Calling all BH-5 and OCZ VX users on DFI NF4... using Vdimm higher tha

[mrlobber]

Hey all who are using any kind of high-voltage loving memory (BH-5, BH-6, maybe also CH-5, Twinmos UTT, OCZ VX et al).

 

I'd like to know how your memory behaves in memtest test #5 IF using 3.5 Vdimm or more (which, of course, is derived from the 5v rail).

 

What I have observed with my BH-5 kit and OCZ VX as well, as soon as I up the Vdimm to 3.5v or more in bios, and loop memtest test #5 for a few loops (10 is enough, you can see it then), memory starts showing memtest errors EXACTLY AT THE SAME MEMORY ADDRESS all the time. Why does it seem suspicious? Because, if memory is not memtest stable, it usually throws out errors at various addresses (maybe in the same range but almost never in the same address, as one chip on the memory is failing, and it is usually the size of 64 Mb now). Moreover, this happens also (my VX is memtest stable up to 262 htt 1:1) on lower speeds, say, 255 htt. The only variable here which seems to impact these errors, is the Vdimm. As soon as I lower it, the errors disappear.

 

So I'd like to know whether your boards and memory exhibit the same "issue" - to see whether this is an isolated incident or not.

 

Please, also if you don't want to run your RAM at those voltages, do a 5-10 minute test, it would be enough. This little test will not hurt your memory, especially if you've got a fan blowing over it (and position a fan over the 5v Vdimm jumper area as well, as it is strongly recommended to avoid overheating).

 

For OCZ VX I'm running the following memory timings (but you can use the best ones you've found for your RAM, if it's a voltage issue, it would pop up anyway):

 

Timings are listed as in A64Tweaker, but DFI bios has all the stuff pretty much in close order anyway:

 

Tcl (cas latency): 2

Trc: 12

Trfc: 16

Trcd: 2

Trrd: 2

Tras: 10

Trp: 2

Twr: 2

Twtr: 2

Trtw: 2

Cas Write Latency: Auto (board doesn't post at any other settings)

Tref (refresh rate): 3072 cycles

 

Read Write Queue Bypass: 16x

CPC: enabled (1T)

Bypass Max: 4x

Max Async Latency: 7ns

Read Preamble: 5ns

Idle Cycle Limit: 16 cycles

Dynamic Idle Cycle Counter: Enable

 

Bank Interleave: enabled

 

 

Thanks & let me know.

[Guest eva2000]

don't have DFI NF4 but try these settings

 

Tcl (cas latency): 2

Trc: 12

Trfc: 16

Trcd: 2

Trrd: 2

Tras: 6

Trp: 2

Twr: 2

Twtr: 2

Trtw: 2 or 3

Cas Write Latency: Auto (board doesn't post at any other settings)

Tref (refresh rate): AUTO

 

Read Write Queue Bypass: 8x

CPC: enabled (1T)

Bypass Max: 4x

Max Async Latency: 7ns

Read Preamble: 5.5ns

Idle Cycle Limit: 16 cycles

Dynamic Idle Cycle Counter: Enabled

[mrlobber]

In fact, my VX does post with Tras 6 (I've tried that), but it doesn't boot, that is, with a black screen and blinking cursor it hangs at the point where it should boot floppy, cd-rom or hdd.

 

Just read some another thread about various bios memory tweaks, this time on ocforums.com: http://www.ocforums.com/showthread.php?t=329627

 

Might want to play more with DRAM drive strength and Tref values when I get home (typing this post from work )

 

However, at Xtreme where I posted this in the DFI nF4 bugs/fixes thread, I have already 2 people reporting quite similar issues.

 

One of them was using BH-5 on DFI nF4, and run into "one address failure" in the 260 range as well.

 

The other had OCZ VX, and had the problem, but note, on a MSI NF3 board (obviously, using a DDR Booster). He provided a nice screenie as well (you can check it out to see what I mean: http://www.xtremesystems.org/forums/showth...3277#post703277)

 

Still insufficient data to make even a preliminary conclusion where the problem might be (if it is a problem at all). Memtest? Voltage instabilities? Memory itself? Something with cpu memcontroller?

 

Myself, I cannot isolate the cause because my cpu is quite weak, as I already said in my "DFI SLI-D first impressions thread". Also, lowering the multi to keep cpu speeds out of the equation, doesn't seem to help, the SuperPi etc crashes are always there if climbing to 250htt (even dorked my Win installation once while messing around with Clockgen). So either my board has trouble keeping even moderate HTT stable (hard to believe anyway, and I can boot in 260, but not fully stable even at 240) or there is something more in it. I'm not yet ready to set 3.7Vdimm and something like 270x6 in bios, and then try to boot

 

So guys, who can, let's do some testing.

[Guest eva2000]

try my settings with tRAS of 10 then

 

weird issues though

 

also you are doing a full test #1-8 loop first right ?

 

use memtest86+ 1.50 from www.memtest.org

[psycho303]

hi there

 

like mrlobber already wrote , i got exactly the same problems with my bh5 at high voltages , is there nobody else who run high voltages on dfi nf4 ?

[bubbers214]

I am also having the exact same issue, I am using 1x512 of OCZ VX and i get errors when i get into the higher volts, even when running at stock ocz speeds.

[Guest eva2000]

bigtoe has said 5v jumper is less stable than using 3.3v jumper with 3.3v rail bump

[Rabbi]

mrlobber, are you able to check with high 3.3vRail.... set it to 3.6v-3.7v and see if 3.5vDimm is then stable

[mrlobber]

mrlobber, are you able to check with high 3.3vRail.... set it to 3.6v-3.7v and see if 3.5vDimm is then stable

 

Unfortunately no. Internal pot of my PcP&C maxes out at 3.42 (measured with a DMM).

 

I wanted to ship the PSU to PcP&C to rewire it to a 24pin one, and maybe ask them to install a higher pots as well (don't want to mod it myself) but now I'm buying ATI X800XL and have run out of funds.

 

Btw, played yesterday with a locally developed benchmark (doesn't matter much what exactly this is), was able to complete it @ 273x10 1:1, 2-2-2-10-12-16, Vcore: 1.45*123%, Vdimm: 3.7v, Vdd: 1.6, cooled by stock heatsink, but subzero outside air (cpu: +10c idle/+29c load, chipset: +8c, pwmic: +2c)

 

However, I'm trouble stabilizing Prime and SuperPi higher than 245htt on ordinary conditions, and this morning, memtest suddenly was not stable where it was before (262x10, 3.4Vdimm, 1.55Vcore), it froze and started giving out errors... we'll see whether some of the components don't intend to die... :mad:

[psycho303]

could be the issue on weak mem controllers somehow doesent like high mem voltages ? mine is 3000+

[psycho303]

However, I'm trouble stabilizing Prime and SuperPi higher than 245htt on ordinary conditions, and this morning, memtest suddenly was not stable where it was before (262x10, 3.4Vdimm, 1.55Vcore), it froze and started giving out errors... we'll see whether some of the components don't intend to die... :mad:

 

mem stil working ?

[bubbers214]

At 3.2 Vdimm (with the 5v jumper on) i was getting erros in memtest on stock speed, when i moved the 5v jumper back over to default pins where 3.2 is the max voltage selectable it was completely error free on the test. And i am running at 235mhz @ 7-2-2-2 on 3.2 error free. There must be either instability with the 5v jumper or my psu isn't powering it enough (although its a brand new 520 ocz so i hope that is not it).