Official DFI LanParty UT X38-T2R(and LT) Discussion/Review/Overclock/Guide Thread

[deadman'sclick]

learn to read

he was only trying to help man altough he did totally miss that bit


on ure prob zsamz only thing i can think of trying, in this order is..

turn off the psu by its switch & give it a few seconds to fully turn off,

then press the power button on ure case or the mobo a few times to get all the power out the mobo etc

remove power cable,

take the battery out move the clear pin over & give it a few min's,

put batt back in, move jumper pin back over & try from there.

has worked for me with a simular prob. i was trying to move my 1900xtx into the secondary slot on my icfx3200 & got just the same thing ure getting, works for a second then black screen.

i was orignaly advised to leave the battery out a copple of hours , but 2 min's was all it needed.

[Rinaun]

learn to read

Calm down skip, thats no way to be responding to someone giving help.

[eva2000]

I found the best advice so far for GTLs was in a completely random thread, the bonetrail board official thread.



Now that I've gotten this beast stable at ddr21000, I'm going to tinker with the gtls and see if I can run with it. Should I really worry about 1.55 vcore (when pushing into 4.0)? 1.6 vcore? My water cooling keeps this 3.8 idling at 25 and running at 45.

i'd only bench at those volts for 24/7 <1.36-1.4v for 45nm for me.

Sub 10s Super Pi 1M @4606Mhz



Super Pi 32M @4564Mhz at 1.648v idle/1.632v load

2x1GB Crucial Ballistix PC2-8500 Tracer @652Mhz 5-5-5-8 at 2.35v bios set!



CPU VTT = 1.28v bios set
GTL = 115/107/88
Vdimm = 2.35v bios set

[Rinaun]

i'd only bench at those volts for 24/7 <1.36-1.4v for 45nm for me.

Thanks for the advice eva, I saw you enjoying your new chip in the other thread. Think you can push it much farther?

[Rinaun]

@Rinaun
I find this method a little too timetaking as you get your max/min numbers ONLY for ONE config of voltages (Vcore, VTT, VNB, Vddr etc.)... Sometimes instability is more a matter of changing voltage(s) one setting, rather than finding "the middle" GTL value... that btw will have to be changed after optimizing main voltages... endless tweaking...

Yea, I'm having a super rough time with my OC. I can get ddr2 1000 stable at 3.7 but anywhere I move up causes OCCT errors on only RAM. I bumped only VTT to 1.4 with appropriate gtls and couldn't get anything either.

[zsamz_]

he was only trying to help man altough he did totally miss that bit


on ure prob zsamz only thing i can think of trying, in this order is..

turn off the psu by its switch & give it a few seconds to fully turn off,

then press the power button on ure case or the mobo a few times to get all the power out the mobo etc

remove power cable,

take the battery out move the clear pin over & give it a few min's,

put batt back in, move jumper pin back over & try from there.

has worked for me with a simular prob. i was trying to move my 1900xtx into the secondary slot on my icfx3200 & got just the same thing ure getting, works for a second then black screen.

i was orignaly advised to leave the battery out a copple of hours , but 2 min's was all it needed.

did all that n more
8 hours of playin with it was enough
cant boot into windows with pci-e
can only boot with a pci card
1 drive got corrupted with so many tries
i even changed the bios chip took out batterie ti'll tomorow
i i got my hands on a 965-dq6 till its fixed or rma
set up the 8800gt n booted straight through windows my primary drive that would lock up on the dfi
so i reinstalled windows i like it 100%

[Rinaun]

did all that n more
8 hours of playin with it was enough
cant boot into windows with pci-e
can only boot with a pci card
1 drive got corrupted with so many tries
i even changed the bios chip took out batterie ti'll tomorow
i i got my hands on a 965-dq6 till its fixed or rma
set up the 8800gt n booted straight through windows my primary drive that would lock up on the dfi
so i reinstalled windows i like it 100&#37;

Maybe its the x38 to 48 flash? or when you switch the chips did you have two different BIOSes? It's suprising none of the pcies worked, I have a friend with that same pcie not working issue on another dfi board(NF4). He ended up buying this board to fix that problem

[zsamz_]

Maybe its the x38 to 48 flash? or when you switch the chips did you have two different BIOSes? It's suprising none of the pcies worked, I have a friend with that same pcie not working issue on another dfi board(NF4). He ended up buying this board to fix that problem

the x48 bios is almost the same just nb voltage increase
its actually better for taiwan board
the fixed bios for china board n taiwan boards got the shaft on volts
they shoulda made 1 for taiwan 1 for china boards
i been back to x38 long ago just for everest voltage temps settings i just set nb volts a little higher
if the x48 bios ever gets better than x38 i will go back to x48

[flexy]

i am back to X48 bios and keep my NB at 1.5 and just dont mess with NB. The reason is that i did some tests w/ same settings...and X48 was stable.

