Hacker News Reader

Comment by liveoneggs | Hacker News Reader

If an old computer can be 100% emulated by a new computer using 1/1000th of the electricity then the old computer should be archived as software and physically recycled/eliminated.

View on HN

Your laptop probably draws, what, 15W. Your old computer probably does not draw 15KW! - yes, I'm sure there will be old computers that do, and more, but since this article is about the Atari 800, I'm going to assume the discussion is about that sort of old computer, rather than something the size of a room that required a reinforced floor and your own personal account manager at the electricity company.

Modern hardware that I've got:

1. Mid-2015 13" Macbook Pro, mostly idle, battery charged, screen fully dimmed: 11W (16W with screen at full brightness)

2. My 2011 i7-2600K desktop: ~100W idle (packed away right now - but it's about that)

I happen to have my old Acorn junk set up right now, so here's some measurements from that:

1. BBC Master 128: 12W

2. BBC Micro model B: 25W

And some other Acorn peripherals. Add these figures to either of the above:

1. Acorn 6502 second processor: 5W

2. 5.25" disc drive: 2W (when in use)

3. Ancient early-80s monitor of unknown provenance: 50W

4. Acorn-branded CM8833: 85W

Looks like my favourite Acorn-branded monitor is actually pretty bad :( - however it's not actually drawing a huge amount more than the 2 x 27" monitor setup I have with my Macbook Pro, even though the pixels per watt isn't quite as good.

In fact: the overall power draw for BBC Master 128 setup vs Macbook Pro setup, both period-authentic, both (if not cutting edge for the time) up to the task of a good range of period-authentic workloads, is about the same! - and even if I was certain 1000x was an overestimate, this is still a bit of a surprise...

If you factor in that your MBP could probably comfortably emulate 100 6502s in real time, 1000x isn’t quite as severe an overestimate. Get the GPU in the game and, even with terrible warp utilization and cache efficiency, you still might be able to hit that 1000x for purely academic purposes on a desktop card with HBM (maybe 2500 real-time-emulated 6502s on a 250w system).

Move to the microcontroller world, though, and 1000x is a vast underestimate. The main core in a current project of mine runs at 24 MHz and draws around 1.2 mW without even entering sleep. This system could probably emulate a Master 128 just fine in real time, for about 1/10,000x the power.

My MBP can probably comfortably emulate 150 x 1 MHz 6502s in real time, but your average 6502-based system is a system, not just a 6502, and the Acorn stuff is no exception. There's a lot more to emulating it than just dealing with the CPU.

The Master 128 runs at 2MHz. For each emulated microsecond you need to (broadly speaking) do at least 2 CPU updates, 2 VIA updates, and 16 bitmap bits/1 teletext glyph. 24MHz = 24 cycles per emulated µsec :( - good luck even squeezing the video output into that, let alone the other stuff. Rather you than me!

(This also ignores handling the FDC, ADC, and ACIA. The devices accessed via the VIA will also take a bit of time to emulate...)

I'm talking more raspberry pi range than MBP, which is also 1000x more powerful. (or more; aka could probaly emulate all of the BBC Micros every sold at once)

a raspberry pi is 5W(?) at 100% cpu I think and can still emulate all of these machines with room to spare.

My macbook pro is only drawing 15W because it also includes a monitor and keyboard and happens to be plugged in. Otherwise I can emulate a data center worth of these machines without drawing ANY additional power ;)

I do tabletop gaming with a group of friends, and have an ongoing debate with one that seem relevant to this.

The debate is that he is of the opinion that tabletop games (particularly ones with complex scoring or recordkeeping) should be computerized. I am of the opinion that they should not, because when you change the form factor, you are changing the game even if all the rules remain the same. We are humans, and as humans, physicality matters to us.

I think that the same thing comes into play with old-school hardware. The hardware itself is important, and no amount of emulation can really replace it.

My friend and I will never agree on this with tabletop games, and that's OK. You and I may equally never agree about old-school hardware, and that's also OK.

At 100% you necessarily include sensory equivalence. Otherwise you still change the experience drastically. Emulators are a fundamental shift in the sensory aspect. Even moving the F-keys around a tiny bit can be enough to drive a user crazy or break a great word processor.

I went to a local video game museum the other day. It was a ton of fun. You simply can't emulate the feel of the Arkanoid controller (a knob which you turn clockwise and counterclockwise to move the platform), or the way games actually looked like on a ye olde CRT (although some might point out that there's decent effort done here and there), not to mention games for vector-based displays.

If. That makes sense only when comparing very old hardware to new, and only because the new hardware is much more power-efficient. But it's not a good argument for perpetuating absurdly short hardware cycles like we have with phones (1-2 years).

I'm specifically talking about things like VAX and old sun and DEC etc

The new computer will probably stop working in <10 years. The silicon will self-destruct in 50. The old one will survive the collapse of global civilization.

Reply on news.ycombinator.com