home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.amiga.applications
- Path: sparky!uunet!usc!cs.utexas.edu!swrinde!sdd.hp.com!hp-col!fc.hp.com!koren
- From: koren@fc.hp.com (Steve Koren)
- Subject: Re: a4000 anims & video
- Sender: news@fc.hp.com (news daemon)
- Message-ID: <Bx8zC5.BLr@fc.hp.com>
- Date: Thu, 5 Nov 1992 14:38:29 GMT
- Distribution: comp
- References: <Bx7ALr.2uC@world.std.com>
- Organization: Hewlett-Packard Fort Collins Site
- X-Newsreader: Tin 1.1.3 PL5
- Lines: 33
-
- Joel B Couch (jbciii@world.std.com) wrote:
-
- > I've heard it said that a 50 MHz 68030 with a 50 MHz 68882 has
- > greater floating point processing power than does a 25 MHz
- > 68040. The reason for this, beyond the higher clock speed,
-
- Only in a few cases, such as transcendental math functions. "Normal"
- floating point math will be faster on a 25 MHz 040 than a 50 MHz 882.
- Consider that for most floating point benchmarks the 25 MHz 040 is
- between 2 and 6 times as fast as a 25 MHz 882. So if you double the 882
- speed, you'll still get between 1 and 3 times the performance on the
- 040, providing you have reasonably good floating point code. (See the
- AIBB results for FP comparisons between 040 and 882 systems.)
-
- If you rely heavily on transcendental math, the fast 882 is the better
- bet. But that is a very uncommon thing. Even 3D rendering engines like
- ray tracers and such use very little or no transcendental math. And
- even programs which seem like they must be doing a lot of transcendental
- math probably spend a fairly small percentage of their time doing that.
- I bet that over 95% of all FP intensive software will be faster on an
- 040 than a 882 which is twice the clock speed. If you *really* need the
- hottest FP performance, the 68K family is pretty far from the cutting
- edge of the performance curve anyway. If FP is your hot item, get a
- workstation of some sort and use it as a compute server for your Amiga.
- You'd need custom software on both sides to make it fly, but if you need
- that much speed, you can probably arrange the software end of it.
-
- Another factor is that 50 MHz is likely to be as fast as 882's ever get.
- But 25 MHz isn't the fastest 040. I believe there may already be 33 MHz
- parts, and I'd expect a 40 Mhz version eventually. (I don't think we'll
- see 50 MHz, but who knows. I also have my doubts about the 68060).
-
- - steve
-