Some encryption algorithms might indeed be a good indicator for overall multi precision integer arithmetic performance.
I suspect that encryption algorithms are a fairly good indicator for deep-zoom fractal performance, but I don't know that for a fact. I do suspect that normal benchmarks are a poor indicator. The main requirement for fast deep-zoom fractal calculations is a fast 64x64 multiply, and fast adc instructions. I don't know how well Intel and AMD compare.
I agree that having other programs installed and running shouldn't matter, as long as they are well behaved. Task Manager can help you find those that are not.
Anyway, I wrote a benchmark program to help answer this question. It uses the same mov/mul/add/adc/adc block that FX uses so it should accurately predict deep-zoom performance in FX, and probably in other similarly optimized fractal programs.http://ftp://ftp.cygnus-software.com/pub/InfprecPerf.exe
Aside: apparently the throughput of FX is not helped much by hyperthreads. I didn't know that.
Here are the results from my four-core eight-thread Sandybridge laptop. It nominally runs at 2.2 GHz but it can Turboboost up to between 2.8 and 3.2 GHz, according to CPU-Z. I think that means that performance is ultimately limited by the four-cycles-per-mul throughput, and that the hyperthreads merely help hit this maximum.
VendorID = 'GenuineIntel'
Stepping = 0x7, model = 0xA, family = 0x6, type = 0
Signature is 0x6A7 (06_0AH)
CPU's rdtsc speed is 2.195 GHz (peak speed may be higher).
Running tests on 8 thread system.
Performance with 1 threads is 0.981 GBlocks/sec
Performance with 2 threads is 1.988 GBlocks/sec
Performance with 3 threads is 2.355 GBlocks/sec
Performance with 4 threads is 2.994 GBlocks/sec
Performance with 5 threads is 3.547 GBlocks/sec
Performance with 6 threads is 3.681 GBlocks/sec
Performance with 7 threads is 3.672 GBlocks/sec
Performance with 8 threads is 3.816 GBlocks/sec
VendorID = 'GenuineIntel'
Stepping = 0x6, model = 0xD, family = 0x6, type = 0
Signature is 0x6D6 (06_0DH)
CPU's rdtsc speed is 3.202 GHz (peak speed may be higher).
On my six-core twelve-thread Sandybridge work machine the results are:
Running tests on 12 thread system.
Performance with 1 threads is 1.149 GBlocks/sec
Performance with 2 threads is 2.268 GBlocks/sec
Performance with 3 threads is 3.263 GBlocks/sec
Performance with 4 threads is 3.961 GBlocks/sec
Performance with 5 threads is 4.803 GBlocks/sec
Performance with 6 threads is 5.447 GBlocks/sec
Performance with 7 threads is 6.123 GBlocks/sec
Performance with 8 threads is 6.365 GBlocks/sec
Performance with 9 threads is 6.452 GBlocks/sec
Performance with 10 threads is 6.506 GBlocks/sec
Performance with 11 threads is 6.591 GBlocks/sec
Performance with 12 threads is 6.667 GBlocks/sec
The system process was consuming about 6% of CPU time, otherwise I would have hit about 7.0 GBlocks/sec.
If you have other results from different CPUs please send them to me. Note that this program is 64-bit only. As somebody observed, 32-bit performance is not really comparable. It will typically be 4-5x worse, so not very interesting.