All of the information in the previous pages has been accumulated from a whole slew of thermal design papers from both Intel and AMD and the profound experience the author has had with designing and operating advanced cooling solutions for semiconductors. And although we feel that Intel, by including the clock throttling feature, has implemented a safety net for high thermal ramp rates, further testing is required to determine whether it affects performance in normal circumstances.



Fig 7. A look inside our test system, featuring the Intel D850GB motherboard and a 1.5GHz Pentium 4 CPU.

For that purpose we'll be testing a standard Pentium 4 system with a 1.5GHz clockspeed CPU under a whole slew of different thermal conditions to determine what is needed to trigger the clock throttling feature and whether we're able to do that under normal circumstances. Naturally the BIOS of the motherboard has the clock throttling feature enabled, as its disabled by default. The following configuration is used for testing:

Pentium 4 system configuration

• CPU: Intel Pentium 4, 1.5GHz
• Motherboard: Intel D850GB, i850 chipset, 400MHz FSB
• Memory: 256 Mb, 2 x Samsung PC-800 ECC 128Mb RDRAM
• Videocard: Elsa Gladiac GTS, GeForce2 GTS, 32 Mb
• Harddisk(s): Quantum Fireball Plus LM, 30Gb
• CDROM: Generic 32X IDE-ATAPI
• Floppy: Generic 1.44 MB
• Case: Cooler Master aluminum midi-tower
• Powersupply: Delta Electronics Inc. 300-watts ATX

For determining whether we can trigger the clock throttling feature we’ll be cooling our Pentium 4 CPU down to –40C with a cryogenic cooling system capable of sustaining that temperature under full load and run three SYSmark 2000 benchmarks, the middle value will be used as a reference for all the others. The next thermal conditions we’ll be looking at is a standard Cooler Master extruded aluminum heatsink with no thermal grease applied and the same heatsink with thermal grease applied. The very same test will be repeated with a EKL AG, folded fin, copper base heatsink.

With all different heatsinks three consecutive runs of SYSmark 2000 will be run and the middle value will be used as the result. Furthermore we’ll also be running the Pentium 4 without a heatsink or unplug the heatsink fan during operation to see whether it keeps on running. Naturally our reference sub-zero cooled Pentium 4 will not suffer from any clock throttling issues due to the extremely low temperatures it is operating at. If any differences in SYSmark 2000 scores occur they are related to thermal issues and if possible to the clock throttling feature, we’ll have to compare between the extruded aluminum and the copper base heatsink to see whether this is the case here.