frezik

Creating a FOSS EV is all do-able right now with off the shelf motors and batteries. Welding a frame would take some skill. How to title it would depend on the local government rules; many states in the US have a kit car designation for this sort of thing, but not all do.

If it's built rigid like a race car with a roll cage, four-point harness, and at least a DOT rated helmet for everyone inside (if not Snell), it could be safer then most cars on the road. If it's not very large, then it's probably safer for pedestrians and bicycles, too.

I don't expect air bags to be viable. It takes a lot of tuning to get them right, and they can be worse than nothing if not done right (they're basically a controlled explosion). However, the race car-like design above, plus helmets, would keep you safer than any air bag. Road cars converted to track cars often disable or remove the air bags. The rules of the event may even require it. They're counterproductive dead weight when you're packed in this way.

Other creature comforts are going to be what you put into it, but keep in mind that many of the things we take for granted in modern cars--A/C, stereos, padded seats, etc.--add a whole lot of weight.

What also adds weight is how many passengers you want to carry at once. Two passengers won't add much weight, but four or more would. All that extra frame material adds up.

Building a traditional frame would take some welding skills. I have just enough welding skills to make some shelves, but anything structural (which my tutor defined as "anything where somebody's life depends on the weld holding") is not something I'm comfortable doing. That is to say, it'll take more than a quick tutorial and a little practice.

However, one interesting possibility is epoxy. Lotus did this for the Elise, and I once tracked down the epoxy manufacturer they use (I'd have to search around to find it again, though). The instructions for it didn't seem to need anything particularly out of reach for a hobbyist (doesn't need a big autoclave or anything like that). Lotus did reinforce certain sections with bolts/rivets. It will take some knowledge to design a frame around this, but it's one time design work by an engineer and then everyone can copy it.

One advantage Lotus had over a welded frame was thinner material. A weld itself is very strong, but it weakens the metal around it (meaning you usually get breaks around the weld, not on it). You have to use thicker metal to compensate for that. Since Lotus was using an epoxy, they could use thinner material for less weight, and it was stronger in the end.

Since it's also getting rid of a whole lot of weight around the frame, the range you get out of those batteries could be extreme. It could also be extremely quick with a modest motor.

This is basically all to say that you can have any three: safe, creature comforts, enough space for passengers, range.

If people bought it at one spec and now it's lower, that could be enough. It would have made the decision different at purchase time.

If you look at pfsense/OPNsense hardware recommendations, it's almost all using chips like the Intel N5105 (10W TDP, though admittedly "TDP" is itself a messy term) or J4125 (also 10W TDP). Using ARM hardware is asked a lot in the community forums, and it's one of those questions that will get you a flamed for not checking Google first. The power usage benefits for switching to ARM just aren't there.

There is the Netgate 1100, which runs ARM on a proprietary build of pfsense. The community has largely ignored it in favor of Intel chips. There isn't much of a price advantage, and the performance is lackluster.

That said, there's lots that you can do with a sub-10W chip, and x86 has nothing modern there.

Personally, I cobbled together an OPNsense firewall out of some old desktop parts I had on hand. Power usage is a bit higher, but not so much that I care. I would like a more viable high-end ARM option, though, just because I don't want x86 to be the only option.

ARM chips aren't better at power efficiency compared to x84 above 10 or 15W or so. Apple is getting a lot out of them because TSMC 3nm; even the upcoming AMD 9000 series will only be on TSMC 4nm.

ARM is great for having more than one competent company in the market, though.

Not from AMD. From the autogenerated transcript (with minor edits where it messed up the names of things):

amd's official recommendation [f]or the cut off now is 1.3 volts but the board vendors can still technically set whatever they want so even though the [AGESA] update can lock down and start restricting the voltage the problem is Asus their 1.3 number manifests itself as something like 1.34 volts so it is still on the high side

This was pretty much all on motherboard manufacturers, and ASUS was particularly bad (out scumbaging MSI, good job, guys).

At the start of this Intel mess, it was thought they had a similar issue on their hands and motherboard manufactures just needed to get in line, but it ended up going a lot deeper.

My issue with surplus server kit at home is that it tends to idle at very high power usage compared to desktop kit. For home use that won't be pushing high CPU utilization, the savings in cost off eBay aren't worth much.

This is also why you're seeing AM5 on server motherboards. If you don't need to have tons of PCIe lanes--and especially with PCIe 5, you probably don't--the higher core count AM5 chips do really well for servers.

On what workloads? AMD is king for most games, and for less price. It's also king for heavily multicore workloads, but not on the same CPU as for games.

In other words, they don't have a CPU that is king for both at the same time. That's the one thing Intel was good at, provided you could cool the damn thing.

Small correction: only surviving that makes desktop/server class chips. Companies like Texas Instruments and Microchip still have US foundries for microcontrollers.

The argument was that while AMD is better on paper in most things, Intel would give you rock solid stability. That argument has now taken an Iowa-class broadside to the face.

I don't watch LTT anymore, but a few years back they had a video where they were really pushing the limits of PCIe lanes on an Epyc chip by stuffing it full of NVMe drives and running them with software RAID (which Epyc's sick number of cores should be able to handle). Long story short, they ran into a bunch of problems. After talking to Wendel of Level1Techs, he mentioned that sometimes, AMD just doesn't work the way it seems it should based on paper specs. Intel usually does. (Might be getting a few details wrong about this, but the general gist should be right.)

This argument was almost the only thing stopping AMD from taking over the server market. The other thing was AMD simply being able to manufacture enough chips in a short time period. The server market is huge; Intel had $16B revenue in "Data Center and AI" in 2023, while AMD's total revenue was $23B. Now manufacturing ramp up might be all that's stopping AMD from owning it.

That's where the lawsuits will start flying. I wouldn't be surprised if they knock off 5-15% of performance. That's enough to put it well below comparable AMD products in almost every application. If performance is dropped after sale, there's a pretty good chance of a class action suit.

Intel might have a situation here like the XBox 360 Red Ring of Death. Totally kills any momentum they had and hands a big victory to their competitor. This at a time when Intel wasn't in a strong place to begin with.

I'll make a keychain out of them.

Yes. Problem is, this is the only way our system of justice allows for keeping companies accountable. They still pay out the nose on their end.

However, in this case, there's a lot of big companies that would also be part of the class. Some from oem desktop systems in offices, and also for some servers. The 13\14900k has a lot of cores, and there's quite a few server motherboards that accept it. It was often a good choice over going Xeon or EPYC.

Those companies are now looking over at the 7950x, noticing it's faster, uses less power, and doesn't crash.

They're not going to be satisfied with a $10 check.