roelof@idca.tds.PHILIPS.nl (R. Vuurboom) (06/08/89)
In article <1217@cbnewsh.ATT.COM> wcs@cbnewsh.ATT.COM (Bill Stewart 201-949-0705 ho95c.att.com!wcs) writes: > My Motorola sales rep has been claiming 5 MIPS for a 25 MHz > 68030 VME board, Our 25 Mhz 68030 board is published at 3.2 _TDS_ mips. 0.7 TDS mips translates to 1 VAX mips. So the equivalent VAX figure for our boards is 4.6 vax mips. We've got good reason to believe that the performance of our board is not lower than that of motorola. I'ld better explain this TDS mips business. Most perfomance measurements are done in what we describe as a "scientific environment": relatively long periods between context switching. In our business oriented environment context switching takes place at a higher rate - about 4000 instructions between a context switch - so the caches are going through a cold start phase that much more often. Over a number of applications we've found the following equivalence: 0.7 mips commercial environment = 1.0 mips scientific environment 3.2/0.7 = 4.6 > ...and predicts 15 MIPS for the 68040 in the fall > (presumably at 33 MHz?) Of course, your sales rep may know (much) more than I do but I'ld be surprised if they achieved 15 mips, a 10-20% "smudge" factor may be applicable here. You're correct in assuming that the figure quoted is for a 33 Mhz version. The version slated for the fall is a 25 Mhz version, the 33 Mhz is slated for some months later (according to my info). > Will Edge be doing a performance upgrade similar to that? Edgcore is taking action which should put them "sufficiently above" the 68040. For obvious reasons I can't be more specific. > Or will the 68040 kill them? Well it won't help things now, will it? :-) No, I don't think so. For two reasons, first off, don't forget that the cpu only represents between 5-15% of a full system (which incidentally is what our customers buy) so even if you were to take out the E2000 and replace it with thin air your savings are at most 10%. (still a good bit but thin air processors are not on the market yet). All things being equal this 5-10% difference in price could be a deciding factor, but all things are not equal, things such as networking, extendibility, X-open compatibility easily overshadow this price differential. I guess I'm just restating here the (somewhat dim-witted) observation that a cpu has to be treated as a system component. And thus a component level optimization (cheaper cpu) can lead to a system level "pessimization". Secondly, the E2000 is multiprocessor _now_. I doubt that anybody is going to have a multiprocessor 68040 system up inside of a year from now (or even whether that would be a good idea. To my knowledge nobodys put a successful multi-microprocessor product on the market up to now). In other words, I think the real comparison - system level performance wise - is a 60 now_mips machine with E2000s against a (possibly) 15 sometime_next_year_mips machine with a 33 Mhz 68040 processor. -- Roelof Vuurboom SSP/V3 Philips TDS Apeldoorn, The Netherlands +31 55 432226 domain: roelof@idca.tds.philips.nl uucp: ...!mcvax!philapd!roelof
aglew@mcdurb.Urbana.Gould.COM (06/10/89)
/* Written 1:01 pm Jun 7, 1989 by roelof@idca.tds.PHILIPS.nl in mcdurb:comp.arch */ /* ---------- "Re: E2000 vs 68040" ---------- */ In article <1217@cbnewsh.ATT.COM> wcs@cbnewsh.ATT.COM (Bill Stewart 201-949-0705 ho95c.att.com!wcs) writes: > My Motorola sales rep has been claiming 5 MIPS for a 25 MHz > 68030 VME board, Our 25 Mhz 68030 board is published at 3.2 _TDS_ mips. 0.7 TDS mips translates to 1 VAX mips. So the equivalent VAX figure for our boards is 4.6 vax mips. We've got good reason to believe that the performance of our board is not lower than that of motorola. I'ld better explain this TDS mips business. Most perfomance measurements are done in what we describe as a "scientific environment": relatively long periods between context switching. In our business oriented environment context switching takes place at a higher rate - about 4000 instructions between a context switch - so the caches are going through a cold start phase that much more often. Over a number of applications we've found the following equivalence: 0.7 mips commercial environment = 1.0 mips scientific environment 3.2/0.7 = 4.6 > ...and predicts 15 MIPS for the 68040 in the fall > (presumably at 33 MHz?) Of course, your sales rep may know (much) more than I do but I'ld be surprised if they achieved 15 mips, a 10-20% "smudge" factor may be applicable here. You're correct in assuming that the figure quoted is for a 33 Mhz version. The version slated for the fall is a 25 Mhz version, the 33 Mhz is slated for some months later (according to my info). > Will Edge be doing a performance upgrade similar to that? Edgcore is taking action which should put them "sufficiently above" the 68040. For obvious reasons I can't be more specific. > Or will the 68040 kill them? Well it won't help things now, will it? :-) No, I don't think so. For two reasons, first off, don't forget that the cpu only represents between 5-15% of a full system (which incidentally is what our customers buy) so even if you were to take out the E2000 and replace it with thin air your savings are at most 10%. (still a good bit but thin air processors are not on the market yet). All things being equal this 5-10% difference in price could be a deciding factor, but all things are not equal, things such as networking, extendibility, X-open compatibility easily overshadow this price differential. I guess I'm just restating here the (somewhat dim-witted) observation that a cpu has to be treated as a system component. And thus a component level optimization (cheaper cpu) can lead to a system level "pessimization". Secondly, the E2000 is multiprocessor _now_. I doubt that anybody is going to have a multiprocessor 68040 system up inside of a year from now (or even whether that would be a good idea. To my knowledge nobodys put a successful multi-microprocessor product on the market up to now). In other words, I think the real comparison - system level performance wise - is a 60 now_mips machine with E2000s against a (possibly) 15 sometime_next_year_mips machine with a 33 Mhz 68040 processor. -- Roelof Vuurboom SSP/V3 Philips TDS Apeldoorn, The Netherlands +31 55 432226 domain: roelof@idca.tds.philips.nl uucp: ...!mcvax!philapd!roelof /* End of text from mcdurb:comp.arch */