The New Celeron Format
In our preview of the socket-370 Celeron 366, our testing was done with a Supermicro 370SBA motherboard, one of the first PGA-370 boards available. From those tests, it's pretty clear that at the very least, a socketed Celeron 366 generally performs where it should - faster than a Celeron 333, and slower than a Pentium II 400. However, that doesn't say too much about the socket-370 format itself, or the motherboards based around them.
A couple of days ago, we received a little surprise - one Socket-370 400Mhz Celeron PPGA (just like the 366 tested, but faster), and an equivalent Celeron 400Mhz SEPP, based on the now-familiar Slot-1 format. This presents us with an interesting opportunity to test the performance of a socketed Celeron vs. an identical slot-1 Celeron, and a 100Mhz FSB Pentium II 400 for good measure.
Of course, there is one major issue here to consider - the motherboard. Differences in the motherboard can skew testing, so equivalent systems must be used. For testing, we were able to get our hands on a Supermicro 370SBA for testing. Our U-tron test system came equipped with a Supermicro P6SBA, the exact same board as the 370SBA, sans socket. Being an overclocking-friendly (or perhaps, future-processor ready) board, we were able to select a 6x multiplier against the 66Mhz bus, effectively using the Celeron 400A in native mode (without the need to overclock the FSB like most tests would require. This allowed to run tests against the Socket-370 with both a Celeron 400A and Pentium II 400.