Date: Thu, 6 Apr 1995 21:59:00 -0700 (PDT) From: "Rodney W. Grimes" <rgrimes@gndrsh.aac.dev.com> To: chuckr@Glue.umd.edu (Chuck Robey) Cc: uhclem@nemesis.lonestar.org, freebsd-hackers@FreeBSD.org Subject: Re: PCI/EISA/ISA performance Message-ID: <199504070459.VAA06199@gndrsh.aac.dev.com> In-Reply-To: <Pine.SUN.3.91.950406211833.10650A-100000@espresso.eng.umd.edu> from "Chuck Robey" at Apr 6, 95 09:22:32 pm
next in thread | previous in thread | raw e-mail | index | archive | help
> > On Thu, 6 Apr 1995, Frank Durda IV wrote: > > > [0]"Rodney W. Grimes" <rgrimes@gndrsh.aac.dev.com> > > [0]On Tue, 4 Apr 1995, Rodney W. Grimes wrote: > > [0]ISA does not have a specified clock frequency, I have seen it running > > [0]as fast as 16Mhz. Most boards die above 10Mhz, but some of the more > > [0]specialized industrial applications boards are spec'd upto 12 or 16Mhz. > > > > Actually, there is an IEEE standard on timing and performance for > > the ISA bus. It came out in the late eighties. I don't have it here > > (we have a copy at work) but I am pretty sure it says 8.33MHz is > > the IEEE spec. > > > > We used this spec recently when fighting Intel over some non-compliance > > on IOW timing. > > I've found this topic, and the way you guys are handling it, incredibly > interesting. Could I ask for one more comment? I have been recommending > friends to buy VLB bus stuff instead of PCI, because of all the PCI > incompatibilities and bugs I've read about, and because my own experience > with my Adaptec 2842 and FreeBSD is so great (thanks, Justin!). Do you > think my feelings are still valid, for anyone not staying as current as, > say, Rod Grimes? I would say your feelings still have some validity to them. Realize that Intel just last month admitted to a bug in the Neptune chip set that I saw back in June of 1994. (There is the problem with running more than two bus master PCI devices on Neptune chip sets, it does not allows occur, and depends on what bus master devices they are and if they all get very active at the same time). They only admitted to the bug after the started shipping Triton chip sets, and claim it is fixed in the Triton! There are still incompatibility issues with some PCI add in boards and some motherboard and/or BIOS implementations. Most of this was solved with version 2.0 of the PCI spec and with boards that are PCI 2.0 compliant. Avoid *any* PCI devices that do not claim to be PCI 2.0 compliant. I still find PCI motherboards that don't like bt946s, some that don't like aha2940s, though this is becomeing rather rare for me. As far as all the hoop-la about Plug-n-Play, well, if you ask me it is a bunch of b*ll sh*t. Though PCI systems are easier to configure than ISA/VLB boards, they all still have the d*mn ISA compatibility crap, and you have to deal with it. Folks, IMHO, this problem won't go away until the ties with IBM-AT compatibility are thrown out the window and things are fixed like they should have been 10 years ago when the 80386 chip came out. (Gee, 10 years after the introduction of a 32 bit VM chip and we still don't have a true to the core 32-bit OS from Microsloth...) Realize VLB had the same types of problems when it was less than a year old (bus masters that wouldn't master, VLB cards that didn't like certain motherboards, etc.) PCI is going through the same growing pains. I steer my customers away from VLB now, since, IMHO, it will be dead by the end of 1995. And since as you say, I am staying on the bleading edge of this stuff, I know what to steer clear of so that my customers don't have problems with the PCI systems I sell. 6 months ago you had to beg me to sell you a PCI system, as I didn't want the support problems. -- Rod Grimes rgrimes@gndrsh.aac.dev.com Accurate Automation Company Custom computers for FreeBSD
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199504070459.VAA06199>