If you can (and want to, incidentally), run it on a SDRAM system of similar CPU spec.... I reckon it'd be a fair bit slower, but that's just my instinct talking...
OK... nothing like an invitation to truly "geek out"... My "similar" SDR system was actually a 1700+ overclocked to 1800+ speed; that shouldn't make a difference. I tried it with both registered and unregistered SDR DIMMs, and also ran the test on a few other systems as well just for grins.
I used a modified version of the program without the output loop, so only the time of the prime calculation was being tested, not the speed of the printf() function. The program was run with an upper bound of 100000000.
Results, from slowest to fastest:
K6-III+ 500MHz, PC-100 SDR - 33.7 secs
Athlon TBird 800MHz, PC-100 SDR - 15.1 secs
Athlon XP 1.53GHz (1800+ rating), PC-133 registered SDR - 13.7 secs
Athlon XP 1.53GHz (1800+ rating), PC-133 SDR - 12.3 secs
Athlon TBird 1.1GHz, PC-2100 registered DDR - 11.1 secs
Athlon XP 1.53GHz (1800+ rating), PC-2100 DDR - 8.4 secs
So yeah, I'd say it looks like it is generally limited by memory performance, except on the 500MHz K6-III+ (which just generally sucks at this test).
This was a mix of Win2K and Linux systems... but I used the g++ compiler on the Windows side as well, to eliminate the compiler as a variable.