** Official ** Gigabyte UD3 P45 Series -- EP45-UD3 / EP45-UD3R / EP45-UD3P

[spinejam]

Just curious, did you guys have to raise vNB or anything else going from 2x2gb to 4x2gb of DDR2?

[Gameman]

well there's definite proof now that these UD3 boards are indeed special....

I tried giving the ol X48 T2RSB Plus a try with the Q9550 and the best it could manage was 450 fsb for 3.6 ghz...considering what i've been benching on this cheap 100.00 UD3L I consider that a thorough and complete embarrassment...
Especially for the great X48 chipset... I gotta say, that was pretty sad...
wel, needless to say i'm gonna be unloading it again cause the only thing that even had a hope of reaching 4.0 ghz would be the Q9650 - the most expensive one....Although I am tempted to try the Blood Iron p45 T2RS Elite with a Quad , even though it seems to be non-existant at the moment in Canada...

I tell ya, when a sub 100.00 board can whip the ass of a 250.00 board...that's pretty sad....

oh well...it's safe to say my next board will be a UD3P instead of the T2RS Elite i was thinking of...

[truehighroller]

my board used to be able to do 1200+mem speeds together with 450+fsb on the 3:4 divider 400 strap. used to get read, write and copy speeds all in the high 9s as well as latency in the low 50s. one day it wouldnt boot, this was a few weeks ago. since then it refuses to boot 3:4 400 strap no matter what i do

Your performance level is the problem perhaps? I don't know what to set that strap to so..

[RAMMAN]

looks like i still have a screenshot from my best everest bench at the time



from what i remember that was a suicide screen, as 468fsb pl8 ddr2-1248 was a bit much for the mch at only 1.2v

edit: its alive! 3:4 400 strap 425fsb pl9 is booting

[spinejam]

windows 7 64bit

[RAMMAN]

my best 24/7 setting so far and most likely the last for my current setup.



again pl8 1.24 mch 0.8 mch ref 2.2 vdimm orthos stable

by orthos stable that means i ran orthos blend for 20 minutes then stopped it manually. it passed 1024k 16 minutes in and im certain the cpu is stable at 3.375ghz so i dont think there is any need to make sure 8k passes. it can play any game at this setting for hours on end so im happy.

i swear 0.8 mch ref is performing magic on this setup even at this low mch core voltage.

as you can see the 400 strap has cancelled out the write speed benefits from the added memory speed, but read speed has gone up 300 and copy has gone up 200 and even surpassed the write speed. latency has gone down 2 1/2 ns as well.

at least now my memory is actually doing something rather than running at a speed and latency even 2gb hynix sticks can do lol.

anyone else find it interesting that 3:4 400 strap has the same max fsb as 5:6 333 strap? 400 strap seems to do wonders for max fsb.

could prove useful to anyone trying for ddr2-1200 450fsb or even ddr2-1333 500 fsb if the memory and board is capable.

i have tried 3:4 200 strap in the past, latency is great but max fsb drops by 30. 266 strap both 4:5 and especially 2:3 doesnt like to scale, and in the case of 2:3 actually becomes more unstable at higher mch voltage.

[RAMMAN]

im happy with the performance im getting atm, but does anyone think its weird that i have 3 kits of memory, none of which can do its rated speed on this board?

2x1gb crucial ballistix pc2-6400(fatbodies) have done ddr2-933 4-3-3-10 2.4v on a p35 ds3p, wont work at anything past ddr2-780 even when clocking in windows on this board, just short of its rated ddr2-800

2x2gb ocz flex ex pc2-9600 (hynix) only boots with 5:6, 4:5 and 1:1 that and 450fsb limitation the rated ddr2-1200 is impossible.

2x1gb team xtreem pc2-10400 (micron d9) clocks high with 5:8 divider, but at its rated speed, timings and voltage (ddr2 1300 6-6-6-18 2.35-2.45v) its only suicide screen stable.

[Gameman]

2x1gb team xtreem pc2-10400 (micron d9) clocks high with 5:8 divider, but at its rated speed, timings and voltage (ddr2 1300 6-6-6-18 2.35-2.45v) its only suicide screen stable.

just curious, what do you consider to be suicide????
I managed 4.3 and am debating whether to pump 1.40vcore and 1.35 through the ram along with 1.48 vtt and 1.75 pll?????...
peoples opinions may differ on this one...
I have had several 45 nm chips and always gave em 1.54,1.45, 1.42 etc and never lost one...even a Quad once and no death to speak of...
or are we talking like dry ice or ln2 here????

thanks

[truehighroller]

My memory does its rated speeds on this board just fine?

[In-Fluence]

Likewise for me too, The dominator GT's I had reached 1120, my chaintech apogee's reach ~1170 and my reapers hit ~1296 - all with PL9 and 1.4vnb max. My board undervolts ram by about 0.2v, maybe the board undervolting could be an issue?

0.8 mch ref is good for me too, along with 0.785 cpu ref, but when going for higher vtt / fsb I had to lower them both down to 0.76.

I personally think that your UD3LR just might be struggling a little and might need more tuning. I assume you've tinkered with CPU/NB skews & CPU/PCI clock drives? These can make a crucial difference if they're off, and can vary from board to board.

[RAMMAN]

just curious, what do you consider to be suicide????
I managed 4.3 and am debating whether to pump 1.40vcore and 1.35 through the ram along with 1.48 vtt and 1.75 pll?????...
peoples opinions may differ on this one...
I have had several 45 nm chips and always gave em 1.54,1.45, 1.42 etc and never lost one...even a Quad once and no death to speak of...
or are we talking like dry ice or ln2 here????

thanks

imo suicide screen means having to set the clock in windows, it does not lockup, bluescreen or reboot but at the same time you cant really do anything with it other than run everest or get into cpu-z. super pi 1m might have been possible but i didnt try it at the time. orthos errors in a matter of seconds, not minutes.

imo there are 5 levels of stability:

suicide screen

super pi 1m

bench stable (being able to get through any version of 3dmark)

game stable

24/7 stable

edit: assuming everything is running cool i always thought 1.4 vcore is the "safe" limit for 24/7 usage and 1.5v+ is the realm of sub zero benching and i dont think raising pll that high is such a good idea, i think high pll kills chips faster than high vcore and high vtt combined.

edit2: i plan on dry ice in the future, if i want to achieve say 4.5ghz by just raising the multi is it really as simple as just raising the multi from 7.5 to 10 and raising vcore to 1.4? or is there other things i have to worry about? i guess vtt cant be to far away from vcore but other than that raising the multi while the chip is under subzero shouldnt make the mch or ram unstable?

[Gameman]

very good...thanks

may try th QX9650 next since it can be found for th Q9650 price these days....

I like the idea of unlocked multiplier....

[andressergio]

very good...thanks

may try th QX9650 next since it can be found for th Q9650 price these days....

I like the idea of unlocked multiplier....

i woudn't mate...QX9650 are severe damaged by VTT, Q9650 can handle lotta punishment...

[Gameman]

ok then thanks...newegg.ca just got some more in

[RAMMAN]

i woudn't mate...QX9650 are severe damaged by VTT, Q9650 can handle lotta punishment...

is it something to do with e0 chips surviving up to 1.6 vtt on p45 mobos while c0 chips degrade past 1.35 vtt? or am i mistaken?

edit: whenever my board coldboots with a strap other than 333 it rapidly starts up, shuts down then successfully boots every time, is that normal?

[RAMMAN]

I personally think that your UD3LR just might be struggling a little and might need more tuning. I assume you've tinkered with CPU/NB skews & CPU/PCI clock drives? These can make a crucial difference if they're off, and can vary from board to board.

i honestly dont know what im doing when it comes to cpu/nb skews and cpu/pci clock drives, so i leave them alone.

edit: bad news, got a lockup after a couple hours of tf2, it is a hot summer afternoon though and it was fine running 4 hours of tf2 this morning.

[In-Fluence]

I would just test to find the best ones for your setup. Lower your fsb a few notches until it's a a few minutes orthos stable, and leave the clock drives at 800,900, and try to tweak your skews - they DO help when increasing the fsb! First try 0,50, then 50,50, then 50,100 then 100,100, and keep moving up like that until you find your most stable point (going beyond 250,250 is likely to get worse again, so not many to test.. 2 or 3 settings might work well for you but try to pick out the best one.

Let us know how you get on

[Gameman]

is it something to do with e0 chips surviving up to 1.6 vtt on p45 mobos while c0 chips degrade past 1.35 vtt? or am i mistaken?

is this for real ??? cause that would yield a lot of leeway

I could probably get 4.7-5.0 ghz on the Q9650 - my likely next chip...

I just wish they made a REGULAR Q9770 with E0 stepping....that would be sweet starting at 3.2 ghz and 400 fsb - it would basically be a Quad-Core E8600... 4.5 ghz would be no problem...

[OLDcomer]

...
I just wish they made a REGULAR Q9770 with E0 stepping....that would be sweet starting at 3.2 ghz and 400 fsb - it would basically be a Quad-Core E8600... 4.5 ghz would be no problem...

X3380 is e0 but it's rather expensive.

[Gameman]

X3380 is e0 but it's rather expensive.

I see it's for "server-use"....thanks for letting me know...never knew about this one....
seems like a Quad Core E8500 @ 3.16...same frequency...

Wonder if the UD3 serious would support the X3380 since it is socket 775...

Actually that might be one to go to after the Q9650...for me....kind of like the next step up and then either the QX9770 or the i7 975 @ 3.33...

or that 3.5 one they are said to be coming out with....

interesting....

[andressergio]

is it something to do with e0 chips surviving up to 1.6 vtt on p45 mobos while c0 chips degrade past 1.35 vtt? or am i mistaken?

edit: whenever my board coldboots with a strap other than 333 it rapidly starts up, shuts down then successfully boots every time, is that normal?

it was known when they first released that some died on more than 1,4VTT or degraded...E0's are more robust chips...inmo

Cheers
Sergio

[Gameman]

it was known when they first released that some died on more than 1,4VTT or degraded...E0's are more robust chips...inmo

that's a load off my mind....thanks!!!!

edit: or at least i hope so....

[RAMMAN]

so the safe limit for c0 chips is 1.3vtt? back in the day the qx9650 held a few world records so it must clock well even without high vtt. btw i got another lockup a few hours into tf2 on a more conservative 450fsb 1:1 on yet another hot summers day. this definitaly has to do with temps as it will run orthos indefinitely but orthos doesnt get the northbridge as hot as games do. the nb hs is definitaly getting to the point where it starts to burn fingers and worse still the ram is getting quite hot as well.

edit: strangely enough i lower mch back to 1.2 and mch ref back to 0.76 and it played tf2 for 5 hours not stop no problem

[Mech0z]

As I have mentioned in this thread my EP45 UD4P rev1 died a week ago, but tomorrow I am recieving a EP45T Extreme Iam borrowing from a friend. How does this clock compared to the UD4P if anyone know? I will use watercooling ofc. But I hope to break my prev cpuz of 4196 with my Q6600.

[Hoss331]

Wonder if the UD3 serious would support the X3380 since it is socket 775..

They support the 3380 just fine.