Path: utzoo!attcan!uunet!lll-winken!lll-lcc!pyramid!octopus!pete From: pete@octopus.UUCP (Pete Holzmann) Newsgroups: comp.sys.ibm.pc Subject: Re: Super EGA cards survey? Summary: Just post relative speeds. There're too many variables! Message-ID: <248@octopus.UUCP> Date: 10 Jun 88 14:14:02 GMT References: <5802@umn-cs.cs.umn.edu> <4364@sphinx.uchicago.edu> Reply-To: pete@octopus.UUCP (Pete Holzmann) Organization: Octopus Enterprises, Cupertino CA Lines: 116 SLOOOOOOWWWW down, folks!!!! [See below] In article <4364@sphinx.uchicago.edu> pre1@sphinx.uchicago.edu.UUCP (Grant Prellwitz) writes: >In article <5802@umn-cs.cs.umn.edu> amit@umn-cs.UUCP (Neta Amit) writes: >>I'm interested in the speed and compatability of several Super EGA Display >>Cards (480 x 640 and 600 x 800). >>Speed -- characters/sec . >> A few weeks back someone (sorry, forgot who) mentioned AWARDSST.ARC >> on Simtel20's pd1: . I tried it on a monochrome >> TrueBlue, and saw 2404 cps. I'm curious to know how EGA Wonder, Video >> 7 and other high-resolution cards do. If you have access to AWARDSST >> (not certain about its accuracy, but does it matter?) and to one >> or more of these wondercards, please post you findings. >ATI EGA Wonder [on 10 MHz/0w.s. AT, unknown BIOS]: > [enh feat enh feat] > enabled disabled >M80 1721cps 2970/2913cps >C80 1741cps 2970cps >CE80 1741cps 2970cps >Zenith [Z171]: 990cps >I think it would be useful to collect all these speeds and post a summary for >SIMTEL20 or something similar. Hold on here! If you run a lot of tests on a lot of machines, as I have, you'll find that there are more variables making up the SST performance result than you can shake a stick at! Just a few that come to mind: - Graphics card, BIOS version on the card, video mode - CPU speed, wait states (on ROM and RAM access), and RAM refresh interrupt overhead - DOS BIOS type and version (Phoenix, Award, DTK, etc etc) - Caching of BIOSes (DOS? Video? Both? it matters!) from ROM into shadow RAM (usually on a '386 machine) - Plus, individual runs will vary their result, as seen in the first numbers posted above. I think the granularity of the SST timing is probably one time-of-day clock tick, which is rather loose (still very useful, but loose). What to do? Well, on any given machine, the RELATIVE speeds of different setups are consistent, and seem to carry over from machine to machine. If somebody wants to collect results, they should accumulate relative speeds (board X is 1.5 times faster in mode A than mode B; board Y is 2 times faster than board Z on the same machine, etc). The Zenith laptop number given above is pretty useless, since there is nothing to compare it with *on the same machine*. The best you can do with these numbers if you aren't comparing relative speeds under controlled conditions (only changing one of the variables) is that you can say 'setup X has faster video performance than setup Y'. In order to duplicate somebody's results, you're going to have to duplicate their entire setup, computer, BIOS, and all! Just as an example, I know somebody with an ATI EGA wonder running on a '386 machine. They get 700 cps out of SST. YUCK! I know that if you put an Everex Microenhancer Deluxe in a 0 wait state 16 MHz '386 machine running Phoenix or Award BIOS and with shadow RAM, you'll get around 15000 cps. Even more if you use the Wizard Deluxe card (fastest EGA card I know of). Obviously, this tells me that there's something wrong with my friend's SYSTEM. Not necessarily his VIDEO BOARD! So far, we've brought his SST number up to 3000 by switching from backfilled 16 bit RAM to 32 bit RAM, and adding shadow RAM for the BIOSes. The hint given in the posted results above tells me I should check to see if Enhanced Features are enabled or not. I also happen to know that his machine is running at 16 MHz with 2-3 wait states. He's not going to get much better performance no matter what with all those wait states! The throughput is *very much* affected by BIOS access speed. I don't have hard numbers handy, but I can tell you the following: BIOSes: Phoenix and Award are both fast; DTK is much slower. Don't know how they compare to others. Boards: Mono-only and CGA-only boards are all much faster than EGA or normal VGA boards. Everex and Wizard are the fastest EGA boards I've seen. Generic EGA boards are next (much slower). EGA Wonder and other cards with extensive emulation are slowest (I think their BIOSes have to do a lot more work). Enhanced VGA cards with VRAM are supposedly the fastest cards of all. I haven't seen one yet, but Video 7 claims their board will run with no wait states at all on video access. We'll see! Non-VRAM enhanced VGA cards are not as fast as fast EGA cards. I guess they just have too much emulation work to do, like the EGA wonder. I've tried Paradise, Everex, Orchid. I've never touched a PS/x, so I don't know how they compare. My recommendation, if enhanced EGA is as good as you need, or even if normal EGA is what you need, is to get the Everex MicroEnhancer Deluxe, or the Wizard Deluxe (costs a little more than the Everex). You can get the Everex, with centronics port, for <$150US if you look. Both cards come with AutoCAD and Windows drivers for their hi-res modes. Ventura Publisher/GEM drivers are also available for both. If your system isn't getting much speed out of the SST test, check your BIOS. Some of the Taiwan clone BIOSes (esp. DTK) are real DOGS. If you have a '386 machine, use shadow RAM to get your BIOS code running faster. You'll be amazed at the difference! Have fun folks! Pete P.S.: If you know somebody (maybe you?) who is located in the southern S.F. bay area and who has access to a logic analyzer, I'd like to hear from you. I need to borrow one overnight to solve a problem in some volunteer work I'm doing [my job: spend as little time as possible earning $$$ so I can spend as much time as possible doing volunteer stuff!] -- OOO __| ___ Peter Holzmann, Octopus Enterprises OOOOOOO___/ _______ USPS: 19611 La Mar Court, Cupertino, CA 95014 OOOOO \___/ UUCP: {hpda,pyramid}!octopus!pete ___| \_____ Phone: 408/996-7746