This is Next-and-the-competition in view mode; [Up]
Date: Sun 03-Jan-1989 19:16:54 From: Unknown Subject: Re: Next and the competition in article <2596@udccvax1.acs.udel.EDU>, anand@vax1.acs.udel.EDU (Anand Iyengar) says: > I recall reading a report from Motorola which says that the 68030 can > perform 50% better than than the 68020 (2 cycles/mem-access vs. 3 for the > 68020). Ideally, yeah, and that's ignoring the data cache effects. If you wanted that '020 to run UNIX, you might add a 68851, which adds a wait state, so in more practical terms, the '030 is probably better than 100% faster than an '020 at the same speed. If you let it be. The NeXT has slower system memory, and so takes advantage of another '030 feature, burst fill of cache lines. This lets it eat 4 longwords in 9 cycles (according to the reports of the NeXT I've read, at least), though writes are probably more like 3 or 4 clocks long. In other systems I've found burst mode to increase system speed by about 20% in practice (you don't always use all 4 longwords), but that depends on the memory speed difference between burst and non-burst. With similar memory systems I'd expect an '030 system to very easily outrun a '386 system, especially running UNIX. However, that may not be the real-world case -- there are a number of high-performance '386 system, using external cache and faster DRAM, that may give NeXT a run for the money, at least in terms of CPU performance. Since many of these are based around the otherwise primitive PC-AT architecture, I'd bet that a suitably equipped NeXT would still come out far ahead in I/O performance, though under UNIX, you might not get to see the difference very often. > Mac OS still doesn't have multitasking -- which the Amiga does (and has for > quite a while I'm still waiting to see what Commodore makes of the 2500. > Could be interesting). Glad you think so! > Anand Iyengar.
Date: Sun 04-Jan-1989 14:40:17 From: Unknown Subject: Re: Next and the competition Running the dhrystone test (version 1.1), our Sequent Symmetry (80386 at 16 mhz) with write-back cache is about 6% faster than my NeXT. I suspect some of the difference is a result of better compiler technology on the Symmetry. >From: herrmann@hpbblb.HP.COM (Andreas Herrmann)
These are the contents of the former NiCE NeXT User Group NeXTSTEP/OpenStep software archive, currently hosted by Marcel Waldvogel and Netfuture.ch.