TR Forums

Any recommendations on RAM coolers. I can't seem to keep my machine stable when all 12GB are active. On reboot, with the bios only enabling 8GB, it runs like a top. I'm thinking a fan for my ram (which is absurdly hot), may get me stable with all 12 GB.

I guess you are using Intel Core i7 (socket 1366)
so if it's in that case, try to check memory support list from you motherboard's website first.
Or test other RAM with your board if it's possible and also test your RAM on other working board.

Before you invest money in something like a RAM cooler, just put a regular 80mm ontop of the RAM and attempt to run all 12 GB that way. If that doesn't work, than no ram cooler will solve the issue.

Doubtful heat is doing this. I'd suspect a voltage bump would help.

May also need to relax the timings too.

I have experimented with the voltage already. And I'm running 1600 memory at 1066 already. The memory is not on the boards officially supported list. But I don't see why that should matter all that much . . . maybe that is the issue though -- I can't change that now.

When I boot a room temperature system, all 12GB are made available to the OS. The machine will not pass a memtest though. I then reboot and only 8GB of RAM are made available by the BIOS, and the system is stable.

I don't think the memory sticks are the problem -- though its a possibility. Based on the behavior, I believe that when the memory is too hot, the bios won't pass it during post. My thought is to increase the airflow, but I don't know how to slap a fan in that area. I've looked, but I can't get it mounted anywhere I can get air to flow. ANy pics oh how someone else has done it are definitely welcome.

I think you need to find out which stick(s) of RAM is/are causing you grief. Go through classic hardware problem solving. Remove all but 1 stick of RAM and run a mem-test and in all likelihood you'll find one or two that may have an issue. You may also want to consider testing them in pairs to verify dual channel operation. Give that a go and see if that turn up any errors. Since 4GB are "disappearing" on you once the system comes up to operating temperature it's likely you'll find the same error when just running whatever sticks are bad (assuming this is the case). I'd have to agree with StuG that just laying a fan over the modules should be sufficient to test the heat dissipation theory. If that yields any results you may want to look into Corsair Dominator coolers and if those are a viable option for your RAM modules.

brsett wrote:

I have experimented with the voltage already. And I'm running 1600 memory at 1066 already. The memory is not on the boards officially supported list. But I don't see why that should matter all that much . . . maybe that is the issue though -- I can't change that now.

It matters. Try an approved memory type and see if that fixes it. Only 1 stick to try and run memtest will give you a good indicator.

TheEmrys wrote:

brsett wrote:

I have experimented with the voltage already. And I'm running 1600 memory at 1066 already. The memory is not on the boards officially supported list. But I don't see why that should matter all that much . . . maybe that is the issue though -- I can't change that now.

It matters. Try an approved memory type and see if that fixes it. Only 1 stick to try and run memtest will give you a good indicator.

I can run one stick of my current stuff. I can run 4 sticks of it. I just can't run 6 sticks. What will running 1 stick of approved memory prove to me?

pixel_junkie wrote:

I think you need to find out which stick(s) of RAM is/are causing you grief. Go through classic hardware problem solving. Remove all but 1 stick of RAM and run a mem-test and in all likelihood you'll find one or two that may have an issue. You may also want to consider testing them in pairs to verify dual channel operation. Give that a go and see if that turn up any errors. Since 4GB are "disappearing" on you once the system comes up to operating temperature it's likely you'll find the same error when just running whatever sticks are bad (assuming this is the case). I'd have to agree with StuG that just laying a fan over the modules should be sufficient to test the heat dissipation theory. If that yields any results you may want to look into Corsair Dominator coolers and if those are a viable option for your RAM modules.

Any combination of the RAM that uses less than 5 sticks works fine. It is only when I use 6 sticks that the problem appears.

Don't rule out the motherboard as a possibility.

Motherboards typically have problems when you trying to drive all its DIMM slots at once. You may have to relax timings to obtain stability and/or increase voltage by a little bit.

Krogoth wrote:

Don't rule out the motherboard as a possibility.

Motherboards typically have problems when you trying to drive all its DIMM slots at once. You may have to relax timings to obtain stability and/or increase voltage by a little bit.

Okay, this has been suggested twice now. Specifically, what timing do you suggest I relax. I'm running at: 8-8-8-24 with command at 2t.

I futzed with the voltage a good bit. Well over the 1.65volts advertised (and accompanying qpi changes). If its heat, voltage will exacerbate the issue. I would love for it to be just a timing issue, but remember, when hot, only 8GB are made available to the OS (even though all 12GB are detected).

