**Official DFI LanParty UT X48-T2R(and LT) Discussion

[Mussels]

why is it to load, it has the same sound dampening and more fans, more fans mean that u can undervolt and get the same flow for less noise and it allows the top to be used for a fan and the bottom for a psu so it will be cleaner.

its the same type of case so im just giving a suggestion since its about the same price

more open areas for noise to escape out of. more holes in the case, the more noise... and the 300 has an extra side fan intake, the top fan hole, and the entire front of the case as an open mesh... its far, far inferior in terms of noise.

all i wanted to know was about the NB coolders and i got my answers. i'll be grabbing the xigmatek NB cooler most likely, as its cheaper.

[sakis_the_fraud]

Thanks for the info mikeyakame!

Driving strength is set for the ICs. 6/8 usually works by default. To adjust you either increase clk drive/decrease dq drive or decrease clk drive/increase dq drive, but you need to change DLL models too when doing this.

On my DDR2 G.Skill 1100mhz 2x2gb sticks I've found the best combo for me is, in the format CK Drv/DQ Drv/Dll 1/Dll 2

3/8/5/5 or 4/7/6/6.

i see that in your working combos the sum of CK Drv and DQ Drv is the same (3+8=11, 4+7=11). do i have to try only numbers that make 11 or just try everything?

Dll 1 and Dll 2 have to be the same?

how do i test if my selection (eg 5/8/5/5) is the best?

MCH Vref are set with respect to the difference default.

Vref1=00h, Vref2=00h, Vref3=80h.

The represent voltage reference points at voltage crossing. You adjust them to the same ratio they are set. ie, Vref1=05h, Vref2=05h, Vref3=85h.

I got it! whatever you add to Vref1 you add it and to the others. Also the numbers are in hexadecimal format.

Are there any settings that "usually work"? I was told to use 23-23-A3.

Rtt = 1/2 ROn
Rtt = 2 * ROn

I am not expert at math but i think that something is wrong here ^

Do i have to change Rtt/ROn in order to be stable at 530?

[mikeyakame]

Thanks for the info mikeyakame!



i see that in your working combos the sum of CK Drv and DQ Drv is the same (3+8=11, 4+7=11). do i have to try only numbers that make 11 or just try everything?

Dll 1 and Dll 2 have to be the same?

how do i test if my selection (eg 5/8/5/5) is the best?



I got it! whatever you add to Vref1 you add it and to the others. Also the numbers are in hexadecimal format.

Are there any settings that "usually work"? I was told to use 23-23-A3.



I am not expert at math but i think that something is wrong here ^

Do i have to change Rtt/ROn in order to be stable at 530?

DLL models follow an order, ie 1 > 2 > 3 > 4 > 5 > 6 > 7.
Micron ICs will generally use DLL1-4. Powerchip ICs seem to like DLL5-7. Samsung ICs like DLL4-6. I dont know about others though.

DLL0 is generic and default, it works with all memory ics usually, but don't expect the fine delays to be anywhere near optimal.

DLL1 has strongest clock fine delay values and relatively weak control/command fine delay.
DLL2 has strong clock fine delay and moderately weak control/command fine delay.
DLL3 has strong clock fine delay and moderate control/command fine delay.
DLL4 has moderately strong clock fine delay and moderately strong control/command fine delay.
DLL5 has moderately clock fine delay and moderately strong control/command fine delay.
DLL6 has weak clock fine delay and moderate control/command fine delay.
DLL7 has weakest clock fine delay values and moderately weak control/command fine delay.

I use 3/8/5/5 because, DLL5 has moderate clock fine delay, so clock drive strength of 3 suits it with my ICs. DQ Drive Strength of 8, because DLL5 has moderately strong control/command fine delay values, so to use higher fine delay PS values, the DQ drive strength needs to be at 8 so weak clock and strong data balance the system.

Weak clock/strong data gives best results with tight performance levels, and aggressive cross clocking. This set up from my testing, works the opposite as PL loosens up. Ie PL7 @ 425fsb is stable, PL8 @ 425FSB gives occasional errors. Moderate clock/Moderately strong data gives best results when using moderate Performance Levels. Ie PL8-9 with aggressive cross clocking with clk drive strength = 4, dq drive strength = 7, dll1/2 = 6, is stable, but PL7 is too tight and throws errors.

That's what I've figured out from testing, it might not be spot on but they seem to reflect the DLL models delay values.
Well there is no exact method for testing what is best, the way i figured mine out was pretty simple actually.

Find a memory frequency that will post but throws errors in Memtest86+, ie sticks are stable to 1180mhz, so try 1190 or 1200mhz. Just don't run OS
Now all you do is go through combinations of drive strengths/dll models, and find the ones that give you the least reported errors. I use test #5 for this (block move), it gives best results.

While doing this make notes for each change, and error count, and repeat. By the end of it you'll have lots of data to go through, and you'll find that when you use the dll/drive strengths that gave you least errors, they will be the best for your memory at most frequencies. If you see Control Fine Delay or Command Fine Delay = 0ps or < 50ps, then your DLL model isn't suited.

MCH Vrefs sort of scale with fsb, but at the same time they won't follow a pattern. Different frequencies respond and react differently. Oscillation due to drive amplitude causes this.

The easiest way to set up MCH Vrefs, is to start from something like 440mhz fsb, and work your way up to your target FSB. This way you'll be able to make small ongoing adjustments and you'll find it much easier to see what works better and what doesn't.

MCH ROn (Resistance On) / Rtt (Termination Resistance) only need adjusting if the clock gen can't keep up or is causing erratic clock jitter/skewing.

For a dual core you may not need to worry about these nearly as much as a quad core, but then again the clock generator on the UT/LT is weak and inconsistent, at least with quad cores. Mine needs MCH Vref and Slew Rate adjustment at only 420mhz FSB.

At 425mhz fsb for example I've found that with my board, the following works right.

MCH Vref1 = 05h
MCH Vref2 = 05h
MCH Vref3 = 85h
MCH Slew Rate = 01h
MCH Ron = 00h
MCH Rtt = 00h
CPU GTL1/2 = 101
CPU GTL0/3 = 87
MCH GTL = 81
Vtt = 1.175v
Vck_gen = 3.60v

The method I use for checking those adjustments is with Linpack. It takes patience, but it's accurate.

I run 2 tests, 10 passes for #1, 6 passes for #2.
This is what my linpack config file looks like, I just name it lp64.in and call it with linpack-xeon64.exe in a batch file.

Intel(R) LINPACK 10.1 Beta2
User Specified Datafile
2                       # number of tests
8000 16500              # problem sizes ( 8 * problemsize^2 ) == memory usage
8008 16508      # leading dimensions ( same as problem size)
10 8                    # times to run a test
4 4             # alignment values ( in kBytes )

Basically what linpack tells you is the time for each test to complete, and the cpu workload for it, as you change things like GTLs, MCH Vrefs, DRAM DLL Clock / Control Skews, etc you'll see the "GFlops" and "time" change.

============= Timing linear equation system solver =================

Size   LDA    Align. Time(s)    GFlops   Residual     Residual(norm)
8000   8008   4      7.126      47.9174  6.260895e-011 3.444035e-002
8000   8008   4      7.133      47.8702  6.260895e-011 3.444035e-002
8000   8008   4      7.152      47.7434  6.260895e-011 3.444035e-002
8000   8008   4      7.143      47.8034  6.260895e-011 3.444035e-002
8000   8008   4      7.146      47.7829  6.260895e-011 3.444035e-002
8000   8008   4      7.143      47.8005  6.260895e-011 3.444035e-002
8000   8008   4      7.141      47.8145  6.260895e-011 3.444035e-002
8000   8008   4      7.141      47.8156  6.260895e-011 3.444035e-002
8000   8008   4      7.144      47.7980  6.260895e-011 3.444035e-002
8000   8008   4      7.065      48.3316  6.260895e-011 3.444035e-002
16500  16508  4      59.525     50.3199  2.468622e-010 3.214697e-002
16500  16508  4      59.424     50.4055  2.468622e-010 3.214697e-002
16500  16508  4      59.618     50.2416  2.468622e-010 3.214697e-002
16500  16508  4      59.295     50.5150  2.468622e-010 3.214697e-002
16500  16508  4      59.906     50.0003  2.468622e-010 3.214697e-002
16500  16508  4      59.725     50.1515  2.468622e-010 3.214697e-002
16500  16508  4      59.552     50.2967  2.468622e-010 3.214697e-002
16500  16508  4      59.322     50.4925  2.468622e-010 3.214697e-002

Performance Summary (GFlops)

Size   LDA    Align.  Average  Maximal
8000   8008   4       47.8677  48.3316
16500  16508  4       50.3029  50.5150

That above is the result of near perfect values for gtls, vrefs, vtt, dll skews.

