I have a number of problems with the article because it think it is overly
simplistic. There are many more problems with PCs besides the CPU and
chipset. There are so many sources of problems that it may be hard to
identify a particular problem with any one component.
I have two nearly identical PCs. One occasionally malfunctions and the other never does. Both PC’s are SOYO
5EHM based. Both PC’s use AMD K6-3/400 CPUs. Both PCs have Voodoo
3/2000 video cards. Both PC’s have Linksys 100TX NIC’s. Both PC’s have
Creative Lab’s Ensoniq sound cards. One PC occasionally hangs on initial
boot and the other never does. It never hangs on the second boot. It
seems as though it requires a little time to warm up.
There is one difference between the two. The memory sticks came from different
manufacturers. I suspect that the stick in the “less than perfect” PC is
marginal. I don’t know; I’ve never bothered to switch sticks to see if the
problem would switch PC’s. The problem did seem to disappear when I
installed a stick of PC133 memory in place of the PC100 memory.
The memory is “generic.” That wasn’t my buying intention. I ordered Micron
memory and the sticks were 3rd party sticks with Micron chips. Micron make
good chips; Crucial is one of the top brands. But do lower grade Micron
chips end up in the grey market? Is water wet?
I really think the problem is the components: resistors and capacitors as
well as the chips. I design and build high reliability electronics
equipment. In almost every case of equipment failure in my line of work, the problem has been
traced to a marginal component. We are very careful about the components
we buy. Every component is required to meet mil-spec standards. Every
component we purchase is screened and tested and we still get bad
components. Sometimes the screening house or the manufacturer is
careless. Sometimes there are hidden and latent defects.
Why would the problem affect AMD more than Intel? I believe the answer is
volume. Most Intel motherboards are made in-house. When Intel buys
components for 10,000,000 motherboards, the manufacturer is going to be very careful not to give Intel a bad
lot. That’s a lot of repeat business to lose. When a mobo manufacturer buys components
for far fewer motherboards, the vendor, probably a distributor, may not
be so careful. The 10,000 resistors the distributor sells to Soyo may have
gone through several re-sellers and the lot information may have been
changed along the way. A batch of 10,000 1000 ohm resistors may have left
the factory as 5% resistors and arrived at its destination as 1% resistors. Some 900
ohm resistors may have been added because the original batch only had
9000. The result is that some of the motherboards may be marginal.
The same process works with any other printed circuit card. The memory
stick could be marginal for the same reason, the sound card could be marginal.
Hardware may not be the only problem. I am sure that you will remember
that several years ago there was a problem with AMD K6-2 cpu’s with an
internal clock speed of more than 350 mhz. The problem was that in Win 95,
Microsoft had used a counter loop as a delay, rather than use the
clock. At 350 mhz, the cpu was ready for the next step before the IOS
handshaking was complete. I wasted nearly 6 hours on that problem. The
next day, when I successfully booted DOS, I began to suspect that Win 95 was
the culprit. A half-hour search of Microsoft Support found the problem and
a replacement driver. The problem was solved and it wasn’t a problem with
the AMD K6-2 CPU.
I do not disagree with your statistics. I don’t have the knowledge base to
question any of your numbers. But I do suggest that the industry may be
much better prepared to serve the guy that makes 80% of the cpu’s as
opposed to the guy that makes 20% of the CPUs out of sheer self-interest.