In my 2007-04-11 blog item, I reported that I had purchased a Sun Fire V210 Server. The configuration I selected has one 1.34 GHz UltraSPARC IIIi microprocessor, 512 MB of main memory, and a list price of 3395 USD.

I thought it might be fun to do a quantitative comparison to a Sun Blade 150 Workstation with one 550 MHz UltraSPARC IIi microprocessor and 512 MB of main memory (no longer orderable). Using a 24041 test case workload on the Sun Blade 150 Workstation and a 61390 test case workload on the Sun Fire V210 Server, the cycles/second speedup of FSS was 1.98. But a clock frequency of 1.34 GHz is 2.44 times faster than a clock frequency of 550 MHz. Furthermore, the UltraSPARC IIIi microprocessor employs a superior microarchitecture as compared with the UltraSPARC IIi (e.g. much larger L1 and L2 caches). Clearly, other factors limited the overall speedup.

It is noteworthy that, using a 24041 test case workload on both machines, the speedup was 2.46. Here, I am comparing the fastest FSS performance ever observed on each computer.

That's enough about logic simulation. A passion of mine is computer arithmetic, particularly floating point. Using a floating point benchmark I wrote that predates Cosmic Horizon, I made some FLoating point Operations Per Second (FLOPS) measurements. The C++ benchmark, called FLOPS3, performs a large number of operations on operands of the long double fundamental type. The output from each of the two Sun products I have been comparing follows.

Sun Blade 150 Workstation:
1.91797e+06 FLOPS
sizeof(long double) = 16

Sun Fire V210 Server:
3.73043e+06 FLOPS
sizeof(long double) = 16

Here, we have a speedup of 1.94, which is less than the clock frequency quotient.