The Sound of silence( cool computers)

[Floid]

Dan wrote:

PSU noise I can live with, but  CPUfan  is to loud.
I looked inside the case and found a 6cm Spire 12V fan and an aluminium heatsink. Would a big slow case fan help?

http://www.zalman.co.kr/english/product/cnps6000Cu.htm

Looks like what I want.
How loud is 20dB? compared to PSU fan?

Pretty quiet, likely.  The thing I don't like about Zalmans is less their mediocre results (not that extra cooling is important, but I like to 'nudge' AMD chips to their highest safe FSB - thus I had my 850MHz T-Bird unlocked to run at 133MHz/266DDR, resulting in a marginal overclock - 866MHz - but that much faster memory access)... than the fact that their fan brackets down to your slot covers, and thus requires one of those ricey side-grilles to function efficiently (and makes it dangerous to run your machine while swapping cards in those slots, obviously).

Personally, I ran my 850, thus-clocked, with a Thermaltake Mini Super Orb, but the noise drove me insane and the fan bearings started going rather quickly.  I replaced it with one of Antec's Molex-style units, which, while not killer in silence, was much less annoying, and performed reliably up until I took the system apart.  (I gather Molex's real designs are even better, as the Antec's fan was sort of permanently molded on to their shroud.)  

This is the Antec I'm thinking of; they were available at CompUSAs and probably Staples, rated up to 1.4GHz or so.  That guy's demonstrating a big DON'T in the installation, though - you should never push down on a heatsink on an exposed-core chip, as the leverage applied is a recipe for cracking; put pressure only on the clip, even if it takes you a while to get it latched.

AMD also has excellent datasheets on their site; the chips are *designed* to run up to 60C or so, depending on model.  They list dissipations and all that, so you can comapre your Duron to the models the various heatsinks around are rated 'up to.'

Soundproofing is isolation which would require a faster airflow, wonder how silent it would be in if i put it in a closet or made a wooden frame around it?

If you're not in the room, I'd hope you can't hear it.  Wood conducts sound, so that wouldn't be much help unless you're adding padding (like one of those old dot-matrix printer enclosures).  People are fond of car audio products, like Dynamat; I've always been interested in trying some Noisekiller or Roadkill (spray-on deadening foam products, rather hard to track down, and a bit pricey).

Does the Hush Mini ITX EPIA M10000 Nehemiah 1000Mhz run without a fan???
Would love to run AROS on a fanless MiniITX board.

Yep.  Via chips are your only option with the Hush products for now.

So it´s possible to run a highspeed G4 without a fan?

No-doubt depends on the G4.  The Beige G3 here (based on the original IBM 750 design; I forget if it had letters after it) has no "fan," but it has a heatsink the size of your fist, two gigantic, loud power supply fans, one that ends up directly over the CPU with the cover closed, and a piece of metal that apparently serves as a thermal bond (attaching the heatsink to the metal of the case) as much as a grounding contact.

Stupid pc engineers why dont they mount the processor on the otherside of  the motherboard then  it would be possible to use the whole side of the tower as a heatsink. And making a powersupply that replaced the top of the pc case with a heatsink would be easy. The outside of the case has an unlimited airflow after all so it wouldn´t need a fast, if any fan.

Last I checked, most people put their machines places like under desks, in racks, etc.  People were fond of killing their Mac cubes by resting papers or CDs atop the 'chimney.'  (Hmm, maybe they shouldn't have made it flat..)  Surfaces can only radiate heat so fast, safety and RFI guidelines have to be met, and so forth.  (Pop open your power supply sometime, and touch one of the heatsinks- in many designs they're hot, and I don't mean in temperature.)  Manufacturers are also mostly interested in ideas that can apply equally to racks and other server-type enclosures, now that desktops are a low-profit commodity game.

Now, some of the recent Shuttles use heatpipes and radiators with large-diameter fans; perhaps you should check out that hardware, and see if it can fit in your case (or if it'd be cheaper to swap your chip into the usual case/mainboard combo they're sold as)...

[Floid]

B00tDisk wrote:
Compaq's Presario desktop PII systems (Slot1) didn't even have a CPU fan - just an aluminium heat sink the size of a PopTart (about 4x6 inches) that was directly in line for airflow from the front of the case to the power supply - that is, air was sucked in the front by the PS fan and blown out the back, and along the way passed over the CPU-HS.

This was standard operating procedure for just about any slot-CPU system made by an OEM.  (I'm not sure if any 'passive' Slot-A Athlons were ever shipped; common for the Intel camp, anyway.)  All was not roses, though, as some chips/OEM designs shipped with undermassed heatsinks and fans that weren't up to the task- bad enough on a desktop, but imagine a lab of machines that'd all flake out as ambient temperatures varied.  (This garnered specifically from a Slashdot post I read a while back from someone fed up with their specific Compaqs, yes.  It's not hard to imagine.)

