Re: Audacity and Gentoo

On Sun, 15 Jan 2006, Ian Rawlings yowled:
> On 2006-01-15, Nix <nix-razor-pit@xxxxxxxxxxxxx> wrote:
>
>> Yeah, but the smaller the fan the faster it has to spin to move a given
>> amount of air :/
>
> No *** sherlock ;-)

Yeah, some friends of mine found that *** makes a very *bad* coolant
(don't ask, an accident with a plumbed-in liquid cooling system in a
sewage plant). Air is far preferable.

>> I don't know how overspecified most machines are when it comes to
>> cooling, but I *do* know that in the summer months, I've had cooling
>> problems with machines here. (No air-conditioning in this house.)
>
> I will be watching for this come the summer months, I only set my rack
> up about a month ago (probably not even that long ago) and am
> expecting to get some overheating from something somewhere.

Ah well, if we shut the Gulf Stream down that problem will go away.

> With the
> house heating on I can open the rear door and feel a little heat
> wafting out but not much,

I get a constant stream of warmth coming from the back of hades (hence
the name). But then it's an Athlon IV, and they're known for running
hot. They're also known for tolerating insane temperatures: the CPU and
mobo fans both failed for a week, the mobo started to smell as the
lacquer or something went, but the CPU just chugged on, unaffected.

hades is actually very quiet; by far the quietest of my boxes. amaterasu
has the four fast 1U-fitting fans in it, and loki has a cheap noisy disk
cooler in. I really must replace these.

Alas, if I put variable-speed fans into amaterasu I fear that the
machine will blink its fan-failure light because they're not drawing
enough current. It still has three of the noisy-*** Sun fans in it:
I'd replace them all but Sun have the (US-style?) swapped earth and live
leads in the fan cabling, which is rather annoying to fix if you've got
coordination as crap as I have. So I'm replacing them as they die. Plus
the grilling at the front of a 1U case might as well be regarded as a
noise conduit into the room...

(If you know of a source of quiet 1U fans cabled the way Netras like
them I'd be overjoyed to know :) )

> and I don't close the rear door completely
> as the wires for my home-made radio time sync clock run out of it to

Bore holes in the door. (Hell, I'm boring holes in the *ceiling* when
I move amaterasu into the loft, if the noise it makes up there is
tolerable. I've got to get power and networking and serial console
leads up there somehow.)

The radio sync clock thing is something I've thought sounded nifty, but
it seems a rather too hardwarey thing for me. (Plus, the radio
interference from all the UTP in my house batters radio comms something
chronic, and recabling is way too much effort for something I can get
just as easily by syncing to a reasonably good time server. In any case,
amaterasu has the OpenPROM bug that leads to it gaining about 15 minutes
a day, so precise millisecond-accurate syncs seem a little over the
top. I have to sync every ten minutes just to keep the time synched to
within a second.)

> the clock box on the top of the rack. Also I bent the rack getting it
> out the back of my landy so it lists slightly to one side meaning the
> door doesn't like closing that much ;-)

Oops. Been there (only I knocked one machine on the case door when
getting it *into* the moving van).

>> Really? My disks downstairs run at ~45C even with 1U fans, and one
>> of my older ones upstairs stabilises at 62C if you don't shove air
>> over it (although with enough air it's cooler than the downstairs
>> disks).
>
> I'm thinking more of the size of the disc as well as the temperature,
> it's a larger piece of kit than just a processor and probably has a
> lower maximum temperature than the processor does. The size means it
> chucks out more heat even though it's at a lower temperature, and
> needs a bit more effort to cool it.

Oh, very true. A few really crude back-of-the-envelope calculations
indicate that you're right: a CPU chip would need to be at about a
thousand C before it chucked out as much heat as a disk running at 60C.
I doubt even an Athlon runs that hot. :)

> They're not normally a problem
> though but when you start cramming several 1U servers on top of each
> other, two to a case,

Well, *that* is crackers. Use 1U cases and set up cooling properly.
This is sort of the *point* of rackmountable cases, isn't it?

(The 2U case I was blathering about buying will have only *one* machine
in it, of course.)