Another observation:

I dont know whether its my settings, or a combination of settings and my hardware. I did a OCCT MIX test 2 hrs stable at one set of settings.
After a few warm-reboots....same setting BSODs OCCT out in 4 minutes.

In general using OCCT if i have problems i am getting rather BSODS and only *rarely* calculation errors.

the weird thing now is that i believe my settings are stable (i mean i let it run 2 hrs OCCT)...and when i do a cold-start (power-button) the same settings which gave me BSOD after 3 mins befoer are now stable again. I always save my stable tested settings to CMOS...so there is no reason that a setting i tested 2 hours the day before all of a sudden becomes instable......BIZARRE.

So i have reason to believe that its advisable after any change in bios settings to do a cold-start and then test..otherwise i get BSODs in OCCT no matter what, even if settings are actual stable.

I still did not solve my problem that bios sometimes resets/hangs with my strap/settings...i already tweaked a lot and changed voltages and losened timings. But as long as i am on 8x442 i have this problem. I think both problems (instability without a cold-start) might be related.

[Neuuubeh]

Guys how do you test for stability?
No matter what I do I always seem to get 1 error on the Prime95 Blend test. All other tests run fine, but I cant get rid of this error.. is that normal lol

[eva2000]

rather than waste hours on occt/prime95, I first use systool pi test quad 16M or dual 32M (for quad core or dual core respectively) if you only have 2GB ram - don't try quad 32M as you'd run of my memory heh.

systool pi tests takes around 4-8 mins to run and should be a good starter.

wrong GTL ref volts or unstable FSB as well as unstable mem settings will reboot system usually.

then you can move onto prime95 tests etc

Thanks for the advice eva, I saw you enjoying your new chip in the other thread. Think you can push it much farther?

on air not much more i don't think as heat is holding it back 78+ C load temps.. water maybe another 100Mhz for 32M i hope

[Praz]

I found the best advice so far for GTLs was in a completely random thread, the bonetrail board official thread.

To completely understanding the reference voltages is complicated. But a basic understanding of what to do with them is pretty easy.

Adjust it in one direction until you get a failure. The failure criteria can be anything, as long as it's consistent. Booting to Windows is a pretty good test. Fine tune it down to one step = pass, the next step = fail. Then adjust it in the other direction and fine tune in again. You'll end up with a min and max number, usually referred to as you margin window. Then you simply pick the number in the middle of the window, and set your final value at that middle point. That will be your best stability point, at the voltages and frequencies that you tested at.

You don't need to do long term stress testing to find the absolute min and max stability points, your not going to be running out there, you just want to use a consistent, and quick test to fine min and max, so you can then calculate the middle point.

As you increase frequency the min and max numbers, your margin window, will shrink down. The min might shrink faster than the max, so you might start failing because your negative margin, adjustment range below default, has collapsed down to nothing. But you might still have positive margin, adjustment range above default. You could raise your reference level to recenter in the margin window that is remaining.

If you jump too far on the adjustment and cause a hard no boot failure, you'll have to pull power from your power supply and move the BIOS jumper to get back going again, so go in small steps

Now that I've gotten this beast stable at ddr21000, I'm going to tinker with the gtls and see if I can run with it. Should I really worry about 1.55 vcore (when pushing into 4.0)? 1.6 vcore? My water cooling keeps this 3.8 idling at 25 and running at 45.

The problem with that method is on the logic 1 of the signal. It's possible to drive GTLREF hard into overshoot and still have stability as long as the resulting ringing is clamped fast enough. This will give a false impression of the midpoint as the indicated upper bound of GTLREF is considerably higher then is being indicated. Also GTLREF overshoot can cause irreversible damage to the processor.

[Jor3lBR]

rather than waste hours on occt/prime95, I first use systool pi test quad 16M or dual 32M (for quad core or dual core respectively) if you only have 2GB ram - don't try quad 32M as you'd run of my memory heh.

systool pi tests takes around 4-8 mins to run and should be a good starter.

wrong GTL ref volts or unstable FSB as well as unstable mem settings will reboot system usually.

then you can move onto prime95 tests etc

on air not much more i don't think as heat is holding it back 78+ C load temps.. water maybe another 100Mhz for 32M i hope

Link to that systool pi software download?

[t_ski]

systool is another software program from W1zzard (ATITool and GPU-z creator). Go to www.techpowerup.com for the file.

[Jor3lBR]

systool is another software program from W1zzard (ATITool and GPU-z creator). Go to www.techpowerup.com for the file.

Thanks T I thought this was a OC tool only (like setFsb) never heard of it having a stability PI quad test that takes 8min to check system

[eva2000]

Thanks T I thought this was a OC tool only (like setFsb) never heard of it having a stability PI quad test that takes 8min to check system

Actually <8min

quad 16m actually uses 1.33GB not 960MB memory



not 100&#37; stable test but it's a starter before moving onto prime95

[Jor3lBR]

Actually <8min

quad 16m actually uses 1.33GB not 960MB memory



not 100% stable test but it's a starter before moving onto prime95

Sweet I have to use this tool!

My old process was to use wPrime>SuperPi>3dMark06>OCCT1hr>Prime95

But my PC is totally disassembled for the new upgrades, I guess this will have to wait.

[t_ski]

You're welcome for the info. I used to use it all the time for my DFI NF4 board, and I would use it for temp monitoring, stability testing, and even for trying to find an easy max OC for the current mem settings.

I didn't have any luck with it supporting my Asus P5W DH, so I kinda forgot about it. I'll have to go and try it out again myself.

[Jor3lBR]

Thats still too close for comfort for me, I need atleast 1/2in between them to make me satisfactory. I use second slot.

What you said above doesnt make sense at all.
If you use 2nd PCI_e slot then the cooler would be very very far from touching anything. So I didn't get what your point was

@Jor3lBR
Did you have to change/mod anything to fit RAM cooler over gfx card ?? There is really NO contact between RAM cooler and gfx card ??
Another photos would be nice to see it closely

The only change you need to do is to apply black tape on the XTC leg that may touch the lower part of the gfx card, it's a precaution to avoid short-circuits.

The black tape is just on that small part of the ocz xtc and it's unnoticeable.

IMPORTANT NOTE:
The OCZ XTC Cooler does NOT fit with original DFI Chipset cooler. So you should try other solution like zsam did.

[t_ski]

Are you using the Danger Den Universal on the chipset:

http://www.dangerden.com/store/produ...&cat=46&page=1

Or the new X38:

http://www.dangerden.com/store/produ...&cat=46&page=1

[Rinaun]

What you said above doesnt make sense at all.
If you use 2nd PCI_e slot then the cooler would be very very far from touching anything. So I didn't get what your point was



The only change you need to do is to apply black tape on the XTC leg that may touch the lower part of the gfx card, it's a precaution to avoid short-circuits.

The black tape is just on that small part of the ocz xtc and it's unnoticeable.

IMPORTANT NOTE:
The OCZ XTC Cooler does NOT fit with original DFI Chipset cooler. So you should try other solution like zsam did.

I was saying I USE the second slot, because I don't want to tape up my expensive rig and rather make it look right.

[Jor3lBR]

Are you using the Danger Den Universal on the chipset:

http://www.dangerden.com/store/produ...&cat=46&page=1

Or the new X38:

http://www.dangerden.com/store/produ...&cat=46&page=1

MPC Universal I don't think the X38 will fit and it's bulky.

[t_ski]

Cool - thanks

[zanzabar]

MPC Universal I don't think the X38 will fit and it's bulky.

u dont have to cover the entire IHS u just need the center section, the swifttech or any standard sized one will work fine

[Neuuubeh]

Actually <8min

quad 16m actually uses 1.33GB not 960MB memory

not 100% stable test but it's a starter before moving onto prime95

hm, weird, when i did this today it completed within 200 secs (i think it was 48 sec per thread) on a x3350 at 3.3ghz . Tho I guess you mightve had stuff running in background. Either that or my brain is playing tricks on me
gonna try again when I get home