Corsair 6400 VDimm Problems

I've been rather disappointed with the matched 2 gig pair (2x1Gig) of this RAM - its standard 6400 stuff, TwinX, CAS5 version.

My setup is as follows, P5B-Deluxe, E6600 and the above RAM.

I've never known RAM NOT to like extra Vdimm before ... which seems to me an odd behaviour ...

At stock SPD speed 800 MHz, 5-5-5-15 - 1.9 VDimm is fine, Orthos stable for hours, but if I increase the Vdimm to 1.95V, won't last more than 3 minutes in Orthos, and above 1.95V, 2.0V, 2.05V, or 2.1V - won't last for more than 40 seconds ....

if I overclock slightly - up to 840MHz, 5-5-5-15 or even 18, 1.9V Vdimm is not enough, but 1.95V then suddenly becomes OK, but again, anything over that is stable for less than 3 minutes. Above 845 - at any voltage, completely unstable - but will boot.

I know I can't expect memory to overclock, but I've never known RAM not to like extra voltage, not enough one would of course expect it to be unstable.

Cooling is not a problem, case has 3 120mm fans, and case is large and uncluttered.

I've done some further tests - to see if I could work out what is going on.

If I set memory to 800 MHz, SPD settings, VDimm to 2.1V - then run Mem Test in dos - it doesn't get past test 5 without errors, test 5 gives 370 errors itself.

If I set memory to same settings, but VDimm to 1.9V - Mem Test passes fine.

I thought these modules were meant to take 2.1V without any problems ? Any idea to why it can't ?

thanks, Mark.

if it was me would prob just rma it my 6400 does 980 all day at 2.1 but anything over 2.3 wont boot just my 2 cents

Plumber4life said:

if it was me would prob just rma it my 6400 does 980 all day at 2.1 but anything over 2.3 wont boot just my 2 cents

Click to expand...

I don't know that someone should be able to RMA something to us that is performing above it's rated spec, just because they think it should perform even further above that spec.

We rate the 6400 modules (which you have) at 1.9V at 5-5-5 latency at 800 MHz.

If it isn't stable there, we'll replace it, no questions asked.

However, from the initial post, Buckster, there are some things I think you should know.

1) The 6400 C5-rated modules you have could be made with any number of ICs. Here's a list, based on revision number:

CM2X512A-6400 XMS6405v4.1 0633067-1 = PROMOS 800MHZ, CL5, 0.09u 64Mx8 ETT
CM2X1024-6400 XMS6405v5.1 0634287-2 = PROMOS 667MHZ, CL5, 0.09u 64Mx8 ETT
CM2X1024-6400 XMS6405v5.1 0631252-2 = PROMOS 667MHZ, CL5, 0.09u 64Mx8 ETT
CM2X1024-6400 XMS6405v5.1 0631252-9 = PROMOS 667MHZ, CL5, 0.09u 64Mx8 ETT
CM2X1024-6400PRO XMS6405v1.1 0610028-1 = Micron, 64Mx8, DDR2-667, CL5, Die Rev D
CM2X1024-6400 XMS6405v5.1 0633171-6 = PROMOS 667MHZ, CL5, 0.09u 64Mx8 ETT
CM2X1024-6400 XMS6405v5.1 0629230-7 = PROMOS 667MHZ, CL5, 0.09u 64Mx8 ETT
CM1024X2-6400 XMS6405 v4.1 0626413-3 = Infineon AENEON, Die Rev B, 64M X 8 Rev B
CM2X1024-6400PRO XMS6405v3.1 0609115-2 = Infineon AENEON, Die Rev B, 64M X 8 Rev B
CM2X1024-6400 XMS6405v4.1 0609102-1 = Infineon AENEON, Die Rev B, 64M X 8 Rev B
CM2X1024-6400, XMS6405 v1.4, 0606096-2 = Micron, 64Mx8, DDR2-667, CL5, Die Rev D
CM2X512A-6400 XMS6405v1.1 0531093-0 = Micron, 64Mx8,2-533, CL4, MT47H64M8FT-37E Die Rev A
CM2X512A-6400 XMS6405V3.1 0609062-1 = Infineon AENEON, Die Rev B, 64M X 8 Rev B
CM2X1024-6400PRO XMS6405v1.3 0610028-1 = Micron, 64Mx8, DDR2-667, CL5, Die Rev D
CM2X512A-6400 XMS6405v3.1 0628012-2 = Infineon 64Mx8, DDR2, AENEON, Die Rev B (90nm)
CM2X512A-6400 XMS6405v4.1 0631250-9 = PROMOS 0.09u 64Mx8 ETT

Keep in mind, this is only 6400, and doesn't include 6400C4, 6400C3, etc.

2) Some of those ICs have integrated voltage regulation, which means that above certain voltages, the memory corrects it down to 1.8-1.9, in order to protect the circuit from out-of-spec voltage. The Micron IC-based models don't have this feature, but some of the others do. Look at your modules for the spec number (XMS6405 rev 3.1, for example) and look it up on that list above to find out which ICs you use.

I'm sorry the memory isn't overclocking to your expectations, but personally I think 840 at 1.95 is fairly decent, a 5% overclock for free. Not all ICs perform the same way with extra voltage.

Some really like voltage (Micron DDR2, Winbond BH/CH DDR1) but, for example, Samsung TCCD modules, back in the DDR1 days, anything above 2.9 was actually counter-productive. The sweet spot for them was 2.7-2.85 depending.

Each IC has its own little preferences. Sometimes they like one thing, and sometimes another.

thanks for the really detailed reply much appreciated - yes I shouldn't expect more, and 840 is pretty decent. I checked the chips and they are these ones:

CM2X1024-6400 XMS6405v5.1 0633171-6 = PROMOS 667MHZ, CL5, 0.09u 64Mx8 ETT

Does that actually mean - that even at stock - the chips are running overclocked to reach 800 MHz @ CL5 ? I've actually just answered my own question - as XMS - means "guranteed to run at say 800" even if module are lower ...

I suppose its just a bit frustrating that say 2 people paying the same price for the same RAM - one could end up with "lesser" PROMOS modules and another the seemingly much better Micron RAM. It explains how some people with the CL5 RAM get hardly any overclockablitity and others are managing DDR 1000. I don't see this with other products, say graphics cards etc, from one manufacturer to the other, they may use difference RAM modules within the reference design, but within one product/one manufacturer they'll always be the same.

One last question - could I ask why the RAM modules are sold as 5-5-5-15, they have 5-5-5-12 on the label on the RAM, and in SPD state 5-5-5-18 ?

thanks again,

Mark.

Buckster said:

thanks for the really detailed reply much appreciated - yes I should expect more, and 840 is pretty decent. I checked the chips and they are these ones:

CM2X1024-6400 XMS6405v5.1 0633171-6 = PROMOS 667MHZ, CL5, 0.09u 64Mx8 ETT

Does that actually mean - that even at stock - the chips are running overclocked to reach 800 MHz @ CL5 ?

thanks again,

Mark.

Click to expand...

The reason we're able to sell the 6400s for considerably cheaper than the 6400C4s or C3s is because we can screen 667 parts internally to see which ones do 800 MHz.

Manufacturers sell their ICs to us and we manufacture modules with them. This is how pretty much everybody does it with very few exceptions. When we go to a vendor and say "We'd like 500,000 ICs that do 800 MHz" they say that'll be $x per IC. Usually 800 MHz are more expensive than 667 MHz ICs. So sometimes we'll buy a boatload of 667 MHz ICs, and screen them internally. With those Promos, we were pretty excited to find that a lot of them did 800 MHz at 1.9V at Cas5. We test them for our 24-hour long RST Pro2 qual on motherboards (for 800 MHz DDR2 we use the P5B Deluxe or the M2N32-SLI Deluxe right now), and then once we're sure that they are stable, we'll sell them.

It's a very tough industry right now, DDR2 prices are insane for everyone, demand has shot through the roof, so we're finding as many ways as we can to keep our customers happy.

Those ICs are rated for 667, but we've tested them to run 800. Some Microns we sell are rated for 667 and we can test them to do 1066 MHz, so it's not uncommon.

All overclocking memory has to be rated to overclock, you know? Heh.

Again, if your system isn't 100% stable at 800 MHz at 5-5-5 at 1.9V, we'll RMA those modules no problem. But there's no guarantee what revision of the 6400s you'll get in return, you might end up with the same revision.

Hope this helps.

it does - thanks - please re-read my post above, as I substantially edited it whillst you replied I think.

Buckster said:

it does - thanks - please re-read my post above, as I substantially edited it whillst you replied I think.

Click to expand...

To answer some of your questions:

1) The JEDEC spec of the part is 5-5-5-15, I believe. I'll double-check that tomorrow when I get to work. We can spec a part to be faster than JEDEC, but we make sure that the SPD is JEDEC spec so it should boot in just about any motherboard. The reason it says 5-5-5-12 is because that specific part will do that. Also, the 5-5-5-18 SPD is a misread by the BIOS. We frequently see the same modules show up with different SPD settings in CPU-Z in different boards. The same module you have might show up as 5-5-5-15 in another Asus board and 5-6-6-14 in a Gigabyte board. The way the BIOS reads the SPD is what CPU-Z and other windows programs, as well as memory benchmark programs like Memtest86+, display when they show the SPD settings.

2) Lower-priced modules are usually sold with whatever the cheapest ICs available that day are. Since memory prices are so volatile, there's a chance that the cheapest ICs one day might be the best performing, it's happened before. But with the XMS26400 part is basically equivalent to a value 800 right now since we do not currently have a value select 800 MHz part. Only it has heatspreaders and is under the XMS line.

With a set of Corsair's 6400, what are the recommended settings for DRAM TRFC, DRAM TRRD, Rank write to delay, Read to precharge delay and Write to precharge delay?

Kuolo said:

With a set of Corsair's 6400, what are the recommended settings for DRAM TRFC, DRAM TRRD, Rank write to delay, Read to precharge delay and Write to precharge delay?

Click to expand...

We do not spec those settings, as the vast majority of motherboards do not allow the user to modify them. You are welcome to experiment with them and post your results, but different motherboard BIOSes will react differently to those settings.