It’s also interesting Apple is cooling both sides of the chip. Is that just for efficiency so the computer is quieter? Intel/AMD chips that put off way more power aren’t cooled from both sides right?
https://techcrunch.com/2022/03/17/apple-execs-on-developing-...
The heat sink in the iMac Pro was where things really started to shift. It keeps the Xeons in those machines reasonably cool while being whisper quiet at idle/low load and reasonably quiet at load. They really kicked it up a notch with the Mac Pro, which remains nearly inaudible at all times, and now that’s carried forward into the Mac Studio.
Really nice work. I’d love to see this kind of innovation in more standardized PCs — the NH-D15 and front fans in my custom built tower do a good job on thermals but their acoustics leave something to be desired, despite Noctua coolers and fans being some of the quietest on the market.
Is there any chance at all that six months from now Dell or HP will wow us with something that is faster, quieter, and as aesthetically interesting as the Studio? Or will they crank out another version of a black sheet metal case that is loud and hot and inexpensive? There have to be people in those companies that want to do something interesting, right? Or do they assume that anybody with money and taste are already out of the Windows market?
Maybe it is all the fault of Windows. There's a pretty good story about the problems of Windows elsewhere on HN right now. The comments there are so damning it makes me think nobody believes in the platform anymore. You know how there's no M in FAANG? Is Microsoft stuck with people not good enough to work elsewhere? There might be no point in building interesting hardware that is ultimately going to run Windows.
My old Intel MBP would sound like a jet engine just by opening Chrome.
This one… I hammer it all day with multiple servers running, JS projects, tons of tabs, Docker containers, many misc. software utilities… not a single noise.
I’ve only heard the fans once while using Handbrake.
I don't see a real difference though. Mine doesn't overheat but the fan kicks in often anyway.
Plugging in an external screen enables the discrete GPU so your energy usage and temperature will go up.
I am planning to get a Mac Studio as the replacement for that iMac, but I would be happier, if easy cleaning of the fans would be possible.
It’s basically never an i9, because once it hits 70°C it starts throttling quite rapidly and then stays stuck between 1.0 and 1.4 GHz clock speed for at least 5 minutes, and what good is an i9 running at 1 GHz?
Their laptop cooling has been absolute dogshit for the past (at least) 10 years.
It’s more of an inadequate cooling combined with Intel releasing CPUs that run just hotter and hotter.
I can’t wait to upgrade to a 16” MBP with an M1 Max (or next gen if they release it by then). There are still some essential tools in my workflow that don’t work all that great on ARM.
Interestingly, sharing an external screen using google meet and running something like npm install will easily throttle the CPU even if the CPU usage never shows anything above 15% of use.
I thought that was interesting, so I looked for it on the video: https://youtu.be/IY0gRMpT4AY?t=878 If there's a difference in blade pitch, it's not visible from that angle. What is pretty clear is that the split isn't exactly halfway down, maybe 55/45 instead of 50/50.
Their marketing photos aren't any clearer: https://pbs.twimg.com/media/FNa32k2XoAMCDiW?format=jpg&name=...
You would probably need to cut an impeller up to see a difference, or xray it.
I would also be interested if someone swaps out the blowers and see what difference the Apple impeller makes. I would be surprised if it's more than 5%. But injection molding makes fancy geometries very cheap, and if you're making the molds anyway you might as well pick up the easy design win. Amortized over however many millions they're going to make, the NRE costs are going to be pennies per unit.
i was going to make the same statement.
is chiplet design the future? lets put together two powerful chip to make one super powerful chip. i guess its one way to beat the Moore's Law
Hasn't AMD been doing it for the past half a decade?
Working near the limits of manufacturability will result in occasional defects. If you only have to have a few eight-core 10x8mm chiplets and a more reliable 12nm process node interconnect die come out defect-free, that's much easier than trying to get all of that on a single giant die with 64 cores and the IO on the same die.
To exaggerate, if each chiplet has a 50% chance of failing, on average you have to make 16 chiplets to get one good multi-package chip. But you'd need (2^8 * 8) = 2048 to get one contiguous super-die. with no defects.
Here's the Pro, with 8 or 10 CPU cores and 14 or 16 GPU cores, depending on die yield issues where they might have to shut off a couple cores:
https://www.apple.com/newsroom/images/product/mac/standard/A...
Here's the Max, with 10 CPU cores and either 24 or 32 GPU cores for the same reason:
https://www.apple.com/newsroom/images/product/mac/standard/A...
And here's the Ultra:
https://www.apple.com/newsroom/images/product/mac/standard/A...
It's literally two M1 Max bolted together with more RAM around the periphery.
If you make an M1 Max wafer there are a lot of ways to spin out 8/24-core Maxes, 10/32-core Maxes, and adjacent 10/32 core Maxes into an Ultra. Nothing in the patent seems to indicate they have to decide which ones to use before testing the wafer...