Discussing the result of Team DFI

The 3DMark Vantage benchmark is one of the more expensive benchmarks around as it requires the overclocker to have both an expensive, well-clocking LGA1366 i7 processor which has 4 cores and 8 threads and also 2 GTX295 graphics cards to reach high results. As we can see at the score submitted to the F1 ranking on HWBot, all teams did exactly what I described above. Thanks to Tones I was actually able to get a couple of cards to test, but had to keep them at stock cooling as they had to be sent back. So, all what was left to do was push the CPU as high as possible and see where it could take me.

So, first of all: the score.


(Click to enlarge)



Comments and remarks

A score of 36285 in 3DMark Vantage is not that bad, but not shockingly good either; currently this score is listed 34th global in the overall Hwbot Vantage ranking. For the given clock frequencies, this score is actually quite good, so I'm happy about the efficiency for sure.

The GTX295 used were the single pcb version which doesn't have a voltage IC on board that allows software voltage modifications as it lacks the necessary I2C bus. There are some simple voltage modifications available, but since this card had to go back to Tones I didn't do them. The clock frequencies of 680/1600/1200 on stock cooling can be considered average for this card.

CPU-wise we had quite some luck as the chip is quite good. We had some issues with the motherboard (see below), so I'm pretty sure that this sample can be pushed over well over 5GHz on LN2 in the future.

The motherboard used, DK X58-T3eH6, definitly held me back when trying to push over 5GHz. The reason seems to be the OverCurrent Protection (OCP) as the board shuts down everytime I come near 5GHz no matter what I try: increasing/decreasing the temperature, voltages or a combination of this always leads to the same shut down issue.

At first, the BCLK formed a limitation as well as I could only go up to 205MHz, even on LN2. This was fixed quite easily by setting the boot bclk frequency to 205MHz. Note that the performance did drop in comparison to booting at 133MHz, for which the reason is unknown. I suspect there's a certain strap change (B2B adjustment maybe), but can't confirm without running a couple of extra tests.

All in all I'm actually quite satisfied with this result. I had only three days to test every components separately and one single benchmark session of 7 hours to test everything in one system, so time was most definitely not on my side. On the other hand I have to say that I should've put more effort in trying out, for instance, the Lanparty UT X58-T3eH8.

The other competitors did a very good effort as well. Especially Team Australia did a fantastic job by breaking P40k with old dual-PCB GTX295 cards. With their 40841, they have one of the highest, if not thé, I've seen so a lot of kudos for that one. Team EVGA tested a new combination of 4 x GTX285, which is pretty much the same technology as the Asus MARS uses; this result also required a lot of determination as it's most definitely not easy to handle 5 (five!) LN2 containers at the same time, so a lot of respect goes out to Team EVGA as well. For more information on that product you can go to the Evga website.

More pictures of the setup













Thread: http://www.madshrimps.be/vbulletin/f39/f1-oc-august-team-dfi-65816/