EVGA 750i FTW Overclock Thread

[Eastrider]

***! Are you trying to make the Northbridge support 8GB of RAM on 4 sticks + a so high overclock?...

Well, In your place I would increase the NB voltage to 1,55 and from there, go dropping... because well, it's so much job for the NB.

Pd: 2xGTX280 + Q6600 at 3,8? Isn't it a bottleneck for the graphs? =S My dual 88GT's are bottlenecked at 3,20GHz, I need ,40 or ,60...

[Slightly skewed]

Well bottlenecks are a funny sort. Even Iggsters rig could be considered by some to be bottlenecked.

I don't like venturing into the red zone too much for voltages guys. A little, ok. Cooling isn't an issue really, unless you run small FFT's in Prime. Otherwise they are sub 65C at full load. That's at 3.8.

Nice bench at 3.85. I've tried to set the RAM settings, clock/voltage and timmings very loose, but still can't get 3.9 stable enough to bench vantage.

[amircesc4]

amircesc4

Frequencies/Timings

CPU: E6750 @ 8 x 451 FSB = 3.61 GHz
RAM: 2x2GB Dominator @ 1066MHz 5-5-5-15 2T

Voltages

CPU Core: 1.42500v (1.3125 VID)
CPU FSB: 1.4v
Mem: 2.1v
SPP: 1.5v

[Eastrider]

Errr... you've got an overclocked Q6600 and don't have the 3D Mark Vantage Physics glitch? Uh, I might be one of the unafortunated guys who make the glitch happens (I can't run 3D Mark because physics became crazy and flood the whole scene of Jude Nail lol, same with GPU Cloth. Curious in that, because is GPU simulation, but well...)

--

I found something sooo useful about GTLVREF Lanes:

To get to correct value, multiply your VTT (AKA FSB Voltage) for a 67% in 65nm CPU, 63% in 45nm. In my case, I did (1,4v x 67% = 0,91sooomanydecimalshere), and that was the result. 0,91, rounding. 0,91 is, obviously, more than 0,90, so we have to round up. NEVER go even a bit down of the result. At least never to go for a starting point.
So, 0,91 means 91mV, 95mV taking the rule I said up there. I putted 95mV in the four lanes, and actually, my PC was stable a sooo long time (6 minutes, record at the date) at 3,60GHz, 1,525v at BIOS, 1,4 in FSB and 1,4 in SPP. The low SPP voltage is the cause of my crashing I guess. I raised it up to 1,45 and I will try now, if not, It'll work with 1,55.

I don't know if it's already posted, but well... will be useful for people who do not want to read 97 pages of topic - and they're not just a few ones

Good luck!

---

Edit: meek, and for the Dual Cores (I noticed you amircesc4 ) is the same, but just with two lanes of GTLVREF, and following the same rule of 45/65nm

[amircesc4]

Originally Posted by Eastrider

Edit: meek, and for the Dual Cores (I noticed you amircesc4 ) is the same, but just with two lanes of GTLVREF, and following the same rule of 45/65nm

2 lanes? which lanes? lane 0 and lane 1?

[Eastrider]

Originally Posted by amircesc4

2 lanes? which lanes? lane 0 and lane 1?

No, lane 0 and lane 2. Or at least, I think that it's that way... I have a Quad and never had a Dual, but I think that if you have a dual, the other lanes don't appear.

Well, that's it, lane 0 (die mvoltage) and lane 2 (address mvoltage).

In real, in some motherboard models it's lane 0 and lane 1... but I think that in EVGA's is L0 and L2... if that doesn't work, just try L0 and L1.

[amircesc4]

i tried everything but it doesnt pass fsb=450....

[Eastrider]

Originally Posted by amircesc4

i tried everything but it doesnt pass fsb=450....

I'm starting to get out of ideas too. I had to decrease to 3,50Ghz, to see if was a "FSB Hole" on the 450FSB. I'm on 437 right now. 85-85-85-85mv, and prime still crashing, even a sooo high voltage. What's the problem!? :S

[Microsis]

Originally Posted by amircesc4

amircesc4

Frequencies/Timings

CPU: E6750 @ 8 x 451 FSB = 3.61 GHz
RAM: 2x2GB Dominator @ 1066MHz 5-5-5-15 2T

Voltages

CPU Core: 1.42500v (1.3125 VID)
CPU FSB: 1.4v
Mem: 2.1v
SPP: 1.5v

Got you added. Stable?

[PizzaMan]

Originally Posted by Eastrider

I'm starting to get out of ideas too. I had to decrease to 3,50Ghz, to see if was a "FSB Hole" on the 450FSB. I'm on 437 right now. 85-85-85-85mv, and prime still crashing, even a sooo high voltage. What's the problem!? :S

First of all, they are called "straps" not FSB holes. Just like your ram timings have to be either loosened or more volts added as you increase Mhz, your NB has timings that change as the Mhz increase. Try jumping 100Mhz on your FSB and try again.

Let me give you an example as how this works. I had a 680i board that had a 1066 strap that end at 1300. At 1301 the 1333 strap started. When I got to around 1220-1250 I really had to start increasing the NB voltage. At 1270 I had to start giving SB voltage. I could take it all the way to 1300, but I had to max out both the NB and SB voltage. When I reached 1301 I could bring both the NB and SB back to default voltage because the timings changed to looser values. With this being said you my wonder about performance. When just like ram, your board performs a little better with tight timings. I got a 80MB/s mem bus increase at 1300 vs 1301.

Sorry for the rant, but I use to believe in FSB holes too. Though, of course there may be boards out there that with max NB and SB voltages they can't even reach the top of a strap and have a hole where it is impossible to post. Haven't played with any yet.