Hmm, mostly water (80%) with some other things (anti-freeze, etc.). There goes the dreams people had of Quads (non-Dual cored), no way two of these units would fit in the current case and still have solid airflow.
Wonder* how the weight difference is between all this extra gear and the previous massive heatsinks?
OK, this confirms what Kasper had posted in the news back in February(?) about heat issues of the new processor. At the time, no one believed that the 90 nm processor could have such massive heat issues. And I think the numbers about power dissipation are misleading, since we don't know how heat per surface unit (and not heat alone) compares between the 90 and 130 nm G5 chips.
The 90nm 970fx generates a typical 50W @ 2.5GHz, the 13nm 970 generates 51W at 1.8GHz. That is a power density of 42W/cm^2 for the 970, and 83W/cm^2 for the 970fx. Considering that active cooling such as this can easily dissipate 700-1000 W/cm^2 it would seem to me that Apple has a great deal of room to grow in the future. A theoretical 3.0GHz 970fx with the same voltage and die size would generate a typical 72W, or 120W/cm^2 which is well within the cooling capability of the system.
The 90nm 970fx generates a typical 50W @ 2.5GHz, the 13nm 970 generates 51W at 1.8GHz. That is a power density of 42W/cm^2 for the 970, and 83W/cm^2 for the 970fx. Considering that active cooling such as this can easily dissipate 700-1000 W/cm^2 it would seem to me that Apple has a great deal of room to grow in the future. A theoretical 3.0GHz 970fx with the same voltage and die size would generate a typical 72W, or 120W/cm^2 which is well within the cooling capability of the system.
Thanks for the numbers Fluffy. So, while the chips have the same power dissipation (much lower per clock in the 970FX), the power density literally doubled. This explains the massive cooling system. And it is good to know how much room it gives to Apple to grow.
The 90nm 970fx generates a typical 50W @ 2.5GHz, the 13nm 970 generates 51W at 1.8GHz. That is a power density of 42W/cm^2 for the 970, and 83W/cm^2 for the 970fx. Considering that active cooling such as this can easily dissipate 700-1000 W/cm^2 it would seem to me that Apple has a great deal of room to grow in the future. A theoretical 3.0GHz 970fx with the same voltage and die size would generate a typical 72W, or 120W/cm^2 which is well within the cooling capability of the system.
Wonder* how the weight difference is between all this extra gear and the previous massive heatsinks?
* Too lazy to look it up
Unfortunately, the only thing Apple mentions is that the new G5s are 44 lbs., which "var[y] by configuration and manufacturing process." (I haven't looked at the PDF files found on that page.) The previous version was given as 42 lbs.
Posting pictures of an already-released product's insides does not, in my mind, constitute a breakage of NDA.
Not unless you post pictures that someone else holds the copyright to. Since they're not shipping yet, I seriously doubt they took the photos themselves, meaning that they likely got them from other sources who may not have appreciated their photos released.
Silicon scaling is dead (for now). Intel knows it, IBM knows it and AMD either knows it or is just too stubborn to admit it.
1. with leakage current the scaling of power requirements is not linear. It gets progressively worse the more current you feed (which in turn necessitates feeding even more current) -> you can't extrapolate linear firgures from 2.5GHz modes as higher speeds need more power.
2. The figures taken from IBM literature are typical thermal design power figures. They do not represent maximum thermal dissipation numbers, which can be twice as much. The cooling has to be built to withstand the maximum thermal dissipation situations peaks as well.
3. The Cooligy figures are for the heat transfer element itself. Total thermal power dissipation that the cooling system can muster is defined by radiator, air flow and surrounding air temp. Apple new G5 designs have a relatively small radiator
Let's face it, everybody has thermal issues at 90n.
IBM and Apple would have pumped out 3Ghz if they had been able to pull it off (personally I don't believe it's just a system controller / whole system thermal issue, it's a chip/scaling issue).
Apple has publicly had to come down on their promise that they can't deliver 3GHz now or for some time. I think this is quite telling.
I think the new machines are looking quite spiffy.
Of course, it would've been nice to get fixed IEEE1394B, SATA II (with NCQ support on the controller), PCI-Express and what not, but there's still time for that.
Still, for me it's another year worth of wait. G4/1.46GHz is fast enough for what it does (non-video graphics work).
So, the wait continues.
PS The Apple-X.Net is off-line and so are the images (at the time of writing this).
I think that simply by getting a 90nm chip into one of their machines is a sign that they've pretty much crossed that hurdle, and the 3.0GHz ramp-up won't be so hard.
Comments
Wonder* how the weight difference is between all this extra gear and the previous massive heatsinks?
* Too lazy to look it up
Originally posted by Fluffy
The 90nm 970fx generates a typical 50W @ 2.5GHz, the 13nm 970 generates 51W at 1.8GHz. That is a power density of 42W/cm^2 for the 970, and 83W/cm^2 for the 970fx. Considering that active cooling such as this can easily dissipate 700-1000 W/cm^2 it would seem to me that Apple has a great deal of room to grow in the future. A theoretical 3.0GHz 970fx with the same voltage and die size would generate a typical 72W, or 120W/cm^2 which is well within the cooling capability of the system.
Thanks for the numbers Fluffy. So, while the chips have the same power dissipation (much lower per clock in the 970FX), the power density literally doubled. This explains the massive cooling system. And it is good to know how much room it gives to Apple to grow.
Originally posted by WebFlits
Source: Apple-X.net
More images
The link is invalid.
Originally posted by Placebo
The link is invalid.
They were there yesterday. I'd say Apple Legal had a busy day on Friday.
CV
Originally posted by FormerLurker
They were there yesterday. I'd say Apple Legal had a busy day on Friday.
Posting pictures of an already-released product's insides does not, in my mind, constitute a breakage of NDA.
Originally posted by chris v
Their whole site is, like, gone. WTF?
Apple sends e-mail to host. Host decides to disable site first, ask questions later
Originally posted by Fluffy
The 90nm 970fx generates a typical 50W @ 2.5GHz, the 13nm 970 generates 51W at 1.8GHz. That is a power density of 42W/cm^2 for the 970, and 83W/cm^2 for the 970fx. Considering that active cooling such as this can easily dissipate 700-1000 W/cm^2 it would seem to me that Apple has a great deal of room to grow in the future. A theoretical 3.0GHz 970fx with the same voltage and die size would generate a typical 72W, or 120W/cm^2 which is well within the cooling capability of the system.
Golly, you are one smart mother ?ucker
Originally posted by IonYz
Wonder* how the weight difference is between all this extra gear and the previous massive heatsinks?
* Too lazy to look it up
Unfortunately, the only thing Apple mentions is that the new G5s are 44 lbs., which "var[y] by configuration and manufacturing process." (I haven't looked at the PDF files found on that page.) The previous version was given as 42 lbs.
Originally posted by Placebo
Posting pictures of an already-released product's insides does not, in my mind, constitute a breakage of NDA.
Not unless you post pictures that someone else holds the copyright to. Since they're not shipping yet, I seriously doubt they took the photos themselves, meaning that they likely got them from other sources who may not have appreciated their photos released.
1. with leakage current the scaling of power requirements is not linear. It gets progressively worse the more current you feed (which in turn necessitates feeding even more current) -> you can't extrapolate linear firgures from 2.5GHz modes as higher speeds need more power.
2. The figures taken from IBM literature are typical thermal design power figures. They do not represent maximum thermal dissipation numbers, which can be twice as much. The cooling has to be built to withstand the maximum thermal dissipation situations peaks as well.
3. The Cooligy figures are for the heat transfer element itself. Total thermal power dissipation that the cooling system can muster is defined by radiator, air flow and surrounding air temp. Apple new G5 designs have a relatively small radiator
Let's face it, everybody has thermal issues at 90n.
IBM and Apple would have pumped out 3Ghz if they had been able to pull it off (personally I don't believe it's just a system controller / whole system thermal issue, it's a chip/scaling issue).
Apple has publicly had to come down on their promise that they can't deliver 3GHz now or for some time. I think this is quite telling.
I think the new machines are looking quite spiffy.
Of course, it would've been nice to get fixed IEEE1394B, SATA II (with NCQ support on the controller), PCI-Express and what not, but there's still time for that.
Still, for me it's another year worth of wait. G4/1.46GHz is fast enough for what it does (non-video graphics work).
So, the wait continues.
PS The Apple-X.Net is off-line and so are the images (at the time of writing this).
Originally posted by Existence
AMD is still at 130nm and is king of the hill in performance. They're moving to 90nm later this year and will break the 4000+ rating barrier.
easier said than done
Originally posted by Existence
AMD is still at 130nm and is king of the hill in performance. They're moving to 90nm later this year and will break the 4000+ rating barrier.
Another trollicious post courtesy of Existence. Why are you still here?
Edit: Nevermind. It's just me.