Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Liquid Cooling Photos
- Thread starter MacRumors
- Start date
- Sort by reaction score
If I had known that it would take filling a Mac with water to get a G5 processor to work in it, I would have bought my new G4 Powerbook MUCH earlier. Thought, I suppose the PB G5 rumors did get me to wait for the 33 Mhz speed bump - not that I notice the difference.
I always think of these things in terms of longevity. That liquid cooling thing looks like it would be the first thing to break.
I always think of these things in terms of longevity. That liquid cooling thing looks like it would be the first thing to break.
MacFan25 said:
JEEZ
You people act like apple just invented something. Liquid cooling has been around for a LONG TIME. Go read up about it for god's sake.
This whole message board is like something from the twilight zone. A bunch of 'computer nerds' who know *jack* about computers.
Its like watching a bunch of monkeys poking at some kind of time travel machine or something.
this thread...
I am reminded by everyone here how much I miss/romanticize my '67 VW Beetle. She was air cooled with fins and had no radiator. A very reliable auto. The engine was in the rear, providing traction. She was also extremely loud and hot. I hope someone can take something from that.
I am reminded by everyone here how much I miss/romanticize my '67 VW Beetle. She was air cooled with fins and had no radiator. A very reliable auto. The engine was in the rear, providing traction. She was also extremely loud and hot. I hope someone can take something from that.
Neodym said:
I'll go out on a limb and treat this as a serious post.
Were Apple to introduce a third line of machines, I think it will have to be either a single-processor G5 tower that's intended as a consumer machine, or a really high-end workstation that actually taxes the liquid cooling system. A iMac-like non-iMac makes no marketing sense whatsoever. You might as well just kill the flat-panel iMacs and sell eMacs alone.
A sphere? And what is this magical computer going to sit on? Do we have to go to a hobby store and buy those cast-iron potpourri stands to stick our computers in, or will apple offer some tripod system that lets you pretend it's a theodolite?
Liquid cooling would probably fit in the current machine if you had a headless design that only connected a monitor. The motherboards already allow for 2GB of RAM, though you have to take them out to get at one of the sticks, and entry-level graphics are the sort of thing that people have been whining about for ages. Since the new PowerMacs didn't get ATI X800 XTs, I find it hard to believe that the X300 entry level card is what you're talking about.
What?
The machine has double the internal volume of the current design, but you think it would add the unnecessary expense and, more importantly, reduced performance of a laptop HD? That's just silly.
Okay, this is now smelling even worse than before. "Wireless Firewire" is a transfer protocol, not an actual technology for moving data. It's going to be built on a standard that has yet to be finalized, known as 802.15.3, which wouldn't have the bandwidth to serve peripherals, nor the power to do so without a separate AC source for the so-called breakout box. Also, unless the LCD screen is powered by ADC, that's another powercord if you're not using an Apple display.
Someone's not checking their facts before they post.
USB2.0 for a single port is 480Mbit/s, FireWire 400 is 400Mbit/s, and yet the backbone for "Wireless FireWire" is planned to intro at 55Mbit/s, which doesn't even begin to feed the ports you've claimed are going to be in a wireless box. This says nothing of feeding the DVI port that you mentioned, but which can't be on the machine since the "housing has no external ports except for a power connector."
Oops.
Actuators to control the machine's position, on top of liquid cooling and laptop drives, not to mention cutting-edge wireless technology that does impossible things? I thought we were talking about the consumer machine, not something that would cost $5000 to build, let alone sell at a profit.
This is the most likely thing that has been said so far.
Pardon me...
BWAHAHAHA!
That's the best thing yet. How in the world are they going to apply and "aggressive pricing" scheme to a machine with even more new technology than the PowerMacs?
The cooling system used in the 2.5GHz G5 looks quite innovative to me. For a while now, PC users have been able to purchase water cooling kits for their PC's, but Apple's liquid cooling implementation is very neatly done and is better than most liquid cooling solutions on the market for Windows machines.
