We're using xwayland-satellite rather than Xwayland directly because X11 is very cursed. xwayland-satellite takes on the bulk of the work dealing with the X11 peculiarities from us, giving niri normal Wayland windows to manage.
xwayland-satellite works well with most applications: Steam, games, Discord, even more exotic things like Ardour with wine Windows VST plugins. However, X11 apps that want to position windows or bars at specific screen coordinates won't behave correctly and will need a nested compositor to run. See sections below for how to do that.
The first revisions were stuff made by qualcomm right? I don't think we have much data on how much customizations they make and where they their IP from, but given how much of the Tensor cores comes from Samsung I think it's safe to say to assume that there is a decent amount coming from some of the big vendors.
It's "made by" TSMC as usual. Their customization comes from identifying which compute operations that want optimized in hardware and do it themselves. And then they buy non-compute IP like HBM from Broadcom. And Broadcom also does things like physical design.
On top of that you don't have that limitation on Android. It's like enterprise IT, where you put up restrictions everywhere on files and then people can upload files to their personal one drive.
It slid into irrelevance as the early internet exploded around it and it became an angryish nerd oasis. Reddit easily outscoped it, and HN had the attraction of VC money sloshing behind the scenes.
Slashdot, the site, still lives as a fossil from 10-15 years ago. It must be popular enough to pay its bills.
That's probably the real reason, but I think their redesign didn't help either.
The /. redesign wasn't as brutal as that which Digg had, but it was certainly something that stopped me visiting so often.
I just looked and saw to my surprise I still have an account there, the last few comments were made in 2014, 2012, 2011. So maybe I did return later after all.
The resistance was the straw, but by then it was already descending from an interesting site frequented by people like Bruce Perens, John Carmack, Wil Wheaton etc, to just more of the same. Taco leaving was another point, I forget if that was before or after the redesign.
There’s was a significant amount of Randian right wing group think too, which tended to spiral away
Ultimately though it was tacos blog, and that type of site doesn’t scale and retain the quality.
"Angryish nerd oasis" sounds like the same thing that happened to Reddit, so perhaps Slashdot wasn't actually uniquely good. Perhaps it will also happen to HN and lobsters.
It's already happening to HN and Lobsters is going through some internal fighting now. Compare conversations 10 years ago here to now and you can see how much angrier the site has become.
I left Slashdot for HN...but I didn't leave Slashdot because of HN. I was frustrated with Slashdot and was actively seeking alternatives. About 2 days after I discovered HN existed, I was done forever with Slashdot.
Among other frustrations (including some really vile comments), I felt like the world was bursting with interesting tech news, and Slashdot was just not keeping up. The publish rate was too slow (maybe 10-13 stories a day), and the %age of stories I found interesting had dropped considerably from a few years previous.
I wasn't a fan of the redesign, but it was content that drove me to seek alternatives.
I left because the people there became worse and worse. There is a side of tech culture I find utterly repugnant, and it gradually became most of the site.
I have no idea if others had the same reason. I hear more about its user interface, but that change didn't bother me.
Maybe the real value of the DGX spark is to work on Switch 2 emulation. ARM + Nvidia GPU. Start with Switch 2 emulation on this machine and then optimize for others. (Yeah, I know, kind of expensive toy).
I think you can get something a lot cheaper if that’s all you want, e.g. something in the Jetson Orin line. That’s more similar to the switch, also, since it’s a Tegra CPU.
Expensive today. But how quickly (years) will these systems lower in value? At least on the Nvidia side of things they can be stacked.. so maybe not so much =/
That is really cool. One thing I have to ask though. Does the Framework have the same problem as other bottom intake fans that collect dust inside the fan?
It really depends on your usage environment, but one good thing is that its probably easier to clean the fans on a Framework Laptop than on most other laptops!
I found this to be the opposite of true. There is no easy way to clean the fan on the fw-13. You either need to take the mainboard out completely and disassemble the housing around the fan, or take the entire radiator off from the top and re-paste the cpu. I resorted to using a pick through the small gap on the top. I would say the fw-13 is the hardest to clean of any computer I've owned lately.
Oh boy do they. I have a fw-13 and I think it is the biggest issue with this machine. Cleaning it is not trivial either, harder than most laptops in fact.
Valve has hired a bunch of FAANG engineers that brought their own toxic hiring practices to Valve. It's only a matter of time before those people promote their culture in that organization if they haven't done so already.
It's not really Linux' fault. Most It's Microsoft that forced this, and most vendors don't know how to deal with this and work proper firmware. I think Framework and Valve fixed theirs. I have a GPD and just found out that the reason I kept getting it wake from sleep was some MS related option triggering an ACPI IRQ 9 sleep wake.
On another note, I actually think that the most important things that work better on the Apple devices is the mic and camera, the rest is somewhat unimportant on the go if you work at a desk.
As a user I don't care whose fault is it. I want my laptop to go to sleep when I close the lid; I want it to stay asleep while the lid is closed; I want it to wake when I open the lid. Only macs seem to be able to do that consistently; I'd be glad to be proven wrong, but over the past decade I haven't found a counterexample yet.
I don't think that's even true with my m3 mbp16. I haven't tweaked with the power settings but I'm pretty sure it is in a connected sleep state when I close the lid; at least when I'm hotspotting my phone will register it as a connected device.
This can be enabled or disabled under the System Settings > Battery > Options > Wake for network access (Always / Power adapter / Never). Or possibly the phone registers it as connected for a while after it sends its last packet?
IIRC, Macs also do tricky networking things to make it faster to come back online from sleep. I'd be curious if the computer is actually sending packets vs. just keeping the address configured and waiting.
I’ve never had a problem with a system76 or tuxedo computers laptop using suspend correctly. If you want it to just work, you may need to buy from a manufacturer who you pay to make it just work. Otherwise you’re comparing a dyi setup to Apple.
I think Intel is partially at fault. On Apple’s Intel hardware, suspend and resume worked, but it was very slow due to the weirdly baroque power management. The M1 MacBooks were a revelation; the screen woke instantly when you opened it.
I suppose this, again, depends on the country. They can get very expensive, especially if you're in a "loser pays" country. And a judge may even then divide the costs more fairly, and you find that your lawyer costs exceeded the monetary value of the judgement you won. That on top of potentially stressing years and years about it.
And then the loser may request a next level court to consider the case, potentially leading to more stressful years and a lot of additional financial risk.
Georgi gave a response to some of the issues ollama has in the attached thread[1]
> Looking at ollama's modifications in ggml, they have too much branching in their MXFP4 kernels and the attention sinks implementation is really inefficient. Along with other inefficiencies, I expect the performance is going to be quite bad in ollama.
ollama responded to that
> Ollama has worked to correctly implement MXFP4, and for launch we've worked to validate correctness against the reference implementations against OpenAI's own.
> Will share more later, but here is some testing from the public (@ivanfioravanti
) not done by us - and not paid or
leading to another response
> I am sure you worked hard and did your best.
> But, this ollama TG graph makes no sense - speed cannot increase at larger context. Do you by any chance limit the context to 8k tokens?
> Why is 16k total processing time less than 8k?
Whether or not Ollama's claim is right, I find this "we used your thing, but we know better, we'll share details later" behaviour a bit weird.
reply