This is my attempt to have every pinout on a PS/2 listed and named. At worst this will solve the problem of someone referring to a DB9 as 'The little port with some holes in it and in a 'D' on its side shape' :). And at best it will provide the information to build cables yourself and hybridize certain cable sets.
See Appendix E for listing.
I have just bought a third-party SCSI cable for my PS/2, and it works. It allows you to connect the 60-pin PS/2 SCSI adapter external connection to any device with a standard 50-pin Centronics connector. So, there are three options for getting this type of connector. By the way, the PS/2 SCSI connector is the same as the one on the RS6000. Of course, the alternative to all this is to use an internal SCSI device, if possible. The IBM internal SCSI connections are the same as those found in internal SCSI devices (the 50-pin rectangular connector).
Buy the IBM cable from your IBM dealer. The part number is 32G4143.
It will cost about $49.
Buy a third-party cable. I bought mine from Storage Solutions [closed
as of 10/28/98]. Their number is (203)325-0035. Mine cost $75 for
a 5' cable. Storage Solutions call the IBM connector a
"60-pin compressed" connector (though they are not really pins - it's really
a kind of edge connector). They also know what you're talking about
if you just call it an RS6000 SCSI
Inmac (1-800-323-6905) also sells them (see their UnixSelect catalog). They call it a Mini-Centronics (60) connector. They charge a bit more than Storage Solutions.
There are probably other suppliers. The key piece of information
is that it is
Make your own. The SCSI connector is available from AMP (1-800-522-6752 or 1-800-526-5142 or (717)564-0100). The AMP part number is 557025-6 (not to be confused with the 557025-5, which is the same connector, but without thumbscrews to hold it in the SCSI port). AMP also calls it a CHAMP .050 Series III Plug Cable Connector. I was unable to find any AMP dealers who could supply this. I had to get it straight from AMP. And it would have been so complicated for them to supply one as a normal order that they sent me one as a free sample. In the end I didn't use it though. It turned out to be almost impossible to solder. It is designed to have each of the tiny SCSI wires forced into a slot that automatically strips the cable. You'd need a special tool to do that. By the way, I bought a cheap SCSI cable and cut one end off, so that I wouldn't have to solder the other end as well.
Thankfully, IBM followed most of the ANSI SCSI standard. Any common SCSI-1 (50 wire) cable will work. Just remove the connector that attaches to the SCSI controller and replace it with a crimp-on 50 pin edgecard connector. Note that Pin 1 is toward the mounting bracket and is on the circuitboard side! The "true" IBM afficianado knows that the SCSI cable exits the controller's connector to the side without components. 50 pin edgecard connector for the IBM SCSI /A, SCSI w/cache, and FW SCSI controller. Available from Dalco and Jameco...Dalco part 40720 .
Any .025 pitch cable with HPDB68 drive connectors will work (called SCSI-3), but the IBM F/W SCSI adapter's Fast/Wide internal port uses the "P" connector. (Molex 71660, AMP 1-557089-2) also called a VMC (VESA Media Channel) connector.
Newark has the AMP equivalent. 1-557089-2, Newark # 97F8813
but OH is zero. Oh-oh...
Use common sense (scarce commodity) when using a "benchmark" program. When it comes to calculating processor performance, some common programs use timing loops. If the loop fits inside the L1 cache, you have a screamer. If not, your system is hopelessly slow. Or is it?
Benchmarks programs are a good way to compare systems and even better to see how changes to a system affect it, however, it is unreliably to compare benchmarks to other benchmarks (even if the programs are the same) unless the same environments are used.
A very good and pretty much the standard benchmark program is COMPTEST version 2.59 is the latest on 10/94). This is excellent to compare system to system and describes how to set up the autoexec and config files to run the test under. This is the control variable and allows apples to be compared to apples. This is a public domain program and probably the best in its class.
Other standards are somewhat variable as to what they will rate your
system as A LOT of hardware manufacturers (esp. video card) are setting
up benchmark 'sniffers' on their hardware to give back excellent numbers
to the common types of benchmark tests. I would like a list of benchmarks
known to be sniffed' for and the hardware that does 'sniff'.
The best way to prevent this is to make new benchmark test constantly and
not to buy the benchmark programs whose creators share what the tests look
for with the manufacturers before they come out. What happens is
X-company is coming out with a new benchmark program, then Hardware R Us
requests the parameters of the testing and figures out what to send back
to the benchmarking software to get FAST but REALISTIC marks. Usually
the benchmark company is affiliated with the hardware company so both benefit.