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

[Ace-a-Rue]

i am hoping this gets resolved on future bios releases...i am a 1000 points behind what the IP35 Pro board will allow for 3D2001 bench...i faced the same going from abit's AB9 QuadGT (P965x) over to their P35 chipset...actually, it got better once i flashed the Pro's boot block...i see DFI has that switch incorporated in their autoexec.bat file....now, it's up to them to improve on (tweak) the performance output.

FIRST IMAGE IS THE IP35 PRO

SECOND IMAGE IS THE DFI LT X38 T2R

[eva2000]

i am hoping this gets resolved on future bios releases...i am a 1000 points behind what the IP35 Pro board will allow for 3D2001 bench...i faced the same going from abit's AB9 QuadGT (P965x) over to their P35 chipset...actually, it got better once i flashed the Pro's boot block...i see DFI has that switch incorporated in their autoexec.bat file....now, it's up to them to improve on (tweak) the performance output.

FIRST IMAGE IS THE IP35 PRO

SECOND IMAGE IS THE DFI LT X38 T2R

got everest bandwidth and latency comparisons between the 2 systems ? I'm finding from testing Gigabyte X38-DQ6 and DFI X38 LT that X38 by default seems to have loosened up some latencies to allow higher memory clocks but they've included in their respective bioses, options to tighten then back up. For DFI X38 those options seem to be these ones marked with *

It could be the case with most X38 boards compared to P35 ?

DRAM Timing
- Enhance Data transmitting: FAST *
- Enhance Addressing: FAST *
- T2 Dispatch: Enabled *

Clock Setting Fine Delay
Ch1 Clock Crossing Setting: More Aggressive *
- DIMM 1 Clock fine delay: 6
- DIMM 2 Clock fine delay: 7
- Ch 1 Command fine delay: 11
- Ch 1 Control fine delay: 8


Ch2 Clock Crossing Setting: More Aggressive *
- DIMM 3 Clock fine delay: 6
- DIMM 4 Clock fine delay: 7
- Ch 2 Command fine delay: 11
- Ch 2 Control fine delay: 6

Ch1Ch2 CommonClock Setting: More Aggressive *

Ch1 RDCAS GNT-Chip Delay: Auto
Ch1 WRCAS GNT-Chip Delay: Auto
Ch1 Command to CS Delay: Auto

Ch2 RDCAS GNT-Chip Delay: Auto
Ch2 WRCAS GNT-Chip Delay: Auto
Ch2 Command to CS Delay: Auto (where cpuz sees 1T or 2T SETTING)

CAS Latency Time (tCL): 4
RAS# to CAS# Delay (tRCD): 4
RAS# Precharge (tRP): 4
Precharge Delay (tRAS): 5
All Precharge to Act: 4
REF to ACT Delay (tRFC): 30
Performance LVL (Read Delay) (tRD): 5 *

Read delay phase adjust: Enter

Ch1 Read delay phase (4~0)
- Channel 1 Phase 0 Pull-In: AUTO *
- Channel 1 Phase 1 Pull-In: AUTO *
- Channel 1 Phase 2 Pull-In: AUTO *
- Channel 1 Phase 3 Pull-In: AUTO *
- Channel 1 Phase 4 Pull-In: AUTO *

Ch2 Read delay phase (4~0)
- Channel 2 Phase 0 Pull-In: Auto *
- Channel 2 Phase 1 Pull-In: Auto *
- Channel 2 Phase 2 Pull-In: AUTO *
- Channel 2 Phase 3 Pull-In: AUTO *
- Channel 2 Phase 4 Pull-In: AUTO *

[Aerou]

wow, really appreciate that post eva! thanks

so the clock-fine delay - does it alter even performance?

[eva2000]

read http://i4memory.com/showthread.php?t=5226 lots of info.. clock fine delay improves memory clocks so in a way yes better performance in memory bandwidth intensive apps through higher memory clocks..

[Ace-a-Rue]

got everest bandwidth and latency comparisons between the 2 systems ? I'm finding from testing Gigabyte X38-DQ6 and DFI X38 LT that X38 by default seems to have loosened up some latencies to allow higher memory clocks but they've included in their respective bioses, options to tighten then back up. For DFI X38 those options seem to be these ones marked with *

It could be the case with most X38 boards compared to P35 ?

i already have adjusted all of those that you have asteriks!...even using PL=6.

it has become apparent that the DFI board will keep you busy with tweak options while the IP35 Pro board is made for the guy who likes the easy way!

[Aerou]

it has become apparent that the DFI board will keep you busy with tweak options while the IP35 Pro board is made for the guy who likes the easy way!

that is kinda apparent, but I wish you could achieve higher performance at the same clocks with the DFI - did you doublecheck all mem subtimings and tryed tightening the options that eva stressed out? I really am considering this board for some quad-fun, but since Xfire aint gonna happen, and if I can have higher performance on a different board (maybe even clock higher) ... that why should I get the DFI? *sigh* the memories of the X38 hype that was ... is X38=DDR3 true then?

[ic3m4n2005]

What do you guys think is suitable voltage on the nb and vtt ?

