DIMM0 slots (everything worked fine then), then I installed another two 1GB RAms into the black DIMM1 slots and since then I have those problems.
The original thesis was that the new memory is bad, but when I placed the new memory into the DIMM0 slots and removed the old memory, everything was fine too. Testing each single RAM in DIMM0/Channel A also produced no problems.

The problem only exists when all 4 slots are filled. So this is not a problem of the RAM modules, but of the fact that 4 instead of just 2 slots are filled!

All four RAM modules have the same spec DDR2 8000 5-5-5-12 and are from the same brand company.

The motherboard is an Intel DG965WH with the newest BIOS with an Dual Core2 6600 running at 2.4GHz

For me, this does make so not sense, but the problem can be repeated -- I have removed 2 GB now and everything is fine again (except that I would need that additional 1.5GB I would get for the work I do on that machine)

If your bios allows, up the mem voltage slightly or loosen the mem timings a little. This is not unusual when running 4 sticks.This can be most efficiently tested for with memtest86+, a small program that can be run from a dosboot cd or floppy. Memtest86+ is also one of the programs on the "Ultimate boot CD" (a invaluable diagnostic tool) under file systems last I checked.


Jim

xgsound wrote:

If your bios allows, up the mem voltage slightly or loosen the mem timings a little. This is not unusual when running 4 sticks.This can be most efficiently tested for with memtest86+, a small program that can be run from a dosboot cd or floppy. Memtest86+ is also one of the programs on the "Ultimate boot CD" (a invaluable diagnostic tool) under file systems last I checked.

I haven't found a way to change voltage, but I have played with the timings -- my experiments did not change anything.
I have used a Windows system test utility for measuring memory speed -- and what is interesting is the the speed is either a full
4880MB/s or exactly one fourth: 1220MB/s, and with all 4 memory modules installed I get the slow speed in most of the tests, while with only 2GB installed (any two of the modules I have) I always get the fast speed.
I am seeing the slowdown when all 4 modules are installed both under Windows and Linux.

I have run the MemTest86 utility that got installed on my hard disk with Linux, but after 5 hours and only 30% passed I cancelled it.

I simply do not understand how the same memory works if I only use modules and obviously does not work correctly when I use all four.
Is this something that others have experienced? Could this mean that my motherboard is damaged? Since 2GB work, no matter which modules I use, it cannot be the RAM, no?

I am pretty much a layman and newbie when it comes to hardware though ...

My bet is also on voltages. One of my friends recently made a similar upgrade from 2GB to 4GB and experienced similar instability and slowness. It turns out that his RAM was significantly under-volted by the motherboard, giving it a mere 1.8V when his RAM was rated to run at 2.2V. Apparently the lower voltage was sufficient to sustain 2 DIMMs with full speed and stability, but not 4.

Your BIOS will almost certainly provide control over memory voltages somewhere, but it might be hidden in an overclocking or advanced section. My next step would be to find out what voltage your RAM is supposed to run at (should be in the manual, if any, or on the manufacturer's website) and compare that to what your BIOS pulled out of thin air.

Many thanks for your support!

My memory modules are Corsair CM2X1024-6400. The label on the module says 5-5-5-12. 800Mhz, 1024MB, XMS6405v5.1
On the Corsair site the same module number is listed as 5-5-5-18 and that is what my BIOS shows too.
Here is the PDF about the module I found: http://corsairmemory.com/products/go.as ... X1024-6400

I have the CM2X1024-6400 and as far as I understand this needs 1.9V -- there is stuff on the internet about people failing to use the TWIN2X1024-6400C4 modules with my MB, but these seem to need 2.1V.
I have also read that the voltage on my MB cannot be changed and it seems it provides 1.8V so that should work, no?

Do you think that getting a different memory will solve this? If there are good chances that this will help I will do it because I NEED the 4GB for my work on this computer and I am really desparate.
I just want to avoid that I buy another 4GB and then the same problems occur.

I've got 4GBs of TWIN2X1024-6400C4, and you're right, it needs roughly 2.1V. With 2GB I managed to drop that down to 1.9V and remain stable, but after my upgrade, I had to go back up to the stated voltage.

I don't know if the difference between 1.8V and 1.9V would be enough to cause the problems you describe, but it is possible. I can't say with enough confidence that buying lower voltage RAM would definitely solve your issues though.

That said, there does seem to be quite a lot of people posting on the web about problems they're having with that particular motherboard and various types/speeds of RAM. On the Corsair forum, one of their representatives suggests that he would only recommend their DHX series for that board - I imagine the appropriate model would be QUAD2X4096-6400C5DHX or 2 lots of TWIN2X4096-6400C5DHX.

One thing you could try, if your BIOS has the option, is running the RAM at a lower clock speed, i.e. 667MHz as opposed to 800MHz. I'm not sure what the performance hit would be for that.

I do not understand those voltage specifications in the first place: the module shows 5-5-5-12 (is the last number, 12, a voltage specification here?) but the BIOS shows the same module with 5-5-5-18 and the PDF file describing the module talks about 1.9V?

I have tried using lower clock speed but this did not help either.

johann_p wrote:

5-5-5-12 (is the last number, 12, a voltage specification here?)

No, those are all timings.

5-5-5-12 is the notation for RAM timings, as Meadows indicates. There's a pretty good guide as to what they are here if you're interested, and no doubt much more ludicrous detail elsewhere on the web. The rated voltage is whatever it says in the manual.

Sometimes the BIOS will get the wrong timings as well as the wrong voltage. If you have control over the timings, I suspect you could set 5-5-5-12 instead of 5-5-5-18. I think the latter is what mine defaulted to, despite my RAM being rated for 4-4-4-12. However, having worse timings isn't going to trigger the behaviour you describe - if anything worsening the timings should aid stability at the price of performance.

That's why you don't get Intel motherboards due to lack of options. In this case to change the memory voltage.

OK .. I am learning here - thanks for all the advice!

I did not choose the MB or the memory myself but trusted somebody who has experience in building computers.

I wonder what the best/fasted/easiest/cheapest way to come to a working system would be: buy different 4GB memory (which? How would I find one where I can expect the whole 4GB to work on my motherboard?). Buy a different motherboeard and use the memory I have? Something else?

I am using the system for work under Linux and I need the 4GB because I deal with lots of data in Java VMs. The system was also built to be a silent as possible and not to use up too much power (I do not need extensive graphics power, for example).

So, any advice about what to do next would be highly apreciated.

Trade the memory out for stuff that is actually spec'd to work at the DDR2 spec of 1.8v.

OK - you have all convinced me
Obviously, the CM2X1024-6400 are *tested* and guaranteed to work at 1.9V and they sometimes *can* work with 1.8V.
The person who recommended that memory to me has 4GB of these modules running just fine in a DP35DP board which is 1.8V only too.
In my case, either 2 of these modules work while with all 4 I experience bandwidth slowdowns. I assume that this is because whatever exact voltage the board provides might go down with all 4 draining current just enough for the RAM to slow down sometimes.

In any case I have decided to do what all of you kindly suggested: pick a correct set of modules.
I ordered the TWIN2X4096-6400C5DHX kit.

EDIT: just installed the TWIN2X4096-6400C5DHX kit and everything works as it should.

Bummer.

That was all a nice theory, but alas, didn't hold up to experiment.

I was happy to quickly: the two first quick memory bandwidth tests (using Sisoft) showed that everything is ok, but after a while of heavy use I experienced that odd sluggishness again.
Carried out more tests and the same erratic symptoms occur: memory bandwidth breaks down on some tests, not on others.

Back to 2GB again, now I have 6GB or useless RAM laying around

I can only guess that this is some very odd problem of the mainboard.

What kind of power supply do you have? Are the voltages reading ok with all four sticks in? You may even need to grab a multi-meter because the BIOS voltage readings aren't always very accurate.