wtallis

21 ms·

481.
by wtallis 11mo ago
Intel's biggest problem has been management remaining in denial about their serious engineering problems, and believing that they'll have things sorted out in another few quarters. They were years late to taking meaningful actio
482.
by wtallis 11mo ago
Most server CPUs require RDIMMs, and while non-ECC RDIMMs exist, they are not a high-volume product and are intended for workstations rather than servers. The used parts market would look very different if there were lots of large-scale s
483.
by wtallis 11mo ago
Those deliberately-missing features in Wayland would have been a good opportunity to instead provide an official API plus an official security/privacy framework. Pretending that the hard problems are out of scope was such a disappoin
484.
by wtallis 11mo ago
FEX is relevant to real use cases because a lot of things that would be GPU-limited on an x86 machine can become CPU-limited when running on DGX Spark. The overhead of x86 to ARM translation, plus the low performance of the ARM cores, plus
485.
by wtallis 11mo ago
Meshtastic already fulfills that, except for the bit about accessing simple text-based websites, because Meshtastic isn't an IP network. The data rates only go up to 21.88 kbps for the fastest preset: https://meshtastic.org&
486.
by wtallis 11mo ago
What do you even mean by "AI agent"? I've head that term used in a few different ways, but none of them are at all close to anything that would benefit from an off-grid ad-hoc dialup-speed communication channel between agen
487.
by wtallis 11mo ago
Reliability carries serious costs in airtime usage and power consumption, and there's always going to be someone who demands that the network support an even higher level of reliability. "Toy network" is highly subjective and
488.
by wtallis 11mo ago
It's unfair to assign that much blame to assign to any one person. I think it's more fair to say that the Meshtastic community as a whole has a problem with people making overly-narrow assumptions about the goals and what use case
489.
by wtallis 11mo ago
Many (if not most) Meshtastic devices don't have a GPS receiver of their own, and also may not be paired with a phone app to supply location. So at least some devices would need to sniff GPS coordinates from traffic on the wrong band
490.
by wtallis 11mo ago
A full scan would take hours . With just the default choice of channel width (250 kHz), spreading factor and coding rate, in the US you'd have to listen to 104 different frequencies, and you'd need to listen for more than a minut
491.
by wtallis 11mo ago
One could try to make an argument that "adoption rate" should mean change in adoption over time, but the meaning as used in this article is unambiguously not that . It's just percentages, not time derivatives, as clearly s
492.
by wtallis 11mo ago
There's a big difference between saying that many of the drivers are missing entirely, and saying that the drivers that exist are alpha-quality.
493.
by wtallis 11mo ago
Steam gaming with FEX is possible on the DGX Spark. The GPU is approximately a mobile 5070 with much less memory bandwidth. The CPU cores are relatively weak, especially after the instruction set translation overhead. There's a lot of
494.
by wtallis 11mo ago
It might be more fair to say that there's simply no standard way to do power management for discrete GPUs in a laptop, or to integrate such power management with Windows and whatever power management it's trying to do. And the lac
495.
by wtallis 11mo ago
> are there any NVME "spinny rust" disks? My recollection is that NVMe had some features added specifically for hard drives. I don't know if anyone ever bothered making a hard drive that natively used NVMe over PCIe; the m
496.
by wtallis 11mo ago
That was never a myth, was it? It was just sound advice that was repeated without the information about which specific use cases it applied to.
497.
by wtallis 11mo ago
It really comes down to which architectures you're porting to. The two biggest issues are big endian vs little endian, and memory consistency models. Little endian is the clear winner for actively-developed architectures, but there a
498.
by wtallis 11mo ago
What dissuaded you from upgrading to Sequoia?
499.
by wtallis 11mo ago
I said "single-thread performance or power efficiency", not "single-thread performance and power efficiency". Though at the moment, the best single-thread performance does happen to go along with the best power eff
500.
by wtallis 11mo ago
> Qualcomm only ever pledged to support Elite processors Link, please. > and perhaps not coincidentally all of the Plus laptops require reversing You're still acting as though Qualcomm has made meaningful contributions to Linux s
501.
by wtallis 11mo ago
If carrying around a little bit of x86 compatibility baggage had enough die size cost to matter for Apple, then Intel and AMD would be pushing much harder to reduce their comparative mountain of x86 compatibility baggage. In reality, th
502.
by wtallis 11mo ago
Based on historical trends, that's never going to happen for any workloads where single-thread performance or power efficiency matter. If you're doing something where latency doesn't matter but throughput does, then old serve
503.
by wtallis 11mo ago
> Intel's consumer processors (and therefore the mainboards/chipsets) used to have four memory channels, but around the year 2020 this was suddenly limited to two channels since the 12th generation (AMD's consumer processo
504.
by wtallis 11mo ago
The performance drop due to SLC caching only applies to writes . Sequential reads (and often, even random reads at sufficiently high queue depth) will still more or less saturate the PCIe link. Most workloads and use cases read a lot more
505.
by wtallis 11mo ago
I don't think your link contains the evidence you think it does. I'm not seeing anything that looks like Qualcomm contributing device trees on behalf of system OEMs, for any of the Snapdragon X products, so I don't see how
506.
by wtallis 11mo ago
> They might even decide that they will be moving that functionality to software and decide to also leverage FEX. That's crazy. Modifying an already-working CPU design to remove hardware features, and modifying Rosetta to implement
507.
by wtallis 11mo ago
COVID forced remote learning to be adopted very broadly, without the usual self-selection effect of families that choose to homeschool when they have a choice. So the observations from COVID don't really support any stronger claim than
508.
by wtallis 11mo ago
Yeah, I saw that. I'm pretty sure that's a statement more about the drivers than the underlying hardware. Open-source drivers often have more limited feature support than the underlying hardware. I doubt anyone is producing WiFi c
509.
by wtallis 11mo ago
Ok, so having clarified the naming, it still looks like you're wrong about which chips are getting driver support upstreamed, because the Snapdragon X Plus parts are (with maybe one exception, IIRC) literally the same chip as the Sna
510.
by wtallis 11mo ago
Why do you assume that dropping support for Intel hardware from the OS will coincide with dropping hardware features that help support for x86 applications? Have you not seen Apple's documentation that states they plan to retain some R

More ›