Just after lunch (Subway sandwich, since you asked), the benchmarking started – most of the teams hadn’t taken time out to have something to eat and instead focused on making sure their system was tweaked as best as possible so that they were ready for battle as soon as the timer started counting down.
The liquid nitrogen was well and truly flowing at this point in time – the teams were filling their thermos flasks every ten minutes or so and they were going through LN2 almost as quickly as you could say ‘gosh, that’s cold!’ The benchmarking was now in full flow and scores were being submitted at regular intervals.
One thing was clear from this point on – there was a lot of frustration and it takes a cool head to remain focused when you are literally running on the edge of stability. Many of the teams were pushing their Core 2 Duo E8600 processors up well beyond 6GHz for the SuperPI 1M phase of the competition.
A few of the teams didn’t bother modifying their graphics cards until after
the SuperPI 1M phase was completed because the GPU speed had no impact on the systems’ ability to calculate PI to a million decimal places. This made the most sense to us, because you didn’t need to control the flow of LN2 onto the GPU as well as the CPU (and in some cases the northbridge too).
One interesting thing I spotted with team Indonesia’s rig was that the north bridge cooler wasn’t actually secured to the motherboard. Instead, the team decided to merely press the mini cooling pot down with the paint brush they’d used to insulate their motherboard – this didn’t look to be the plan originally though, as they were attempting to fit their mounting mechanism but ran out of space for both CPU and northbridge coolers.
In the end, it was team Indonesia who came up trumps in the SuperPI test, with a time of 7.5 seconds – that’s apparently a new record according to a statement sent out by MSI following the event! Their closest rivals were team Greece, who recorded a time of 7.687s, which ended up around 4.5 percent slower than the Indonesian time.
That wasn’t the best time that Greece recorded though, as the team managed to get below 7.5 seconds at one point but the benchmark is not valid unless it is screenshotted with three CPU-Z screens open at the same time as the benchmark. Upon opening the third CPU-Z screen, the Greece machine cold bugged and black screened, which meant that the result could not be counted – it was gutting for the team, but rules are rules as they say.
After a short break, the second 50 minute countdown started and this time it was time to find out who could hit the highest 3DMark 2001 SE score in the allotted time. LOD hacks were in full force and the scenes looked anything but normal – this was all about getting the highest score possible and image quality was the last concern. You can see the kind of image quality they were using in these tests in the middle screenshot above – that’s the Nature test from 3DMark 2001, if you were wondering.
The Taiwanese team from Coolaler’s website was the only team to not directly cool the graphics card with liquid nitrogen, but that didn’t stop them from using LN2 in their makeshift reservoir. The teams that hadn’t already modded their graphics cards did that and fitted their GPU cooling pots ready for some frosty 3DMarking.
The session was taken into overtime and another 50 minutes was allotted to the 3DMark test – in this time, the lead changed hands a number of times. South Africa, Taiwan Coolaler and Indonesia had all reached the highest score along the way, but then Greece came from nowhere in the dying minutes after a generally frustrating hour and a half of benchmarking with no luck.
Gprhellas and Gorillakos, with the help of the Turkish team whose system had coldbugged for a second time meaning they were out of the competition, managed to get their system tuned pretty well and had a couple of good runs through some of the parts of the benchmark where there can be huge variance. It was a great example of teamwork and good to see the teams working together even in the heat of battle.