Stephen - Re: "Are they pretty even-handed?"
No.
Slater, as founder of the MDR, has been axing Intel for 8 or 9 years. His editorials, for years, predicted the demise of the clunky x86 architecture by the elegant, fast, clean RISC chips.
Alas, that never happened.
A few years ago he became disillusioned with the lack of RISC's success and Intel's de facto "win" by scaling x86 CPU performance far faster than almost all RISC chips (Alpha is an exception). And commercially, Intel won the whole ball game.
With the advent of Cyrix and AMD copy cat chips, Slater and his puppet Linley Gwennap have found another dagger that they can use to stick at Intel. In fact it was Slater and MDR that came up with the "PR" rating for Cyrix's 6x86 chip to help hide the fact that it's clock speed could not keep up with Intel's chip.
As a confirmation of this, when MDR did the first evaluation of the 6x86, they presented bench marks of the 6x86 in bar graph form showing how it outperformed the Pentium chip by a wide margin. They showed these results in a bar graph that had a vertical axis that, instead of starting at 0, had an origin of 1 and full scale of about 1.1.
They then plotted relative speeds of the Cyrix and Pentium chips - with specific benchmarks. These showed a small bar for a Pentium chip (of negligible height) and a big whomping bar for the Cyrix - extending near full scale. They DIDN'T elaborate that the difference in performance was less than 10% - NOT 100%!
This is a standard trick of statisticans/mathematicians wishing to slant the relative results of two sets of data. A CHEAP trick , I might add.
Here's the URL for that hatchet job - check it out for yourself:
chipanalyst.com@039513nhhdlh/mdrlabs/labsr_6x86_3.html
From an a "Newsbytes" Article in 1996:
"MicroDesign Resources, Sebastopol, Calif., sponsor of the Microprocessor Forum, has created a new arm, MDR Labs, which will administering the P-rating. In terms of performance, according to MDR Labs, "On Winstone 96, the 6x86 ranges from 1 percent to 6 percent faster than the corresponding Pentium."
However, on CPUmark16 and CPUmark32, which are synthetic tests, the 6x86 was shown "to be 3-6 percent slower than the corresponding Pentium. The exact cause for this discrepancy has not been determined, but it believed by IBM and Cyrix to be the result of deviations in these benchmarks from typical Windows application behavior," the labs concluded. Other tests in the benchmark execution included WinBench 96 Disk WinMark and WinBench 96 Graphics WinMark.
Michael Slater, president of MicroDesign Resources, said in that, in general, "In our tests, Cyrix 6x86 processors clearly outperformed the Pentium in every class." "
Basically, Slater will damn Intel with faint praise when he has to. If he has the opportunity, he will politely show how Intel's competition is so much superior. If that competition falls flat on its face, Slater will most assuredly negelct to bring this to his reader's attention.
Paul |
