Kentsfield survives tropical torture test

THE NEWLY-UNVEILED Core 2 Extreme QX6700, the first quad-core dual-die PC CPU, is a hot chip - both in terms of performance and demand for it, as well as, literally, its heat.

After all, two Core 2 6700 dies put together will more-or-less create double the warmth when operated at the same original speed - or 130W in this case, same as my old Pentium XE965 (which has only two slower cores on the same 65nm process, to be fair).

The bundled Intel heat sinks have never garnered a great reputation in matching these expensive and fast CPUs - if I was Intel retail group I'd bundle at least an equivalent of CoolerMaster or Thermaltake high-end copper model, if not the Zalman 9700, with such a $1,000+ chip.

Anyhow, what happens when I decide to make the best of Singapore's year-round hot and humid weather, perfect for sauna-style steaming, in my notoriously poorly-ventilated lab room, and turn off the airconditioning - letting the hapless QX6700 hang on for dear life at 35 Celsius room temperature on that Intel bundle heat sink, and yet overclocking it for a good measure? Here's how it went.

The system, based on Intel's D975XBX2 "overclocker friendly" mainboard and fast, low-latency Corsair XMS5400CL3 memory (running at DDR2-533 CL 3-2-2-6 at 1.8 volts), was set up on a clean white table within like five minutes, and, for good measure, I plugged in the fast (and just as hot) HIS IceQ 3 X1900XTX card - the one whose Arctic Cooling gadgetry allows smooth 700/1600 GPU/memory operation (see photo). A 500W MGE power supply with LCD display showing power consumption came in handy to compare the power usage in different configurations and loads as well.

Start first with the default settings of 2.66GHz/FSB 1066 operation, the system ran up to 72C CPU temperature within few seconds after entering BIOS setup, all that with extremely high fan noise, comparable to a belching public bus departing from a stop.

The system, nevertheless, went on and handled the WinXP installation all the way without a hitch. After installing all the usual drivers and utilities, I ran Sandra 2007SP1, and right after starting the CPU benchmark, a nasty Intel Monitor message propped up "CPU temperature 81C, exceeding the set limit" - uuuh, this was awful, I never had such message on default CPU speed, even with Pentium XE965. Anyway, the heat sink turned really hot to the touch by then - but all the Sandra benchmarks completed fine at the end. 3DMark06, though, failed at the CPU benchmark stage, until I set the sensor threshold target to 88C - an awfully hot temperature. At that point, 3DMark06 benchmarking was completed, two rounds of two runs each (see table down).

Now back to the BIOS - knowing the previous FSB overclocking sensitivities of 975XBX boards, I started first by upping the multiplier from 10 to 12 to raise the CPU clock towards 3.2GHz with the same 1066MHz FSB. The BIOS hardware monitor temperature jumped to 78C - you don't need to guess that Windows didn't even load, the system froze right after storage detection stage, even with most liberal memory and voltage settings.

In the BIOS itself, I could manage to set the system to 3.3GHz CPU (FSB1200 or 300x4, and multiplier 11), after overclocking the FSB and north bridge by 0.125 volts each, and still get the system to restart OK, but without going back into Windows in any of the cases. Obviously, it was time to stop steaming, and turn on the air conditioning.

Close the windows, and cool the room full blast - 23C temperature. What a difference - now, even with this crappy heat sink, 3.2GHz with 1066 FSB (at the very same 1.2875 volts for the CPU!) ran cooler than the 2.66GHz previously (acording to the Intel monitoring utility, the CPU was 1 - 2 degrees cooler during the benchmark runs). All initial tests completed beautifully, except for greater power consumption: while, in full-load 3DMark06 run, the system took up to 216W at 2.66GHz, the number jumped up to 235W at 3.2GHz, all other things being same - so, it tells you that upping the CPU clock those 20 per cent will take an extra 20 Watts of power, too.

Here's the table:

After this, I tried again my favourite 'recommended' setting of 3.33GHz CPU core with 1.33GHz FSB, keeping that same FSB multiplier of 10, however, no matter what and how hard I adjusted all the settings, it just didn't go ahead any further. I guess this will wait for the weekend Round 2 of the test, where the Kentsfield is covered by a much higher end cooler - and so is the northbridge... Otherwise, there is always the MSI 975X Platinum board to try.

Overall, still, quite an impressive showing for such a hot CPU even in the hostile climate - after all, processors, unlike most Western European humans, prefer freezing cold Iceland to hot Ibiza beaches.

The Sandra synthetic scores are roughly double, with FSB sharing having minimal, if any impact. And finally, with so many cores on hand, one could finally justify that crappy software-based on board sound - simply, let one core do that when listening to your favourite MP3, the other three cores won't really bother. µ