386>486 Processor Upgrade? Didn't Happen as Claimed!

Bill.Tetlow@Colorado.EDU

We have started to receive calls from users about the desirability of using the Cyrix 486 processor upgrade chips for 386 machines because these chips are retailing for only a few hundred dollars. According to a glossy brochure I received via a direct mail marketing campaign, it reads, in half-inch high letters, "One-Chip Microprocessor Upgrades. 486 Power is yours Now! SAVE THOUSANDS ON THE COST OF A NEW MACHINE! INSTALLS IN MINUTES . . . "

Next on the page, a colored box includes the following statement: "The 386 to 486 Upgrade Microprocessor requires no other modifications to your system's hardware, software, or operating system." And elsewhere on the page:"The Clock-Duty Cycle Correction compensates for variable clocks on 386 motherboards."

Well, there are two full pages of color illustration and virtually breathtaking claims and assertions. For example, "Replace one chip -- in just minutes -- and enjoy the new generation speed on your current PC." I think by now you are salivating at the prospect of "breathing new life into your tired old PC" and are ready to spend some hard-won budget dollars for an office upgrade.

Well, just hold on a moment, folks. Let us relate our experience at the CNS DISC Center and let you be the judge of these advertising claims.

We have two IBM PS/2 model 70 machines, one at 16 MHz and the other at 20 MHz, that are perfect candidates for the upgrade; so we decided to test the claims. If accurate, we would truly squeeze much more useful life out of these networked machines.

After installing the chip, we first discovered that the new chip with its "top hat" heat sink installed to cool the chip, was sufficiently taller to prevent re-insertion of the PS/2 RAM memory upgrade card in its original slot. Now that slot becomes useless unless a half-size card occupies it, but we could adjust by moving the two existing cards to the other two slots. We realized that will cause a CMOS configuration error but felt that condition can be accommodated easily. Incidentally, the accompanying Cyrix manual provided excellent step-by-step instructions.

Next we re-assembled the system and re-booted, only to get a "225" error code, which in IBM jargon translates to "system board memory error." At first we thought that this was related to the re-positioning of the RAM expansion board, but a search of the Cyrix manual appendix C, entitled "Trouble Shooting," revealed two astonishing statements that make the advertising claims pure hyperbole. I will quote in full so that you can fully understand our anger at the deceptive advertising by the distributor:

"The following systems have displayed various BIOS related problems, some of which are corrected with a new or updated BIOS:

AST Power Premium - 25 MHz (no BIOS upgrade available, this system is incompatible with the CX486DRx2 Upgrade Microprocessor)

IBM PS/2 Model 70 - 16 MHz (New BIOS available from Cyrix which eliminates the 225 error caused by slow RAM)

NCR 920" (page A-6)

Well, it turns out that this "few minute upgrade" is going to be put on hold until a new BIOS can be obtained (free of charge) from Cyrix. Since we are into the intriguing exceptions to the claims of trouble-free installation, let us read on to see what other "gotchas" lie ahead. Next heading is "Math Coprocessor Incompatibilities" and we discover:

" . . . Intel 387DX coprocessors manufactured before 1991 and other non-Intel coprocessors will actually send the READY signal to the microprocessor before it is ready to accept an operand. This is not a problem for the 386DX microprocessor because the 387DX has enough time to get ready by the time the operand is sent. Application problems occur after the CX486DRx2 is installed because the 387DX does not have the extra time to get ready when the Upgrade Microprocessor sends the operand. This problem is eliminated by replacing the 387DX Coprocessor . . . ."

Lo and behold! Looks like we are going to have to either buy the replacement Coprocessor from Cyrix or remove it entirely from the system. No changes to our hardware, eh?

The third section of the troubleshooting Appendix C is entitled "Software Timing Loops" and the next nine pages give instructions on how to solve these problems. The opening sentence of this section states, "Loaded TSR's and device drivers may cause apparent system failure when the internal cache of the Upgrade Microprocessor is enabled."

Now all of this makes the statement, "The 386 to 486 Upgrade Microprocessor requires no other modifications to your system's hardware, software, or operating system," nothing more than a bald-faced and outright lie!

I was inclined to consider that the lie comes from an overzealous and deceitful distributor until I began seeing other articles in personal computer magazines such as "Light a Fire under your Old 386" (PC Computing, August 1994, volume 7, Number 8, page 56), which states "Installing the Cyrix Upgrade is painless . . . . After swapping the chips, install the caching software, and the job's done . . . ." The article does point out that their tests show a 35 - 65 percent speed increase and not the 100 percent plus that some readers will expect, but no mention is made of any incompatibilities.

Well, our bottom line is that maybe we had a fluke upgrade situation that will apply to only a small percentage of users, but we cannot settle that issue without extensive further testing. In addition, for reasons cited in a previous article on this subject (Digit, July/August) we know that the performance boost will not match that of a new system designed around the 80486 processor but is a reasonable expenditure and upgrade for some users.

Upgrade your processor? Caveat Emptor! And for those of you who believe that it will be trouble-free, I am tempted to "make book" and cover all bets myself; it would be like winning the Colorado lottery!