Julian Wolfe wrote: Can you elucidate on "stabilize/speedup NEC V20 chips"? I thought I had considerable experience working with 8088 clones (8086/NEC V20/NEC V30) but I've never heard of such a procedure (or need)... -- Jim Leonard (trixter at oldskool.org) http://www.oldskool.org/ Help our electronic games project: http://www.mobygames.com/ Or check out some trippy MindCandy at http://www.mindcandydvd.com/ A child borne of the home computer wars: http://trixter.wordpress.com/

Julian Wolfe wrote: Found it. I was somewhat disappointed: mov al, 74h out 43h, al ; Timer 8253-5 (AT: 8254.2). mov al, 8Fh out 41h, al ; Timer 8253-5 (AT: 8254.2). mov al, 2 out 41h, al ; Timer 8253-5 (AT: 8254.2). int 20h ; DOS - PROGRAM TERMINATION This is the standard "reduce DRAM refresh rate so CPU has more leftover cycles" trick. It has nothing to do with V20's at all, it works on any XT. I was rather hoping it was something V20/V30-specific. -- Jim Leonard (trixter at oldskool.org) http://www.oldskool.org/ Help our electronic games project: http://www.mobygames.com/ Or check out some trippy MindCandy at http://www.mindcandydvd.com/ A child borne of the home computer wars: http://trixter.wordpress.com/

Things I could find that were effected by it were old Sierra games. They were fairly resource-intensive. Autosketch is also faster, as is F19 stealth fighter.

On 11/11/06, Josef Chessor <josefcub at gmail.com> wrote: I remember spending $20 for a 68010-L8 to get that 5% advantage on an Amiga 1000 (one-instruction loop cache was the relevant feature). Back in the day, 5% for _free_ would have been sweet (I never knew about this hack when I had an XT). -ethan

On 10 Nov 2006 at 10:36, Cameron Kaiser wrote: Depends on how much nerve you have, but typically about 10 percent or less. I thought the overclocking craze on the PC-AT was a lot more interesting, myself... Cheers, Chuck

On 10 Nov 2006 at 14:23, Julian Wolfe wrote: There were stories of some folks clocking the 286 as high as 20 MHz. I think IBM didn't want to be saddled with the support issues that overclocking entailed, so they modified the BIOS to make the clock speed check. If you knew someone with a PROM burner, it was easy to patch out, however. Or you could kick the clock up after POST. Some early Taiwanese PC clones had a jumper-selectable Turbo mode that simply switched oscillators (or crystals) to a higher frequency. Very crude--the time-of-day clock would run way too fast without a TSR to scale things back. That brings up an interesting question. Who was the first to apply the word "turbo" to a CPU? Cheers, Chuck

On 10 Nov 2006 at 18:03, Jim Leonard wrote: But this was on a regular old 6MHz AT chassis. At least one outfit advertising in the back of PC Week (IIRC) offered a little kit of oscillators. You kept plugging in faster and faster clocks until your machine quit working. A very dangerous way to live. Cheers, Chuck

This reminds me.., I had a .., hey! wait a minute... I *still* have a DX2-66 that I could run with a clock of 40Mhz during the winter..., when I kept the office temperature at 58 deg.. During the summer (it was un-airconditioned) I had to drop back to 33, else the CPU would lock up after 5-10 minutes of operation. John

--- Chuck Guzis <cclist at sydex.com> wrote: Or 'sploded altogether. Ah the aroma! Somebody disconnect the battry from the smoke alarm. LOL LOL ____________________________________________________________________________________ Yahoo! Music Unlimited Access over 1 million songs. http://music.yahoo.com/unlimited

On 11 Nov 2006 at 17:04, Chris M wrote: We're talking REAL IBM AT's--model 5170. "Clones" didn't incorporate the speed-checking code in the BIOS. A few folks forget the PC/XT 286 model 5162 that was actually a bit faster (because of 0 WS DRAM) that the original 5170. I don't know if the 5162 ever incorporated any speed-checking in the BIOS. Cheers, Chuck

No it doesn't, for the reason you've suggested. A 5170 mainboard will not fit in an XT case. The XT-286 mainboard is electrically quite similar to the AT (5170) mainboard, but there are differences. One obvious difference is the RAM -- the AT mainboard can take a maximum of 512k (either in 256K*1 chips in later versions or some odd IBM 128K*1 devices that are actually 2 64K*1 chips with strange pinouts piggybacked in earlier versions). The XT-286 board is designed to hold 640K as 2 256K*9 simms, 4 64K*4 chips and 2 64K*1 chips (parity on that last 128K). There's also a problem with some adapter cards being too tall to fit in the XT case. For example only ther later version of the hard/floppy disk controller will fit. -tony

