Return to BSD News archive
Received: by minnie.vk1xwt.ampr.org with NNTP id AA219 ; Fri, 29 Jan 93 17:00:53 EST Xref: sserve comp.unix.bsd:10560 comp.sys.ibm.pc.hardware:42604 Newsgroups: comp.unix.bsd,comp.sys.ibm.pc.hardware Path: sserve!manuel.anu.edu.au!munnari.oz.au!sgiblab!spool.mu.edu!yale.edu!ira.uka.de!math.fu-berlin.de!informatik.tu-muenchen.de!roell From: roell@informatik.tu-muenchen.de (Thomas Roell) Subject: Re: ATI Mach32 and Xfree86 or XS3? In-Reply-To: johng@informix.com's message of 27 Jan 93 05:20:56 GMT References: <1993Jan26.162704.13442@usasoc.soc.mil> <1993Jan26.220906.26671@netcom.com> <1993Jan27.052056.23103@informix.com> Sender: news@Informatik.TU-Muenchen.DE (USENET Newssystem) Organization: Inst. fuer Informatik, Technische Univ. Muenchen, Germany Date: Thu, 28 Jan 1993 08:24:33 GMT Message-ID: <1993Jan28.082433.8404@Informatik.TU-Muenchen.DE> Lines: 59 >The ATI Mach32 is based upon ATI's mach32 chipset. >for speed I suggest acquiring a 928 based card which is faster than the >ATI MACH32 by as much as 34% overall graphics performace, even on the ISA :-) The number are correct, but my benchmarks say that EISA has no major advantage over ISA (VLBus is a different story). Since the only thing that differs between ISA and EISA is that transfers can be 32bit instead of only 16bit, and a S3 chipset or a Mach32 are programmed almost exclusively over io-ports to use the graphics engine, you hardly won't see any major performance increases (please note, the ISA and EISA bus speed are the same ...). >Any info on cards that are using the 928 that are available? There will be real soon a BIG number of them available. But right now it is my understanding that nobody can actually deliver boards in large quantities. Another issue is that some 801 boards are that fast (STB PowerGraph X-24), that it doesn't make THAT big difference in performance which you would actually see. Granted that the difference is 30%, but given those high benchmark numbers, that doesn't really matter. The problem is that it depends on what you are doing. If you only to 1024x768 and could live with a lower benchmark number (which is still higher than for the Mach32), go for a 801. If want to do more than 8bpp or things like 1280x1024 get a 928 board, which is upgradeable to 2MB VRAM at least and has a Bt485 or TLC34075/76 RAMDAC on it. But you can be sure that this will cost you around $800. >Why? It was my understanding that the local bus was specifically intended >for things like video cards that are twiddling bytes back and forth with the >cpu, as opposed to massive memory transfers. The local bus is generally way faster than the ISA/EISA bus no matter what you are doing. On a dumb framebuffer with no intelligenze on it where the CPU has to do the drawing you'll see major performance increases due to the bigger memory bandwidth. But accelerators like the S3 chips are intelligent, and there you have to do everything over the IO-Bus, or at least only with 16bit memory accesses. Hence if you compare the ISA bus with the VLBus and see numbers like 2.7MB vs 12MB for IO-ports, keep in mind that the VLBus number was created by using 32bit wide accesses. If you now think about the fact that you'll be using only 16bit accesses it looks more like 2.7MB vs 6MB. This is still very high you might say, but then take into consideration, that the next limit is the drawing speed of the graphics chip. And currently there are only very few cases where you could feed the 928 faster over the ISA bus than over the VLBus, since the CPU has to wait for the 928 to finish it's last operation. If I were to buy a motherboard and a graphics board, I would go for ISA/VLBus instead of an expensive EISA design. - Thomas -- ------------------------------------------------------------------------------- Das Reh springt hoch, e-mail: roell@sgcs.com das Reh springt weit, #include <sys/pizza.h> was soll es tun, es hat ja Zeit ...