The Linpack memory performance curves in the article make this whole thing a bit of a mystery and raise more questions than answered. Either the test is somehow invalid or VIA has managed, at great expense, to design a brand new kind of kluge - the cache memory on this chip.
The article mentions it, but does not expand on it. It would be interesting to see the companion integer memory test also. Maybe the FPU i/o is choking down this FPU memory test, but it sure looks like the cache system is not fully enabled. The memory performance is about on par with a Winchip 266. Why?
If everything VIA claims is inside this chip is functional, the curves should be almost identical with the AMD on the left side of the graph. Instead, it is less than a 1/5 what it should be. ????