• Please review our updated Terms and Rules here

8088-II 8Mhz vs. 286 8Mhz is not same?

M

musicforlife

Experienced Member
Joined
Jan 10, 2018
Messages
255
Location
Finland
I have got this very fancy full set of Hyundai Super-16T EGA setup with HCM-1420 EGA monitor including all the manuals, drivers and utilities.

The computer contains 8088-II processor which you can manually set between 4.77Mhz and 8Mhz. The computer itself was made to be PC/XT compatible.

However when I wanted to replace my other 8Mhz machine with similar specs, I found out that all the games I am playing in my 286 8mhz machine are slower and lagging in Hyundai although the processor speed and the amount of free conventional memory is the same in both. Also both contains modern EGA card with 256kb memory and I tested it with same dos version.

Why is that? What makes this difference? I know the 286 computer is AT vs. XT but the games are working. Would it make any difference to replace 8Mhz with 10Mhz as I don't use the 4.77Mhz option but is that possible when considering the computer was made for the speed switch? Would it cause compability issues?

Here's a brochure of the machine http://www.1000bit.it/ad/bro/hyundai/Hyundai-Computer&Peripherals.pdf My unit contains one floppy drive and 20Mb hard drive.


Fun fact; while the computer is PC/XT, the keyboard layout is identical to IBM AT keyboard even with all the leds while XT keyboards didn't contain leds.
 
Last edited:
An 8088 uses an 8 bit path to memory while an 80286 uses a 16 bit path to memory. So right there, the 80286 running at the same clock speed is going to have a performance edge. The 80286 also has more instructions; code that is aware that it is on an 80286 could use those extra instructions. The implementation of the standard instructions might be more efficient.

I vaguely remember that an 80286 running at 6Mhz was supposed to be about 3x faster than an 8088 running at 4.77Mhz. The numbers depend on the code, but this is why you are seeing a difference; they different CPUs.
 
Their are at least 2 reasons why the 286 is faster than the 8088 at the same clock speed. First the 8088 is 16 bit internally but 8 bit externally. So access to memory ,disk drives and video is 8 bit. The 286 is 16 bit internal and external. So twice the data width to memory and drives and video. The 286 has some improved instructions that make it run more efficiently at the same clock speed.

Looks like mbbrutman answered while I was typing but he said pretty much the same thing.
 
Bottom line:

Never confuse CPU clock frequency with speed. I recall a quip by one of the 6809 designers that if they'd have known that CPUs were to be judged by clock speed, they would have put a waveguide on the 6809 for clock input.
 
According to the Landmark benchmark, even the fastest Turbo XT is still slower than the slowest AT.
V20 @ 12 MHz is measured as equivalent to 5 MHz AT - while the original AT was 6 MHz.
 
Plus, in addition to bus-width issues and additional instructions, some things are just implemented more efficiently on the 286. Multiplies, for example, are way faster.
 
There's more logic elements in 286, compared to 8088.
So, on a 286, some instructions are fully implemented in hardware, while on an 8088, they are executed via microcode.

Edit: transistor counts:
8088 - 29,000
80286 - 134,000
Majority of the extra transistors must've gone to the new instructions and protected mode support, though some were spent on improving the old instructions.
 
Last edited:
Anyone got any ideas whether changing the processor to 10Mhz would be possible? I'm afraid that the speed switch mechanism was implemented somehow in hardware level that it wouldn't work and also I'm wondering whether it is possible to break something when trying to use 10Mhz chip?
 
musicforlife said:
Anyone got any ideas whether changing the processor to 10Mhz would be possible? I'm afraid that the speed switch mechanism was implemented somehow in hardware level that it wouldn't work and also I'm wondering whether it is possible to break something when trying to use 10Mhz chip?
Click to expand...

Possible, yes. Change CPU and change the crystal. It might be easier in the Hyundai if it has split clocks so that only the CPU crystal needs changing while retaining a slower crystal that handles the rest of the system. Some video cards for example expect a standard clock from the system and won't produce correct video frequency with a replacement crystal. Also, if you push the bus too fast, expansion cards may no longer function. You will have to closely examine the Hyundai motherboard to determine how much work will be necessary to upgrade the clock and still have a functional system.
 