On 11 Nov 2006 at 18:03, Chris M wrote: The XT 286 did not use the same motherboard that the AT did--it would have been physically impossible. As it was, the XT/286 was more than a little cramped under the hood--a couple of the slots could take only short cards. The other problem was that original "stock" AT expansion cards were too tall to fit in the otherwsie XT-high chassis. It was a bizarre beast, but a good performer for the money. IIRC the memory used in the XT 286 was a bit odd also. The original 5170's were 6MHz products. I'd heard that IBM originally intended to ship 8 MHz units (certainly 8MHz 80286 CPUs were available in 1986), but the idea of shipping a system that was at least 5 times faster than the 5160 was nixed by marketing--I don't know if the story's really true, but it would be typical of IBM of the time. I suspect this was behind the later incorporation of a speed check in the later 6 MHz BIOSes--ant hat IBM shortly thereafter began shipping 8 MHz versions of the PC/AT. FWIW, this intentional "dumbing down" to meeat a price point was an old game. Others could verify this, but I believe that the only difference between the CDC CYBER 72 and 73 was the addition of a wait state or two. I'd heard that CE's could sometimes be bribed to remove the offending jumper for customers. When I was at Durango, we'd spec-ed our models for 7 and 14 MB Rodime hard disks and set the pricing accordingly. Not too long thereafter, Rodime discontinued the models and begin substituting (for no addional cost) 10MB and 20MB drives. Marketing asked us to artificially "clamp" the hard disk capacity in the OS so that customers wouldn't get more than they paid for. Only recentlly, I pulled out the CPU from an Toshiba Infinia 7230 (which used the Intel AN430TX mobo) and discovered that it was a 233 Mhz MMX unit, yet Toshiba had left the CPU clock at 200MHz on the mobo because that's what the model was advertised as. I haven't checked the 7201 that I have (ostensibly 160MHz, but it wouldn't surprise me if a faster CPU were under the hood). Cheers, Chuck

On 12 Nov 2006 at 1:24, Jim Leonard wrote: Under a compute-bound load, many seem to think it's pretty close to 5x. For example, see: http://www.pcguide.com/ref/cpu/arch/perfSpecifics-c.html http://www.ultratechnology.com/bench.htm I'm not quoting the Dhrystone benchmark results here, because while the sampling is large, there's just too much of an apples-and-oranges feel about the results. I wish I could find some early iCOMP results for the 8088 and 286. I submit the proposition that the relative improvement in performance between any two consecutive IBM PC models never exceeded that obtained between the 5160 and the 5170. I recall running some very compute-heavy benchmarks from an 8 MHz 8088 and being startled by how much faster they ran on a 6 MHz 80286. I believe that benchmarks tend to have a first-cousin relationship to unvarnished mendacity, but given the combination of faster I/O and processing speed, I think an 8 MHz 5170 comes very close to being almost 5 times as fast as a 4.77 MHz 5160. Perhaps one of the list members can verify this. I've got XT and AT "clones", but that might not give an accurate reflection of performance on IBM hardware. Cheers, Chuck

On 12 Nov 2006 at 13:31, Jim Leonard wrote: That wasn't my claim; I said that "had IBM decided to release the 5170 as an 8 MHz machine..." when comparing with a 5160. Right there, you start off with a cycle time that's nearly twice as fast. If anyone has a 5150/60 with an 4.77MHz CPU clock and an 8 MHz 5170 handy I'd be interested to hear the results of SI and SPEED numbers. I can probably do it if no one volunteers. Cheers, Chuck

Chuck Guzis wrote: Yes, but twice != five times as fast. They didn't improve every single opcode's time... the core ones like ADC went from 3 cycles to 2... MOV involving registers stayed the same... JMPs are only one cycle faster if jump isn't taken... etc. A lot of the esoteric opcodes like XLAT and AAD were dramatically improved, as were MUL/DIV but those aren't exactly basic building blocks of all applications. I would suggest you ditch Norton SI and instead use MIPS.COM which is much more granular and has defaults for exactly your test case machines. -- Jim Leonard (trixter at oldskool.org) http://www.oldskool.org/ Help our electronic games project: http://www.mobygames.com/ Or check out some trippy MindCandy at http://www.mindcandydvd.com/ A child borne of the home computer wars: http://trixter.wordpress.com/

Chris M wrote: I overclocked a ZX Spectrum by replacing the 14MHz crystal with a 17.73MHz one. Of course everything is derived from this, including video timings, so I needed to tweak my monitor up proportionately, and run a cassette player a little bit faster so the tape speed matched. It made Manic Miner extremely difficult. Gordon.

--- Cameron Kaiser <spectre at floodgap.com> wrote: were both horizontal and vertical scanning rates affected? Presumably so if the dot clock itself got bumped up. ____________________________________________________________________________________ Sponsored Link Mortgage rates near 39yr lows. $420k for $1,399/mo. Calculate new payment! www.LowerMyBills.com/lre

On 10 Nov 2006 at 13:55, woodelf wrote: Back in the day, fast static RAM was fairly expensive. Of course, you could do this today, but where's the fun in that? Cheers, Chuck

On 10 Nov 2006 at 19:29, Teo Zenios wrote: In the PC world, I don't know. But SRAM was fairly popular in the S- 100 world, even with its higher power dissipation. My old MITS Altair had 2 boards of 4K TI DRAM and it would drop bits if you looked at it crossways. When I replaced those with some SSM static RAM boards (91L02s), I was much happier. Cheers, Chuck

The P112 has static RAM also. If you store something on a RAM disk then reset the system, you find the files are still there! Pretty cool. ----- Original Message ----- From: "Teo Zenios" <teoz at neo.rr.com> To: "General Discussion: On-Topic and Off-Topic Posts" <cctalk at classiccmp.org> Sent: Friday, November 10, 2006 7:59 PM Subject: Re: Need Docs for XT RAM Board...

As a sort-of aside, the HP110 and 110+ (Portable and PortablePlus) machines were 8086-based laptops running MS-DOS (but not at all PC-compatible at the BIOS or hardware levels) with static RAM. Row upon row of 6264s (8K byte static RAM chips -- my Portable+ has, IIRC, 112 of them in SOIC pacakges. My 110 has rather fewer, but all in DIL packages, there's a PCB the size of the base of the machine containg the CPU, ROMs, a bit of support logic, and loads of RAMs -tony

Chuck Guzis wrote: The V33A should be even quicker, no? Was it still pin-compatible with the 8086? Peace... Sridhar

On Nov 10, 2006, at 12:46 PM, Jim Leonard wrote: Oh my. I'm sorry for jumping in, but I got VERY excited about V20 chips back in the day. They truly seemed like something for nothing! The V20 is a drop-in replacement for the 8088, basically a re-implemented 8088 that consumes less power and is a bit faster at the same clock rate. I don't recall exactly how much faster, but I seem to recall maybe 25%. It was certainly enough to be noticeable. On a 4.77MHz system, Norton's SI (system information) would give you a relative rating of (if I recall correctly) 1.7, where the 4.77MHz 8088 was 1.0. It wasn't really a 70% increase; that was due to the imperfections of the Norton benchmark. The V20 also has a 8080 emulation mode, and if you had the appropriate boot disk, you could run CP/M-80 on a V20-equipped PC. You had to execute a special instruction (again I don't recall the details...these are twenty-year-dormant neurons) to kick it into 8080 mode, and this was done by the CP/M bootstrap. It was a neat scheme! The V30 is basically the same thing, but is a drop-in replacement for the 8086. Though the technical details escaped me, I remember very well the feeling of excitement when I got that first V20, plugged it into my no-name 4.77MHz PC/XT clone, and noticed how much faster Turbo Pascal compiled my stuff. :-) -Dave -- Dave McGuire Cape Coral, FL

--- Tony Duell <ard at p850ug1.demon.co.uk> wrote: I read something in an early BYTE stating that as time went on, the dip switches on the pc's mobo had little consequence on the systems ability to determine the amount of ram installed. Am I in error to state that the dip switches are read by code? Perhaps someone can confirm, deny, elaborate, denounce... ____________________________________________________________________________________ Sponsored Link Mortgage rates near historic lows: $150,000 loan as low as $579/mo. Intro-*Terms https://www2.nextag.com/

On 8 Nov 2006 at 17:01, Fred Cisin wrote: It's one of those endearing things about the 5150 that I like. The 5150 actually needed NO DIP switches--the hardware designers must have had a very low opinion of the software people. Consider: the first bank, memory size, describes available memory only up to 64K (on the original 5150). The number of available diskette drives can be determined easily by software (note that there was no "drive type" setting, as all drives were assumed to be 5.25" 320/360/160/180K types). NDP presence is again easy to determine with software. Presence of a CGA or MDA is easy enough--and if a CGA was detected, it's easy enough to start things in 40x25 mode--a switch to 80x25 can be made in the autoexec.bat file easily enough (or the SCREEN command in ROM BASIC). Cheers, Chuck

Yes, the DIP switch settings can be read in software, and for most of the switches, all the affect is the value read by the software. But IIRC the memory size swtiches are also wired to the memory address decoder PROM, and will thereforee control the address range where the mainboard memory data buffers are enabled, etc. In othor words, if you only have (say) 64K on the mainaboard, you can add memory up to 640K on expansion cards _but_ you'd better set the DIP switches on the maiboard correctly or you'll end up with a bus contention (beteen the buffers on your memory expansion card and the mainboard data buffer) for, say, the addtress range 64K -- (256K-1). Incidentally, that address decoder PROM contains 4 addressing tables. IIRC, in som order, they are : 256K (with 64K*1 chips in all 4 banks) 640K (with 256K*1 chips in banks 0 and 1, 64K*1 in banks 2 and 3) 64K (with 16K*1 single-rail (4816s) in all banks) 512K (with 12kK*1 chips in all banks). The 2 DIP switches basically select how many of the banks will be used. -tony