[zsamz_]

i'm back on my x38
gonna do crossfire soon
i tested most of the boises yesterday
i strongly suggest the 602 beta bios
to be quite honest all the newer ones are crap compared to this one
all the others i could only do 470x9
i'm @ 510x9 now
unless you have a 4870x2 get newest

[exemay]

Hi, i got my x48 LT T2R board a few days ago, had alot of issues getting it to work, mostly due to the fact that i needed to update the bios as im running a 4870x2....and i had no spare graphics card to use in order to update it but its all good and working now!!

i just got one little query about it though the most ive been able to take my Q6600 to is 3.21 ghtz....(356 x9) im on a PA 120.3, DCC 18W and swiftech GTZ, ive prime95'd for 6 hours today and not exceeded 69.....my max core temps are: 69 67 62 63

my question is to why i can not take it any higher, when i go past my current multiplyer it just wont boot....i put the voltage up to 1.43 but it just wont boot......any reasons why???

EDIT: im aiming for something near FSB 400 so i can get my Q6600 to about 3.4!
any suggestions welcome!

many thanks

exe

[zsamz_]

exemay: set voltage to auto and use special add 110%+

[exemay]

exemay: set voltage to auto and use special add 110%+

i set it to auto and the CPU VID special add to 110.15%

it turned on jut fine, but when i went to stress test it just crashed....is there anything else to do?

[zsamz_]

i set it to auto and the CPU VID special add to 110.15%

it turned on jut fine, but when i went to stress test it just crashed....is there anything else to do?

just go higher until you get the voltage you want
enable vdroop
set pcie to 100
dissable everything in cpu section but not the cores
and set the right gtl settings if you want to use less volts
chek your memory

[Mussels]

here to ask the masters for some more advice

Just got myself a xeon E3120.
I can run 475 FSB stable (thats not the issue), however cranking the multi up my stability goes away. I would like some advice for voltages, GTL ref, and so on that i should use to help get some higher OC's out of this beast of a CPU.

[zsamz_]

here to ask the masters for some more advice

Just got myself a xeon E3120.
I can run 475 FSB stable (thats not the issue), however cranking the multi up my stability goes away. I would like some advice for voltages, GTL ref, and so on that i should use to help get some higher OC's out of this beast of a CPU.

the need to find the right gtl settings to clock it higher
takes time you gotta test
dont be scared to go 1.5v on nb
i never use below 1,6v

[Mussels]

the need to find the right gtl settings to clock it higher
takes time you gotta test
dont be scared to go 1.5v on nb
i never use below 1,6v

alright, NB voltage will be going up.

Any advice on the GTL? I've never had a board before with GTL settings, and all of the guides google gets me only talk about quads.

[zsamz_]

alright, NB voltage will be going up.

Any advice on the GTL? I've never had a board before with GTL settings, and all of the guides google gets me only talk about quads.

clunk has a nice guide for gtl settings i suggest u read it
http://www.clunk.org.uk/forums/revie...ng-review.html
read anantechs review too
http://www.anandtech.com/mb/showdoc.aspx?i=3294&p=14

[Mussels]

clunk has a nice guide for gtl settings i suggest u read it
http://www.clunk.org.uk/forums/revie...ng-review.html
read anantechs review too
http://www.anandtech.com/mb/showdoc.aspx?i=3294&p=14

thanks man, i'll be busy for a few days reading and tweaking now.

edit: yep, putting all the voltages back to stock except CPU and NB, and now she's stable at 4.2GHz

[zsamz_]

thanks man, i'll be busy for a few days reading and tweaking now.

edit: yep, putting all the voltages back to stock except CPU and NB, and now she's stable at 4.2GHz

uw
man thats a nice cpu

[zsamz_]

i think ijust discovered somethin lol
after putting my zigmantech 1283 on the nb my board would only do 8.8.
played with it for a few hours n thought there was no hope dead board
then i lossened the screws a bit and the damn thing booted

[repsol_23]

Can someone explain how I go about making adjustments to MCH ROn/Rtt. I only ask because I have noticed that my fsb/cpu clock has been all over the place. It has been like this awhile, but I never knew how to fix until seeing this mentioned in an earlier post.

"MCH ROn (Resistance On) / Rtt (Termination Resistance) only need adjusting if the clock gen can't keep up or is causing erratic clock jitter/skewing."

My mobo has been fairly stable with a FSB=500, this puts my x6800 @ 4 ghz and my ram a 1000 MHz but I would like to eliminate these odd changes to my FSB. Thanks in advance.

[mikeyakame]

Can someone explain how I go about making adjustments to MCH ROn/Rtt. I only ask because I have noticed that my fsb/cpu clock has been all over the place. It has been like this awhile, but I never knew how to fix until seeing this mentioned in an earlier post.

