I’m bleeding, making me the victor!

Right now Intel and AMD have less to fear from Apple than they do from Qualcomm – the people who can do what they need to do with a Mac and want to are already doing that, it’s businesses that are locked into the Windows ecosystem that drive the bulk of their laptop sales right now, and ARM laptops running Windows are the main threat in the short term.

If going wider and integrating more coprocessors gets them closer to matching Apple Silicon in performance per watt, that’s great, but Apple snatching up their traditional PC market sector is a fairly distant threat in comparison.

Rotovap on a flytrap.

The problem is that the private sector faces the same pressures about the appearance of failure. Imagine if Boeing adopted the SpaceX approach now and started blowing up Starliner prototypes on a monthly basis to see what they could learn. How badly would that play in the press? How quickly would their stock price tank? How long would the people responsible for that direction be able to hold on to their jobs before the board forced them out in favor of somebody who’d take them back to the conservative approach?

Heck, even SpaceX got suddenly cagey about their first stage return attempts failing the moment they started offering stakes to outside investors, whereas previously they’d celebrated those attempts that didn’t quite work. Look as well at how the press has reacted to Starship’s failures, even though the program has been making progress from launch to launch at a much greater pace than Falcon did initially. The fact of the matter is that SpaceX’s initial success-though-informative-failure approach only worked because it was bankrolled entirely by one weird dude with cubic dollars to burn and a personal willingness to accept those failures. That’s not the case for many others.

NASA in-house projects were historically expensive because they took the approach that they were building single-digit numbers of everything – very nearly every vehicle was bespoke, essentially – and because failure was a death sentence politically, they couldn’t blow things up and iterate quickly. Everything had to be studied and reviewed and re-reviewed and then non-destructively tested and retested and integration tested and dry rehearsed and wet rehearsed and debriefed and revised and retested and etc. ad infinitum. That’s arguably what you want in something like a billion dollar space telescope that you only need one of and has to work right the first time, but the lesson of SpaceX is that as long as you aren’t afraid of failure you can start cheap and cheerful, make mistakes, and learn more from those mistakes than you would from packing a dozen layers of bureaucracy into a QC program and have them all spitball hypothetical failure modes for months.

Boeing, ULA and the rest of the old space crew are so used to doing things the old way that they struggle culturally to make the adaptations needed to compete with SpaceX on price, and then in Boeing’s case the MBAs also decided that if they stopped doing all that pesky engineering analysis and QA/QC work they could spend all that labor cost on stock buybacks instead.

Another perovskite hype piece. You’ll know that they’ve got something that’s commercially viable once they’re making these sorts of efficiency claims and not omitting information about cell degradation.

Given that the first perovskites studied had lifespans that could be measured in minutes, this is great progress, but the fundamental problem is that as a class of materials they just don’t want to exist outside of an inert atmosphere. Without significant progress in stability and encapsulation materials, they’re more of a research curiosity than a viable real-world PV tech.

A combination of resting on their laurels during AMD’s lost decade, and failure to retain competitive process technology during the extended gestation and ultimate failure of their non-EUV 10nm node. The arrogance of taking their foot off the gas and assuming nobody would ever catch back up to them backfired hard.

The reverse. OceanGate saw how planes were being built and said, “let’s do that for submersibles!” even though in airplanes, composites are subjected to <1 atmosphere of tension loading and <2g aerodynamic loading, whereas their submersible was going to be subjected to >400 atmospheres of compression loading, and a much more corrosive environment.

Composites in aircraft have a fairly long and uncontroversial history, and there’s nothing inherently wrong with them in that application. The biggest problem with composites is what happens with them at the end of their service life. Finding ways to recycle them without compromising safety is a good thing, and if it weren’t for Boeing having such a damaged reputation at the moment I think nobody would bat an eye.

The average American house on a basement will have something like 40 m^3 of concrete in its foundation. If all of it could be utilized, that’s still ~12kWhr of storage capacity. Nothing to be sneezed at.

Any time you see perovskite-based cells mentioned, you can assume for the time being that it’s just R&D. Perovskites are cool materials that open up a lot of neat possibilities, like cheaply inkjet-printing PV cells, but they have fundamental durability issues in the real world. When exposed to water, oxygen, and UV light, the perovskite crystals break down fairly rapidly.

That’s not to say that the tech can’t be made to work – at least one lab team has developed cells with longevity similar to silicon PVs – but somebody’s going to have to come up with an approach that solves for performance, longevity, and manufacturability all at once, and that hasn’t happened yet. I imagine that when they do, that will be front-and-center in the press release, rather than just an efficiency metric.

This is a reference to upscaling algorithms informed by machine learning a la Nvidia’s DLSS – seems like AMD is finally going to add the inference hardware to their GPUs that will let them close that technological gap with the competition. I’m guessing it won’t come until RDNA5, though.

This is actually becoming somewhat commonplace. For example, in many cutting-edge cancer therapies, blood is drawn from the patient, processed in tissue-culture suites on site to extract the patient’s immune cells and sensitize them to some marker expressed by their specific cancer cells, and then the modified immune cells are returned to the patient room and transfused back into their bodies. It’s not cheap per se but it’s something that most top-tier cancer centers can do, and to do the similar process of extracting stem cells, inducing them to transform into pancreatic islet cells, and transplanting those into the patient’s pancreas isn’t that big of a jump – and it’d be cheaper than a lifetime of insulin in any case. It also points the way towards treating other kinds of organ failure without the risk of rejection, too.

Data center cooling towers can be closed- or open-loop, and even operate in a hybrid mode depending on demand and air temps/humidity. Problem is, the places where open-loop evaporative cooling works best are arid, low-humidity regions where water is a scarce resource to start.

On the other hand, several of the FAANGS are building datacenters right now in my area, where we’re in the watershed of the largest river in the country, it’s regularly humid and rainy, any water used in a given process is either treated and released back into the river, or fairly quickly condenses back out of the atmosphere in the form of rain somewhere a few hundred miles further east (where it will eventually collect back into the same river). The only way that water is “wasted” in this environment has to do with the resources used to treat and distribute it. However, because it’s often hot and humid around here, open loop cooling isn’t as effective, and it’s more common to see closed-loop systems.

Bottom line, though, I think the siting of water-intensive industries in water-poor parts of the country is a governmental failure, first and foremost. States like Arizona in particular have a long history of planning as though they aren’t in a dry desert that has to share its only renewable water resource with two other states, and offering utility incentives to potential employers that treat that resource as if it’s infinite. A government that was focused on the long-term viability of the state as a place to live rather than on short-term wins that politicians can campaign on wouldn’t be making those concessions.

The Wikipedia article for these little monsters describes the males aggressively fighting over females mid-mating, to the point of killing some as they attempt to tear them away from one another, and then squeezing the eggs out of their dead bodies to fertilize them… Gonna guess it’s the same one.

I’m a lab planner, and sometimes getting researchers to describe what sort of containment device they need for a given process is like pulling teeth.

• Chemical fume hood? That’s a hood.
• Class II, Type B2 BSC? Also a hood.
• Class II, Type A2 BSC? Believe it not, hood.
• Laminar flow bench? Yep, that’s a hood too.
• PCR dead air box? Somehow also a hood.

Like, surely you’re not doing BSL-2 work in a LAF? Please tell me you’re not doing that.

I exaggerate – but Magic Rock is doing booming business installing strings of natural gas generators at Buc-ee’s across the state, and I’m currently dealing with an institutional client who wanted to provide backup power for a satellite campus, and didn’t even stop to consider battery-backed PV on the way to asking for a natural gas generator farm.

It’s not a coincidence that Texas is a hotbed of development for “microgrid” systems to cover for when ERCOT shits the bed – and of course all those systems are made up of diesel and natural gas generator farms, because Texans don’t want any of that communist solar power!

I’ve got family in Texas who love it there for some reason, but there’s almost no amount of money you could pay me to move there. Bad enough when I have to work on projects in the state – contrary to the popular narrative, in my personal opinion it’s a worse place than California to try and build something, and that’s entirely to do with the personalities that seem to gravitate to positions of power there. I’d much rather slog through the bureaucracy in Cali than tiptoe around a tinpot dictator in the planning department.

The only link I am aware of is that Intel operates an R&D center in Haifa (which, it happens, is responsible for the Pentium M architecture that became the Core series of CPUs that saved Intel’s bacon after they bet the farm on Netburst and lost to Athlon 64). Linkerbaan’s apparent reinvention of the Protocols of the Elders of Zion to the contrary, the only real link seems to be that Haifa office, which exists to tap into the pool of talented Israeli electronics and semiconductor engineers.

Historically AMD has only been able to take the performance crown from Intel when Intel has made serious blunders. In the early 2000s, it was Intel commiting to Netburst in the belief that processors could scale past 5Ghz on their fab processes, if pipelined deeply enough. Instead they got caught out by unexpected quantum effects leading to excessive heat and power leakage, at the same time that AMD produces a very good follow-on to their Athlon XP line of CPUs, in the form of the Athlon 64.

At the time, Intel did resort to dirty tricks to lock AMD out of the prebuilt and server space, for which they ultimately faced antitrust action. But the net effect was that AMD wasn’t able to capitalize on their technological edge, Ave ended up having to sell off their fabs for cash, while Intel bought enough time to revise their mobile CPU design into the Core series of desktop processors, and reclaim the technological advantage. Simultaneously AMD was betting the farm on Bulldozer, believing that the time had come to prioritize multithreading over single-core performance (it wasn’t time yet).

This is where we enter the doldrums, with AMD repeatedly trying and failing to make the Bulldozer architecture work, while Intel coasted along on marginal updates to the Core 2 architecture for almost a decade. Intel was gonna have to blunder again to change the status quo – which they did, by betting against EUV for their 10nm fab process. Intel’s process leadership stalled and performance hit a wall, while AMD was finally producing a competent architecture in the form of Zen, and then moved ahead of Intel on process when they started manufacturing Zen2 at TSMC.

Right now, with Intel finally getting up to speed with EUV and working on architectural improvements to catch up with AMD (and both needing to bridge the gap to Apple Silicon now) at the same time that AMD is going from strength to strength with Zen revisions, we’re in a very interesting time for CPU development. I fear a bit for AMD, as I think the fundamentals are stronger for Intel (stronger data center AI value proposition, graphics group seemingly on the upswing now that they’re finally taking it seriously, and still in control of their destiny in terms of fab processes and manufacturing) while AMD is struggling with GPU and AI development and dependent on TSMC, perpetually under threat from mainland China, for process leadership. But there’s a lot of strong competition in the space, which hasn’t been the case since the days of the Northridge P4 and Athlon XP, and that’s exciting.