Does Anyone know or have an idea about how much hotter the 2.5Ghz G5 chips run? I know that the 130nm 2Ghz G5 chips actually ran at nearly 1/3 to 1/2 the temperature of a 3.4Ghz P4. I would not be suprised if the operating temperature of the new 2.5Ghz G5s are actually still cooler running than a P4 or for that matter even the Xeon chips. I have read that the 90nm 2.0 Ghz G5 chips in Xserve run at an even cooler temperature than the 130nm 1.8Ghz G5 chips. So with this in mind I think the LCS in the new Dual 2.5Ghz was not out of neccessity and the machine could have been air cooled albeit at the expenese of increased noise. I am hoping the LCS in the 2.5Ghz MP G5 is just a precursor of things to come, 2.5Ghz could have been done without LCS perhaps the LCS in the 2.5Ghz is a testbed that will prepare us for 3Ghz+ chips that would actually require it. All of this is speculation though until we actually find out the exact operating temperture of these new 2.5Ghz G5s. I think we are stuck at 2.5Ghz today not solely because of heat considerations but probably more so for very poor yields for 3.0Ghz G5 chips. I am quite confident that by the end of the year or by sometime in Jan 05 we will see 3.0Ghz machines announced and or shipping.
don't confuse temperature with heat
There are really two somewhat related issues here. Temperature (degrees) and heat (watts).
The PPC970 series emits somewhat less heat (the specs for the 2.5 have not been released) than the top P4/Xeon systems - probably in the 1/2 to 2/3 range.
The 90nm PPC970fx, however, is a much smaller (in area) chip than the P4/Xeon. That means that the heat is much more concentrated - so the temperature is potentially much hotter. (Consider a 25 watt light bulb vs. a 25 watt soldering iron - the light bulb is much cooler due to the larger surface area, even though the heat is the same.)
Temperature, therefore, depends on the heat sink. Since silicon chips have a recommended temperature range, I suspect that that running temperature is probably pretty similar between the PPC970fx and the Pentiums - they'll both have good enough heat sinks to avoid getting too hot.
I suspect that Apple's use of the extreme cooling system is because they are overclocking the IBM chips. (It's happened before with Apple, right?) The chip might not be rated at 2.5GHz, but if kept cool enough some of them might work well enough.
One reason for this suspicion is that Apple is using 2.0GHz in the Xserve G5, not faster. IBM is also quoting 2.2 GHz for a future JS20 Blade. It's unusual to have such a gap in the specs of the chips, unless the 2.5 GHz are really overclocked 2.2 GHz.
This quote is great: (http://www.apple-x.net/modules.php?...&POSTNUKESID=87662d781995aaafc88835539bba2c50)
Seems to me to be a pretty extreme measure - especially if you need to take hazardous waste precautions when working on your computer!!
Little Endian said:
There are really two somewhat related issues here. Temperature (degrees) and heat (watts).
The PPC970 series emits somewhat less heat (the specs for the 2.5 have not been released) than the top P4/Xeon systems - probably in the 1/2 to 2/3 range.
The 90nm PPC970fx, however, is a much smaller (in area) chip than the P4/Xeon. That means that the heat is much more concentrated - so the temperature is potentially much hotter. (Consider a 25 watt light bulb vs. a 25 watt soldering iron - the light bulb is much cooler due to the larger surface area, even though the heat is the same.)
Temperature, therefore, depends on the heat sink. Since silicon chips have a recommended temperature range, I suspect that that running temperature is probably pretty similar between the PPC970fx and the Pentiums - they'll both have good enough heat sinks to avoid getting too hot.
Little Endian said:
I suspect that Apple's use of the extreme cooling system is because they are overclocking the IBM chips. (It's happened before with Apple, right?) The chip might not be rated at 2.5GHz, but if kept cool enough some of them might work well enough.
One reason for this suspicion is that Apple is using 2.0GHz in the Xserve G5, not faster. IBM is also quoting 2.2 GHz for a future JS20 Blade. It's unusual to have such a gap in the specs of the chips, unless the 2.5 GHz are really overclocked 2.2 GHz.
This quote is great: (http://www.apple-x.net/modules.php?...&POSTNUKESID=87662d781995aaafc88835539bba2c50)
Seems to me to be a pretty extreme measure - especially if you need to take hazardous waste precautions when working on your computer!!
&RU said:
I think the reasoning is that the current case has the fans mid-case, and so you'd want the radiator as close to the fans as possible. The fact that it blows back through the mechanism is an unfortunate side effect, but the alternative would have been to relocate the fans to the rear of the case and make them pull instead of push air, which might have been less efficient and certainly would have been louder.
At least this way, the radiator is cooled with ambient air, not air which has already been heated by the board (minor, minor thing).
I'm confused by the two heat pipe/radiators in the back of the setup (or, visually, on the lower left on the largest picture posted in the .sit file above). I can only assume that what they're sucking heat from couldn't be mounted on the other side of the board and so couldn't be cooled by the liquid cooling system.
That is an exceedingly long limb.thatwendigo said:
I think it'll be something like here. I mean, what with the inductive power and wireless firewire, why have it touch anything at all?thatwendigo said:
Terribly bad Photoshop hack attached...
Well, maybe Apple has "Wireless FireWire 2.0" ready - perhaps it's done via a laser link. Wouldn't want any nasty cables.thatwendigo said:
Well, you have to hand it to him, if they sold such a thing at under US$2K, I'm pretty sure Apple'd sell millions of them. I'd buy it for the wireless power supply, the advanced wireless 800Mbps+ technology, the floating sphere, and the liquid cooling system. I might be a bit put off by the small hard disk, but those laser communications devices do take up a lot of room.thatwendigo said:
Attachments
AidenShaw said:Seems to me to be a pretty extreme measure - especially if you need to take hazardous waste precautions when working on your computer!!
A lot of things in a computer are considered hazardous waste. Batteries, CRT tubes, and circuit boards are all considered harzardous was by the EPA.
pjkelnhofer said:
Hazardous for disposal, not hazardous to work on.
Q: How can you tell a PowerMac repairman?A: He's the one with the static strap on his HAZMAT suit !!
edgarj said:
I'm one of those that doesn't tinker with my computer. It's good to know that Apple has built a solid system.
As aafuss1 commented that the cooling system in the G5 2.5 is more innovative than those found in the PC kits. Just what I would expect from Apple.
I also agree that the AppleCare is a necessity.
benpatient said:
Liquid cooling for the pc has been around for ages since I can remember its only in the last 2 years companies have imergered selling specific components.
Apple said 3GHz, so where is it? I said they would run into cooling problems and that they would not make it and the mac boys were hopeing for 3GHz++.
As for the cooling system its just heatpipes nothing special, they were just adapted to handle more heat, so I dont know why apple calls it a "sophisticated liquid cooling system" and they designed it, err no. Dont compare this to properly built liquid cooling built by the likes of me and others, it doesnt compare in terms of performance and versatility so what if they have to br refiled but the cooland HAS to be changed simply because it impairs performance. (Alluminium and copper will react all you can do is slow it down)
Water is also out with the 3M Fluorinert coolant coming into use (higher performance, non conductive, non corrusive and doesnt need to be changed ever).
KLEOS
Wow. It's been a while since I saw so many backwards ideas in one post.
I guess you missed IBM's announcement that they had more significant problems going to 90 nm than they anticipated. The "3 GHz by next summer" milestone was abandoned a while back.
No, it's not a heat pipe system. A heat pipe system doesn't have a pump. This system has a pump. It's a closed-loop cold-plate system.
It's sophisticated, in part, because it uses both liquid cooling and air cooling at the same time. Check out the photos from the service manual. It's a dual system that can, undoubtedly, handle a great amount of heat with minimal airflow over the heat exchangers.
It's also sophisticated because it's entirely sealed and modular. No leaks, no muss, no fuss.
You got that wrong, too. Fluorinert isn't "coming into use." It's been around for years and years. But because it's so expensive, it's only used for immersion cooling, where the electronics are actually dunked into a tank of liquid. (See the Cray 2 and the T90 for examples. They're basically aquariums filled with fluorinert, with circuit boards suspended in them.)
For obvious reasons, immersion cooling is unbelievably expensive and difficult to maintain. It's only suitable for very large supercomputers... and even in that arena, it's disappearing. Consider the brand-new Cray X1, one of the fastest commerically available computers in the world. It uses cold-plate cooling, not immersion cooling. And the coolant is just water. Tap water, right out of the building's water supply. (No need for antimicrobial agents because the water doesn't remain in the system. It circulates in from the municipal water supply, gets warmed by the machine, and circulates back out to the municipal supply again.)
No, there's nothing revolutionary about Apple's cooling system. But there's something pleasingly evolutionary about it. It'll let us run much hotter CPUs in our G5's without making them noisier, which is a big plus.
army_guy said:
I guess you missed IBM's announcement that they had more significant problems going to 90 nm than they anticipated. The "3 GHz by next summer" milestone was abandoned a while back.
army_guy said:
No, it's not a heat pipe system. A heat pipe system doesn't have a pump. This system has a pump. It's a closed-loop cold-plate system.
army_guy said:
It's sophisticated, in part, because it uses both liquid cooling and air cooling at the same time. Check out the photos from the service manual. It's a dual system that can, undoubtedly, handle a great amount of heat with minimal airflow over the heat exchangers.
