mb@pertec.UUCP (mike burg) (11/24/84)
I had a friend that replace the crystals with faster one and it total burned the 8086 CPU chip and terminated the service contract with IBM. Mike Burg decvax!randvax!trwrb!pertec!mb
wizard@intelca.UUCP (Kevin Supinger) (11/30/84)
Sometimes I really wonder about the amount of misinformation that is propagated by this network. But as far as speeding up the IBM AT I would like to set the record straight As a point of reference I happen to own one of the "GEMs" and have actually tried the speed up. It is very simple to do and was performed the second day after recieving my AT. It has run for over 3 months now without a glitch. You replace the socketed crystal with a 16 MHZ crystal. Carefully pull the 80286 pin grid array out of the socket. Replace it with an 8 MHZ 80826 . If an 80287 is used in the system it must be the same speed or a special adapter card should be used with an 8284 clock generator so that a lower clock speed for the 80287 is provided. You might note that the 80287 was designed to be asyncronous with the 80286 clock. The observent student may also examine the AT tech ref manual to note that the IBM AT was designed for 8 MHZ 80286 operation. This doesn't affect the disk controller or the memory ,which is the proper speed. Current programs execute at 25% of the original XT speed. This really enhances the AT. It would be very nice in the future if the net would look at INTEL as the shining star it really is. These views are my own and do not reflect on Intel
lauren@vortex.UUCP (Lauren Weinstein) (12/02/84)
Of course, if you play games with the timing of your PC or AT, you're going to be up the creek with any programs that do any sort of timing, especially communications programs of various sorts. In fact, any program with timeouts will be burned. You can't expect vendors to know how to deal with people's "off the wall" crystals. --Lauren--
scott@opus.UUCP (Scott Wiesner) (12/04/84)
> Sometimes I really wonder about the amount of > misinformation that is propagated by this network. > Current programs execute at 25% of > the original XT speed. This really enhances the AT. Whew!! That hot 286 is a real speed daemon eh? Let's see, an 8Mhz 286 runs programs with 25% of the speed of a ~5Mhz 8088. And I suppose the 386 will run at a blinding 25% of the speed of the 286 which is 25% of ..... or did you mean to say programs execute in 25% of the TIME of the XT? > The observent student may also examine the AT tech > ref manual to note that the IBM AT was designed for 8 MHZ 80286 > operation. > It would > be very nice in the future if the net would look at INTEL as > the shining star it really is. Uh, right. And I wonder why the AT is designed to use the 8Mhz part when they ended up using the slower one. Could it be that our "shining star" didn't deliver what was promised WHEN it was promised? Seriously, I do appreciate knowing the fast cpu can be used without flakey side affects. -- Scott Wiesner {allegra, ucbvax, cornell}!nbires!scott
guy@rlgvax.UUCP (Guy Harris) (12/04/84)
> It would be very nice in the future if the net would look at INTEL as > the shining star it really is. When they come out with a CPU chip that 1) isn't an IAPX-432 and 2) supports a large address space conveniently and efficiently, I'll think about it. (Note: the 80286 is not that chip. Period.) Guy Harris {seismo,ihnp4,allegra}!rlgvax!guy
ee161abt@sdcc13.UUCP ({|stu) (12/04/84)
> Of course, if you play games with the timing of your PC or AT, you're > going to be up the creek with any programs that do any sort of > timing, especially communications programs of various sorts. In fact, > any program with timeouts will be burned. You can't expect vendors > to know how to deal with people's "off the wall" crystals. > > --Lauren-- Programs written for the PC are already gonna show this effect on the AT, so what's the diff? If programmers used the timer values available from the 8253 interrupt handler, chances are timeouts wouldn't be "burned." Any program that uses timing "loops" is going to worthless in a concurrent environment anyhoo. - stu
gnu@sun.uucp (John Gilmore) (12/04/84)
> The observent student may also examine the AT tech > ref manual to note that the IBM AT was designed for 8 MHZ 80286 > operation. > ... It would > be very nice in the future if the net would look at INTEL as > the shining star it really is. If Intel is such a shining star, howcum IBM is not shipping the 8MHz 80286's that they designed for?
wizard@intelca.UUCP (Kevin Supinger) (12/05/84)
This is a question better asked of IBM. Even SUN could use a good processor
lauren@vortex.UUCP (Lauren Weinstein) (12/06/84)
The timing problem is not so simple. Dealing with the diffs between the AT and PC are easy -- you just have an environment variable specifying the machine type. But you can't really rely on the interrupt timer either. Some "look-alikes" are running it off the main system clock, which means it gets changed along with all the other system timing. But even worse, functions that, for example, are checking for input from a com line and time out after N seconds if there isn't any, often involve much more code than the simple call to check the interval timer. That code is part of the timing process, and can't even be expected to behave in a linear fashion in all cases with clock changes. The bottom line is that you pretty well have to tweak up timing functions to work on a particular machine if you need other than rather gross timings. Any software that tried to directly use the interval timer in a multi-tasking environment would probably be in even worse shape, since the OS itself would probably be using various parts of the timer for its own internal uses. That's why such OS's would typically provide "sleep" calls for programs to use to avoid such problems. But I'm not talking about multi-tasking now, just the normal single-task situation. --Lauren--