Horrible systems overall though; they finally (?) abandoned the whole PCI-slots-on-riser-card with the later Presario desktops and just stuck the slots on the MB (900mhz and up).  I'm talking about the systems they provided for businesses, now, not home systems.  Never touched a non-business Compaq...

Believe it or not, the riser design is a standard - NLX or LPX, I forget which.  Deskpros were the business line, Presarios were consumer (and, after the launch of the initial PS/1 - iMac looking box, and a really neat looking 5x86? LCD lunchbox that I can't seem to find mention of anywhere nowadays, became mostly Generic Super-Socket-7-Motherboard in Generic Minitower Case... I guess they had a few with Generic Slot-1 Motherboard, too.  Someone gave me a really awful model from the Presario's transition period from 'luxury crap' to 'generic crap,' a hefty desktop with nifty integrated speakers and amp, stock with a K5 or K6 of some sort, driven by the worst/first SDRAM chipset Intel ever made for that Socket.)...  Risers would be great if the way they bolt them down didn't negate all the advantages.

After working with those for a while, you do develop something of a fetish for Torx screws.  (Gotta love the way they provide spares!)

[Floid]

B00tDisk wrote:
On the underside of each heatsink frost had formed, and yes, Virginia, if you stuck your finger to it and held it there your fingertip would freeze onto the HS!

Those were likely using Peltier elements.  They push heat across a gradient, given an electrical boost- cold side gets cold, hot side gets hot, and you'd better have a fan on the hot side to avoid burning the thing out (and/or having the waste heat 'leak' back to the thing you're trying to cool).  Things are:

1. They use a fair bit of electricity, 40W or so.
2. They can only pump so much heat; putting an underpowered unit on your chip would be like wrapping it in fiberglass insulation - worse than using a passive heatsink with good conductivity.  Sure, you're seeing it make frost in *air* (where the thermal load is effectively zero), but press a lightbulb against it for 20 minutes and see if the temperature at the interface is still that cold.
3.  It's making frost.  Frost is water; at warmer temperatures, frost is dew.  A Peltier running with enough power to cool as good/better than a 'passive' heatsink and fan will be getting cold enough (especially on parts not contacting the hot chip) to condense water.  This means you have to go nuts caulking and foaming your mainboard, socket, and cooling system.  The expensive active cooling systems from Kryotech use (used to use? I dunno if Kryotech's still in business) refrigerant and compressors, as in your freezer, and they had some sort of heating strip around the edge of their contact pad to avoid condensation.

This from a guy who killed his Socket 7 board with drippage from a few years of Peltiering -- and never got a solid overclock for the effort.  K6-2s had other limiting factors; cooling isn't the only thing that governs the highest stable clock of a chip.

They were, IIRC, "golden orb" fan/hs combos.

Golden Orb, IIRC, is Thermaltake's brand name for a now-ancient series of Socket 370 coolers, famous for looking bling, performing okay, and crushing Athlons to death before Thermaltake tried to redesign the clip, at which point more Athlons were crushed to death and they started marketing them as Socket-370 only.

Of course, anyone could pick up some Peltier elements, and thermal-epoxy them to the bottom of their favorite (cheapest) heatsink or fan.  It'd also be pretty easy to store your heatsinks in your beer cooler (or even better, one full of dry ice), occasionally swapping out a fresh one to awe the rubes.

[Floid]

For what it's worth, this is a proper Golden Orb.  A passive cooler making *frost,* not *dew* on a desk in Florida?  Erm.  I seem to recall something about dry ice/liquid N2 snake-oil demos brought up in the 0v4rcl0ck4r circles a while back, but it's been a long time since that (and since the Golden Orb would ever have been adequate)...

Remember, a heatsink that feels cold *under load* isn't doing its job, and/or means your fan is overkill.    I know the Orb design(s), and the base was pretty massive, while the cute fins - especially on the Super Orb - didn't really soak up or transfer the heat that well.  However, a block of metal that hefty could easily stay cold for hours after a night in a freezer (or a night with your PC off), allowing for such sales tricks, and providing more than adequate performance for the casual user.  (My crappy Antec provided slightly higher numbers, with half the fans and a thousand less RPMs than my Super Orb, and the added advantage/cheat of its blow-on-the-chip-package design.  However, the temperature stopped creeping upwards as the box gained uptime.  That said, I've no fear for AmigaOnes with the Tt bling, since the wattages involved are nowhere near what we're talking with the older AMDs.)

Heck, if they were selling Peltiers, there's your answer- stack them on the cold sides of the peltiers during show setup, and they're probably chilled through by the time the rubes arrive.  How many people are going to stop back multiple times over the day... and of those, how many are going to point fingers, vs. laughing with them?

Anyway, the big huge problem we had with them was end-users trying to treat them like towers by putting them under their desks on their sides.

Problem: The HS was heavy enough to gradually drag the CPU out of the horizontally mounted socket; the system was never intended to be put this way.  Convincing the end users otherwise was a royal pain in the ass.