It's also sophisticated because it's entirely sealed and modular. No leaks, no muss, no fuss.
army_guy said:
You got that wrong, too. Fluorinert isn't "coming into use." It's been around for years and years. But because it's so expensive, it's only used for immersion cooling, where the electronics are actually dunked into a tank of liquid. (See the Cray 2 and the T90 for examples. They're basically aquariums filled with fluorinert, with circuit boards suspended in them.)
For obvious reasons, immersion cooling is unbelievably expensive and difficult to maintain. It's only suitable for very large supercomputers... and even in that arena, it's disappearing. Consider the brand-new Cray X1, one of the fastest commerically available computers in the world. It uses cold-plate cooling, not immersion cooling. And the coolant is just water. Tap water, right out of the building's water supply. (No need for antimicrobial agents because the water doesn't remain in the system. It circulates in from the municipal water supply, gets warmed by the machine, and circulates back out to the municipal supply again.)
No, there's nothing revolutionary about Apple's cooling system. But there's something pleasingly evolutionary about it. It'll let us run much hotter CPUs in our G5's without making them noisier, which is a big plus.
Fluorinert is in use and is cheap in small quanities Armari UK quoted me £54 for 500ml when I bought some and you dont need to add anything to it like with water, there is no submerging. The advantages are non-conductive, higher performance, non corrusive and the coolant doesnt react will copper and allumium combinations.
If I remember isnt this the cooligy system?
Has apple implemented ant redundantcy with the pumps, I always have at least 2 pumps regardless and I can hotswap.
If I remember isnt this the cooligy system?
Has apple implemented ant redundantcy with the pumps, I always have at least 2 pumps regardless and I can hotswap.
army_guy said:
According to Calculator.app, £54 is $98, and 500 ml is a pint. Bottled water costs about 12¢ per pint. Water from the tap is so cheap by the pint that I can't even express it in terms of cents. The fractions are meaningless.
In other words, even bottled water bought retail is is eight hundred times cheaper than fluorinert.
I wouldn't call that "cheap in small quantities." I wouldn't call that cheap in any quantity at all.
Jeff Harrell said:
I agree with just about the entire post, but, on this point - as a minor thing - I do think that, in addition to the pump system, there are two smaller heat pipes - at least that's what it looks like to me, attached to the heat sinks (not to the radiator). It looks like they are separate and not liquid cooled.
Of course, the pictures provided, while wonderful, aren't complete enough for me to be certain.
Jeff Harrell said:
As a rough guess, I think the system would take maybe 250ml of fluid or so, which would run maybe US$50. Not that horribly pricey, but 1/60th the price of the entire system. Too big a chunk. The water is probably ultrapure, so maybe it costs, oh, $1. At most. If shipped from the other side of the planet.
So I think Apple went with the cheapest solution that worked. Good for them. As far as dual pumps (army_guy) are concerned, I doubt that it would make much of a difference. Looks like you'd have to take the whole thing out anyway. One pump is simpler, cheaper, and sufficient. It's all self-contained, so I'm guessing that there will be very, very few maintenance issues. Redundant pumps would, cost aside, be nicer. But they likely aren't necessary. At least not on a dual-2.5.
Little Endian said:
The original 130nm 970 2.0ghz chip runs at 51 watts typical, and the 90nm revision 970fx 2.0ghz runs at 24.5 watts typical, 49 watts maximum. It could easily be the case that the 2.5ghz is running at a similar wattage to the original processor, but this presents a whole other problem.
The 970fx is produced on a 60mm^2 die, as opposed to the original 970's hefty 118mm^2 die, giving less than half the area to dissipate the heat. If we figure this on a wattage to area ratio, we come up with figures of:
130nm 970 2.0ghz (118mm^2)
--0.43watts per mm^2 (typical wattage of 51)
--0.86watts per mm^2 (maximum wattage of 102 - I'm assuming this figure from the typical)
90nm 970fx 2.0ghz (60mm^2)
--0.41watts per mm^2 (typical wattage of 24.5)
--0.81watts per mm^2 (maximum wattage of 49)
The coolings needs at the same clockrate remains the same with the lowered wattage, but if you force it to climb, then you see something else. Let's say that the 970fx at 2.5ghz is the same as the 130nm part at 2.0ghz:
90nm 970fx 2.5ghz (60mm^2)
--0.85watts per mm^2 (typical wattage of 51)
--1.7watts per mm^2 (maximum wattage of 102)
Suddenly, you have twice the heat per millimeter to be getting rid of! Here are numbers of the Northwoods and Prescott P4s, just for comparison:
130nm Pentium 4 Northwoods 3.4ghz (145mm^2)
--0.6watts perr mm^2 (typical wattage of 89)
--1.23watts per mm^2 (maximum of 178)
90nm Pentium 4 Prescott 3.4ghz (122mm^2)
--0.84watts per mm^2 (typical wattage of 103)
--1.68watts per mm^2 (maximum of 206)
Those who are astutely paying attention might notice that the 970fx and Prescott have similar wattage per mm ratings, but that the 970 is a vastly smaller core that still does some pretty impressive computing. The size is right at half the other chip's die.
Most likely the LCS is intended as a proof of concept, showing that Apple can create a system to cool the next generation of processors. As the process is ironed out and improvements are made, I expect to see the 970fx's heat profile inch down in wattage, but it's not there just yet. Interesting that a 90nm 2.0ghz chip uses two cores on a latency-prone SMP framework to equal the 3.4ghz Pentium 4 at the same heat expenditure but fewer watts overall.
---
Edit: Somehow, it's appropriate that my 666th post be one about the die size and heat of the Pentium 4.
thatwendigo said:
I totally agree (oh crap, this is like the second or third time I have agree with thatwendigo in the past week - maybe I am losing it). I don't understand why every one saying that the 2.5GHz G5 needs the Liquid Cooling. I suspect that it could have been cooled with the same heatsink and fans as the previous G5's. I think they learned there lesson with "WindTunnel" MDD G4's and wanted to keep noise down before it became a problem.
I agree with those that say it is a little scary to spend your money on brand new technology, remember that while Apple's LCS is new to them, the technology itself has been used in computers for quite some time now. I for one am excited to see Apple using it because it of the future possibilities it opens up.
jsw said:
Why?
- Interference. The more EM you involve in a design, the more potential for crosstalk and other issues, especially within enclosed spaces. This is the basic principle that's causing problems at the 90nm process size for everyone and their cousins. Additionally, I bet it would make it extraordinarily hard to use any 802-series wireless technology, because an inductive powersource would have to use a conductive surface that would, by its nature, be an EM source. Then there's the shielding for your RAM and drives, since you'd be pumping a nominally magnetic current all over the place...
- Magnetism. As noted above, that globe is suspended by magnets. What do hard drives use to write? Magnets? Before someone points out that the PowerBook and iBook use magnetic latches, I'd like to add that a magnet strong enough to lift If you extrapolate screen sizes and weights from the iMac tech specs page and the Apple Cinema Display page, you come up with about 12 pounds of display and 10 pounds of computer. That means you'd need an electromagnet that would stably lift 20 pounds of computer. Good luck on not causing problems with that.
- It's not currently possible to completely wirelessly interface 2 FireWire 400 ports, 3 USB 2.0 ports (that's 2160Mbit/s, or roughly twice the 150MB/s speed of wired SATA, right there), along with the audio and video-out. This. is. ridiculous.
I repeat: 2160Mbit/s for the peripherals alone. That's more than 40 times the bandwidth of current 802.11g, and doesn't take into account anything other than firewire or USB. Also, unless you're running some kind off power cord to this mystial breakout box, you'd need a battery in it to provide the same kind of powering options you get from ports directly on the machine.
This is just stupid.
Apple would go out of business if they tried to sell a machine like that. It wouldn't compete performancewise with a PC, it would be hideously overcomplex, and using way too much unproven technology. If people think that Apple's repair bills hurt now, imagine what happens when your liquid cooling system breaks down and leaks all over your induction power system, causing an electrical fire. Jesus.
Also, I don't know where you're getting this "wireless power supply" nonsense. No matter how it's done, there will always be wires, even if they're from some kind of base station to the wall.
thatwendigo said:
It must not have been apparent, but that post of mine was ENTIRELY tongue-in-cheek. Sorry! I was just kidding with the obviously off-the-wall stuff that was predicted.