Image Map Image Map
Page 5 of 6 FirstFirst 123456 LastLast
Results 41 to 50 of 53

Thread: VAX 7000-640 CPU frequency

  1. #41
    Join Date
    May 2018
    Location
    Lancashire
    Posts
    260

    Default

    The PC revolution was also helped by the 'portability' of the software from both expensive and cheap hardware.

    Company IT buyers bought IBM at first, then HP/Compaq etc on a professional level, then staff bought cheap clones at home and took all the disks from work to run at home. Everyone had word perfect then MS word at home. Even office machines were full of copied software.

    Certainly, I wanted a PC at home because it became the obvious choice. Remember it influencing buying decisions in work.

  2. #42

    Default

    Quote Originally Posted by vol.litwr View Post
    Wow! I have 7 entries for the VAX family in my benchmark table. Other entries:

    <SNIP>

    PDP-11 - 7
    Let me see what I can do about this one! Here are some numbers for the EIS version of the RT-11 benchmark on a PDP-11/34A:

    100 0.50
    1000 36.65
    3000 321.43
    4182 822.70

    This machine is an 11/34A w/ a full 124kW of RAM running RT-11 5.01 (XM monitor). I'll try to get the non-EIS benchmark for you in a bit.
    Last edited by elb; July 28th, 2019 at 01:34 PM.

  3. #43

    Default

    And here are the numbers for the non-EIS version:

    100 1.42
    1000 122.35
    3000 1090.27
    4728 2703.13

  4. #44

    Default

    Quote Originally Posted by elb View Post
    And here are the numbers for the non-EIS version:

    100 1.42
    1000 122.35
    3000 1090.27
    4728 2703.13
    Thank you very much. However I have only been able to partially update the tables because I need more data about the PDP-11/34A CPU. I could find that it is the KD11-EA but what is its frequency? I have checked the manual, it gives about 400 ns for memory access time, so I assumed that CPU has ≈5 MHz frequency. Is it correct?
    I also do not completely understand the results for large number of digits. The EIS 4182 digits is impossible and has to be rounded up to 4184 and the result (822.70s) looks implausible - it should be about 625s for this number of digits. The non-EIS 4728 digits looks ok and 2703s looks good too. So I can think that EIS data is a bit incorrect. The pi-calculator can get more than 6500 digits using the PDP-11. Did you try it with 6000 digits or more?

  5. #45

    Default

    Quote Originally Posted by vol.litwr View Post
    Thank you very much. However I have only been able to partially update the tables because I need more data about the PDP-11/34A CPU. I could find that it is the KD11-EA but what is its frequency? I have checked the manual, it gives about 400 ns for memory access time, so I assumed that CPU has ≈5 MHz frequency. Is it correct?
    I also do not know the answer to this question, and have not been able to find a good answer. I believe that this is because it is a heavily microcoded CPU and the clock speed is not particularly informative, so DEC did not advertise it as such. I also believe that it is about 5 MHz based on other things that I have read. It is on my TODO to do some probing around on the CPU boards and confirm this, but perhaps someone else here knows the answer.

    Quote Originally Posted by vol.litwr View Post
    I also do not completely understand the results for large number of digits. The EIS 4182 digits is impossible and has to be rounded up to 4184 and the result (822.70s) looks implausible - it should be about 625s for this number of digits. The non-EIS 4728 digits looks ok and 2703s looks good too. So I can think that EIS data is a bit incorrect. The pi-calculator can get more than 6500 digits using the PDP-11. Did you try it with 6000 digits or more?
    You are correct, I made a data entry error and it is 4812 digits, not 4182 digits. Hopefully that makes more sense! I used 4812 and 4728 because it gave me a message like this:

    Code:
    NUMBER OF DIGITS (UP TO 4812)? 4812
    I had assumed that this was limited by (e.g.) available RAM (since the NOEIS version was slightly less, and the binary image for this was slightly larger). I was running under the XM monitor, I could probably get a few more digits by running under the SJ monitor if this is the case -- but I would be surprised to see 6000. This particular machine has maximum possible physical ram, which perhaps requires more tracking structures in the OS itself, and a large TT driver to handle multiple serial ports as well as some loaded modules that can possibly be unloaded (or I can try booting from a floppy with a simpler configuration).

    I won't be able to look at this until near the end of the month, unfortunately, but I'd be happy to revisit at that time.

  6. #46

    Default

    Quote Originally Posted by elb View Post
    I also do not know the answer to this question, and have not been able to find a good answer. I believe that this is because it is a heavily microcoded CPU and the clock speed is not particularly informative, so DEC did not advertise it as such. I also believe that it is about 5 MHz based on other things that I have read. It is on my TODO to do some probing around on the CPU boards and confirm this, but perhaps someone else here knows the answer.

    You are correct, I made a data entry error and it is 4812 digits, not 4182 digits. Hopefully that makes more sense! I used 4812 and 4728 because it gave me a message like this:

    I had assumed that this was limited by (e.g.) available RAM (since the NOEIS version was slightly less, and the binary image for this was slightly larger). I was running under the XM monitor, I could probably get a few more digits by running under the SJ monitor if this is the case -- but I would be surprised to see 6000. This particular machine has maximum possible physical ram, which perhaps requires more tracking structures in the OS itself, and a large TT driver to handle multiple serial ports as well as some loaded modules that can possibly be unloaded (or I can try booting from a floppy with a simpler configuration).

    I won't be able to look at this until near the end of the month, unfortunately, but I'd be happy to revisit at that time.
    I hope that you can find the quartz which gives the frequency for you system. It should be somewhere among other chips.

    Indeed 4812 sounds quite ok. An EIS program allows calculating more digits because a non-EIS version requires additional code to emulate MUL and DIV and so it is a bit larger and capable to produce less digits . Did you try to run RT-11SB or SJ with your iron? It has to give the most number of digits. A man used RT-11SB with his microPDP-11/83 and could get more than 6500 digits. Anyway thank you for your help.

  7. #47
    Join Date
    Feb 2011
    Location
    NorthWest England (East Pondia)
    Posts
    2,246
    Blog Entries
    10

    Default

    Quote Originally Posted by vol.litwr View Post
    I hope that you can find the quartz which gives the frequency for you system. It should be somewhere among other chips.

    Indeed 4812 sounds quite ok. An EIS program allows calculating more digits because a non-EIS version requires additional code to emulate MUL and DIV and so it is a bit larger and capable to produce less digits . Did you try to run RT-11SB or SJ with your iron? It has to give the most number of digits. A man used RT-11SB with his microPDP-11/83 and could get more than 6500 digits. Anyway thank you for your help.
    Its not true that there will be a crystal that's at the fundamental frequency of of the system. Often the master clock runs a lot faster and is used to generate overlapping clock signals. A severe example is my BabyBaby where the master clock runs at over 100Mhz but the bit rate is 100Khz and a word access (its a serial machine) takes 360 miliseconds.

    Looking at the PDP/11 34 manual (the 11/34a runs at the same speed,)

    http://www.bitsavers.org/pdf/dec/pdp...nual_Dec76.pdf

    the machine at uses multiple phases and the length of a micro cycle is either 180 or 240ns so 5.5 or 4.1Mhz but its microcoded so a real instruction might take multiple steps...
    Dave
    G4UGM

    Looking for Analog Computers, Drum Plotters, and Graphics Terminals

  8. #48
    Join Date
    Sep 2019
    Location
    Zurich, CH
    Posts
    27

    Default

    Quote Originally Posted by vol.litwr View Post
    Thank you very much. However I have only been able to partially update the tables because I need more data about the PDP-11/34A CPU. I could find that it is the KD11-EA but what is its frequency? I have checked the manual, it gives about 400 ns for memory access time, so I assumed that CPU has ≈5 MHz frequency. Is it correct?
    It's usually not helpful to look at memory access times to find CPU frequency on PDP-11 systems. Memory was usually on the Unibus, which is asynchronous. So memory could have all kind of access speeds, and have no relation to the CPU speed.
    That said, unless the CPU have cache, it might then get stalled quite a lot of memory is slow. But different memory technologies had different access times. Core memories also, for example, are usually slower reading than writing (because any read has to be followed by a write).

  9. #49

    Default

    Quote Originally Posted by bqt View Post
    It's usually not helpful to look at memory access times to find CPU frequency on PDP-11 systems. Memory was usually on the Unibus, which is asynchronous. So memory could have all kind of access speeds, and have no relation to the CPU speed.
    That said, unless the CPU have cache, it might then get stalled quite a lot of memory is slow. But different memory technologies had different access times. Core memories also, for example, are usually slower reading than writing (because any read has to be followed by a write).
    Thanks. So this frequency is rather an elusive matter. Maybe only IBM mainframes are more cryptic.

  10. #50
    Join Date
    Feb 2011
    Location
    NorthWest England (East Pondia)
    Posts
    2,246
    Blog Entries
    10

    Default

    Quote Originally Posted by vol.litwr View Post
    Thanks. So this frequency is rather an elusive matter. Maybe only IBM mainframes are more cryptic.
    On any big machine with multi-ported memory it becomes an irrelevance for real workloads. For example you could bring a Honeywell L66 or DPS8 to its knees by writing long blocks to tape. The tape controller needed to be supplied with data at the rate it was written to tape, it therefore always had priority access to memory and could lock the CPU out. The result was you ended up with a CPU that was 100% busy but executing no instructions per second because it was locked out of memory.

    The ICL 2900 has a similar architecture to the L66 and IBM 360 and I presume it behaves in the same way...
    Dave
    G4UGM

    Looking for Analog Computers, Drum Plotters, and Graphics Terminals

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •