sliken

41 ms·

571.
by sliken 4y ago
Not surprising. M1 pro -> m1 max = same number of cores at the same speed. So only if you are doing something CPU intensive and not cache friendly would you notice the memory bandwidth doubling. M1 generation -> M2 generation is a p
572.
by sliken 4y ago
You can run any process on any core and allocate local memory to that core. What is missing? Sure you can't hide metadata in the unused bits of a pointer, but that didn't seem particularly critical for making good use of a NUMA m
573.
by sliken 4y ago
Heh, I was running DOS and Deskview, windows 3 was out of coming soon. I jumped into Linux with a 386sx-16, not even a FPU. I think it had 4MB ram, eventually upgraded to 16mb. Your ideas remind me of the Sun MAJC CPU. Claimed to be desi
574.
by sliken 4y ago
I've not see any evidence of power issues. Saw a minor miss that they had some ray tracing functionality in the GPU, but ended up being more power hungry than planned, so they disabled it. But the chip with the same cores is doing fi
575.
by sliken 4y ago
Well the tools are there, and the hardware is there as well. Early AMD epycs had multiple chiplets, each with it's own memory controller. However that means that even on a single socket different chiplets would see different memory la
576.
by sliken 4y ago
There's a ton of stuff not upstreamed. There's work being done on upsteaming, but the changes are not trivial. Some of the things I've heard discussed: * All distros default to 4k pages and that's set at kernel compile
577.
by sliken 4y ago
I've been writing micro memory benchmarks and have been rather surprised how hard something as simple as quantifying latency and bandwidth under multicore loads can be. The memory hierarchy is getting ever more complex. Cacheline siz
578.
by sliken 4y ago
> 100000x lower unit production volume than GPUs means higher unit costs Two points. Nvidia's RTX 3000 series (3060 Ti, 3080, and many other flavors) ships 6 or more flavors per generation. The related silicon has names like the G
579.
by sliken 4y ago
Or just post a magnet link to a torrent.
580.
by sliken 4y ago
Heh, I didn't have to imagine. Got a GigE switch (expensive at the time) and 8 of the 164LX motherboards. Worked fine, but aged quickly since annual performance increases on the PC were substantial, unlike today. Then HP released (19
581.
by sliken 4y ago
Standard operating procedure seems to be, promise the world, update the maps with lies, then announce a stock buy back plan with the profits.
582.
by sliken 4y ago
True, but those are typically rather low density/large plots, which makes them much easier to serve with a satellite based service.
583.
by sliken 4y ago
Well trick is, decent iGPUs need better than 128 bit wide DDR5 @ 5200 Mhz. So apple's solution is a small CPU package with ram on board that lets them do 128, 256, 512, and 1024 bit wide memory interfaces. All but the 1024 bit wide a
584.
by sliken 4y ago
Actually, they did. Intel promised aggressive delivery schedule, performance ramp, and performance. The industry took it hook, line, and sinker. While AMD decided not to limit 64 bit to the high end and brought out x86-64. Sun did a port
585.
by sliken 4y ago
Anyone know exactly config is needed for training, presumably the $300 is on some GPU heavy instance on ec2? Is $300 of Ec2 p4de.24xlarge which is $40.96 an hour? Or maybe 7-8 nodes for an hour? Something else?
586.
by sliken 4y ago
I read a review of a prototype BMW mini. The turbine (fixed RPM) + batteries seemed like a great combination. Plenty of HP for highway cruising, plenty of battery+AWD for great acceleration. Turbines are efficient, crazy small for the HP
587.
by sliken 4y ago
Except no cars do that. That's called a serial hybrid, which never shipped, unless you count oddballs like the BMW i3, with the optional "range extender".
588.
by sliken 4y ago
Mostly that two full drivetrains are expensive, heavy, and labor intensive to install. You end up with the worst of both worlds. BTW, the volt was advertised/promoted as a serial hybrid, which was a really interesting idea ... right u
589.
by sliken 4y ago
I hate snaps, doubly so when they appeared on a LTS release and some /dev/random sillyness broke simple things like booting. However this seems pretty silly when what actually changed is the default OS installs will not have flatp
590.
by sliken 4y ago
Dunno, the earth often has a 140F degree range across it's surface, hell often 100F or so at the same point just different latitudes between 0 and 35,000 feet. Doesn't seem that different for a tidally locked planet, sure the vari
591.
by sliken 4y ago
Heh, dunno. Once you get used to Linux moving to a new Linux box is easy. Some people use ansible, puppet, git, nix, docker, or even just a list of packages to install. On apple there's no package management for the entire system, su
592.
by sliken 4y ago
There's quite a bit of activity getting the rk3588 working in the mainline kernel. People having current mainline kernel+patches working today, and getting those patches mainstreamed has been making steady progress. Some places to wat
593.
by sliken 4y ago
Heh, or maybe: the orange zero orange pi plus 2e orange pi 3 lts orange pi 4 lts orange pi 5 In any case the rk3588 is available with 4x the PCIe, 2x the ram, 2-8x the performance, and ethernet and storage are over PCIe (not
594.
by sliken 4y ago
I've had similar issues, but bought a NanoPi R6S with 8GB ram. Got ubuntu 22.04 going on the first day, and it's a nice little server. It's crazy faster than a Pi4, stable, and runs cool. I paid for the $20 CNC machines ca
595.
by sliken 4y ago
I bought rk3588 nano pi r6s a few months ago, running ubuntu 22.04 with a custom kernel. Been watching progress, and it's been quite active. Out of tree patches work with newer kernels, various upstreaming efforts are underway, and v
596.
by sliken 4y ago
Well depends, Apple has been refining their arm chips for many generations now. Generally the Arm competition (non-apple) hasn't caught up to Apple yet. CPUs are amazingly complicated. A major new CPU designs can take 3-5 year, consum
597.
by sliken 4y ago
Heh, I do follow Linus Tech Tips, and shared their hate of subscriptions and almost bought a Ubiquiti g4 door bell they recommended because I didn't want a subscription, didn't want my video uploaded to a cloud, and wanted to keep
598.
by sliken 4y ago
I've had a google home since early on and was really impressed how well it worked. Dramatically better than apple or amazon at the time. You could ask it complex questions and it would go find results for you, not just "We found
599.
by sliken 4y ago
Nice, I was right ... up to yesterday. Now it's fixed. Things are moving quickly.
600.
by sliken 4y ago
I hate the idea of requiring 6 random binaries from 6 random companies just to get the basics working like tabbing windows, resizing windows, and reading drives. Doubly so when they are closed source and free. How are the developers making

More ›