**Official DFI LanParty UT P35-T2R Review/Overclock/Guide Thread**

Quote:

---

Originally Posted by Seanie's Show

I know exactly how you feel, my G0 stepping quad posts and boots to windows vista at 3.8-3.9ghz, but I am struggling to get it stable anything over 400mhz FSB on this board (3.6ghz), upto 400mhz I have the following settings and will blend prime all day and all week with my crucial ballistix 667mhz at 1000mhz:

CPU = 1.45v, RAM = 2.23v, FSB VTT = 1.38v, NB = 1.45v, SB = 1.15v, GTL's all currently Disabled, Clock Gen = 3.45v, mem is running at 4-4-4-12 @ 1000mhz.

However even at 401 FSB everything becomes unstable, I just dont get it, i've tried bumping up voltages etc and it doesnt seem to make any difference, im not willing to run the NB at 1.6v 24/7 as it already gets hot enough (47oC with a fan on it whilst blending).

I was testing memory earlier in case that was the problem, and the memory was stable no matter what dividers I used all the way upto 1000mhz 4-4-4-12, some of this testing was done with the board and the CPU at stocks, others required the FSB moving up and the multiplier on the CPU dropping to keep the CPU at stocks, but once again, the minuite I pushed the FSB over 400 it was failing memtest again, I think the setting I had on the ram at the time was 333/800. at 400FSB it was at 960mhz so I pushed the FSB to 416 to put the ram at 999mhz (1000mhz) and like I said it was failing memtest.

Below is a kind of review I just typed up for myself to keep myself in check, so I knew what I had tried and what I had'nt, feel free to download it and have a look, then hopefully someone might be able to answer both of our issues.

---

The primary mistake you are making is leaving the GTL ref's on disabled, even with VTT point of 1.38 and suitably applied GTL's you would see more success. The GTL's need to operate within a margin of VTT, typically around 1-2% within 67% of VTT, anything below or above this margin will likely fail to hold steady (unless you have a real freak processor)


A side effect of Cas 4 is that VNB and Vdimm will need to be higher than Cas 5 clocks. Check Kris's article on CAS related overclocking margins over at the Tech Repository.

Cas 4 on DDR 667 ram, which is not binned as Micron D9GKX, will need quitwe a voltage bump in some cases to scale, bear this in mind...

The 266 strap and divider combo's will operate Performance level 6 on this board upto around 463 fsb (266/667 divider). The 333 strap will give a margin of around 10-15 fsb over the 266 strap for absolute stability (333 strap loses latency and fsb overhead is a little low to claw back the deficit), however I feel that the 266 strap is the sweet performance ratio for Quad core clocking on this board (if not benching for extreme cpu mhz).

1.5vnb will get you 400fsb ddr 1200 Cas 5 on this board. Cas 4 gives the NB a harder time, so one has no choice but to bump VNB to provide suffcient overhead for the lower latency. 450fsb will require 1.61-1.65 VNB @ Cas 5 , with suitably low VDimm on good ram. 400 Fsb should come with VTT of around 1.50v (using the appropriate GTL ranges) and upwards, or less if you are prepared to put in the time to study how much VTT/GTL your setup can get away with.

It's just a matter of you picking a range of performance and voltages within the boards abilities at a point you are comfortable with (there will have to be compromises in both directions.

regards
Raja

Thanks for that, the image below shows how its going at the moment, the mem is 667 CAS3 stock, micron based, not completely sure what chips are under the spreaders as ive never had them off, but from 800mhz to 1000mhz they are completely capable of CAS4, anything over 1000mhz requires it to be reduced to CAS5.

Thanks for the GTL info, I know about the 67% rule based on VTT however based on my current VTT of 1.38 that would mean I only need 95 across all 3 GTL Refs, however one thing I have noticed of recent is most people end up putting alot more GTL to the NB than the 2 cores GTL adjustments ie 115-115-155, is this NB GTL based on a different % to the CPU GTL. sorry about this question but this is the first board I have owned where GTL ref has played such a big part, my last DFI was the Expert 939, everything beyond that was ASUS.

Upto 450fsb matching VNB and VTT levels is quite effective. Hence the 1.61 VNB setting with a VTT of 1.6 (1.58 real), falls in at a nice 115/115/115 or so, close enough.

You may find that matching up with lower VTT/VNB/GTL's work, however, you will approach ratios of FSB where under 1.5VNB just won't hold the bandwidth.

I cannot in any way beat the Prime stable @ 3.6 with any other combination of settings for equivalent CPU mhz (prime stable that is). The only way to do so would be to run cas 4 with a very good pair of sticks that can run near 1100 or so with less than 2.27V (within warranty specs), at 450 fsb Cas 4 (on the same dividers), the VNB is likely to need another boost.

Hence why I thought 8x450fsb 2.1vdimm @ Cas 5 was so effective and gives superior performance to ddr 1200 cas 5 @ 400fsb on the 9x multi.. 1.61 on the NB will not put as much strain as prolonged use of DDR2 over 2.25v, 2.1vdimm really is sweet for the performance the 450fsb point gives on the 266 divider.

Once my Cas 4 ram gets back from RMA I will be in a position to watercool the CPU for in excess of 3.8ghz stable, then I shall be looking for around 430fsb @ Cas 4, to see what kind of voltages are required to hold the board/cpu/mem down...

regards
Raja

Good Morning guys:) While we are on the subject "Wow" Monster O.C Potential with this board! Reminds me of the Expert board but once you get over that learning curve (watch out) its one Bad Mutha F@cka.....and I don't mean SHAFT either....:D

Currently I am at 3320mhz X8 Multi @ 415 FSB Ram: 333/667 ~ 1:1 @ 1.9v 1.325 vcore , CPU VTT 1.36v ,NBv 1.45v , GTL V (Enabled) 110 - 110 - 110


How does all of this look to the onlookers (you know who you are) does there appear to be good balance in my Voltages any thing to raise / lower I know its all done according to individual basis but please scrutinize :D

Has anyone else experienced issues with their 3.3v line?

When running stability Testing with Prime around 3600 Mhz the 3.3v line will drop off to 0v. I did not notice this at first because I was using Everest to record readings. After adding Smart Guardian this became apparent with bios and Everest readings showing the same.

The 2 - PSUs I have seen this happen on were the primary Corsair 620W and then I popped in an old Power PC and Cooling 510.

Since this has happenned on 2 Power Supplies was wondering if this is a bios issue (no big deal), will try on another PSU to confirm (no big deal) or a MB issue which would be a big deal to make sure I get an RMA from newegg for a replacement.

Empty

Nope.... holding my own over here:D

Quote:

---

Originally Posted by empty1969a

Has anyone else experienced issues with their 3.3v line?

When running stability Testing with Prime around 3600 Mhz the 3.3v line will drop off to 0v. I did not notice this at first because I was using Everest to record readings. After adding Smart Guardian this became apparent with bios and Everest readings showing the same.

The 2 - PSUs I have seen this happen on were the primary Corsair 620W and then I popped in an old Power PC and Cooling 510.

Since this has happenned on 2 Power Supplies was wondering if this is a bios issue (no big deal), will try on another PSU to confirm (no big deal) or a MB issue which would be a big deal to make sure I get an RMA from newegg for a replacement.

Empty

---

SGuardian has issues reading correct temps and voltages, the rails are likely fine. I tested Vdimm droop under load and could not find the variance that was shown in SG.

regards
Raja

Quote:

---

Originally Posted by raju

SGuardian has issues reading correct temps and voltages, the rails are likely fine. I tested Vdimm droop under load and could not find the variance that was shown in SG.

regards
Raja

---

Bios also shows 0v on the 3.3v line.

@raju........Good morning:) hey hows about a look see at my post above and give me you're opinion please? Also whats going to fair better in this board my G-Skill HZ's or my Crucial Ballistix Tracer (PC2 8500) ?

Quote:

---

Originally Posted by Brother Esau

@raju........Good morning:) hey hows about a look see at my post above and give me you're opinion please? Also whats going to fair better in this board my G-Skill HZ's or my Crucial Ballistix Tracer (PC2 8500) ?

---

Hi man,

AFAIK both the HZ's and Crucial use Similar Micron D9, with possibly the Crucial 8500's being a better bin. A cross comparison on your end should confirm which of the 2 sets scale better..

For the most part I would encourage you to use the 266 strap dividers. Either toy with cas 4 clocking under ddr 1100 (depending on how good a vdimm you can get away with). Or use Cas 5 for over ddr 1100 clocking. This will immediately give you a bandwidth and latency boost (use enhanced addressing and enhanced data transmitting to fast).

Of course this will require increasing your currently low Vdimm/VNB/GTL settings...

One thing that may be worthy of merit to you is scaling down the 3.6ghz clock I posted at the club to a lower divider if you don't want to take the hit in voltages. THe OCZ Flex is micron D9 based too, so should allow some room for similarities...

regards
Raja

Sorry for the offtopic right here.... but...

Can anyone of you please have a look at this?

http://www.abload.de/thumb/dfi_ut_p35_t2r_247ni.jpg

Can anyone poste please whats written down on the IC marked in red? Can't identify it from the picture. The text written on the IC is needed for a Vcore mod. And, YES, i need more vcore. No discussion about that. :D

There's a related thread over at the mods section if anyone cares.

tia ;).

One other thing for a 400fsb performance point is to run the 266 strap with the memory at 1200 cas 5 @ performance level 5, this will almost give a repeat performance overclock (not quite as low in access latency, perhaps lower mb/s in copy and write speeds) of the 450fsb clock posted at the club, assuming you can get it stable. VNB again will need 1.61 or so for this level of thru-put. @ 3.2ghz this gives around 50.2ns access latency which is ok at the fsb, bandwidth is around 9200mbs, copy/write speeds both coming in around 7800mbs. Still hard to match the 450fsb point, whichever way I turn, it comes back to that for quads.

regards
Raja

Quote:

---

Originally Posted by empty1969a

Has anyone else experienced issues with their 3.3v line?

