What we have here is a failure to communicate.
Yes, you are right, Brian, the L2 cache is on the motherboard in K6-2 systems. It was late last night, and I guess I didn't make myself clear. The point I was trying to make was that the K6-2's L2 cache on the motherboard already does an adequate job. It's not as good as a backside cache (off-chip or on-die), but it's certainly much, much better than nothing (e.g. cacheless Celeron).
I'm basing my very biased opinion on both the results of Xeon on Winstone 98 and AMD's unwillingness to push the K6-3 to an earlier release date. First, the Pentium II Xeon displayed only a sliver of a performance advantage over the regular Pentium II when it comes to desktop benchmarks like Winstone 98, even with its full-speed, lower latency L2 CSRAM cache. It's not as great a speed increase as in the case of the K6-3, but it does tell me that diminished returns do exist. Second, AMD has let the K6-3 slip to the final days of Q1, perhaps even letting it slide to Q2. Why? Either they've found that the performance advantage of the K6-3 isn't worth the extra cost, or they're pulling resources toward the K7, or both.
Now when the K6-3 is released and we see the benchmarks, perhaps I'll be proven wrong. But if I had to make an honest guess, this would truly be my guess, whether I was working for Intel, AMD, or PC Magazine.
Tenchusatsu |
