John Archdeacon (jarch++at++gemtech.com)
30 Apr 1996 10:44:48 -0800
Michael Jones wrote the following regarding the proposed OpenGVS Benchmarks to the Performer news group:
> Well, considering how little your long OpenGVS advertisement had to do with
> Performer ... I'll go ahead and forward it (to info-perfomer) since you say it's OK.
Clarification: I was *not* sending out any advertisements as I respect the spirit of a technical news group such as info-performer; Gemini was simply responding to a couple of emails from *Performer users* that brought up questions about cross platform solutions and an OpenGVS *direct reference* from the United States Army.
Michael Jones also wrote:
> The GVS versions only speak to GVS speeds on the machine.
> They do not measure the machine. That would be like a ISV PHIGS
> company ranking workstations on drawing a car model based on their PHIGS
> implementation... There is no need for GVS code, for binaries, etc. Just the
> databases and example information so that people porting the tests to other
> environments can make the test measurements fairly.
First, OpenGVS does not use PHIGS as you know; it uses OpenGL for machines which natively implement it (SGI workstations, Intergraph Pentium workstations, Pentiums with 3DLabs GLINT accelerators, and so on). As you also know, OpenGL was invented by SGI and we use it because it is *good* and it offers users (OpenGVS and others) great *portability*. An OpenGVS implementation on any OpenGL machine is *very* close, making OpenGVS (and the proposed benchmarks) an ideal environment (in our minds) for evaluating *relative* performance between different systems since the underlying implementation, databases, and application software are essentially *identical* (note: we do conditionally use OpenGL extensions on systems like RE2 or Integraph if those extensions improve graphics performance...). Although we believe OpenGVS to be extremely efficient with a proper utilization of the underyling OpenGL graphics hardware, we recognize and have never claimed that these benchmarks represent the final word in performance f
or any system, just *relative* performance where the popular and well recognized performance of a single-pipe, single-CPU RE2 represents our normalized 1.0 metric in a realistic "workload" (not simply spinning tea pots or single car models). This simple foundation using common workload is the *basis* of this proposed benchmark (at least Version 1.0).
Second, we are not proposing rendering a car model; we are proposing real-world benchmarks where the OpenGVS is used (not just a database) so one *can* measure relative performance easily since we have near *identical* workload. In case you haven't had a chance to read the final benchmark proposal (which should help clarify the intent of the paper), it has been published on the web at:
http://www.gemtech.com/rwbench/proposal.html
Michael Jones wrote:
> I have accepted the head-to-head test that you proposed.
> Are you backing out of it now? I don't see why you would do that after making
> such a big splash in the Performer mailing list with your original announcement.
> I hope that I misunderstand your new position. I urge you to follow through
> with the plan to make the databases and eyepoint/vehicle information avaliable
> as your test suite...
First, this is not a contest between OpenGVS and Performer; sorry you see it that way. Again, just to make sure we are communicating, Gemini will also immediately *stop* responses to your private Performer news group when there are no more open forum questions/responses made to Gemini and our products. I offered to do this in my last response as you know...
Second, Gemini has *not* changed to a new position as you have suggested. The version 1.0 benchmark suite as proposed (again, the paper is on the web) will be designed and implemented using OpenGVS. Any "version 2.0" extensions to the benchmark suite can and will be addressed later and SGI's (and other) comments or suggestions can be considered later. Thanks for your suggestions though.
Cheers!
John Archdeacon
This archive was generated by hypermail 2.0b2 on Mon Aug 10 1998 - 17:52:49 PDT