Another prime example being the 5170 with 8MHz vs. the 5162 with 6MHz both with 80286 cpus - while one would think it's a no brainer that the 5170 is faster due to the higher clockspeed, it's actually the 5162 that often comes out on top because the RAM has 0 waitstates.

But yes, there's a lot more to "speed" than just the clock frequency, waitstates, buswidth, attached/used peripherals adding a lot to the general performance. Other than that others pretty much hit the nail head on.
 
Shameless plug: For anyone curious to compare CPUs clock-for-clock, download TOPBENCH and do Database->Compare and you can directly see how two different systems compare in terms of speed (and why).
 
We might do some benchmarks to see the real differences. But maybe they are already available somewhere. We can do those:

5150 with 8088 or Olivetti M19 with 4,77 Mhz
IBM AT 286-8 or Olivetti M28 with 80286-8 (I still look for one of these)
Some turbo XT with 8088 at 8, 10, 12 MHz
Olivetti M24 /AT&T 6300 with 8086 (16 Bit) at 8 Mhz
Olivetti M24SP and Olivetti M240 with 8086 at 10 Mhz (I have these)

We can repeat these benchmarks after replacing 8088 by V20 and 8086 by V30. Additionally I have some machines with V40 from Olivetti, but they might not have speed advantages compared to V30 at the same clock speed.

But which benchmark to use? Landmark? Norton SI? Checkit 3? Or something "real"?
 
Heh, that's one question that hasn't changed in perhaps more than 50 years.

To misquote an oft-misattributed chestnut: There are lies, damned lies and benchmarks... :)
 
1ST1 said:
We might do some benchmarks to see the real differences. But maybe they are already available somewhere. We can do those:

5150 with 8088 or Olivetti M19 with 4,77 Mhz
IBM AT 286-8 or Olivetti M28 with 80286-8 (I still look for one of these)
Some turbo XT with 8088 at 8, 10, 12 MHz
Olivetti M24 /AT&T 6300 with 8086 (16 Bit) at 8 Mhz
Olivetti M24SP and Olivetti M240 with 8086 at 10 Mhz (I have these)

We can repeat these benchmarks after replacing 8088 by V20 and 8086 by V30. Additionally I have some machines with V40 from Olivetti, but they might not have speed advantages compared to V30 at the same clock speed.

But which benchmark to use? Landmark? Norton SI? Checkit 3? Or something "real"?
Click to expand...

TOPBENCH was created to answer these questions -- and in fact, already has. There are over 200 systems in its database that clearly show which systems are faster than others, and why. I wrote it to obsolete the other benchmarks you just mentioned.
 
1ST1 said:
But which benchmark to use? Landmark? Norton SI? Checkit 3? Or something "real"?
Click to expand...
Trixter said:
TOPBENCH was created to answer these questions -- and in fact, already has. There are over 200 systems in its database that clearly show which systems are faster than others, and why. I wrote it to obsolete the other benchmarks you just mentioned.
Click to expand...
https://xkcd.com/927/
 
BTW, it should be noted that - contrary to what Landmark says - according to Topbench, late turbo XTs can be faster than early ATs:
topbench_000.jpg
That "11 MHz" must be a typo/bad measurment, safe bet it's actually 12 MHz - pretty common for those "Juko Super Turbo" boards.
I've seen XTs running at 4.77, 7.16, 8, 9.54, 10, 12 MHz, but never a 11 MHz one.
 
Again, benchmarks are, as they always have been, extremely subjective.

Is a Turbo XT faster with disk I/O than a slow PC AT? How about scrolling the CRT using rep movsw instructions? Floating point computation?

Do you see what I mean?
 
Xacalite said:
That "11 MHz" must be a typo/bad measurment, safe bet it's actually 12 MHz - pretty common for those "Juko Super Turbo" boards.
I've seen XTs running at 4.77, 7.16, 8, 9.54, 10, 12 MHz, but never a 11 MHz one.
Click to expand...
One thing I wonder about those 12+ MHz V20/V30 machines, is do they actually run the ISA slots at the full clock speed? Some cards are already quite flaky running at 9.54 MHz.

And if you really want the fastest XT-class CPU, the Olivetti Quaderno subnotebook and HP OmniGo 100 and 120 palmtops have a 16 MHz NEC V30HL.
 
You must log in or register to reply here.
Back
Top