>> That's... disturbing. A board built with MAME in mind?
>
> Yep, it even has "shifted" keys to allow you to control mame. If I

Nifty! Personally I find MAME very hard to use because the keybindings
are so baroque. The source is fun to read, though, if only because those
old machines are often so peculiar. (Plus the 6502 and Z80 and stuff
like that are blasts from the past for this child of the microcomputer
age...)

> is a P3 800 and it manages fine, although vector graphics are much
> slower now after I upgraded to mame 0.99, I'm thinking of downgrading
> it again to 0.65 or thereabouts. The earlier version ran games like
> Asteroids, Battlezone, Space Harrier and Double Dragon at full speed
> on a P450, my P800 can't quite do it any more and has dropped a lot on
> vector games.

I'm a bit surprised that *vector* stuff slowed down: it was generally
the most computationally trivial to implement. I guess the quest
for total accuracy has costs...

>> *boggle* the *speed* has reduced? What on earth? Is this some sort of
>> elaborate patent-evasion thing or something?
>
> Not far off. The official aim of the MAME project is to emulate the
> hardware itself, and not to make playable games.

.... and this is what I use it for (perhaps I'm unusual in treating
it as a fun read).

> The playable games
> are a side-effect of the hardware emulation. In this manner, they
> regard a hack that makes the game playable without emulating the
> hardware properly as a temporary fix to make the already emulated
> hardware testable, then when time permits they revisit the hack and
> make a proper hardware emulation of that bit of the game, which slows
> it down.

Well, of course. Do it *right*. (Also, a lot of things really depend
on precise timings, undocumented instructions and so on, so the more
like the original hardware you can make it, the better. You might
slow some things down and make some others that never worked right
suddenly start to function.)

> manufacturers know that it exists (MAME devs have conversations with
> them on occasion) but overlook the copyright violation as it's not
> costing them anything and keeps interest in their brands alive.

How... unexpectedly sane of them (I really never expected that of the
secrecy-wild Japanese ones).

> MAME has hardware emulations of things like 3DFX cards but won't use a
> real 3DFX card in a machine,

Now that really *is* peculiar. Probably also trivially fixable (even if
distributed as a separate patch). I mean, what would be a better
emulation of the hardware than the hardware itself?

>> We're thinking of putting it on polystyrene blocks and seeing what that
>> does.
>
> Should help a lot, especially if it catches fire ;-)

Oh, that's what all the fibreglass in the loft is for. (Cold water tanks
catch fire all the time, donchaknow.)

>> Passive-cooled -> <400MHz, and probably <200Mhz, as even my P233 got a
>> case hot enough to scorch when the fans failed. The last machine I owned
>> that I could truly run fanless was esperi, my 486.
>
> Yes, but I'm talking mini-itx motherboards here with cool-running x86
> clones

Bah, if you want cool, go ARM.

> up to 600MHz with fanless heatsinks, and 1GHz and above with
> heat pipes.

A term I don't know. How do you pipe heat? A solid copper tube?

>> They Intermezzo people are all working on Lustre now, which *rocks*, but
>> doesn't quite serve the same goals as Intermezzo did (as in, it's meant
>> for big, big, *big* distributed filesystems, how many petabytes do you
>> want with that?)
>
> Hehe, I might have a look at that. Especially if it can be used to
> turn unused hard disc space on my machines into a shared storage
> area. I have lots of machines with 20 or 40 gig discs doing nothing
> but holding a 3 gig operating system.

Well, if you want to waste space, in a month or so you should be
able to stick Recant on it. Say what you like about database-backed
filesystems, space-efficient they aren't.

(However, I hope that the identical-block-merging optimization I'm
implementing, combined with the sharing of storage between different
versions of the same file, and the automatic compression of large rows
implemented in PostgreSQL might win some of the space back. Alas I have
to throw that space away again on indexes and block-read-and-write
frequency statistics and on forward-propagation of old-yet-frequently-
read data, because unless I blow space on it it's going to be dog slow
as well as bloated. But all my competition is really slow as well: *no*
version-control systems are especially nippy.)

--
`Logic and human nature don't seem to mix very well,
unfortunately.' --- Velvet Wood
.

Quantcast