Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!tut.cis.ohio-state.edu!ucbvax!vax1.physik.fu-berlin.dbp.DE!falk From: falk@vax1.physik.fu-berlin.dbp.DE Newsgroups: comp.sys.transputer Subject: Re:ATW Message-ID: <52:falk@vax1.physik.fu-berlin.dbp.de> Date: 10 Jan 90 19:17:00 GMT Sender: daemon@ucbvax.BERKELEY.EDU Organization: The Internet Lines: 76 kevin (ramsiri@blake.acs.washington.edu) writes: >Where can i get accurate specs on the >ATW... and where/for how much.. can i >buy one? I have an ATW800, so I might somewhat clarify the situation: Price (count #1 = DM2.7) DM 14500 + Tax. (Dec 89) Specs: 1 T800-20MHz (4MByte-70ns), 1 MC68000 (1MB Atari ST with all ports); 1 Videoboard (1MByte, 1260x960-4bit, 1024x768-8bit, 512^2-24bit, Graphics proc. with peak-rate of 128 MPxl/s); 40MB SCSI drive, prepared for 2nd one, MS-DOS 720KByte 3'5 floppy. Helios Netversion (without C compiler). X-Windows v11.2 (only server and most important xclients. this version does not seem to use the graphic proc., performance below) Add On's: 8 to 24 MByte main-mem with 1Mbit DRams, up to 3 farmcards (a 4 T800-20 with 1MB each) in box. (Ethernet TCP/IP and Berkeley filesystem announced for spring) Software: all Helios (compiler [C,Pascal,f77,..], C-debugger, par-C, TDS, TeX. Graphics packages, Ray-Trace, flight sim., neural nets ... in progress). Performance: Dhrystone: 2500 (bad!). with Helios `Accelerate`=4000. completely on-chip I expect 6000. [Perihelion C v1.1] Floats: 0.3 - 1.7 MFlops, depending on compiler etc. (for f77 there is an optimizing pre-processor.) X-Windows: I hereby publish my X-Bench results of Dec89: Machine Server Planes line fill blit text arc cmplx xstones R ----------------------------------------------------------------------------- Siemens 9733-203 11.2 8 61216 14878 19142 24635 566337 18836 23218 b SonyNews 1850 color 16514 14477 25208 21656 598290 37124 20481 a DECstation 3100 8 78115 6913 7292 79193 231587 16143 15735 a Sun 3/50 purdue 1 14644 11489 16639 15312 43379 9745 14160 a Sun 3/160 11.2 1 11536 10618 9021 11250 233545 10784 11035 b ATW800 i4 mode0 11.2 4/8 21407 10207 4100 31565 347937 10327 11028 d Sun386/250 (tuned) 8 29034 4434 6605 25437 44181 9215 10161 c Sun 3/50 11.3 1 10000 10000 10000 10000 10000 10000 10000 a ATW800 i4 mode1 11.2 8 21119 7777 3515 29631 334269 8889 9433 d NCD16 XTerminal 1 6366 5612 11771 15312 10455 7516 8469 a DEC gpx 11.2 8 4835 7892 5710 30937 390212 5490 8250 b Visual 640 XTerminal 1 5532 3554 4893 8020 28662 3032 5124 a SPARCstation 1 color 4877 1227 3006 9739 189817 12287 3424 a Sun386/250 8 1942 445 1319 3191 44825 4751 1291 b ----------------------------------------------------------------------------- References: a) iX Multiuser Multitasking Magazin 4/1989 page 19. b) Siemens Munich by SP-4's Hacker-clan (authors of X-Bench). c) Kiezmann in Berlin. d) Falk Langhammer at FU Berlin. I do not believe that the current server has much blitter support. => big improvements awaited... Dealers: Atari USA, GB, D, ... for D: Goeken Atari Postfach 1213 D - 6096 Raunheim (I do not have any relationship with Atari, I just bought an ATW.) Summary: I am quite satisfied with my machine, provided that all the minor problems will be solved in the near future, i.e. a serious file-system, more tools, Ethernet support (TCP/IP, NFS), improvement of X-Window scrolling speed and cheaper farmcards. Helios, of course does have good network facilities to other Helios nodes via link. But still there are machines without any link.. The ATW addresses to the workstation market rather than the number- cruncher card market. A single transputer ATW is slightly outperformed by the DecStation 2100 at the same price, but it has much more chances not to be outdated in 3 years. (my first contribution to this forum) Ciao, folks. Falk Langhammer, a physicist. falk@vax1.physik.fu-berlin.dbp.de