Comment by schoen | Hacker News Reader

Comment by schoen | Hacker News Reader

Parent Full thread

I suspect in the very early days of cooperative computing, one could genuinely argue that there were "idle cycles" that were going to waste if a computer was powered on and not computing something. But this became more and more inaccurate throughout the 1990s as new power management technologies became a part of computer designs, and is completely wrong now. Phenomena like Bitcoin mining and feeling your laptop get warmer depending on what it's doing show that easily, as does your room-heating experience.

I'm still curious about the extent to which the claim of idle compute power going to waste was true when various projects were set up. For example, the press release about the Cooperative Computing Awards (I've forgotten whether this was 1998 or 1999) says

> However, the computer industry produces millions of new computers each year, which sit idle much of the time, running screen savers or waiting for the user to do something.

https://www.eff.org/press/releases/eff-offers-cooperative-co...

The main thing I can think of that would have supported the concept of idle cycles going to waste was

https://en.wikipedia.org/wiki/HLT_(x86_instruction)

> [the CPU-idle instruction] was not specifically designed to reduce power consumption until the release of the Intel DX4 processor in 1994.

I participated in DESChall and distributed.net's RC5 challenge and I remember being told, and telling others, that we were only using resources that would otherwise have gone to waste. Were we already greatly mistaken about that in 1997?

I would think of compute resources "going to waste" in three separate groups:

- The baseline electricity the computer is using. This will get used and produce no S@H work either way (in an era where our computers didn't go to sleep)

- The added electricity to do S@H work. This certainly was not idle, but also was a relatively small amount relative to an entire household's use.

- The hardware itself.

I would say that S@H were making use of hardware that was "going to waste" (this is CAPEX the project didn't need to spend), was having volunteers donate the added electricity, and could be said to be making use of the baseline electricity inasmuch as it would be used either way, and if S@H were to set up their own compute cluster, they would also incur similar overhead.

I don't know what system efficiencies were like in the late 90's, but I would totally believe someone if they said "idle" was 40% the power consumption of "max usage". Using that number as an example, "donated electricity" gets (1/0.6=) 1.6x as much compute-per-watt as if it was donated to run an equivalent, dedicated, machine. Maybe you can convince yourself that you were using "idle cycles" and donating on top of that?

That's definitely not the way I thought of it at the time, but that seems like a useful way to describe it!

I may be wrong but back in 90s an idle processor used up just as much electricity as an actively computing one.

That's really part of what I was asking about. The claim from Wikipedia about Intel's HLT instruction shows that this started to change a lot by 1994, but I don't know if there are other similar considerations.

Before 2002 (AMD's Cool'n'Quiet) and 2005 (Intel's SpeedStep) CPUs didn't do any frequency scaling for power savings. When the CPU wasn't asleep it was running at it's full voltage and clock speed. While the HLT instruction kept let them 'idle' they only used marginally less power than when active.

So running distributed.net or Seti@Home was using unused processing capacity for a marginal increase in power draw. If you ran them overnight with your monitor (CRT typically) your computer was drawing less power than if you were sitting up all night on IRC or playing Quake. This changed once frequency scaling was commonplace on desktops.

If you use an electric radiator for heating, might as well use that electricity to compute something worthwhile !

For anyone that isn't aware, yes, they really are equally efficient at heating!

Reply on news.ycombinator.com