This is what I was thinking - RAM guys have been stacking chips for a hell of a long time. First as actual package stacks on the DIMM - I remember seeing some very dense modules in the DDR2 days with stacked packages; and then as you correctly point out: Samsung has been layering dies inside the package for years as a microscopic version of stacking chip packages.

It's also why DIMMs have heat spreaders on them now.

My guess is that they aren't targeting performance, but rather making a lower power system-on-chip that really is a fully-featured system-on-a-chip and incorporates lots of low-power and low-heat peripheral crap like I2C / serial / USB / SATA in addition to RAM, flash storage, NIC, etc. - put the highest wattage bits on the top for direct interface with the heat spreader, and stack the other stuff below with some thermal magic in the sandwich to move as much heat from the lower layers to the edges as possible so you aren't adding to the thermal load of the CPU core from below.

This kind of thing could be really cool in the low-power embedded / industrial controller space where nobody is looking for laptop performance out of a chip. But you are trading one complexity for another: instead of having to use a lot of geometric area to mount and connect all the peripherals to the CPU, you end up with a shitload of thermal management problems for a very compact system without the geometric area requirement.

Unfortunately, that geometric area really helps with the thermal problems.

This was my thought too - if you have essentially 6 layers of silicon with insulators between to create 600% of the transistor density, you're also consuming >600% of the power (nothing is ever 100% efficient) and therefore producing >600% of the wattage to dissipate without a corresponding size in radiative surface area.

How do you not cook the center of the cube when we're already throwing 70W into a single chip the size of your fingernail? Maybe central heat pipes that each layer hooks up to, running vertically through the die? And how much area per layer do you lose to that, at what increased manufacturing complexity (read: cost and reduction in yields)?

I'm sure those are all answerable engineering questions to present if the value is there. And my guess is that since it's very obvious that stacking chips is a sure-fire way to increase transistor area, that the value hasn't been worth the added complexity to solve the inherent problems because die shrinks were always cheaper and easier to do... right up until they aren't.

It's good that someone is asking the question and showing that it can be done. But I'd wager [paywall so couldn't RTFA to confirm] that they aren't exactly stacking up the highest performance Xeon or Epyc chips 41 high and running them at full throttle.

Or more reasonably: Russia or China "accidentally" dropping / dragging an anchor across it's path and crippling the US Government cloud-based operations for weeks by fucking over their prime cloud contractor.

You don't have to think too hard to figure out what Amazon will do with it.

They have massive datacenters at either end, in Loudon County, Virginia (US-East-1, GovCloud) and Cork, Ireland.

This is going to be their own private backhaul connecting regions in North America and Europe.

They're already planning on drilling a tunnel for hundreds of miles under an ocean. I think handling a bay that is tens of miles across won't be a problem.

More than that, apparently they forgot about the eMate and the iBook lines of low-cost notebooks Apple has made over the decades.

You don't know anything about how taxes in Oregon work.

Hint: we have a property tax that pays for stuff at the county level. 9-digit construction cost datacenters are worth quite a bit of property tax to counties that are largely agricultural and would otherwise have a low tax base.

But do go on only considering one slice of the governmental revenue pie without looking at the much larger slice sitting right next to it.

Read what I wrote again, and then try again.

Hint: the city's budget is 20x what it was 15 years ago. They aren't spending that on fireworks and glossy election mailers - they're building infrastructure and services for people who live there.

You're assuming that all land-based cable routes are equal in cost and time to create. That would be a silly assumption, as obtaining grants of right-of-way through various legal jurisdictions is a collection of bespoke bureaucratic and political processes, any one of which that fails dooms the whole project.

It would make sense for them to either lease some dark fiber that already exists that gets them to a suitable piece of property to start their tunnel bore, or to minimize the amount of jurisdictional nonsense they would have to contend with to derisk the project.

Example: Loudon County, VA > Montgomery County, MA > Prince George's County, MA > Anne Arundel County, MA > Queen Anne's County, MA > Atlantic Ocean.

You aren't going to do much better than that - you can skip basically any and every city and only have to transit the Chesapeake, but you're drilling under a friggin ocean so I don't think that's going to be much of a problem.

Running cable on land requires getting right-of-way and property easements, which requires government consent. The more government jurisdictional boundaries you cross, the more consent you need from various political organizations that may or may not be willing to help them out.

It would follow that they would try to minimize that as much as possible in order to cut down on all the bureaucratic noise and squabbling between various states, cities, counties, etc.

My guess is that they either chose a path that minimizes all of that, or they're leasing pre-existing dark fiber to some property they were able to get on the cheap on the coast that meets their requirements, and then they start their tunnel bore from that location.

People seem to be overthinking various aspects of this, and dramatically underthinking the things we already know are the case: the more jurisdictions you cross, the more legal bullshit you have to deal with from grasping politicians always looking to squeeze an extra nickel for their campaign fund, or make a name for themselves by standing up to the big bad bully megacorporation. It wouldn't surprise me if they optimized on a route that currently dodges all incorporated cities / townships and just stays in county jurisdictions in order to keep any grasping motherfuckers from trying to pass a "transient data tax" or some shit to stick it to AWS.

I remember just 12 months ago an orange guy in a badly fitting suit promising to bring prices down, and a bunch of fools believed him and voted for him.

Someone should ask those folks how that's working out for them.

Feel free to start your own version of Slashdot with whatever nationalist bend you like if you're not happy with how this place is being run.

I'm pretty sure you've gotten your money's worth out of Slashdot.

We're talking about AWS's us-west-2 region.

You can't just pull up stakes and move that. It was put there for three very good reasons: cheap land, plentiful hydropower from the Columbia River, and also loads of nice cold water for cooling from the same river. And, if they're serious about nuclear power it's very close to both the Columbia River Nuclear Generating Station and Hanford.

It's been there for a decade and is home to a double-digit percentage of all the stuff AWS runs. Context matters.

You didn't even read the summary, which specifies that someone who used to be a walmart employee is making 3 real estate deals a month now based on the local growth. Or that the city's budget is 20x what it was 15 years ago.

Does that sound like "nothing trickles down" to the people who actually live there?

Try at least reading the summary - it's right there, you don't even need to click anything.