• Please review our updated Terms and Rules here

XT booting times

D

dongfeng

Veteran Member
Joined
Nov 16, 2003
Messages
1,348
Location
London, England
I have noticed that one of my XT's takes longer to boot than the other. The actual counting of the memory is noticeably slower.

The fastest XT has 640kB on the motherboard and the second (Jan 86) BIOS.

The slower one also has the 256-640 motherboard, but the total memory is only 256kB (four banks of 64kB chips). The rest is on an expansion card. For some reason this system has the first XT BIOS.

The memory in both of the XT's is rated at 15ns.

Why is this? Is it due to the second machine having the expansion card? But the memory counting is still the same speed even with the expansion card removed.

It doesn't bother me, but I am quite curious to know why! I could understand if the machines had different speeds of memory, but they don't.

Both the motherboards were made in 1986.

Could it be the earlier BIOS is slower?
 
I've timed from the second the first memory account appears on the screen until the second it reaches 640K.

The first XT takes 23.96 seconds. The second XT takes 46.61 seconds!
 
The memory is probably more like 150ns, not 15ns. Static RAM is in the 15ns range .. DRAM back then would have been just fine at 150 or 200ns. (A 4.77Mhz machine give you a 213ns clock cycle.)

XTs also don't have the concept of local bus. So memory on an expansion card is just as fast as memory on the motherboard. Everything runs at 4.77Mhz.

Have you neglected something obvious, like that the faster XT has been modified with a faster clock crystal? Why don't you run a quick benchmark program on both and find out?
 
Sorry, my typo. I meant to say 150ns. It was the memory that ended in -15!

I am pretty sure the benchmarks are the same. I'll go check now...

The faster XT does have an 8087... if that would make any difference to boot times?
 
I only had CHECKIT to hand, so I used that.

Fast XT:

CPU: 344 Dhrystones
Video: 637 CPS (IBM CGA card)
Maths: 112.6 Whetstones (8087)

Slow XT:

CPU: 344 Dhrystones
Video: 520 CPS (Hercules CGA card)
Maths: 6.7 Whetstones (no 8087)

It would be interesting to see how fast the "fast" XT would boot with the 8087 disabled.
 
I also noticed that the 640kb motherboards with 85/86 BIOS count the memory much faster. My guess would be that a wait state is needed in order to properly access the memory on the card. What would be really interesting is to see what happens when you remove the 384kb from the motherboard, and install a sixpack plus. Do you think that the '86 BIOS would be able to auto-sense this change and adjust the wait states? Or perhaps it would generate some errors.
 
dongfeng said:
...It would be interesting to see how fast the "fast" XT would boot with the 8087 disabled.
Click to expand...

The 8087 creates an "electronic whirlpool" for the 8088 to use, allowing it to boot the system quicker...

Either that or the 8087 counts memory better...

For both of these theories I think I am about 10 days too soon...
 
Anonymous Coward said:
I also noticed that the 640kb motherboards with 85/86 BIOS count the memory much faster. My guess would be that a wait state is needed in order to properly access the memory on the card. What would be really interesting is to see what happens when you remove the 384kb from the motherboard, and install a sixpack plus. Do you think that the '86 BIOS would be able to auto-sense this change and adjust the wait states? Or perhaps it would generate some errors.
Click to expand...

I have just tried that. Changed the switches on the mainboard to have only bank 1 (256kB) installed and the remainder up to 640 on an expansion. No change.
 
Guys, some basic research would have told you that the 8087 is completely not an issue during the boot. Nothing in the BIOS uses it or cares.

The CPUs are the same. The most likely issue is that the newer BIOS is not erasing memory as it goes, or counting 64K at a time. The older BIOS might be erasing memory, and it definitely counts 16K at a time. You really need to read the BIOS listing to know for sure what is going on.

The WS theory is possible, but I doubt it. Wait states are generally used by IO devices. If both versions of the BIOS count by 16s and don't do anything funny like erase memory then you will need to get a book on the ISA bus to research the wait state theory. If it really were a WS issue, it would be more like 20 to 25% slower when accessing memory, not 2x!
 
XT Post speeds

XT Post speeds

It might be worthwhile to check out the BIOS dates and then look at the listings to see how the POST memory tests may differ.

To get the BIOS date:

c:>debug
-d ffff:5 L 8
-q

This will print the BIOS date; my bet is that the machines have different BIOS and that we'lll find the difference in the appropriate listings.

The benchmark you ran suggests that the chips themselves are running at parity. Faster-spec memory chips won't run any faster until the clock cycle is shortened.
 
The fast XT has the January 1986 BIOS (the second revision) and the slow one, although having a 1986 mainboard has the first revision XT BIOS (1983).

I'm not sure why it has the older BIOS, unless it was swapped in the past (Jorg? It was your old board! :p )

I'll have a look in the XT Technical Reference I have, I think it is the First Edition (1983).
 
Last edited:
I think you need both BIOS listings to do a comparison. At least on the planet I come from ...
 
I do know that...

Anyone got the listing for the Jan 86 BIOS, or indeed a late XT Technical Reference? Searches on the Internet have come up blank, although I did find several PDF Technical References but they were all the early one.
 
You should be able to get the IBM tech ref from my site, try ftp.oldskool.org in /pub/misc. Also, I have the last XT bios (1987 if memory serves) -- I can disassemble it for you if you like, just let me know.
 
You must log in or register to reply here.
Back
Top