hedgehog

30 ms·

541.
by hedgehog 2y ago
Adding to this: If the default is too slow look at the more heavily quantized versions of the model, they are smaller at moderate cost in output quality. Ollama can split models between GPU and host memory but the throughput dropoff tends t
542.
by hedgehog 2y ago
As part of that Haswell brought FMA support which was a boon to those of us doing a lot of multiplication and addition (made those workloads twice as fast).
543.
by hedgehog 2y ago
This is more or less what we do but recently the Jellyfin iPad client has become very flaky. My suspicion is the JS app's memory usage has grown to a point where it's causing the web view embedded in the app to run out of memory i
544.
by hedgehog 2y ago
Seems ok from here in iPhone Safari.
545.
by hedgehog 2y ago
As another commenter points out their strategy appears to be to focus on HPC clients where AMD can focus providing after-sale software support around a relatively small number of customer requests. This gets them some sales while avoiding t
546.
by hedgehog 2y ago
That's a misleading number, knowing that he was a lauded prof at MIT we can expect that he was significantly wealthier than the median Mass. resident and with that significantly higher life expectancy.
547.
by hedgehog 2y ago
On a documentation site, photo gallery, etc, it is very reasonable to use something like htmx to implement navigation without triggering full page loads.
548.
by hedgehog 2y ago
It's interesting that for years Safari stored page screenshots in its history to allow a "coverflow" view and there wasn't broad concern.
549.
by hedgehog 2y ago
Tigray region and Mynamar are two earlier candidates.
550.
by hedgehog 2y ago
Living on Earth surface all that's necessary is to bring in outdoor air and reduce it to say 300ppm, currently that's only a 120ppm gap.
551.
by hedgehog 2y ago
I've moved to almost entirely remote dev workflow for similar reasons. Out of curiosity how much RAM do you need per build thread?
552.
by hedgehog 2y ago
There are server workloads like maybe high concurrency web apps that might benefit. Back in the days of single core CPUs HT made desktops feel a lot better not just because of throughput but (my speculation) because it reduced typical sched
553.
by hedgehog 2y ago
Likely true in some cases but are these the bottlenecks on a laptop? I've never experienced HT on a multicore CPU being more than maybe a 40% increase in throughput and no real help on latency or efficiency. Implementing HT isn't
554.
by hedgehog 2y ago
What workloads benefit enough from HT to justify it in a laptop?
555.
by hedgehog 2y ago
Looking at a map Seattle is roughly the same latitude as Zurich and Budapest.
556.
by hedgehog 2y ago
PM2.5 is around the hardest size to filter, I don't understand the physics but both larger and smaller are actually easier to remove from the air. Given that 2.5 is a useful estimate of worst case performance for a given setup.
557.
by hedgehog 2y ago
I remember being really impressed by a tour guide I worked with in the early 2000s who ran his business from a Nokia Communicator that ran more or less the same software but with a cell modem included for e-mail. Very slick.
558.
by hedgehog 2y ago
I don't know how they actually did it but conceptually I thought about it as traversing the map's segments depth first in 2D starting at one edge of the viewport & the currently occupied polygon. Simple, and allows freedom lik
559.
by hedgehog 2y ago
I'm curious how they can hit this price. Stretch 3 is $25k ( https://hello-robot.com ) but for much simpler hardware.
560.
by hedgehog 2y ago
Interesting, I'm curious if anyone will add support for milling. Being able to lay down coarse layers and refine just the external surfaces seems like a path to fast prints with fine surface finish all on one machine.
561.
by hedgehog 2y ago
Two different things, Apple provides a pretty long window of OS updates for each device but doesn't ship updates for older OS versions. For example the 1st gen SE shipped with iOS 9 which has not been getting updates for a few years, b
562.
by hedgehog 2y ago
Google and Apple have fairly different platform software update strategies so I was curious how the support windows line up. Android 6 was the last supported release on devices like the Nexus 5 and Xperia Z3 from 2013 and 2014. iOS 9 was th
563.
by hedgehog 2y ago
I don't know about the rest of it but EV sales continue to increase, at least up to 2023.
564.
by hedgehog 3y ago
The sensors are there and report on CAN bus, I just don't know enough about how the nav systems are integrated to know if they actually use the data.
565.
by hedgehog 3y ago
The car probably has odometry and inertial sensors to do position updates while GPS is unavailable. That's actually how some of the original car nav systems worked even pre-GPS in the 80s.
566.
by hedgehog 3y ago
In my experience it's over cable CarPlay is reliable if the head unit is already booted and the phone is unlocked when plugging in. If either of those things aren't true then no guarantee the phone will connect after the booting &
567.
by hedgehog 3y ago
For consumer Ryzen to pencil out it would require a cluster of APU-equipped machines with the model striped across them. Given say 16GB of model per machine and 60GBps actual memory bandwidth @ $500 it's favorable vs A100s if the softw
568.
by hedgehog 3y ago
I'm curious how the newer consumer Ryzens might fare. With LPDDR5X they have >100 GB/s memory bandwidth and the GPUs have been improved quite a bit (16 TFLOPS FP16 nominal in the 780M). There are likely all kinds of software pr
569.
by hedgehog 3y ago
That is the range of ages for which they found an increase but the biggest increase is in the 30-34 age subgroup. The report is not hard to skim, the overall picture is rural populations particularly in southern states are dying much more o
570.
by hedgehog 3y ago
Ollama (which wraps llama.cpp) supports splitting a model across devices so you get some acceleration even on models too big to fit entirely in GPU memory.

More ›