"MCH ROn (Resistance On) / Rtt (Termination Resistance) only need adjusting if the clock gen can't keep up or is causing erratic clock jitter/skewing."

My mobo has been fairly stable with a FSB=500, this puts my x6800 @ 4 ghz and my ram a 1000 MHz but I would like to eliminate these odd changes to my FSB. Thanks in advance.

The method I use is changing them little by little and noting down what results i get.

Add 01h to each one at a time, it's the best way.

ie. Rtt = 01h, Ron = 00h -> Rtt = 00h, Ron = 01h -> Rtt = 01h, Ron = 01h -> Rtt = 02h, Ron = 01h -> Rtt = 01h, Ron = 02h -> Rtt = 02h, Ron = 02h...

And so on. It's something that will take many many bios changes to figure out

The only other helpful info I can give you is this.

MCH Ron (Resistance ON) determines how long the pins of the Receiver are held at Rtt or at the Pull Down Driver. Basically if you imagine your clock waveform, it determines the period where the clock wave is at constant voltage.

...______.............._
../.........\............/
_/...........\______/

That is a poorly drawn clock waveform, the straight section is set by Ron.

MCH Rtt (Resistance Termination) determines how fast the rising clock edge slew rate will be, and also how much ringback there is on both the rising clock edge and falling clock edge.

Higher Rtt can push more ringback on the falling clock edge while also using higher Ron can delay the falling clock edge so gives the ringback time to stabilize.

Thats all I can tell you really without an oscilloscope!

[Gameman]

Hi folks... I'm back...

I realize this is sort of off topic but does anyone know if the LanParty series x38 and 48 boards will ever support these cpu's

E5300
E5400



Can't find it anywhere on the site and I don't wanna risk blowing 100.00++ on one and find out I'm up **** creek.........

Thanks

Specifically the E5400 for the LanParty X48 T2RSB Plus, cause this is way cheaper than the an E8400 or E8500....

Thanks

[chuck4456]

more open areas for noise to escape out of. more holes in the case, the more noise... and the 300 has an extra side fan intake, the top fan hole, and the entire front of the case as an open mesh... its far, far inferior in terms of noise.

On the contrary.
I have two 300s that (fully fannned) are much quieter than my P182. The P182 has a degree of resonance once you close it up, even with the soundproofing.
At least that's been my experience.

[Gameman]

Don't worry, obviously those cpu's are too new.... I gotta new question though...

Can I get an E5200 to 4.0 GHZ on air and not have it blow up or do I need a huge air cooler like the Thermalright Ultra Extreme????

I read that it takes roughly 1.46-1.52 Vcore to get there stable so I was just wondering because when I tried 1.53 volts for my E2160 to be stable at 3.4 GHZ the machine heated up hotter than the gates of Hell...

It climbed to 85 degrees and then buggered off...

Also just since I'm here - in your opinion is 30 degrees Celcius at 1.39 volts GOOd for a Northbridge????? (currently @ 3.2 GHZ Stable)


Thanks for your valuable insight...

You guys are the best.....

[Shiranui Gen-An]

I downgraded my BIOS to 602 Beta as it was highly recommended, and fiddled with settings until I was able to get 500FSB 600MHz PL8 with Clock Crossing settings at Aggressive and this is all I got for bandwidth


By sabishii_hito

[zsamz_]

I downgraded my BIOS to 602 Beta as it was highly recommended, and fiddled with settings until I was able to get 500FSB 600MHz PL8 with Clock Crossing settings at Aggressive and this is all I got for bandwidth

duh
btw you can do pl7 with 1.63v on nb
i been sayin all along its the best bios
i tested every other bios
dfi should scrap every bios n start with 602 beta
i just wished they support the x48 as much as the p35

i'd move along to ut p45 but would cost 300can$ with shipping n taxes
motherboardpro has it for 160us$ shipped but ups or fedex not an option for us canadians unless we wanna pay a 50$++ customs fee

[Shiranui Gen-An]

I took out the CSX Diablo D9GKX and put in my newest Ballistix D9GMH 16FD3 kit and tweaked the settings, PL8 with pullins@EAEEE/EAAEE and Clock Crossing set to More Aggressive. Bandwidth is looking better but still not as good as others I've seen.

[Shiranui Gen-An]

duh
btw you can do pl7 with 1.63v on nb
i been sayin all along its the best bios

I can't get PL7 to boot unfortunately.

[zanzabar]

I can't get PL7 to boot unfortunately.

try the phases with e-a-e-e-e/e-a-a-e-e on pl8

thats the jams with that board it performs close to pl7