GTX Titan Quad-SLI & Rampage IV Extreme Break New 3DMark 11 Records

[Mankz]

All three 3DMark 11 present records have been smashed by Andre Yang wielding four Titan’s and a Rampage IV Extreme!
The four GeForce GTX Titan’s were ran between 1.4-1.45GHz with memory running at 7.1GHz (effective), while the Core i7-3970X ran at just over 5.7GHz. Naturally everything was super-chilled using LN2.

http://rog.asus.com/220222013/overclocking/gtx-titan-quad-sli-rampage-iv-extreme-break-new-3dmark-11...

[Raja]

Lol, now that's a cool story. There's always a window of opportunity on a new set of cards to be high up in the rankings with conventional cooling. Then things change...

[tabuburn]

Weird how Nvidia would not allow the vendors to modify the reference PCB. When you look at the area of the PCB around the power connectors, you can clearly see that there is a placeholder for another 8-pin PCI-e power connector there. If they had added that 8-pin, I could almost say that it would equal the GTX690 in performance or at least come really close.

[Raja]

tabuburn wrote:
Weird how Nvidia would not allow the vendors to modify the reference PCB. When you look at the area of the PCB around the power connectors, you can clearly see that there is a placeholder for another 8-pin PCI-e power connector there. If they had added that 8-pin, I could almost say that it would equal the GTX690 in performance or at least come really close.

The power phases are the limiting factor more than an extra 8 pin. Someone already blew a power phase with a volt mod to one of these ref cards.


If over-current at the input were the problem, all that would happen is that the connectors would get hot and we'd see some thermal breakdown at the junctions. The power phases blow way before that happens on these cards and that's not because they are "current starved" 🙂

The extra power connectors are added to meet safe current handling at the power connectors and interconnects. The connector itself is not an active device that limits current on its own accord. Any passive electrical connection will continue to pass current until such time there is a thermal failure. So adding an extra 8 pin, would not help the card OC higher, unless there is sufficient current handling at the VRM and assuming the core is not getting into the realms of clock limitations from other issues.


-Raja

[tabuburn]

Raja@ASUS wrote:
The power phases are the limiting factor more than an extra 8 pin. Someone already blew a power phase with a volt mod to one of these ref cards.


If over-current at the input were the problem, all that would happen is that the connectors would get hot and we'd see some thermal breakdown at the junctions. The power phases blow way before that happens on these cards and that's not because they are "current starved" 🙂

The extra power connectors are added to meet safe current handling at the power connectors and interconnects. The connector itself is not an active device that limits current on its own accord. Any passive electrical connection will continue to pass current until such time there is a thermal failure. So adding an extra 8 pin, would not help the card OC higher, unless there is sufficient current handling at the VRM and assuming the core is not getting into the realms of clock limitations from other issues.


-Raja

Forgot about the power phases. Looking at the reference PCB though, it looks like there's still a little room for a couple more power phases.

[Menthol]

Raja,
Destroying our unrealistic fantasies with engineering reality.

[chrsplmr]

My wife would Kill me. (slowly) ... Glad you guyz can play in this areana .. wow. Amazing. thnx.c.

[fr3d]

I thought world record was 6ghz?

[Mankz]

fr3d wrote:
I thought world record was 6ghz?

8.79Ghz 🙂

[Raja]

The WR in question is related to the benchmark scores with the GPUs. 🙂

[Retired]

About it:

t's quite clear now: the reference PCB of the Nvidia GeForce GTX Titan is just not strong enough to push the Titan core to the maximum of its capabilities. It's a shame and some would even say a disgrace that Nvidia dares to release such a powerful GPU soldered on something so weak. In addition, disallowing any manufacturer to build their own custom version of the card makes it damn near impossible to get the most out of this USD $1000 graphics card. Yes, that's right: no Lightning, no Matrix, no Classified.

We've already seen the engineering wonders of the K|ngp|n/TiN team as well as AndreYang/Shamino. Today, our attention goes to the team from Brazil, formed by Rbuass and Schenckel Bros. With their own version of the GTX Titan zombie, they managed to take home the single GPU 3DMark Vantage record, outscoring K|ngp|n by 360 points. For that, the team from Brazil uses a Core i7 3960X clocked at 5.6GHz and a GTX Titan at 1750/1820 MHz. Nice work!

I am rbuass, from Brazil, and admire and respect so much ASUS overclockers, and also is my prefered motherboards.
I used an EPower from EVGA... because I have no information to make a Zombie with the Matrix or even DCU II (also Lightnings)...
I need information to plug the 3.3 volts and how to plug the card.
I know there are lots of risk, and I don't go claim for RMA for sure... is for my own risk.
Please, share information about how to use a Matrix or DCU 2 Zombie...
The problem is that we are working only 1 card... because we have only 1 Epower... but we would like to compete also SLI.
Sure... Shamino is one main inspiration to me...
Let me know if is possible.... also, you can see what we are doing at hwbot (rbuass).
I have no support... and we are doing alone.... so... I will appreciate if you can help me with that informations.

Best wishes and thanks in advance