news@bmers58.UUCP (news) (01/24/90)
The local dealer (Lynx) did a side-by-side comparison of the GVP and CBM accelerators on Saturday. The test involved using Scupt Animate 4D to wire-frame -> rough ray-trace -> some finished product. With a stock 68000 Amiga the exercise took 8 1/2 hours 100 With the 2620 board 1 1/2 hours 17.6% With the 2630 board at 25 MHz about 40 minutes 7.8% With the GVP 3001 board at 25 MHz a little (maybe 5%) faster 7.4% Now GVP is shipping with 28 MHz and by summer will be shipping with 33 MHz. Obviously, the speed won't scale linearly but it should get even faster. Does anybody know if CBM is going to up their clock at all? Dave? Some thoughts: o You pay a fair bit for the added performace of the GVP board, largely because of the expensive page-mode RAM. See prices below. o It's only a single test and one that is very floating point intensive at that. I would really like to have seen some other tests. In particular, I'd like to see their effect on disk I/O and on multi-tasking; these are the subjective things that the end-user actually feels *while* he's using the box. Letting the box sit in a corner is only partially useful. o I'm still leery of GVP because of the way the advertise their controller as "DMA to an on-board 64K cache" That's not the spirit of DMA!! Now maybe their engineers are more ethical than their marketing types... Prices: CANADIAN! 2630 board with 2M of 32bit RAM CDN$2295 additional 2M for 4M on-board CDN $500 (estimate) daughterboard for additional 2/4MB not released 3001 board with no RAM CDN$1750 with 4M CDN$3795 ($2295 separately) with 8M CDN$4595 ($1000 separately) I'd be partial to the CBM board with these figures. Any comments? Keith Hanlan Bell-Northern Research, Ottawa, Canada 613-765-4645 uunet!bnrgate!atreus!keithh or (via bitnet) keithh@bnr.ca
eapu030@orion.oac.uci.edu (Jason Goldberg) (01/24/90)
In article <1271@bmers58.UUCP> uunet!bnrgate!atreus!keithh (Keith Hanlan) writes: >Now GVP is shipping with 28 MHz and by summer will be shipping with >33 MHz. Obviously, the speed won't scale linearly but it should get >even faster. Does anybody know if CBM is going to up their clock >at all? Dave? There new adds are back to 25 Mhz. > >I'd be partial to the CBM board with these figures. > >Any comments? Particularly if you consider the A2630 offers Unix compatabilty, and the ability to add a 64 MEG daughter board, and the A2630 can be a great solution for the majority of the people. Not to mention, an AT contorller that supports only two device is kinda rinky-dink, in my mind. -Jason-
hue@netcom.UUCP (Jonathan Hue) (01/24/90)
In article <1271@bmers58.UUCP> uunet!bnrgate!atreus!keithh (Keith Hanlan) writes: >Some thoughts: > o You pay a fair bit for the added performace of the GVP board, > largely because of the expensive page-mode RAM. See prices below. I had heard it was nibble-mode RAM. Page-mode RAM doesn't make sense here, the idea behind a nibble mode RAM is to put out an (row and column) address once and grab the next three locations without putting out an address (just a column address strobe), and that's what the '030 does in burst mode. I also heard $1495 for 4M. 1Mx1 Page mode DRAMs are under $10, I thought nibble-mode wasn't much, if any, more expensive. Seems like they're luring you in with this $800 '030 board, then once your locked in, they stick you with an outrageous price for a 32-bit memory board. -Jonathan
bear@bu-pub.bu.edu (Blair M. Burtan) (01/24/90)
Sounds to me like CBM's 030 system comes out on top. This would make getting the 2500/30 system at discounted prices well worth it. Opinions? Horror stories?