Page 11 of 11 FirstFirst ... 7891011
Results 101 to 102 of 102

Thread: MicroPDP-11 - some questions

  1. #101


    Quote Originally Posted by commodorejohn View Post
    ... evidently there was a major improvement in microcode efficiency between the KDF11 and the DCJ11, even without the increased clock speed and cache (again, if I'm reading this correctly!)
    The improvement was indeed vast. Real experience gains of 2x to 10x could be attained by swapping an 11/23 CPU out for a J11 CPU - often without changing any other hardware.

    Examples of this could be seen in our CAD PCB Layout system, and later on a VENIX based development system. Both were 11/23 CPUs swapped out for one of my KDJ11-A boards, before FPJ11.

    • The CAD system was easily 3x as fast with the J11, and many operations were 4x. [I had to literally pry my CPU out of their hands to get it back]

    • The VENIX system, was an older development system being abandoned in favor of 386/486 PCs for firmware development. Swapping it's 11/23 CPU for one of my J-11s resulted in so much improvement that there was a slight advantage to using the VENIX system over a department of PCs.

    Ultimately, the PCs won out because of collateral reasons having to do with the "Networking" direction of the product line, and not the pure engineering tasks.

    But it did cause re-evaluation.

    These were both examples running on QBUS.

    UNIBUS to QBUS comparisons tended to add an additional gain of 2x or more when the system was an 11/24. Our 11/24 based "Full Price" system was handily outperformed by the "Mid-Price" 11/23 system [based on a VT103] when a KDJ11 was installed instead of the 11/23.

    Additional performance gains were attained by swapping system disk types for newer ones. [Same OS and product build]

    We did do an amusing evaluation one evening that would have been completely absurd for most of our customers. We configured our "Full Price" system with it's UNIBUS video adapter on our 11/70 to take measurements.

    The 11/70 with MASSBUS connected RM03s was 2x faster than the 11/73 with RQDX2 MFM disks. Moreover, the 11/70 showed almost no impact when "loaded" with traffic, as compared to the 11/73.

    To this day, I would still like to perform this evaluation of that old system and software on a PDP-11 being emulated on a PC using SIMH or ERSATZ-11.

    One of these days, I'll get around to it.

  2. #102


    Quote Originally Posted by mnbvcxz View Post
    MIPS may not be a perfect measure of processor speed, but it is better than nothing which is what replaced it.
    Sorry, didn't see this until now.

    You're right, in that cross-platform yardsticks of the day weren't standardized. The practice of needing them was only catching on at the end of the 80's, and really gathered momentum once POSIX was adopted. [about the time when the term "MIPS" came into common use]

    Until then, migration of an application to higher performance platforms often required tremendous pain. The success of the PDP-11 family approach, to re-spin hardware and get gains from one generation to the next without requiring software to be re-built to take advantage of the newer hardware, was a fundamental reason for DEC's success. (Many other vendors did not accomplish this)

    Likewise, the "measures" tended to families. Until an OEM was so unhappy with it's present platform that it was willing to re-engineer their product, often from the ground up to move to a new one - was the common yardstick needed. [Marketing excepted]

    With the advent of UNIX, retaining major portions of an application's engineering was a possibility. More than any other factor, this would assure UNIX's longevity. I find it ironic that UNIX first succeeded on PDP-11s, and perhaps it's no coincidence.

    DEC's approach for PDP-11's failed badly at the juncture of VAX migration for a large percentage of OEMs. Once DEC offended these companies, they had little reason to re-engineer to VAX over another platform, particularly when faced with the cost trends of PCs vs VAX. Ultimately, many would take the leap and deal with negative performance impacts, because of DEC's stance. If they persisted a year or two, they could be rescued by intel's evolution of the PC engine.

    Hardware of the 90's was to see adoption of standards that drove the world forward and influenced development from then on. By the end of the 90's, no hardware vendor could succeed without adherence to performance measurement standards. This was not so for the 80's and unheard of as an issue, prior to that.


Posting Permissions

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