Finally found out one reason why I can't run 3.4+ Vdimm on DFi nF4

[mrlobber]

To make a long story a bit shorter:

Already back in the February, ever since I own my DFI nF4 board, I was always stopped by the fact that as soon as I tried to run more than 3.4 Vdimm, memtest froze in time of 2-3 minutes, or gave out tons and tons of errors even at much lower speeds. Therefore my VX remained Memtest stable up to around 260mhz, no more, where they were carried with the maximum 3.4 Vdimm I was able to feed them.

After trying almost all the official and beta bioses, various timings suggested for VX & Winchester clocking both by OCZ boys as well as by others (nothing changed really), at first I thought my PSU (PcP&C) was guilty until I modded it to 24pins by disassembling the rails from one of the native SATA connectors. That didn't help at all as well. Than I thought my CPU is crappy (and it is indeed, with those 2300 mhz maximum Prime stable speeds), but just a couple of days ago I noticed some corellation elsewhere. 3.4 Vdimm started to give out similar behaviour. So I thought a little... and guess what? Recently the spring has came in my country as well - and ambient temps, naturally, are rising. So I did a quick test. Set 3.4Vdimm and 260htt - memtest stable speeds back in February. Memtest froze in the 3rd pass of test #5. I took off the side panel, rebooted Memtest run for 120 passes, with errors each 6th-7th pass. I installed the side panel back (with system running memtest). After 4 passes, memtest froze. So the conclusion: it's heat which holds me at the place... simple as that... where was I earlier, huh? :stick:

And that's with my case's 2x120mm fans, and 2x92mm high speed ones installed separately over the memory and the "hot" corner of the 5V mosfet area. Don't know what else much can I do to cool the 5V vdimm mosfet area better. Any ideas (xtreme are velcome as well :) )? Right now, I'm only thinking of building a custom case with the option to install a bunch of fans all over the hot place...

[blander]

To make a long story a bit shorter:

Already back in the February, ever since I own my DFI nF4 board, I was always stopped by the fact that as soon as I tried to run more than 3.4 Vdimm, memtest froze in time of 2-3 minutes, or gave out tons and tons of errors even at much lower speeds. Therefore my VX remained Memtest stable up to around 260mhz, no more, where they were carried with the maximum 3.4 Vdimm I was able to feed them.

After trying almost all the official and beta bioses, various timings suggested for VX & Winchester clocking both by OCZ boys as well as by others (nothing changed really), at first I thought my PSU (PcP&C) was guilty until I modded it to 24pins by disassembling the rails from one of the native SATA connectors. That didn't help at all as well. Than I thought my CPU is crappy (and it is indeed, with those 2300 mhz maximum Prime stable speeds), but just a couple of days ago I noticed some corellation elsewhere. 3.4 Vdimm started to give out similar behaviour. So I thought a little... and guess what? Recently the spring has came in my country as well - and ambient temps, naturally, are rising. So I did a quick test. Set 3.4Vdimm and 260htt - memtest stable speeds back in February. Memtest froze in the 3rd pass of test #5. I took off the side panel, rebooted Memtest run for 120 passes, with errors each 6th-7th pass. I installed the side panel back (with system running memtest). After 4 passes, memtest froze. So the conclusion: it's heat which holds me at the place... simple as that... where was I earlier, huh? :stick:

And that's with my case's 2x120mm fans, and 2x92mm high speed ones installed separately over the memory and the "hot" corner of the 5V mosfet area. Don't know what else much can I do to cool the 5V vdimm mosfet area better. Any ideas (xtreme are velcome as well :) )? Right now, I'm only thinking of building a custom case with the option to install a bunch of fans all over the hot place...

rise up your vdimm ;)
3.4v is the hardest setting for the mosfet

[mrlobber]

rise up your vdimm ;)
3.4v is the hardest setting for the mosfet

Where did you get that and why should it be so? :confused:

[LowRun]

I would suppose this is because the mosfets have to dissipate the excess from 3.4v to 5v but i know near nothing about this stuff.

[mdzcpa]

Where did you get that and why should it be so? :confused:

Resistance builds heat. To get from 5v down to 3.3 takes resistance.

[HARDCORECLOCKER]

To make a long story a bit shorter:

Already back in the February, ever since I own my DFI nF4 board, I was always stopped by the fact that as soon as I tried to run more than 3.4 Vdimm, memtest froze in time of 2-3 minutes, or gave out tons and tons of errors even at much lower speeds. Therefore my VX remained Memtest stable up to around 260mhz, no more, where they were carried with the maximum 3.4 Vdimm I was able to feed them.

After trying almost all the official and beta bioses, various timings suggested for VX & Winchester clocking both by OCZ boys as well as by others (nothing changed really), at first I thought my PSU (PcP&C) was guilty until I modded it to 24pins by disassembling the rails from one of the native SATA connectors. That didn't help at all as well. Than I thought my CPU is crappy (and it is indeed, with those 2300 mhz maximum Prime stable speeds), but just a couple of days ago I noticed some corellation elsewhere. 3.4 Vdimm started to give out similar behaviour. So I thought a little... and guess what? Recently the spring has came in my country as well - and ambient temps, naturally, are rising. So I did a quick test. Set 3.4Vdimm and 260htt - memtest stable speeds back in February. Memtest froze in the 3rd pass of test #5. I took off the side panel, rebooted Memtest run for 120 passes, with errors each 6th-7th pass. I installed the side panel back (with system running memtest). After 4 passes, memtest froze. So the conclusion: it's heat which holds me at the place... simple as that... where was I earlier, huh? :stick:

And that's with my case's 2x120mm fans, and 2x92mm high speed ones installed separately over the memory and the "hot" corner of the 5V mosfet area. Don't know what else much can I do to cool the 5V vdimm mosfet area better. Any ideas (xtreme are velcome as well :) )? Right now, I'm only thinking of building a custom case with the option to install a bunch of fans all over the hot place...

:D So what's Your PWM & chipset temp ?

:toast:

[coop]

Resistance builds heat. To get from 5v down to 3.3 takes resistance.
exactly,... You could also try to use a fan that has more pressure. Not one that has more cfm, but the same size (dimension wise) fan you have now, or even smaller, but that puts out more air. Of course it will be louder.

[Jasonxxx]

Is there a heatsink "edit, just look at pic of one" mounted on this mosfet?, and if so, how is the contact between them? Is there TIM or ???? (I believe I remember someone saying they were soldered to the board, but with what between the two?)

I believe a fan with better air pressure is a good idea, the Delta three blade fans seem to carry a decent amount of pressure...

[EMC2]

FYI... best cooling for the Vmem FET area is by setting a fan to blow across the area (i.e. across the surface), if possible from the right edge of the MB, 2nd best from the top edge. A large portion of the heat is dissipated via the MB copper (traces/planes), so blowing air down onto the surface from above the MB isn't as good ;)


--- Jasonxxx ---

Yes, the HSs are soldered to the MB, but they are secondary cooling. The lowest thermal resistance for the type packages used for those ICs is thru the metal tab which is soldered to a copper area on the MB (usually the Drain on a FET).

[Jasonxxx]

I have a couple of 92mm three blade deltas sitting here that blow ~80cfm w/ ~10 mm H²O pressure, you think that using this fan blowing across the MB would be good enough to cool the mosfet and the ram (vx4000)?

[EMC2]

For VX, at high V... nope. For TCCD, yes.

I would suggest one blowing across for the FETs and one blowing down over the DIMMs for the VX. :toast: You could probably rheostat or 7V the one blowing across the FET if it blows in from the right side.

[The Mofo]

So you could hot clue a fan to the edge of the mb facing the FET's?

[EMC2]

You could... but I wouldn't. Best results would be with the fan around an inch or two away from the edge of the board (for why, you can do an experiment, set a fan on a table. carefully hold you finger in varying spots in front of the fan up close and then a bit further back. You'll notice "dead spots" and "low spots" very close to the fan that fade away a bit further out)

You could attach it to the MB tray that way... or if you have holes in your MB tray farther out than where you are going to put the fan, you can run something like piano wire from the corner MB standoff, thru the bottom screw hole of the fan, and then attach the other end to a screw/standoff behind the fan to hold it in place. Attachment methods fall in the "innovation" category and can be done a bazillion ways... I always like using a method that a) isn't permanent and b) is easily changed/moved/removed.

[mdzcpa]

Attachment methods fall in the "innovation" category and can be done a bazillion ways... I always like using a method that a) isn't permanent and b) is easily changed/moved/removed.

I like to use a "dab" of sticky seal string to hold my fans in place. Easily removable and it sticks to just about anything. Great for mounting fans in locations where there is no convenient conventional mounting spots.

http://www.xtremesystems.org/forums/attachment.php?attachmentid=28471&stc=1

[EMC2]

Haven't ever tried seal string on fans... by itself it holds your fan securely Mike?

*envisions fan falling if big foot accidently hits case* :lol:

[Ailleur]

Heres what i did
Ugly but hey, it works and im not dating my pc
http://www.alienhax.us/uploads/Picture%20006.jpg

[computerpro3]

I like to use a "dab" of sticky seal string to hold my fans in place. Easily removable and it sticks to just about anything. Great for mounting fans in locations where there is no convenient conventional mounting spots.

http://www.xtremesystems.org/forums/attachment.php?attachmentid=28471&stc=1

you just solved my problem of figuring out how to mount my 40mm fans there and on my mushkin :toast: :woot:

[Monkeyyy]

what size and what kind of fans do you guys use? Also, is there anything in the way of cooling the mem on a DFI Ultra-D MB (floppy/IDE cables) if you put a fan on the side of the MB?

A little off topic: One last thing, is there any room left over on the side of the DFI inside the Antec SLK3000-B?

A lot off topic: Since I'm asking questions like a little B***, anyone here try doing a CPU air duct mod on the Freezer 64 (I'm trying to design both an intake from the side and an exhaust to a blow hole, I just haven't fully thought of how I'm gonna isolate the heatsink from crossflow, Thinking of something that looks like an extended 3-D Omega)? On a plus note, if I ever actually survive this next week and put the system together I will be taking lots and lots of pics for the XS people.

[brassen]

This may be a really dumb question, but where exactly are the memory mosfets?

[pancake]

by the side of cpu with heatsinks on

[vlad]

I've used seal string but find it will pull off in time with fans 80mm & larger but i've had real good luck with 3M adheasive tape strips. They are 1"x3" and they hold forever. I hung a 6"x3' strip lights under my kitchen cabnits using four strips and they've never come off on their own. :)

[saaya]

when using the 5v rail feed jumper for vdimm the extra current pretty much gets "cut off" and dumped to the ground, that mosfet is leaking a lot of current to reach 3.4v vdimm wich results in a lot of heat, while when your at 3.6v its cutting off less current and dumps less current to ground wich means lower temps.

please report back if increasing vdimm helped :)

[brassen]

by the side of cpu with heatsinks on

Oh okey, I've always thought them to be on the corner. Time to move the fan then :woot:

[FireDragon]

But the thing with that is that the ram will then produce more heat a the voltage and it may be to much voltage for the ram like my UTT dosn't like anything above 3.4-3.5 volts...no matter the speed...so reall it will be in effect producing "roughly" the same amount of heat just in a differnt arrea...

Dragon

[mrlobber]

when using the 5v rail feed jumper for vdimm the extra current pretty much gets "cut off" and dumped to the ground, that mosfet is leaking a lot of current to reach 3.4v vdimm wich results in a lot of heat, while when your at 3.6v its cutting off less current and dumps less current to ground wich means lower temps.

please report back if increasing vdimm helped :)


Now the shortened story gets a bit longer :D

To answer your question, saaya, not at all. At 3.6Vdimm (and 3.5, and 3.7), memtest errors out immediately (all the 3.4+ Vdimm settings seem to exhibit the behaviour also at lower ambients), and freezes in time of 30 seconds to 1 minute. These high Vdimm memtest failures happen also at lower htt speeds (~230) where I'm even Prime stable in Windows (of course, if setting Vdimm lower, to 3.3, for example). As you see from the thread title, heat might be one reason, but there are obviously others which are completely unclear to me even after 2+ months of board use. I've measured the voltage at ram slots with a DMM while set 3.7 in bios - it was 3.7 there as well, so it doesn't fluctuate.

In fact, I noticed the problem in 3 days I started using my board, and was the first to post about it on dfi-street. However, people had much "simpler" troubles then, like usb not working etc, so nobody did pay much attention to any voltage problems (and I had too much variables to sort out before I would start blaming hardware or whatever). While on my own testing, I think I saw at least a couple of guys trying to run high Vdimm with Winnies, VX, and 20pin PSU's, and they were successful while I was not. Then AndyOCZ posted on bleedingedge and other places that DFI 4v memory settings will not work if using 20pin PSU's (however, they're at least partially working for me). Modding my 20pin to 24pin didn't give anything new in this area. Heat seams to impact stability, but only at the 3.4Vdimm setting. The higher ones fail immediately in memtest even if the system is freshly started after a period of being switched off, and has not warmed up yet.

So here I'm now. Failing high Vdimm and running out of ideas, other than trying a different cpu (CBBID 0451 right now), and a native ATX 2.0 PSU, both of which are out of my budget at the moment. Luckily, today finally I was able to borrow some TCCD from my friend who runs Ultra-D with 3200+ (watercooled, though, while I'm on air) and 20pin Fortron, and was able to reach 300htt with them. We'll see what they do with my present setup.

[FireDragon]

If you are using VX then the higher voltage wont help only hinder ata certain point...

Dragon

[EMC2]

--- brassen --- VMem FETs, upper right corner, the HS all by itself.

--- mrlobber --- If you can splice wires, you can make a proper 24-pin adapter... take the extra 12V, 5V, and GND wires in the 24-pin adapter and wire it to one of your 4-pin HD connectors. That way you are supplying the extra 12V, 5V, & GND pins with individual wires all the way back to the PS, just like a real 24-pin cnx ;) If your PS has a spare SATA power cnx, you can pick up all voltage for all 4 extra pins (+12V, +5V, +3.3V, & GND). Just be sure you get the right rails to the right spots!!

[mrlobber]

--- mrlobber --- If you can splice wires, you can make a proper 24-pin adapter... take the extra 12V, 5V, and GND wires in the 24-pin adapter and wire it to one of your 4-pin HD connectors. That way you are supplying the extra 12V, 5V, & GND pins with individual wires all the way back to the PS, just like a real 24-pin cnx ;) If your PS has a spare SATA power cnx, you can pick up all voltage for all 4 extra pins (+12V, +5V, +3.3V, & GND). Just be sure you get the right rails to the right spots!!

EMC2, by meaning "modding 20 to 24pin" this is exactly what I did (reread my first post). By no means I'd run anything like a 20 to 24pin adapter. But as I already said, that didn't change anything.

[EMC2]

You mean I am suppose to read the entire thread? :p: My bad :hitself:

In that case, combined with your success with the TCCD and your results with case cover removed < since now I have read 1st post :lol > you are indeed pretty much limited to temp related topics or a winnie that doesn't like the higher Vmem (possible, but...) Already commented on fan position earlier (from side), so... Don't know what your case looks like, but if tight on side but a drive bay same height as Vmem FET area... semi-extreme would be cobble fan and duct to push external-to-case air over Vmem FET area. :shrug: