Live data from Hacker News

A cycle-accurate IBM PC emulator in your web browser

martypc.net

21–26 of 26 posts

Re: A cycle-accurate IBM PC emulator in your web browser

#21
post #14

Earlier quoted context omitted.

Temperature? Really? How does that work?

If you program a register at a moment and in a way that causes two signals to "collide", the result effectively depends on transistor behavior. That in turn can be temperature dependent. For an example on the PC see https://int10h.org/blog/2023/03/cga-6845-crtc-phantom-vsync-...

Amazing link. Thanks for sharing.

Where can I find more writeups like this?

Re: A cycle-accurate IBM PC emulator in your web browser

#23
post #3

Pretty incredible! I’m on mobile right now so I can only comment on the demo that runs automatically, which I understand isn’t the _point_. :) More about the demo: https://www.vogons.org/viewtopic.php?t=89435 (For those unfamiliar with “demo” in this context, see https://en.m.wikipedia.org/wiki/Demoscene ) I look forward to checking out the features you mention on a proper computer.

Thanks for pointing people to Area 5150. Context for others: This, and its predecessor 8088 MPH, uses every single trick in the book and many not written down anywhere to squeeze that kind of performance out of a real 1981 IBM PC 5150 with IBM CGA card. Thus they only run correctly on that specific hardware, not a clone computer or card. "Regular" software emulators also have a tough time. Showing off MartyPC with Area 5150 is thus two technical feats combined into one.

Re: A cycle-accurate IBM PC emulator in your web browser

#24
post #18

Earlier quoted context omitted.

One of my neighbours in Sydney sold a board he made in his garage, that gave you an extra ten slots, which I filled with way too much RAM. Sort of like a homebrewed 5161, before the 5161 existed.

That's kind of neat. Did he just extend the bus in a "raw" fashion or put buffers on it? (The downside of the 5161 was any memory device in the 5161 had an extra wait state thanks to the speed of the buffers, which were necessary to deal with the capacitance on the cable to the expansion unit.)

They were buffered, from my poor memory. So was slower when you exceeded mainboard memory, but you could load entire tapes into memory, which let you do things no one else could. Faster processing than disk/tape access.

Re: A cycle-accurate IBM PC emulator in your web browser

#25
This appeared on the fp the same day i spent 15 hours trying to get various old/weird OSes installed. Xerox Star (viewpoint), os/2 warp 4, serenityos, debian 9, and where is freebsd 2.2 disc 2, might i ask?

commenting for posterity, and to say the web emulator is very slick, i have some dos diskettes to try out.

Re: A cycle-accurate IBM PC emulator in your web browser

#26
post #14

Earlier quoted context omitted.

Temperature? Really? How does that work?

If you program a register at a moment and in a way that causes two signals to "collide", the result effectively depends on transistor behavior. That in turn can be temperature dependent. For an example on the PC see https://int10h.org/blog/2023/03/cga-6845-crtc-phantom-vsync-...

That was a fun rabbit hole. My first computer (8-bit, 4k RAM, 0.8 Mhz) had a video chip related to that 6845, the 6847, which was (sadly) a somewhat lower cost, less capable, less interesting chip.
Post reply on HN