January 25, 2020 at 8:31 am #40011ParticipantD-Dan@d-dan
Nope, just computer history. Pre BBC micro, was “The Computer Programme”
One of the most powerful computers in the world could process at 50 Million Instructions per second. (This was 1982).
Less than 40 years later, my own CPU (Ryzen 7 1800X) can do 304,510 Million instructions per second. So my desktop is more than 6,000 times faster than one of the fastest computers in the world 40 years ago. How I wish I would be alive in another 40 years to see how far we get (chances of me reaching 95 are slim).
Arch Linux, on a Ryzen 7 1800X, 16 GB, 6 (yes - 6) HDs inc 2 SSDs, 4 RPi 3Bs + 1 RPi 4B - one as an NFS server with two more drives, PiHole (shut yours), Plex server, cloud server, and other random Pi stuff. Nice CoolerMaster case, NV GTX 1060 6GB, and a whopping 32" AOC 1440P monitor.
January 25, 2020 at 10:33 am #40016BlockedRichard@sawbomanForumite Points: 6,508
But, do the modern devices feel that much faster?
I cut my digital teeth reprogramming a Xilog80 powered 6~32 position training simulator. The device probably used an NEC version of the chip running at a breathtaking 0.58 MIPS. The programs had been written in assembler, but we did not have access to a proper terminal on which to write in assembler. We had to use a 6 byte window and poke HEX data through that. The program was written or rewritten out long hand, transcribing it into HEX ready for entry. I learned to save often and then run any tests. That was in about 1979.
After that, the PC I then used was an 8 MHz 8088 powered item. I have no idea of its operations per second, at the time it just did all the magic that I needed, mainly spreadsheets. Though, I claim to be one of the few able to use a terrible word processor of the time for data processing. I think it was called Multimate and modelled on the Wang word processor. Work did not want any nasty cobbled together hand written programs, though that was soon changed, thank goodness.
I have no data on the MIPS of my present CPU, Google let me down, but I doubt that I seriously exercise its capabilities these days.January 25, 2020 at 10:30 pm #40021ParticipantD-Dan@d-danForumite Points: 2,586
I dunno, I cut my teeth on an IBM 1966 and later a 2966 mainframe using COBOL, and we had to leave it to the operators to run what were glorified databases overnight.
Now, I could do the same on my desktop using just SQL in seconds.
Arch Linux, on a Ryzen 7 1800X, 16 GB, 6 (yes - 6) HDs inc 2 SSDs, 4 RPi 3Bs + 1 RPi 4B - one as an NFS server with two more drives, PiHole (shut yours), Plex server, cloud server, and other random Pi stuff. Nice CoolerMaster case, NV GTX 1060 6GB, and a whopping 32" AOC 1440P monitor.January 25, 2020 at 10:54 pm #40023BlockedBoris@borisForumite Points: 1,544
I remember learning COBOL 74 way back in the early-70s when as a trainee accountant I was seconded to our IT department for 12 months because of the results of an aptitude test all the trainees had to do.
I can recall very little of it now, even though I wrote a couple of programs that were still being used 15 years later.
Never trust an atom - they make up everything !January 26, 2020 at 12:02 am #40024ParticipantD-Dan@d-danForumite Points: 2,586
COBOL was practically English, once you figured out the data block, code block etc.
Arch Linux, on a Ryzen 7 1800X, 16 GB, 6 (yes - 6) HDs inc 2 SSDs, 4 RPi 3Bs + 1 RPi 4B - one as an NFS server with two more drives, PiHole (shut yours), Plex server, cloud server, and other random Pi stuff. Nice CoolerMaster case, NV GTX 1060 6GB, and a whopping 32" AOC 1440P monitor.January 26, 2020 at 11:57 am #40032ParticipantWheels-Of-Fire@grahamdearsleyForumite Points: 5,365
MIPS were never a very good measure of performance, especially between processors with different architectures. The main problem is that its not a measure of actual work done. A 64 bit CPU is going to shift 8 times as much data per instruction as an 8 bit CPU would for example and a duel core CPU will execute twice as many instructions per cycle IF the instructions can be executed in parallel. The other problem is choosing which instructions you are going to measure because they all take different numbers of cycles and just taking the average doesn’t tell you very much. An add instruction may take 1 cycle and a multiply 20 but how many add instructions does it take to do a multiply ?
- You must be logged in to reply to this topic.