Thanks for the warning, actually -- if I ever get back to adminning the Deskpro-Under-the-Desk, I'll think twice before putting it in a towerizing bracket.  That said, the SECC clips on my particular model seemed firm enough that it'd take a good earthquake to accomplish same.

LPX.  The riser cards were just something else to break, and break they did...hate 'em.

They're "interesting" to pull in and out, yeah.  On my unit, the AGP card would come with; if they'd made the rest of the cardcage removable in similar fashion, it would've actually made things easy.  Probably more a boon to their manufacturing, as they could doubtless stick populated risers in before clipping the mainboard to the case... or maybe they couldn't, seeing as it *is* Compaq.

After working with those for a while, you do develop something of a fetish for Torx screws.  (Gotta love the way they provide spares!)

Yeah, I always thought that was "polite" of them.  The later Deskpro TORX screws were dual-slotted; you could either use a flathead screwdriver or a Torx screw.  Did you acquire a "Compaq Tool"?  (Multihead screwdriver with a bajillion different sized Torx, flat, and Phillips heads).[/quote]Hm, I've met the dual-slotted, but IIRC they didn't cut/mold the slot all the way through, so you'd still need the Perfect Flathead Screwdriver Never Invented.  I had a cheezy 'PC Toolkit' I got one holiday, but it served fine - it's thin-slotted Phillips screws that always cause pain.

Right now I'm using an old Intellistation, which is nice in a different way - IBM used little bolts made of a similar Indestructibilium, with slotted heads, about 6 or 7mm high... and all the 3.5" bays have come with blue rubber damping grommets for the screws, while the 5.25" bays pop out *easily* with a single unbolting, and the goofy molded faceplates pop off to reveal equally-aesthetic standard sized bays.  Good luck getting the case panel off, though -- it's screwless, *if* you remember you're supposed to flex-and-shove it over its latch with the little handle provided -- and it'd require some major disassembly to see how they rigged up the built-on RFID hardware!  (Disableable in the BIOS.)

[Floid]

Waccoon wrote:
Frankly, I've been regretting sticking with AMD ever since I bought it.  If AMD doesn't get their act together, cap their cores, and start working on intelligent throttling like Intel, I'm jumping ship for good.

They have with the Opteron, and given the Danish? suit over crackable cores, seems like they'll be using it for the Athlon64 as well... especially since the first A64s will be in the Opteron pinout (with a pin snipped for whatever reason).  Meanwhile, throttling/safety-shutoff is available, it's just that mainboard makers are responsible for implementing the appropriate reaction to output from the thermal sensor included in recent cores; it's supposed to be enough to save your butt if your fan fails.  Everything I've read says the Opteron's (rather low) dissipation numbers are much more 'worst case' than prior figures, and perhaps lower than the competing Intel chips', but AMD does want the board/system builders to leave room for any future faster, hotter chips.

Which chip are you running?  You have to keep track whenever you go shopping, be it AMD or Intel; whenever any company makes the final push for an old core or process, the dissipations go to hell (and admittedly, AMD's been there for a while, but I gather Barton? was much improved, at least at the initial speeds.) ... admittedly, AMD's been in that unsavory position for the past year, given the delays on x86-64.  Looks like dissipations on the recent 32-bit ("consumer") chips are hovering around 50-60W nominal, with 70-something peak.  A look at the latest P4s shows the 2A at 52W/68W peak, on up to 82W/70W peak??? for the 3.20C; admittedly, "Thermal Design Power" has some sort of different meaning in that camp.  I have no idea if those are dogs or not, as I don't follow Intel, but it doesn't look too pretty.  Thing to remember is, the AMDs are designed to tolerate some heat -- as long as it's below the maximum, there's no reason to freak out over getting to down to 30 or 40C.  I'd cough up the current number, but I don't feel like re-retrieving the PDF, ugh.

For those who asked:
Athlon XP Datasheets -- Yes, you want the real datasheets; the thermal info is buried in section 7.1ish in them.  Durons and Opterons are linked from that page, over on the right.

Intel's P4 Datasheets are over here.  Section 5.1ish.

Meanwhile, if the P4 is ever throttling during normal use, you've got problems there.  Both companies keep shoving out Wunderchips somewhat impractical for normal use (a problem not limited to CPUs anymore, given the first GeForce FXen)... I'd suggest hanging behind the pack, picking off the 'sweet spots' once they hit the $50-$150 mark, depending on application.  (Of course, my last personal investment was my venerable 850, and keeping the upgrade path open is harder when courting Our Lady Intel of the Hundred Sockets.)

Best advice, even if I'll get flamed?  Examine your needs closely (as in: stop shooting for 30C), and try a vendor other than Thermaltake for a while.  They're the Abit of the cooling world, much loved for no apparent reason... and meanwhile, some Reg or Inq article pointed out that AMD are making their retail heatsinks suck less.