SuperPI calculates the number PI
to 8 Million digits in this raw number crunching benchmark. The program allows the user
to change the number of digits of PI that can be calculated from
16 thousand to 32 million. Our benchmark is set to 8 Million digits and 22 iterations.
The benchmark is measured in seconds; lower results are
better.
Now that's a smooth
graph! Looking at these benchmark results, it becomes obvious how
processor performance plunges as the clock speed drops. The 800MHz test
wound up almost three times slower than the stock speed test, which makes sense
since it was running at exactly a third of the speed.
Conclusions to this Underclocking
Experiment
One thing that emerged as a
definite trait of the underclocked system was predictable
performance. As processor speed was reduced, performance dropped off in a
predictable curve. The fact that processor-intensive applications like
SuperPI suffered more than the gaming benchmarks in our tests makes sense, since
the video card we used was not being underclocked in any way and was still
shouldering some of the burden of the 3D benchmarks. When the video card
was taken out of the equation, the performance drop was essentially equivalent
to the level of underclocking.
As processor speed dropped
below 1GHz, the law of diminishing returns kicked in and the decreases in
performance at each speed downstep became larger. On all the
benchmarks, the curve becomes more pronounced at the lower speeds,
indicating that underclocking further would most likely drop
performance to unacceptable levels.
If we
take the numbers literally, underclocking an Athlon 64 4000+ to a third of its
original speed will result in a ~7 degree drop in temperature under load, a 30-50% drop
in performance in video card assisted 3D applications and a 66% drop in processor performance. On the
bright side, the system would still be quite usable as a standard Internet/word
processing/general work computer.
So could you repeat this
experiment at home? Certainly, and you don't need to use the same
equipment seen here either. The Scythe NCU-2000 is only one of several
brands of passive heatsink available, and you can also use an active heatsink
with an alternate fan if you want a quieter system. Remember that case
ventilation is essential if you are trying to create an all-passively cooled PC
though.
Find out about this and many other reviews by
joining the Weekly PCstats.com Newsletter today!
Related
Articles
Here are a few other articles that you might enjoy
as well...
1. Intel Pentium 4 3.2E GHz Prescott Processor Review
2. Intel Pentium 4 3.2GHz Extreme Edition Processor Review
Super Pi
RSS Newsletter
