Image Map Image Map
Page 1 of 5 12345 LastLast
Results 1 to 10 of 46

Thread: PDP-11 π benchmark

  1. #1

    Default PDP-11 π benchmark

    I have a project - http://litwr2.atspace.eu/pi/pi-spigot-benchmark.html. I got a bit surprising results. PDP-11/70 (1975) and PDP-11/83 (1988 ) has the same speed. So I can conclude that KB-11C at PDP-11/70 is clocked to 18 MHz. Is it true? It looks like information about PDP-11/70 frequencies was never published.
    BTW I continue to gather results for other PDP-11 models under RT-11 and Unix. Help is welcome. Some training for old hardware is the best cure for rust and dust.

  2. #2

    Default

    Quote Originally Posted by vol.litwr View Post
    PDP-11/70 (1975) and PDP-11/83 (1988 ) has the same speed. So I can conclude that KB-11C at PDP-11/70 is clocked to 18 MHz.
    I remember testing an 11/73 against an 11/44. Similar speed results for many programs, however the 11/73 CPU is much faster than the 11/44 if you are getting all cache hits, but much slower if you are getting cache misses. So while the balance of performance was in many cases the same, the speed of the CPU and memory subsystems differed a lot.

    I would be very willing to believe the same of your comparison between the 11/70 and the 11/83. They have different CPU technologies, different cache size and organization, different memory interconnects, different memory technology, etc.

  3. #3
    Join Date
    Nov 2014
    Location
    Chicagoland
    Posts
    234

    Default

    Quote Originally Posted by vol.litwr View Post
    So I can conclude that KB-11C at PDP-11/70 is clocked to 18 MHz. Is it true? It looks like information about PDP-11/70 frequencies was never published.
    There is a timing generator for microprogram control on the KB-11C. Timing pluses T1-T5 occur at 60ns intervals from a 33 MHz clock. An R/C maintenance clock or external pulse can also be used.

    Using raw clock speed to compare the PDP 11 types is challenging because each architecture had differences in how each implemented memory, instructions, cache, etc.

    The PDP 11/34a had clock cycles either 180ns or 240ns long, varying by the instruction type. So depending on the actual work load and processor, your mileage may vary .
    Last edited by wa2flq; February 12th, 2017 at 09:20 AM.

  4. #4
    Join Date
    Aug 2010
    Location
    Silicon Valley USA
    Posts
    871
    Blog Entries
    4

    Default

    Quote Originally Posted by wa2flq View Post
    There is a timing generator for microprogram control on the KB-11C. Timing pluses T1-T5 occur at 60ns intervals from a 33 MHz clock. An R/C maintenance clock or external pulse can also be used.

    Using raw clock speed to compare the PDP 11 types is challenging because each architecture had differences in how each implemented memory, instructions, cache, etc.

    The PDP 11/34a had clock cycles either 180ns or 240ns long, varying by the instruction type. So depending on the actual work load and processor, your mileage may vary .
    The 11/70 (and 11/45, same basic design) run the microprogram engine at 150ns per microcycle. The fastest instruction (eg, like a MOV R1,R2 in the cache) takes 2 cycles, or 300ns. If you are looking at the microcycle clock period of 150ns the 'clock' rate of the CPU is then 6.7MHz, even tho the master timing generator runs at 5X this frequency.

    The DEC J-11 Control Chip spec (on bitsaver's) indicates that the maximum J11 clock rate by design is 20MHz (50ns period) and a basic microcycle consists of 4 clock ticks, or 200ns. Like the 11/70 analysis above, even tho the master timing generator runs at 20MHz, the effective CPU clock rate is 1/4 this, or 5.0MHz.

    NEW INFO: Reading the 11/94 user guide, it appears that the J11 can do the cached MOV RX,Ry in just one microcycle, due to aggressive execute/fetch overlap. So the MOV Rx,Ry that takes two microcycles (300ns) on the 11/45-70 takes just one microcycle (200ns @ 20MHz, or 220ns @ 18MHz) on the J11 CPU.

    The fastest hardware PDP-11 ever (still?) I believe is still the PDP-11/55, a 45 with 150ns bipolar memory, so it effectively runs with a 32KW cache.

    So, sadly, it looks like the 11/94 with (effectively) all cache will beat out the 11/55 by about 33% (300ns vs 200ns, 100ns faster).


    Of course running SIMH on a fast CPU now will blow this design away, but that really doesn't count now, does it?

    Don
    Last edited by AK6DN; February 12th, 2017 at 11:47 AM.

  5. #5
    Join Date
    Dec 2010
    Location
    Seattle, WA
    Posts
    2,018

    Default

    Quote Originally Posted by AK6DN View Post
    The fastest hardware PDP-11 ever (still?) I believe is still the PDP-11/55, a 45 with 150ns bipolar memory, so it effectively runs with a 32KW cache.
    How does that memory speed compare with the onboard 2MB or 4MB on an 11/93,94 KDJ11-E? The memory there is supposed to be fast enough that no cache is used. I can't easily find details on how fast the memory actually is.

  6. #6
    Join Date
    Jul 2009
    Location
    Boulder , Colorado USA
    Posts
    3,109

    Default

    Hi All;

    There is some (I think) minor difference between the 11/45 and the 11/55 in the CPU Circuitry, but the 11/45 can use Core or MOS Memory, where as the 11/55 is for mostly MOS Memory, and when putting Bipolar in either then the speed is Greatly Increased compared to MOS Memory..
    The big Difference, besides word size for the 11/45/55 and the 11/70 is with the 11/70 the Cache fetches two Words at a time, where as the 11/45/55 only fetch one Word at a time, thereby making the Memory fetching twice as fast in the 11/70..
    The Timing board in all three Machines is the same speed..

    THANK YOU Marty

  7. #7

    Default

    Thank your very much for precious information. 16.5 MHz looks like a correct frequency of 11/70 for me.
    I'm still curious to get some π-benchmark results. BTW I used PDP-11/70 at LIVING COMPUTERS MUSEUM + LABS.
    Quote Originally Posted by Marty View Post
    The Timing board in all three Machines is the same speed..
    Sorry, I do not quite understand. What is the timing board? Is it quartz?
    Last edited by vol.litwr; February 13th, 2017 at 07:41 AM.

  8. #8
    Join Date
    Jul 2009
    Location
    Boulder , Colorado USA
    Posts
    3,109

    Default

    Hi All;

    "" Sorry, I do not quite understand. What is the timing board? Is it quartz? ""

    The timing board is about a Crystal 33.3 Megahertz.. It is the Speed of the CPU Boards in all three Machines..

    THANK YOU Marty

  9. #9

    Default

    Great project congrats, would love to see the Altair included .

    Quote Originally Posted by vol.litwr View Post
    I have a project - http://litwr2.atspace.eu/pi/pi-spigot-benchmark.html. I got a bit surprising results. PDP-11/70 (1975) and PDP-11/83 (1988 ) has the same speed. So I can conclude that KB-11C at PDP-11/70 is clocked to 18 MHz. Is it true? It looks like information about PDP-11/70 frequencies was never published.
    BTW I continue to gather results for other PDP-11 models under RT-11 and Unix. Help is welcome. Some training for old hardware is the best cure for rust and dust.

  10. #10
    Join Date
    Apr 2017
    Location
    Switzerland
    Posts
    295

    Default

    Where can I get a binary image for RT-11? I would like to benchmark my PDP-11/Hack.
    For more information about my projects see https://www.5volts.ch

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
  •