I just got a dually 800 G4 cpu, but it came sans heat sink. I'm replacing my Quicksilver's original single 733 (OC->800), but single CPU sink won't work with the duals.
Sooooo, I reworked the single cpu heat sink into a dual cpu heat sink. Simply put, I cut out the middle and replaced it with a larger block of copper. Details to come . . .
dan k
better keep an eye on your thermals
Using different pieces for the heatsink isnt always a great idea, because the thermal energy then has to jump another gap. And copper and aluminum against each other without the proper coating will corrode each other (electrollisis). Having the copper block and the aluminum heatsink both lapped will make a big difference in how much of that energy makes it to the fins of the aluminum heatsink.
details page here
Good point, especially as this is a so-far unproven homebrew mod. I'm going to pick up a cpu thermometer so I can keep an eye on temps, plus it'll be a good thing to be able to actually measure the results of further mods.
Agreed, the more gaps the heat must jump increases the potential for problems. However, I've got a nice tight fit between the two parts, plus any tiny gaps are filled with heatsink goop.
I don't think electrolysis is an issue, the two parts have a layer of goop between them.
Lapped? You mean smoothed flat? Yes, both parts are quite flat and smoothed with fine wet-n-dry paper.
dan k
I suggest you at least try the modified heatsink on the lower-heat single CPU module to see if it works. It's better to lose the old one than to lose the new, more powerful one.
best way to do it:
get a 1/8" thick piece of aluminum plate, cut and drill it to fit over the two CPUs and mount it to the CPU board with the 4 heatsink holes. then get some adhesive cooling goop and glue two single-proc heatsinks to the plate.
or use one single proc heatsink if its a huge one.
Only in the presence of an electrolyte of some kind containing ions, such as H2CO3, H2SO4 , water etc...
So a physical dry contact will not do alot...
remember: too much goop actually hinders efficiency. use as little as possible.
to make sure you use the right amount, put a little glob of it on one end of the copper block, and scrape the glob over the entire length of the block with a razorblade. itll fill only the necessary places and improve heat transfer. dont press too hard or youll scrape it all off. you want to see an almost invisble layer of the stuff on the block
Yeah? Tell that to the copper-aluminum contact areas inside an airtight sealed fiber optic distribution node. All it takes is the slightest amount of humidity in the air to make them corrode. Yes it will corrode MUCH faster if salt or something is around, but the basic water in the air is enough to do it. Its probably not a huge deal in this because you arent conducting electricity, only heat. I was just posting it up there as an FYI.
Your probally correct about the fiber optics etc as I have no experience in that field.
Even so, many comercial heatsinks have a copper base and alluminium fins. The tiny amount of water present in the air could theoretically case corrosion, but it would be minor if visible... Remember, there is little contact with the air between the copper and alluminum due to the air excluding heatsink goop.
The only problem with electrolitic corrosion that ive seen is inside a cheap watercooling setup, which had an alluminium water block and copper fittings. The block was litererally eaten away. Adding some anti-corrosion additive to the water the block stopped being eaten. Btw, anyone know how that stuff works??
Ahh for water cooling yeah it would be bad.. Inside your hot water heater is an aluminum rod, this acts as an anode, while your copper pipes int he house act as the cathode. The aluminum rod should be replaced every few years if you have real hard water (namely well water). The anti-corrosion stuff most likely just lower the PH lever of the water, to try and prevent the corrosion from happening again.
The heatsink got pretty warm so I added a second fan. Now it gets less hot. Damn, I'm too lazy to describe this new detail, have to snap a pic to illustrate.
I ran iTunes playing Handel's Messiah and Visualizer (!) all day Christmas with nary a hiccup so I guess it must be cooling well enough.
dan k
The best test to overclocking that I know of is to run dnetc for as long as you can. If yu can run for a week solid without a crash, then your most likely fine. If your running OS9 though, your stuck doing other things.
As for monitoring the temp of your procc. i use some software called "Temperature Monitor". Just search for it at www.versiontracker.com and you'll find it. It's free, and it displays the CPU temp in a few different ways.
The software only works on Aluminum Powerbooks, G5's, iMac G4/G5's, and possibly MDD's. Other G4's do not correctly show temps, and G4's after the 7410 don't even have internal temperature sensing (the above listed machines that do work have external thermisters for getting the temps).
But only the 750. My iBook doesn't have the Temp. Lead going into the Proc. I think my 040 PB 190cs Also had a Temp lead that I could have it displayed with the right program. That is how I first knew that Taking temperature from the inside of the Processor was possible.
EDIT: Just tried it on the iMac G3 (PPC 750) DV and it doesn't have temp. leads.
Funny, It works on my PIsmo [scratches head] And it uses the Same version of the Proc. Also, the same speed
Which iMac? Teh original iMac (233-333) did. The later slot loads all had 750CX's in them, not 750L's. The 750L (aluminum core G3made by mot) had correct temp reading. (copper 750's did not, like the 333 beige with the 1MB L2 did not show temps correctly, although it would show a temp).
Under The Hardware overview, it says:
Machine Model: iMac
CPU Type: PowerPC 750 (83.0)
Number of CPUs: 1
L2 Cache (per CPU) 512KB
Memory: 320MB
Bus Speed: 100 MHz
Boot ROM Version: 4.1.9f1
PCI/AGP Cards
Type | Bus | Slot
ATY,Rage128Pro Display AGP
Display
The Config is as follows
400/500MHZ 750 CPU
CD/DVD
iMac DV
Airport
The L2 Cache has been overclocked to 200MHz, up from the Original 160MHz. (I have seen a difference between the cache speeds.
Sorry 'bout that. Im using it on my 17" PowerBook G4 and my MDD Dual G4 and it works on both, so i guess i forgot about it's limited compatibility.
to my page.
dan k
Your quake 3 scores are pretty low for a dually. Are you sure you have r_smp 1 set?
I really haven't been going for outright Q3A speed, I'm more interested in quantifying the value of various system changes. I don't really play Q3A much, and in any case frame rates over 40fps or so is more than adequate for decent gameplay.
I'm curious of course what sort of speed is 'normal' for similar systems. My now-dual-800 QS has 512MB ram, 7200 rpm/8MB cache HDs and a NVidia Ti4600 display card, running latest OS X release.
By comparison, my similarly specced 1.5gHz P4 with 768MB and a Ti4600 running win2k-SP4 is much faster, IIRC it gets speeds around 130fps (with no tweeks.)
dan k
The speed jump you saw was just fromt he L3 cache. To take advantage of that 2nd processor, go into the cfg file and do a search for smp and change the zero to a 1, then save the file. The cfg file is located in ~/library/application support/quake3/config.cfg
Ah ha, thanks! K, that did bump it up a bit, speed zoomed from 74fps to 107fps. That's about where I'd expect it to be based on my PC's results, taking into account of course superior wintel NVidia drivers.
dan k