wtallis

28 ms·

601.
by wtallis 1y ago
From a compute perspective, GPUs are mostly about fast vector arithmetic, with which you can implement decently fast matrix multiplication. But starting with NVIDIA's Volta architecture at the end of 2017, GPUs have been gaining dedi
602.
by wtallis 1y ago
The Embedded DisplayPort standard has had the panel self-refresh feature since 2011, and the partial update feature since ~2015. I found a press release from Parade in early 2017 for a TCON supporting the partial refresh feature. I don&#x27
603.
by wtallis 1y ago
The problem is that there's not even a hypothesis for how light reflecting off an e-ink display could be easier on the eyes than light emitted from an LCD, unless the LCD is using PWM dimming of the backlight and thus flickering. I&#
604.
by wtallis 1y ago
> That's impossible because Apple has bought up most of TSMC's 3nm production capacity. You could try to approximate by comparing Apple M4 Max against NVIDIA B300 but that'll be a very one-sided win for NVIDIA. Have you no
605.
by wtallis 1y ago
Until recently, consumer, workstation, and datacenter GPUs would all share a single core design that was instantiated in varying quantities per die to create a product stack. The largest die would often have little to no presence in the c
606.
by wtallis 1y ago
The WiFi silicon vendors seem to base their SDK/reference software on OpenWRT (albeit often badly-outdated versions), so almost everyone selling a WiFi box ends up using some variant or fork of OpenWRT. It's been a long time since
607.
by wtallis 1y ago
GPU vendors have come to the realization that the in-kernel driver needs to be open-source, but the userspace portion can be closed-source. There's just really no good reason to accept a design where outdated closed-source drivers coul
608.
by wtallis 1y ago
Could you provide any specifics? Are you wanting a dumbed-down UI that hides most of the functionality, or are you wanting the existing functionality to be exposed to the user differently?
609.
by wtallis 1y ago
The CPU in such switches is far too slow to do any data plane operations, so performance is entirely due to the hardware switching. Replacing Ubiquiti's OS with OpenWRT just gets you a different management interface for configuring t
610.
by wtallis 1y ago
As far as I can tell, GCC supports compiling multiple versions of a function, but can't automatically decide which functions to do that for, or how many versions to build targeting different instruction set extensions. The programmer n
611.
by wtallis 1y ago
In general: consumer cards with very bad FP64 performance have it fused off for product segmentation reasons, datacenter GPUs with bad FP64 performance have it removed from the chip layout to specialize for low precision. In either case, th
612.
by wtallis 1y ago
IIRC, there were several smartphone SoCs that dropped 32-bit ARM support from most but not all of their CPU cores. That was straightforward to handle because the OS knows which instruction set a binary wants to use. Doing anything more fine
613.
by wtallis 1y ago
> answered complaints about Windows being bad for development by embedding an OS that is good for development and calling that an innovation Well, they actually did it twice in recent history: WSL 1 was relatively innovative, integrati
614.
by wtallis 1y ago
> It’s really the hardware block size that matters in this case (direct I/O). That value is a property of the hardware and can’t be changed. NVMe drives have at least three "hardware block sizes". There's the LBA si
615.
by wtallis 1y ago
Please consider the possibility that some proposed features should not exist . The objections to many of Chrome's features are fundamental , not aesthetic, or complaints about nuances of how it's implemented. Many people outsid
616.
by wtallis 1y ago
You're still dodging the issue. Your article title accusing Apple of an "assault on standards" is implicitly treating Google's proposals as a fait accompli that Apple is resisting, which is not at all what the situatio
617.
by wtallis 1y ago
That article also does not appear to have anything to say about the validity of "standards" that are nothing more than Google's feature creep for web browsers. At some point, you need to actually defend the idea that a web
618.
by wtallis 1y ago
Out of the box, a minority of computers have a third button to click: it's pretty much just desktop mice not from Apple. Laptops and Apple mice require a power user to jump through hoops to have a way to express a middle click that&#
619.
by wtallis 1y ago
Pop-up ads from the dark ages of the web largely weren't modal dialog boxes. For the most part, clicking to dismiss them, or clicking to bring the window you cared about back to the foreground was at least as easy as getting rid of &qu
620.
by wtallis 1y ago
In the case of five disks of the same capacity, RAID6 or RAIDZ2 only gets you 20% more capacity than btrfs RAID1. That's not exactly a huge disparity, usually not enough to be a show-stopper on its own. There are plenty of scenarios wh
621.
by wtallis 1y ago
Tiering didn't go away with the migration to all-SSD storage. It just got somewhat hidden. All consumer SSDs are doing tiered storage within the drive, using drive-specific heuristics that are completely undocumented, and host software
622.
by wtallis 1y ago
> with BTRFS you mount the array through one of the volume that is part of the array and not the array itself I don't think btrfs has a concept of having only some subvolumes usable. Either you can mount the filesystem or you can&#
623.
by wtallis 1y ago
> but it's silly to act like it's a paragon of stability in the kernel. And it's dishonest to act like bugs from 15 years ago justify present-tense claims that it is constantly eating people's data and is a bad joke.
624.
by wtallis 1y ago
If you're not interested in a multi-disk storage system that doesn't have (stable, non-experimental) parity modes, that's a valid personal preference but not at all a justification for the position that the rest of the feat
625.
by wtallis 1y ago
If you live on the coast, following the details of the forecasts and warnings is extremely important. But my experience living a little over 100 miles inland taught me that the hurricane category is a useful predictor to some extent: anyt
626.
by wtallis 1y ago
I don't think you've successfully demonstrated that those assumptions are bad . They're simplifying assumptions that are necessary to have a "basic" economic theory in the first place—one with tractable and under
627.
by wtallis 1y ago
Does the whole filesystem need to be marked as unstable if it has a single experimental feature? Is any other filesystem held to that standard?
628.
by wtallis 1y ago
BTRFS does have stable, usable multi-disk support. The RAID 0, 1, and 10 modes are fine. I've been using BTRFS RAID1 for over a decade and across numerous disk failures. It's by far the best solution for building a durable arr
629.
by wtallis 1y ago
The NVMe spec is freely downloadable and sufficient to write a driver with, if your OS already has PCIe support (which doesn't have open specifications). You don't need any vendor-specific features for ordinary everyday use, so it
630.
by wtallis 1y ago
> The biggest cost is still energy consumption, water and maintenance. Are you saying that the operating costs for inference exceed the costs of training?

More ›