84 Points Short – My 3Dmark06 Adventure
84 Points Short – My 3Dmark06 Adventure
On the 17 September 2007 I stated the followingTalk about pride before a fall – October has come and gone and the new year is less than two weeks away and still I have not achieved that mark. And going into December it looked highly unlikely that I would ever get close and once my Ultra was sold the likelihood of achieving that elusive 20K was well not going to happen till the next generation of graphics cards is released. Those words would keep haunting me each time I benched 3Dmark06.Originally Posted by 1Day
Well this morning the new owner arrived to collect his graphic card. I was not only saying good bye to an outstanding graphic card I was also bidding farewell to any chance of making my boast fact. But life has a tendency to be unpredictable and this day is just one such crazy day. Dawud and his wife are the curve ball that has made a very sad day into one that has been an incredible rollercoaster ride.
As some of you know I have had my Ultra graphic card advertised in the classified section on two forums, I have also advertised locally in the area I live in and Dawud a local business owner was the man who decided he wanted that card. He and his wife Farida came round this morning to collect the card. I have no idea what they were expecting but my benching lab was certainly not it. Chaos is a polite way to describe first impressions I am sure. A whole heap of water cooling kit hanging from hooks on the walls with a multitude of power cables and other PC type cables snaked around two the different colored Tygon tubing makes a bizarre montage. My two desks are littered with hardware, tools and these obscene benching pots, that taken with my benching rig which looks like something that Dr Frankenstein would be very at home working on.I guess they wondered just what sort of household they had walked into. :ph34r:
But surprisingly they were more interested in what all the stuff was for and what it did. So a half-hour later I was still answering questions and not getting anywhere quick I decided to show them both rather than explain further. I ran 3Dmark06 using their graphics card, this served two purposes first it clearly showed them both that the card worked and is an excellent example of what benchmarking is all about. As most of you know the benchmark has lots of lovely graphics and then a very CPU intensive bit in the middle for contrast and is quite pleasant to watch. They were both quite taken with the whole benching thing and I mentioned that I was very sad to see the card go because I felt that I had some unfinished business with the card. Next thing I know I am instructed to complete the business I have outstanding with the card.
Now you have to understand something, you do not get over 900 MHz on the clock by wishing it so. You have to do stuff to the card and you have to freeze the card. So after another half hour of explaining just what was needed for us to do as they had suggested I was amazed to hear that they were still both as keen as mustard. Dawud was dispatched to find 10 Kilograms of dry ice (sadly not the 3mm pellets as those are special order items) and I began to attach the variable resistors onto the card. While my new assistant and team mate insulated the TEK9 graphics pot in preparation for our little adventure. It was a kind of weird feeling having people around when I benched as I have never benched with anyone before, my kids are normally asleep when I bench, but it was great to have someone chop the dry ice into mush for me. Once the slurry mixture of bashed up dry ice and acetone was ready I booted up into a fresh OS install and the benching began.
The first eleven attempts at the benchmark failed and I could see that my new team mates were getting a bit worried, but I explained that it was the CPU that was failing and not the graphics part of the equation. It was at that point I realized that if this couple was willing to do the unthinkable and risk a card that costs more than most people earn in a month then I would have to be prepared to risk my CPU too. So once more into bios and I upped the voltage to the core to 1.8 volts and pushed the VTT up to 1.94 – do not try this at home folks unless you have lots of frozen stuff attached to your CPU. Setting the PCI frequency to 115 and my ram volts up to 2.4. I had never benched 3D at this FSB before so it was uncharted waters for me.
First run was good but not good enough to be a PB. But at least the benchmark completed. All three of us were happier; it was at this point I asked Dawud if I could crank his card up to the max, at this point it was still running 702/1828/1221 clocks. Ignorance is bliss I guess because he quite happily said “go for it” and I did.
I had only ever been able to run AM2 at 918 MHz on the GPU before, but today I was able to run 06 time and time again as I searched for that elusive perfect benchmark run. These poor people nearly had a heart attack when I stuck a piece of planking into the dry ice in the pot and leant on it each time during the last part of the benchmark you get a work out while benching let me tell you.
Normally 10 kilograms of dry ice will last me the whole day, say ten hours of benching when I use only a CPU pot. But both the quad core and the GPU are very hot. So before I knew it 10 kilograms of dry ice were nearly gone. And all I had to show for it was this score.
19286
Dawad saw we were running out of dry ice and in a flash was gone and returned fifteen minutes later with a fresh box and began the task of mashing it up with a hammer and mixing acetone to make slurry type mixture. As soon as he was done and we had had a bite to eat – which he had thoughtfully provided as well we started again. Failure after failure to improve on our score was getting me down. But my two new team mates seemed not to even notice or really care and were having a blast. After a while it was clear that we had hit a platue and not much improvement was possible as things stood. Going back into bios I bite the bullet and set voltage onto the CPU to a max of 1.96 volts with the FSB at 500 and ram settings of 4-4-4-11 pushed to the limit of stability (4:5 effective 625 MHzat 2.5 volts). PCI frequency was jacket up to 155 and away we went. First run died during Canyon so I upped the voltage on the ram to 2.7 volts and bingo the next run was a real good one and here it is. Not the jackpot true but darn close.
19916
Sadly the very next run frooze and killed the OS and toasted the hard drive partition, the price you pay for an insane PCI frequency and SATA I guess. But the score is worth it to me. I am hoping to install a fresh OS and try again a little later when it is cooler as both these two new converts seem up to it. :blink:
Reply With Quote
Bookmarks