DFI LanParty P35-T2R shows great potential for quad core clocking

[freecableguy]

This post is short and sweet, with a single screenshot showing the amazing potential that the DFI LP P35-T2R has regarding FSB overclocking for Intel quad core CPUs. This is not the top, just a small intermission while I probe some more motherboard voltages and make some more tweaks. Important to note, like every board we test here at The Tech Repository, this board is not a special pick from the retailer or reseller, there are absolutely no modifications to the board, and no voltages are applied that could be considered unsafe or potentially damaging to any components.

As you can see we had little to no problem reaching a staggering 475MHz FSB with an Intel Q6600 G0 ES quad core processor. The secret is in the right combination of VTT, CPU PLL, and GTL reference voltages for Cores 0/1, Cores 2/3 and the NB. More detail to come soon enough.

-FCG

[Praz]

Hopefully I'll see similar results later this week.

[Raja]

OK been strudying the GTL table scale for this board, looking at the use of the offset algorithm,

DFI appears to have provided 0.0015mv scale adjustments (per 1 digit of GTL) around the 67% mark of VTT.

For every 0.1v increase over the stock VTT 1.2 the GTL scales are offset (increased) by around 0.004 ~ 0.0055mv.

No really plausible explanation for this other than to keep the scale under the 200 mark, which in itself is not an outstanding justification.. Well it's beyond me anyway..

Of use to the overclocker is the fine tuning around the 67% point of VTT, knowing a corresponding 5 gtl digits are equal to around 0.0075mv or so (0.0015mv per digit);

With a rise of 20 GTL digits equating to an increase of ~ 0.03mv...

Now I just need a board, lol...

Seriously though, GTL tuning requires perseverance to get right..

Seems DFI have this one bang on though, reaching 475fsb with a Quad this quickly is a good sign...


regards
Raja

[thorgal]

As I have told Kris already, I reached 480-485FSB with both my Quads (QX6700 B3 and Q6600 G0) on the Asus Blitz Formula, without any GTL-ref tuning. I'll provide proof of the quad prime QX6700 run 480x6 asap.

My question : any idea how Asus is able to pull this off ?

I have my DFI board in now, and see how well that one does with the same CPU's.

Edit : the curious thing was that for me 490 seems to be a wall that I cannot get past, both cpu's react the exact same way : 489 : can run about everything except prime, 490 : no go. Tried all kind of VTT/NB/PLL combinations, but no luck. Anyway, I'm not complaining about this kind of performance.

[eva2000]

I think it's just a matter of the motherboard being the right fit for certain cpus in Asus' case ?

my Q6600 GO L725A903 on Asus Blitz Formula does 493FSB easy now on corsair nautilus 500 water cooling.

[Raja]

Originally Posted by thorgal

As I have told Kris already, I reached 480-485FSB with both my Quads (QX6700 B3 and Q6600 G0) on the Asus Blitz Formula, without any GTL-ref tuning. I'll provide proof of the quad prime QX6700 run 480x6 asap.

My question : any idea how Asus is able to pull this off ?

I have my DFI board in now, and see how well that one does with the same CPU's.

Edit : the curious thing was that for me 490 seems to be a wall that I cannot get past, both cpu's react the exact same way : 489 : can run about everything except prime, 490 : no go. Tried all kind of VTT/NB/PLL combinations, but no luck. Anyway, I'm not complaining about this kind of performance.

Does the board overclock all 4 cores over 450fsb? I had heard that over a certain point of fsb 2 of the cores are actually clocked lower on the Asus boards..

regards
Raja

[thorgal]

Originally Posted by Raja

Does the board overclock all 4 cores over 450fsb? I had heard that over a certain point of fsb 2 of the cores are actually clocked lower on the Asus boards..

regards
Raja

Good question, how exactly can I make sure of this ?

[Ryderocz]

One thing to do would be to open Task manager > Performance tab and make sure you see 4 CPU Graph's when you are over 450 FSB.

We found the limit on the P5K was around 455 -462 after that...only 2 cores in Task manager, which was a bummer.

CPU-z and everything said we were running 500 * 6

[eva2000]

Originally Posted by thorgal

Good question, how exactly can I make sure of this ?

run wprime and cinebench R10 as well







some ES sample quad cores had that issue of 2 cores clonking out as well so make sure you try with retail quad core cpus

[thorgal]

Here's mine. QX6700 and just did Cinebench Q6600 to make sure.

So far the hijacking of FCG's DFI thread