[Ace-a-Rue]

that is kinda apparent, but I wish you could achieve higher performance at the same clocks with the DFI - did you doublecheck all mem subtimings and tryed tightening the options that eva stressed out? I really am considering this board for some quad-fun, but since Xfire aint gonna happen, and if I can have higher performance on a different board (maybe even clock higher) ... that why should I get the DFI? *sigh* the memories of the X38 hype that was ... is X38=DDR3 true then?

my memory is lower end since it was an 8 gig kit (PC6400) for $265...STILL...i have it at DDR-1003 and i can go to DDR-1050(ish) but still i cannot get inside the 1000 point difference with 3d2001 score from the pro board.

that said, this is still an upgrade for me since it has less Vdroop, 8 phase digital PWM power and 2 PCIe 16x working slots that allow me to put my 8x Areca raid card in and WORK properly!...some people can not see any difference in speed but i saw a slight increase in burst speed (630 to 660 mb/sec) in HDTach on the Areca card working between 4x slot or 16x slot (8x actual speed).

my assessment is like yours...the X38 was meant (designed) to do better with DDR3 memory...that is where the real gains are to be gotten.

[Johnny T.O.]

Hello! Does this motherboard run stable without any fan on northbridge chipset? I want to buy Thermalright U120E and I know that it can be only in one position. What do you all think about thin fan like Sunon KDE1207PTV3 on chipset? Will TRUE stand in any position on motherboard?

[Ace-a-Rue]

i have the TRUE mounted and i am using a 38mm thick fan on the heatsink...it is mounted on the NB side and so, you won't need any fan mounted on the NB since that fan is way strong enough to cool that plus the TRUE.

[Aerou]

did anyone test 400+ FSB and running ahhd in AHCI mode on the onboard (ICH9R) controller? thanks

[Jor3lBR]

Anyone here tried to OC with 4 x 1gb or 2 x 2gb is better to achieve higher clocks?

R.

[Ace-a-Rue]

i think this board offers better opportunity to clock 4 x 1gb sticks higher with all the DRAM settings in the bios.

[Jor3lBR]

i think this board offers better opportunity to clock 4 x 1gb sticks higher with all the DRAM settings in the bios.

Thanks! I thought that also.

I can't wait for the system to arrive so I can start to burn it and give some feedback for you guys.

For now I'm on the 3rd page of this Thread! Lot's of reading to do till New Year.

[intothefantasy]

seriusly i would really love to see some q6600 or yorkies oc in action on this board...i think eva has dominate the whole memory clocking thing

[Jor3lBR]

Finding it hard to follow my results ? well Merry Christmas guys => http://i4memory.com/showthread.php?t=5226

Thats great Eva, thanks it will be easier for the new comers to read all.

What are your best results for 4Gb (4 x 1GB) Ballistix Tracer 8500? Have you tried that yet?

[zsamz_]

Anyone here tried to OC with 4 x 1gb or 2 x 2gb is better to achieve higher clocks?

R.

4x1gig 1200mhz is no problem with this board @ 5-5-5-15 but i like 5-4-4-15 @ lower clock n lower voltage
the pic is sticks

[sofarfrome]

seriusly i would really love to see some q6600 or yorkies oc in action on this board...i think eva has dominate the whole memory clocking thing

I can tell you that you will be at the mercy of the FSB wall of the Quad CPU. I have tried 1 Q6600 and 1 x3210, both G0. The Q6600 has a FSB wall of 480 and the x3210 is around 485, depending on how much vcore you want to apply. Increasing CPU PLL and NB vcore doesn't do much with quads except increase the stress on the board. I can get the Q6600 to run around 450FSB with the 9x multi but it is not prime95 stable. Hell, it isn't even 32M Pi stable. But I am not done trying yet. These are just my initial findings. I am sure the QX9650 will work better than the 2 I mentioned but I certainly am not going to spend $1000 to find out.

[OnDborder]

Too bad the vcore only goes to 1.60
I like an option for more..

[Jor3lBR]

I can tell you that you will be at the mercy of the FSB wall of the Quad CPU. I have tried 1 Q6600 and 1 x3210, both G0. The Q6600 has a FSB wall of 480 and the x3210 is around 485, depending on how much vcore you want to apply. Increasing CPU PLL and NB vcore doesn't do much with quads except increase the stress on the board. I can get the Q6600 to run around 450FSB with the 9x multi but it is not prime95 stable. Hell, it isn't even 32M Pi stable. But I am not done trying yet. These are just my initial findings. I am sure the QX9650 will work better than the 2 I mentioned but I certainly am not going to spend $1000 to find out.

How did you set the jumper for that Q6600 G0 on the dfi x38? 1333 or 1066fsb?

[Praz]

Too bad the vcore only goes to 1.60
I like an option for more..

It does. It's called CPU VID Special Add.

[lazy]

i'm idling 32c on nb n pwm never went higher than 36c on both just got some new 40mm fans for pwm i'll see if it made a difference soon

what do you use to measure the temps for those areas?

Also what is a good fan to place on the nb?

[CapFTP]

limits for my Tracer8500....not like Eva, but quite satisfied

[sofarfrome]

How did you set the jumper for that Q6600 G0 on the dfi x38? 1333 or 1066fsb?

I let the CPU didctate it (default).

[CapFTP]

for those who like GKX's on this mobo and 2x3850 Crossfire (stock freq and cooling)



nice to see there's a PL auto-adjusting on bank A and B....