Edit: For the record, this is Corsair XMS3 pro memory supporting 1600. It is what is suggested by TR. I regret using it. I have already tested out one obviously bad stick and sent that back. Also, I haven't ruled the motherboard out, but I don't want to change that out very much at all, as I have work I need to do on the system, and can do with 8GB of RAM even if 12GB would be better.

brsett wrote:

Krogoth wrote:

Don't rule out the motherboard as a possibility.

Motherboards typically have problems when you trying to drive all its DIMM slots at once. You may have to relax timings to obtain stability and/or increase voltage by a little bit.

Okay, this has been suggested twice now. Specifically, what timing do you suggest I relax. I'm running at: 8-8-8-24 with command at 2t.

I futzed with the voltage a good bit. Well over the 1.65volts advertised (and accompanying qpi changes). If its heat, voltage will exacerbate the issue. I would love for it to be just a timing issue, but remember, when hot, only 8GB are made available to the OS (even though all 12GB are detected).

Edit: For the record, this is Corsair XMS3 pro memory supporting 1600. It is what is suggested by TR. I regret using it. I have already tested out one obviously bad stick and sent that back. Also, I haven't ruled the motherboard out, but I don't want to change that out very much at all, as I have work I need to do on the system, and can do with 8GB of RAM even if 12GB would be better.

I haven't had much experience with DDR3. I would suggest to stick with JEDEC-spec timings.

BIOS is current? BTW, I am using this http://www.newegg.com/Product/Product.aspx?Item=N82E16835704001&cm_re=corsair_fan-_-35-704-001-_-Product because there is more room for it than this http://www.newegg.com/Product/Product.aspx?Item=N82E16835181008&cm_re=corsair_fan-_-35-181-008-_-Product [heatsink is crowding the RAM] which I also own.

RAM sticks almost never have heat issues, unless extremely closely packed in an enclosed space without air flow or ventilation while being greatly over-volted (which is pretty much a hypothetical scenario). All the extra RAM cooling products are marketing gimmicks trying to get more money from consumers. I once saw an infrared camera shot of an OC'd computer, and the area where the RAM sticks are is by far one of the coolest spots out of the entire setup.

It's probably a motherboard compatibility issue or bad stick problem. Try every one of the 6 sticks one at a time to rule out bad stick(s), also try motherboard manufacturer approved RAM to see if the problem persists. Make sure you have the timings and voltages set as per RAM specification.

If you really want to prove us right, then just open the case up, grab a big fan and just blow the fan into the case.

You haven't mentioned what motherboard you're using, but just picking a triple channel board at random (Asus P6T SE) and looking at the qualified memory configurations, it looks like there are a lot of instances where a particular 2 GB module is qualified for use only with 3 of the slots filled; the same module is not qualified if you fill all 6 slots. Corsair 2 GB 1600 happens to be one of the types of module where this restriction applies (at least for the Asus P6T SE).

Yes, I am a little puzzled as to why downclocking to 1066 didn't fix it... but FWIW Asus seems to believe that using certain 2 GB modules in 12 GB configuration is a no-no.

Jd007 wrote:

RAM sticks almost never have heat issues, unless extremely closely packed in an enclosed space without air flow or ventilation while being greatly over-volted (which is pretty much a hypothetical scenario). All the extra RAM cooling products are marketing gimmicks trying to get more money from consumers. I once saw an infrared camera shot of an OC'd computer, and the area where the RAM sticks are is by far one of the coolest spots out of the entire setup.

It's probably a motherboard compatibility issue or bad stick problem. Try every one of the 6 sticks one at a time to rule out bad stick(s), also try motherboard manufacturer approved RAM to see if the problem persists. Make sure you have the timings and voltages set as per RAM specification.

Not true anymore, DDR2 and DDR3 can get pretty toasty, especially the non-JEDEC spec stuff. RDRAMs were more infamous for heat issues due to its powering scheme. Usually the airflow of the standard ATX form factor "wind tunnel" is enough to keep JEDEC-spec DDR2/DDR3 happy.

I have 4 x 1GB sticks of 667 DDR2 and they got hot enough to give me a burn so I got a RAM cooler. I was not having stability issues I just lie to keep things cool.

So from my experience, DDR2 dimms do get hot. I assume DDR3 are similar enough. Does that influence stability? maybe, it certainly does for other chips in a computer, cpu, gpu, northbridge...

Is heat causing Krogoth's issues? It should be considered, but I'd bet that the 12GB configuration is not certified for a reason!

Flying Fox wrote:

If you really want to prove us right, then just open the case up, grab a big fan and just blow the fan into the case.

How many hours of memtest do you suggest I stand there holding the fan? I suppose this is the thing to try at least before worrying about getting a RAM cooler though.

I'd suggest you put the fan on the ground, don't tire your arms like that, and do it over night.