jakobson14

It's just not currently very fast, yet.

It depended a LOT on the workload.

valley? 20%

glmark2? 91%

vkmark? 11%

To some degree this is probably just the vulkan driver and compiler being really, really immature. I expect that the gap will become a LOT closer in the next 6 months, or whenever they get past implementing vulkan 1.3 in the compiler and can put some work into cleanup.

Say what you will about openCL, but openCL is a WAY BETTER IDEA than ROCm. AMD are getting their asses handed to them in card support right now because ROCm requires you to compile your application all the way down to shader binaries. That makes applications supporting their cards VERY non-portable. Nvidia on the other hand supports PTX which is portable across a wide range of cards and drivers.

OpenCL mirrors this by allowing you to write compute kernels in normal shader languages to be compiled by the driver. It doesn't matter what card you want to run something on, with rusticl ALL cards can run the same code, including across vendors. AMD, nvidia, intel, raspi, apple AGX, doesn't matter. You can even run it on top of zink in case all you have is a proprietary vulkan driver.

Humanity, from now until the sun explodes, will only get one chance to check mars for life. Right now the first samples are still waiting in tubes to get picked up from the martian surface. Musks' solution to the problem of contamination is to fire as many rockets as he can as fast as he can and hope to contaminate mars completely before anyone can red tape his ass.

We owe it to all future generations to make sure he never lands a single payload before the question can be thoroughly resolved.

There was an article I saw a while back about a "young entrepreneur" who used his natural-born wealth to buy up a couple homes, then used rent collected on them to progressively buy up more. Eventually he owned the entire neighborhood and could charge whatever rent he wanted. It's a micro scale version of what's happening in cities like london and vancouver.

But if you think this will stem the tide of LLM hallucinations, you're high too. LLMs' primary function is to bullshit.

In chess many games play out with the same opening but within a few moves become a game no one has played before. Being outside the dataset is the default for any sufficiently long conversation.

When did Chernobyl break ground? 1972, operated by lying soviet bureaucrats.

When Fukushima Daiichi break ground? 1967

When did Three Mile island break ground? 1968

There have been a half-dozen whole generations of nuclear reactor developed since then, drastic reductions in the amount of waste they generate through advancements in nuclear chemistry, and now a whole new paradigm of SMRs.

All in pursuit of greater safety from a bogyman that is nothing but an echo of the past. It's like the Mill architecture people, still trying to design a CPU to beat the itanic when the rest of processor design has moved on.

Expecting that you can trust a computer you've never even seen, let alone one you *know* is a VM with a whole other untrusted OS under it, is a fool's errand.

If you need this level of trust, you should be on-prem. No excuses.

that's the part you're conveniently leaving out.

Indeed, if amazon paid entire divisions of people to sit on their hands in areas they've decided not to pursue (and thus don't generate revenue from) then it would be very hard for them to succeed at remaining solvent.

Maybe you should read the above quote again.

Doesn't matter how much value you generate for The Company. They're perfectly happy to take it from you for free if they can get away with it. Without bargaining power and leverage, either through unions or scarcity of labour, there's no natural limit to how little a company will pay it's employees.

Used to be that knowledge workers were scarce, now everyone is getting degrees. The scarcity of the engineering labour pool that meant people could get good jobs without a union is gone. If you want a good job, get a union.

It doesn't matter how much value you generate for The Company. If you don't have strong labour (eg. via unions) then there's absolutely no reason for a company to compensate employees fairly. Companies are not only happy, but DEFAULT to raising prices AND lowering wages. Max profit.

Opportunistic price increase.

That said, this seems like a very transparent "just give us a bunch of money" request.

doesn't matter if you use your position to fuck over your suppliers and workers, so long as consumers don't see a price increase. this particular action may not be an example of trust behaviour, but amazon overall is perfectly moulded to the consumer welfare legal environment so I can see this as having that verbiage leaking through

There are a fuck of a lot of rando x86 laptops, and you need to record a different profile for every single one of them. That means sitting down in front of every single laptop model you want to ship a profile for and running tests with a measurement-grade microphone and equipment.

Marcan did that for the handful of M-series mac laptops that exist. Good luck physically testing even 0.01% of the x86 laptops that exist.

That means sitting down in front of every single model with decent measurement equipment and running a test. This doesn't scale, especially not for a distro with limited resources.

If you keep careful track of the audio going through the speakers and the energy in each frequency range, you can figure out exactly what the temperature is at any given moment and make sure it stays under the limit. However, this requires a ton of math.

Alternatively, you can just put a hard cap on the volume, such that no combination of frequencies could ever overheat the speakers. This is safe, but sounds like crap. It's a far, far less nuanced way to do things and results in restricting how hard you can drive the speakers far more than is actually nesisary.

If a distro is expecting you to install onto the rando asus gaming laptop you got at staples? not going to work.

The thing about M-series apple macs is there are only like 10 models total to record profiles for. You can't realistically do that and ship profiles for every x86 laptop under the sun.