• Please review our updated Terms and Rules here

Who has the slowest computer on the forum?

S

silcreval

Experienced Member
Joined
Mar 15, 2009
Messages
73
Location
UK
I'll kick things off with my Microtan 65. Its clocked slightly slower than most 6502 based machines at 750KHz. This was a trick to avoid noise on the graphics output.
 
How about a Science Fair Microcomputer Trainer?
TMS1100 @ 400KHz, ROM programmed with a kind of low-level interpreted language.
Display on a seven-segment and 8 LEDs.
You can type in programs and execute them.
Enough for very basic games, or to play tunes.
Planning on bringing 1 or 2 to VCF Midwest next week
 
If anyone has something like a Kenbak-1 or replica that's also a serial machine, that despite a 1mhz-ish basic clock pulse frequency has an average IPS of under a thousand.

But I guess that points to the problem of how you measure "slowest computer", because the input clock to the CPU is almost meaningless unless you're comparing computers with the same hardware. For instance, a 6502 at 750khz is certainly probably going to be slower than an Apple II or Commodore PET, which run the same CPU at about 0.9 and 1mhz respectively, but how is it going to compare to a TRS-80 Model I running a Z-80 at 1.77mhz? The 6502 is designed so it can do a bus operation every clock cycle, but the Z80 only accesses memory every three or four clock cycles (depending on if it's an instruction fetch or a date/operand read/write operation), which is why you often see people claiming that a 1mhz 6502 is the equivalent of a 4mhz Z80, and herefore implying your MicroTan 65 should be faster than the TRS-80?.But then it's not that simple, because the number of machine cycles different instructions actually take are all over the map on the two CPUs and the Z80 simply has many more instructions, some of which are the equivalent of a significant loop of code on the 6502, so depending on what you're trying to do and how it was coded the Z80 might actually be able to give the 6502 at least a run for its money if they're clocked similarly...

Of course now I'm waiting for someone to slap their Digi-Comp I down on the table while yelling "CHECKMATE!".
 
Here is a reason to run a computer slowly: a simple diagnostic CPLD can capture the address and data bus of a slowly running computer then transmit the address and data values via high speed serial port. This way every instruction execution can be captured without complex storage and recall diagnostic hardware. I have designed such diagnostic module consisted of single 64-macrocell CPLD (EPM7064S) for RC2014 computer that clocks RC2014 at 7Khz and serially transmit address&data at 460k bps.
Bill
 
The one between my ears. Clock rate on the order of milliseconds.
Besides, clock speed related to performance is at best, a false premise. Consider a CPU with a serial ALU running at a clock speed of 4 times the speed of a comparable parallel ALU CPU.
I believe that one of the 6809 architects stated something to the effect of "If we thought people were judging performance based on clock speed, we would have built the 6809 with a waveguide".
 
Chuck(G) said:
The one between my ears. Clock rate on the order of milliseconds.
Besides, clock speed related to performance is at best, a false premise. Consider a CPU with a serial ALU running at a clock speed of 4 times the speed of a comparable parallel ALU CPU.
I believe that one of the 6809 architects stated something to the effect of "If we thought people were judging performance based on clock speed, we would have built the 6809 with a waveguide".
Click to expand...
Is it Turing complete?
 
I believe that I have the slowest computer. Its my instruction speed accurate FPGA implementation of the Manchester Baby or SSEM. It runs at a clock speed of 100Khz and is a serial machine with a 32 bit word. Each memory access or arithmetic operation takes 360 micro seconds per word, so each instruction needs four of these giving an instruction time of
1.2 milliseconds. It can't add, just subtract, so each add uses five of these giving an add time of 60 milliseconds. You can read about it here:-

hackaday.io

BabyBaby or Extremely Small Experimental Computer

I am both a Vintage Computer Enthusiast and a Volunteer who demonstrates the Baby replica at the Museum of Science Industry in Manchester. I wanted to learn more about the Baby and FPGA CPU implementations. The result is this implementation of the Manchester Baby in VHDL on a Diligent Nexys 2...
hackaday.io hackaday.io

.. although I expect its still fastrer than Chuck
 
g4ugm said:
I believe that I have the slowest computer. Its my instruction speed accurate FPGA implementation of the Manchester Baby or SSEM. It runs at a clock speed of 100Khz and is a serial machine with a 32 bit word. Each memory access or arithmetic operation takes 360 micro seconds per word, so each instruction needs four of these giving an instruction time of
1.2 milliseconds. It can't add, just subtract, so each add uses five of these giving an add time of 60 milliseconds. You can read about it here:-

hackaday.io

BabyBaby or Extremely Small Experimental Computer

I am both a Vintage Computer Enthusiast and a Volunteer who demonstrates the Baby replica at the Museum of Science Industry in Manchester. I wanted to learn more about the Baby and FPGA CPU implementations. The result is this implementation of the Manchester Baby in VHDL on a Diligent Nexys 2...
hackaday.io hackaday.io

.. although I expect its still fastrer than Chuck
Click to expand...
wow - thats going to be hard to beat.
 
I have a Commodore 8032 PET on the workbench with a crystal oscillator that doesn't work - so this is a DC 6502. I am still waiting for it to execute the "LDX #$FF" instruction at address $FD16... Does that count?

Dave
 
A friend of mine had a collection of electromechanical computer logic relays possibly from before the time of electronic computing. He made an addition module with them but told me that it was abysmally slow, so an entire computer using that technology would win outright no doubt. I once made a binary counter just using cold cathode trigger tubes in a series of linked flip-flops. It was pretty to watch working but not very practical given that the 400 volt DC supply had to be kept very stable within narrow tolerances for it to work correctly. It is amazing that the oldest working digital computer, the Harwell Dekatron Computer also known as WITCH, uses this type of technology reliably although dekatrons were made to closer tolerances than the extremely cheap trigger tubes I used and WITCH also uses hot cathode tubes and relays.

I would ask whether my replica Honeywell 200 counts as it is taking an incredibly long time just to bootstrap it out of my mind into the physical world. However this autumn (That's "fall" to you cross-pondlife) I intend to complete the register increment/decrement logic module, so it will certainly count then assuming that I got the design right.
 
I forgot to mention my Dataplex machine, which is not actually a computer because it can't do arithmetic but is an early word processor using magnetic cards for offline storage. It is however to some extent programmable. Its basket electric typewriter wore out terminally years ago and I dismantled it but I still have the processor with its two integral card drives. I mention this device because the internal memory is made up of long bucket-brigade style serial shift register ICs which constantly circulate the data, only the input and output ends of each bucket brigade being accessible, so the processor has to wait for the right byte to come around to process each one and moving data from one place to another is therefore incredibly slow. I seem to recall that each such register IC in the machine holds 256 bytes serially as 2k bits. This method of working was adequate for a word processor working at the speed of a slow basket typewriter, so not even a daisy wheel or golf ball type, but far too slow for any other operation. Probably the only documentation for it now is the original patent, a reference to which I may have somewhere, but maybe someone else knows better.
 
I probably have a single-board something-or-other that is slower, but my MITS Altair 680b boxes are clocked at 500 khz.
 
You must log in or register to reply here.
Back
Top