When running stability Testing with Prime around 3600 Mhz the 3.3v line will drop off to 0v. I did not notice this at first because I was using Everest to record readings. After adding Smart Guardian this became apparent with bios and Everest readings showing the same.

The 2 - PSUs I have seen this happen on were the primary Corsair 620W and then I popped in an old Power PC and Cooling 510.

Since this has happenned on 2 Power Supplies was wondering if this is a bios issue (no big deal), will try on another PSU to confirm (no big deal) or a MB issue which would be a big deal to make sure I get an RMA from newegg for a replacement.

Empty

---

Everest is the core of the problems-i had it.Disorganize most of the sensors-after using of Everest even CPUZ shows strange readings

I just have to say that " THIS BOARD ROCKS" :D couldnt be happier, actually Ill be happier if I dont have to use the cpu header for my board to work hehe... This is my first LanParty and it wont be the last.. Thanks DFI and Oskar Wu... worth every penny spent....

http://ocidb.com/albums/userpics/10062/DSC00750.JPG

http://ocidb.com/albums/userpics/100...iu9u097097.PNG

Quote:

---

Originally Posted by luihed

I just have to say that " THIS BOARD ROCKS" :D couldnt be happier, actually Ill be happier if I dont have to use the cpu header for my board to work hehe... This is my first LanParty and it wont be the last.. Thanks DFI and Oskar Wu... worth every penny spent....

http://ocidb.com/albums/userpics/10062/DSC00750.JPG

http://ocidb.com/albums/userpics/100...iu9u097097.PNG

---

You aint shettin Brother...heheh ....D.F.I for the Win once again!:bows: :woot:

Spent a few hours yesterday, trying to get memtest to run. tried both 170 and 165.

The best I managed with my Gskill 6400 GBHZ's was 600 mhz. 1:1. It actually required a boost in clockgen of all things to get it running.

At stock 800 mhz, forget about it. Manually set subtimings. Everything loosened up and relaxed. Vdimm up to and around 2.2v. One stick of ram. No joy at all.

I know these sticks run fine on other bits of kit, so something isn't right.

About to return this board, but anyone else here having issues with running memtest?

RLM

I have the same Ram RLM sounds like another Duff board to me!:(

Quote:

---

I have the same Ram RLM sounds like another Duff board to me!

---

You're able to run memtest okay then?

By the way have responded to your post on i4memory regards GBHZ subtimings .:)

Cheers

RLM

Quote:

---

Originally Posted by raju

Upto 450fsb matching VNB and VTT levels is quite effective. Hence the 1.61 VNB setting with a VTT of 1.6 (1.58 real), falls in at a nice 115/115/115 or so, close enough.
Raja

---

Ah I had set gtl's at 115/115/155, but that may have been a typo. I thought that it was unusually high given the approx x 0.67 rule. 155 NB GTL worked though :shrug:

I have moved on to trying 450x8 and it seems to be going well so far.

Thumbs up to DFI, 450fsb prime stable with a quad core on air is awesome.. :D When i get my WC gear, it will be time for me to get off my ass and setup that hwbot account for our Aussie team i4memory.com hehe

Quote:

---

Thumbs up to DFI, 450fsb prime stable with a quad core on air is awesome.. When i get my WC gear, it will be time for me to get off my ass and setup that hwbot account for our Aussie team i4memory.com hehe

---

:up: :D

RLM

Guys is running High CPU VTT settings supposed to reduce the amount of Vcore needed and does it do harm running VTT up high? My CPU VTT is @ 1.40v NBv @ 1.50v GTL V @ enabled / 110/110/110 "Do these GTL volts seem right looking at my voltages?"
3410MHZ @1.37v 427 FSB

Some GSkill PC2-6400 HZ settings here to try out

Quote:

---

Originally Posted by Brother Esau

Also whats going to fair better in this board my G-Skill HZ's or my Crucial Ballistix Tracer (PC2 8500) ?

---

Quote:

---

Originally Posted by Brother Esau

does it do harm running VTT up high?

---

I too would like to know this. I hope some of these dead boards we are seeing isn't because of prolonged high CPU VTT. I believe Raja and Praz have indicated that it's ok. but still, I think the stock value is 1.2v, so running a 33% overvolt to 1.6v is a big jump...

As briefly discussed a few pages back. Running too low an NB and VTT with the quads can actually cause more problems than solve them.

I actually managed to corrupt my OS running too low a VTT/NB on this board.

A bit different to my Abit Quad GT, in that running a lower VTT around 1.3v with the QX 6700 gave me better stability.

RLM

Yea, I remember reading that but...... I would like to know what D.F.I's response to my question is or at least some educated explained answers from the dudes with the hook up from D.F.I that have tons of time in Analyzing this board:cool: Praz , RGone, eva, FCG , Tony, Andre, etc ,etc sh*t even Oskar himself might come join this little party that revolves around his baby....Hint....Hint:yepp: