mpogue@dg.dg.com (Mike Pogue) (05/01/90)
In article <38407@mips.mips.COM> mash@mips.COM (John Mashey) writes: >I believe you can find similar effects. For example, unless I'm mistaken, >a DG AV 410 costs >= an AV 310, but (according to most recently published >SPEC data), has a lower SPEC rating. Correct. The AV410 has basically the same processor and memory, but provides additional capabilities that the 310 does not have: VME bus (2 6U slots) expandable graphics, up to 24b/p optional Z buffer capability optional second CPU Because of additional buffering involved in supporting all these additional options, access to memory is slightly slower, resulting in a slightly higher SPECmark rating for the less configurable machine. This is more than made up for (at least on the SPEC programs) when you add a second CPU. (Of course, this gives you more SPECthruput, but not more SPECmark performance. In most cases, where you are running more than one task, this is not important, and the machine will give the performance you would expect with 2 processors). Mike Pogue Data General I speak for myself, not my employer....
Publius@dg.dg.com (Publius) (05/02/90)
In article <38407@mips.mips.COM> mash@mips.COM (John Mashey) writes: > >Seriously, one of the two machines has considerably more expandability >in memory, I/O capacity, and additional CPUs, and you pay for all of that. >This is hardly news, and in fact, if you study your own (i.e., DG pricing), >I believe you can find similar effects. For example, unless I'm mistaken, >a DG AV 410 costs >= an AV 310, but (according to most recently published >SPEC data), has a lower SPEC rating. > >In these days when the CPUs aren't much different from top to bottom, >other differences are what cost. >-- Thanks for your reply, especially for your excellent and insightful comments. You are absolutely right. Customers not only pay for the benchmark numbers. They are willing to pay much more for things which are not that tangible, such as expandability and compatibility, with good reasons. As the industry standard computing systems grow bigger, more sophisticated, and more popular, these systems should be evaluated NOT just as fast stored-program calculators. Can the system grows with customers' need? What kind of support is provided by the vendors? Do they hold your hands or just dump it on your doorstep and you and your brother the guru would get it to run your applications? Do they use industry standard I/O buses or use a proprietary I/O bus in the system? These should be factors based on which the systems should be evaluated, in addition to the benchmark numbers. As the comparison between DS5400 and DS5810 examplifies, even the highly respected SPECmark numbers only tell a small part of the story. We have seen that old Dhrystone get used beyond proper context. Let us make sure that SPECmark does not get mis-used in the same manner. -- Disclaimer: I speak (and write) only for myself, not my employer. Publius "Old federalists never die, they simply change their names." publius@dg-pag.webo.dg.com
gerry@zds-ux.UUCP (Gerry Gleason) (05/02/90)
In article <38407@mips.mips.COM> mash@mips.COM (John Mashey) writes: >In article <414@dg.dg.com> publius@dg-pag.webo.dg.com (Publius) writes: >>Maybe I am just being naive, but I just don't understand this. >>A DECsystem 5400 has the SPECmark of 11.8, and the base listed price >>is $30,000. A DECsystem 5810 has the SPECmark of 11.3, and the base >>listed price is $75,000. Why is a system that has a lower SPECmark >>selling for TWO AND A HALF times of the price of a system with a >>higher SPECmark? Can anyone give me an answer? >Pay for the weight. :-) Something like AT&T's 3b line, the 3b2, 3b5 and 3b15 were all base on the 32100 chip set. The 3b2's were basically desktops with a more or less minimal I/O bus, but the 5 and 15 were meant to be higher performance systems, and had more "advanced" I/O systems (and of course more room for boards, disks, etc.), but advanced almost always turns out to be just complex with longer development cycles for follow ons. Pretty soon not only were the 3b2's smaller and cheaper to make, but they were faster as well. It seems that the industry needs to keep learning this lesson. Gerry Gleason
cory@3letter.MV.COM (Cory Kempf) (05/04/90)
pchen@dg.dg.com (Paul Chen) writes: >Maybe I am just being naive, but I just don't understand this. >A DECsystem 5400 has the SPECmark of 11.8, and the base listed price >is $30,000. A DECsystem 5810 has the SPECmark of 11.3, and the base >listed price is $75,000. Why is a system that has a lower SPECmark >selling for TWO AND A HALF times of the price of a system with a >higher SPECmark? Can anyone give me an answer? A computer is more than just a chip... The biggest difference between the two systems is not the CPU, but the BUS. If I remember correctly, the 5810 is supposed to produce 18 VUPS, while the 5400 is supposed to do 16 VUPS. The 5810 is supposed to have a MUCH higher I/O bandwidth than the 5400, as well as the possibility of ADDING a second (and in the future 3rd and fourth) processor card to generate 36 VUPS (or 54, or 72). Ultrix needs to be fixed before more than 2 CPU's can be present. The Bus gives out at 4 cards. The same box with VAX processor cards can handle six CPU boards. +C -- Cory Kempf I do speak for the company (sometimes). Three Letter Company 603 883 2474 email: cory@3letter.mv.com, harvard!zinn!3letter!cory