Can't run 800 MHZ DDR2 in Dual Channel

[khaz]

Using a Gigabyte DQ6 mobo, I can't get an overclock with ram at 800 MHz. Actually I can't get ram to run at 800 in dual channel period. I've tried supertalent 2.2v ram and when that didnt work I got some mushkin's rated at 1.8v. Running at 800 MHz in dual channel I can boot and get into windows, but dual prime95 fails after a couple of minutes (usually on transition from test 1 to test 2).

http://www.newegg.com/Product/Product.asp?Item=N82E16820146118

I can OC to 3.0 gigs with ram in dual channel at 667 MHz and 3.2 gigs at 800 MHz in single channel (I just pull a stick really) both on stock voltages. Both OCs will run dual prime95 forever. The Mushkin ram is 5-5-5-12 and I losened timings to 5-5-5-18 to be safe. I set timings and 1:1 manually. I have upped vcore, vdimm, mch, and fsb in uncountable combinations to no avail. 'Auto' memory settings with cpu at stock FSB doesn't work either.

I tried the ram in different slots (still shows as dual channel) and both individually pass memtest86. Any ideas? I figure if dual channel is on the northbridge then maybe the northbridge needs voltage, but any time I up MCH even 0.5 the system gets unstable. CPU is watercooled and temps are not a problem. Northbridge has the Gigabyte heatsink sans cover with a fan strapped to it. Airflow in case is excellent.

Any ideas on where I go next?

System details:

Intel C2D E6600 2.4 GHZ CPUz
Gigabyte P965-DQ6 mobo
Mushkin DDRII 800 1GB*2
Western Digital Raptors 74GB*2 in raid 0
Sapphire ATI x1950xt
Samsung SH-S183L 18x SATA Lightscribe DVD
Thermaltake Toughpower 700W Modular PSU
Sunbeam Rheostat +2 CCFL +4 LED Fans
Koolance PC-720BK Watercooled Case + HD/GPU/CPU Block

 

[AdamBomb42]

Is the bios up to date?

 

[khaz]

Yeah, I'm running bios version F9, released January 2007 and is the latest. I've also tried F7 and F8.

 

[rdhood]

Hmm... you didn't say:


You said both pass memtest86+ individually. What happens when you run memtest86+ with both of them together?

Have you tried boosting ddr2 voltage +.1, +.2, +.3??? (i.e. memtest86+ both together with boosted ddr2 voltage)?

Have you tried relaxing timings even more (5-8-8-20)?

Have you tried both memory banks?


It's still not clear to me if this is the memory. You need to be able to run each stick individually at 800Mhz in memtest86+ , and then run them together at 800Mhz before you ever even think about trying to run dual prime95.

good luck

 

[khaz]

Both pass memtest86+ individually at 800 MHz, but I haven't tried them both together. I will do that next.

I have tried boosting vdimm as high as +.3 on the mushkin, which is rated at 1.8v, and +.4 on the supertalent which is rated at 2.2v.

I have not tried more relaxed timings. I'll give that a shot.

I have tried all the slots on the mobo if that's what you mean by memory banks.

-------

Also, I've RMA'd the mobo over this issue and the replacement is doing the same thing.

 

[Mondoman]

Since you're water cooling the CPU, how are you cooling the radiators next to the CPU? If you use an OEM-type CPU cooler, does that help?

 

[khaz]

Well the case has the following fans in addition to watercooling on the cpu, gpu, and raptors:

2x80mm fans in the front blowing in
2x120mm fans on top set to blow down through the radiator
2x80mm fans in back blowing out
1x40mm strapped to the NB

So there is no active cooling in the immediate vicinity of the heatpipes or the other heatsinks. I've never tried the oem cpu cooler and I didn't pull the heatpipes/sinks to replace whatever gigabyte uses (paste/pads) with arctic silver. I suppose I could put a 120mm on the door and have it blow right down at them and see if that helps. If worst comes to worst, I could pull the heatpipe assembly and put better goo on.

Since I'm not really overvolting anything, is heat a likely culprit still on an overclock?

 

[PCD]

2x80mm fans in the front blowing in
2x120mm fans on top set to blow down through the radiator
2x80mm fans in back blowing out
1x40mm strapped to the NB

So actually you have over-pressure in the case(or what it's called) + you are blowing the hot air from the radiator INTO the case.... Doesn't sound right to me.

 

[khaz]

Ya I see what you are saying about the pressure. That didn't really occur to me. I tried the fans on top both ways, and I get better water temps by blowing cooler ambient air through the rad dead spots and all than I get by pulling hot air from inside the case across the rad. For all that, flipping the fans actually didn't increase internal case temps.

And as an update, the sticks individually and all slots pass memtest86+ at 800 MHz but fail with two sticks. At this stage I'm thinking its the mobo's memory controller. What are the chances of me having two bad memory controllers in a row? Or maybe thest ram sticks are just on the 'finicky' list for this mobo.

 

[khaz]

By fail I mean hang ... I've never actually gotten error reports with this mobo. When it fails, it just hangs.

 

[Mondoman]

Well the case has the following fans in addition to watercooling on the cpu, gpu, and raptors:
...
2x120mm fans on top set to blow down through the radiator
...

I assume you mean a radiator for the water cooling, not the two copper radiators immediately to the right and above the CPU (as shown in this MB photo: http://www.tech-hounds.com/review27/fullGigabyte965PDQ6.jpg ).

So there is no active cooling in the immediate vicinity of the heatpipes or the other heatsinks. I've never tried the oem cpu cooler...

Since I'm not really overvolting anything, is heat a likely culprit still on an overclock?

Sure; even without any nonlinearity in the response, doing more clock cycles of work in the same time period will result in more heat.

OK, I think I have a better idea of the setup now. Some things to think about:
1) I doubt very much that you have good airflow inside the case. The normal airflow path is in at the front/bottom of the case and out at the back/top of the case. However, you've got the 120mm fans at the top, which are pushing hot air down into the case. If they are just blowing case air, not fresh air from the outside, they are just heating up the inside of the case. I'd measure your case interior temp at a couple of places; with good air flow it shouldn't be more than 5C or so above ambient (room) temp, and with excellent flow it shouldn't be more than 2C above ambient.
Possible solutions: move the radiator out of the case and/or add at least one 120mm fan blowing out the upper back of the case.
2) The copper heat radiators are arranged around the CPU for a reason -- the OEM CPU fan blows air down onto the MB and out in all directions, cooling surrounding components as its secondary function. Many people who go to aftermarket coolers don't take this into account. Given the physical arrangement of those radiators and the fact that you're doing OCing, even excellent case air flow may not be enough to cool them. They are directly cover the PWM heatsinks, so check your PWM temp to get an idea of how hot that region is getting.
Possible solutions: I'd mount something like a 50mm fan to each radiator (on the CPU side), blowing away from the CPU. Alternatively, just mount an OEM-style CPU cooler.

 

[joex444]

I'm looking at this thread and also the one I have, Memtest86+ Hangs, and can't help but see some similarities.

I have a newer DQ6, rev 3.3, and with 2x1GB DDR2-800 in dual channel, Memtest86+ will hang reporting no errors. The two sticks work fine individually, just like the OP has stated.

I have a Tuniq Tower 120. During these tests I had that fan cranked up all the way. In my case I have a 120mm fan blowing out on the back, and theres another 120mm fan on the top, which I think blows in. There are no front fans. When I feel the DIMMs after testing the heatspreader is slightly warm, but not hot at all. Since this is in an Antec P180B, the PSU is seperated and the HDs are also seperated. The only heat sources are the CPU, which isn't making the heatsink anything but slightly warm, the DIMMs and the 8800GTS, which gets warm. The fans should be enough to deal with this, and I leave the front door open and can feel air being sucked inside at a noticeable rate.

I think that whatever is causing the OP's problem is also causing my problem, and vice versa.

 

[khaz]

Yes I was referring to the radiator for the watercooling and not the copper.

I've been doing extensive testing all day. Memtest86+ even hangs at my 333x9 overclock with ram at 667 which I assumed was fine as I can pass dual prime95 at that setting, but I had never tested being focused on the 800 MHz point. I'm going to try a very low multiplier to see if I can pass at 800 MHz at 1:1 without overclocking or 'auto' to get it there.

 

[khaz]

I can't run 800 mhz in dual channel even on a 6x multiplier (for a 2.4 ghz cpu). However, I can clock the cpu all the way to 3.6 ghz on one stick at 800 mhz.

At this point, I think I can say I either have a heat problem with the chipset or the memory controller is flaky on the mobo. I might run an experiment where I leave the door off and blast a room fan at the chipset.

 

[Mondoman]

You have all 3 of the MB power connectors plugged in, right (8-pin near I/O shield, 24-pin main connector, and 4-pin Molex next to RAM)?
What kind of PWM temps are you reading?

 

[khaz]

All 3 power connectors are connected, and I disconnected and reconnected them just to be sure. At 333 FSB (the stable OC) PWM Temp 1 reads at 27-29C and Temp 2 reads at 13-16C in Speedfan during dual prime95. At 400 FSB (stable for superpi only) PWM Temp 1 reads at 30-33C and Temp 2 reads at 15-18C in Speedfan during superpi 8m. Prime95 takes it down after a couple of minutes put PWM reads Temp 2 reads at 26-28C before it hangs.

 

[Mondoman]

Well, I'm stumped at this point. I'd still try going with an OEM CPU cooler, just because it can affect a number of different components.