When clock speed is equal more cache equal more performance. But in the AMD64 architecture, the clock speed have more influence than cache size. Intel P4 suffer much of cache size reduction (see Celeron performance).
And AMD choose to offer different version of the same chip, because they have better yields this way! So, if a chip have defective cache, they deactivate the defective part and they can sell it with a lower PR.
This PR rating business didn't seem as confusing when you could at least count on it to have a close relationship to clock speed within the AMD line of CPUs.
The problem with clock speed rating is that there is much more than the core speed that affect CPU performance. It's not without reason that Intel is switching to model numbers too. Because they want to slowly destry the "MHz-Myth" that Intel build in the past.
When you check today's Intel line-up, there is a lot of speed grade in different CPU family, so how to make sure people will understand that the Dothan 2.0GHz is faster than P4C 2.4GHz or a Celeron 2.8GHz??? By model number, the higher the model number, the higher the performance.
AMD respect their rating and they don't want to push it to the limit... If they were doing this we would have model numbers like 3175+ or 3250+ and this would add complexity for no reason. In today's cpu, difference of 1 or 2% don't mean much.
So for your CPU choice, it's up to you? Do you want a bit more performance and lower cache at stock speed or a bit less perf. for more cache? you have the choice!
choosing between Athlon 64 3200+ with 512K or 1024K of cache, is like choosing between Athlon XP 2500+ (Barton) or Athlon XP 2600+ (t-bred), you want clock speed or cache size?
--
Asus A7N8X / <font color=green><b>AMD Sempron 2800+</b></font color=green> (tbred @ 167x12)
Kingston DDR333 2x256Megs
<font color=red>Built by ATI Radeon 8500LE 128Megs</font color=red> @ C:275/M:290