Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
pedrocr
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
10 ms
·
91.
▲
by
pedrocr
4y ago
AlphaZero uses Monte Carlo Tree Search not MinMax. So it does use a neural network for the search. Modern Stockfish is the one that moved to having a neural network for position evaluation while keeping MinMax with AlphaBeta and heuristics
92.
▲
by
pedrocr
4y ago
Are you perhaps confusing 18A with 18nm? 18A is 1.8nm and is scheduled for the second half of 2024.
93.
▲
by
pedrocr
4y ago
Nothing guarantees that. clone() may still be provided by the kernel but nothing guarantees it will still be fast and not just a compatibility retrofit. Particularly since most users will move on automatically through libc.
94.
▲
by
pedrocr
4y ago
So ARM Macs in OSX can now or soon have true Vulkan support with just a custom userspace app code targetting the kernel interface directly?
95.
▲
by
pedrocr
4y ago
Are no brand lightning cables not a thing too?
96.
▲
by
pedrocr
4y ago
Last time I used LineageOS while you got the Android updates on most devices the underlying kernel and drivers were stuck at whatever the manufacturer originally shipped, and that included a bunch of security problems on most phones. Androi
97.
▲
by
pedrocr
4y ago
We've had flying drones in operation for a long time now. I wonder why the same hasn't been done to tanks as well. The latency is even less of a problem.
98.
▲
by
pedrocr
4y ago
Everyone is missing the actual point of the quote. The UK is at the top end in terms of online penetration in grocery and even there it's true that there's "no way to compete with supermarket efficiencies". Grocery retai
99.
▲
by
pedrocr
4y ago
This exists for big refrigerated warehouses. By using an appropriate liquid with the phase change tuned for the working temperature you can shift the electricity usage to times with lower cost. That can be because you have your own generati
100.
▲
by
pedrocr
4y ago
Does the core area math check out on that? Is the x86 instruction decoding area a small enough part of today's cores that having a chip that includes both a RISC-V decoder and a x86 one not a big penalty? It's also not clear that
101.
▲
by
pedrocr
4y ago
That seems something in the ballpark of my favorite wiki software: https://github.com/gollum/gollum Edit and view pages as a normal markdown wiki. But the backend is just a git repository of markdown files so you can a
102.
▲
by
pedrocr
4y ago
We could use more benchmarks about idle power usage for that. I'm speccing a new NAS and those are the watts you end up paying 24/7, not power usage at peak. You'd have to be running constant compiles or be a very avid gamer
103.
▲
by
pedrocr
4y ago
The difference is much starker in this generation in particular. It's 60% to 160% more power to get a 5 to 10% benefit. Intel was left behind by AMD and started competing by cranking the power up. In this generation AMD responded in ki
104.
▲
by
pedrocr
4y ago
RAM works on a wide set of bits per transfer so if you're using ternary the mapping will waste very little. Encoding 64 bits in ternary only wastes half a bit in the mapping for example which is less than 1% waste and thus irrelevant.
105.
▲
by
pedrocr
4y ago
An insight I've found very useful in many areas is to separate cases where you're satisfying requirements versus optimizing. If you're satisfying, a single logical argument for a decision is enough. If you're answering t
106.
▲
by
pedrocr
4y ago
Why get those instead of modern Ryzens that are much faster at the same power levels?
107.
▲
by
pedrocr
4y ago
There's no adequately priced ECC UDIMMs because the market is so small. Servers use registered so it's a very small niche. But for the gaming overclock application it would actually be a great fit. Because ECC gives you a great ea
108.
▲
by
pedrocr
4y ago
Usually it's 13A from a Shucko plug. They're rated for 16A but you're not supposed to run that continuously so it's 13A to be at 80% of maximum for a continuous load.
109.
▲
by
pedrocr
4y ago
I've been trying for a week now to get Windows to properly suspend and stay suspended on a Thinkpad X1. Even after changing the BIOS to "Linux" mode (S3 Sleep), fiddling with drivers and the registry, and issuing powercfg com
110.
▲
by
pedrocr
4y ago
I wonder if Apple would have ended up better off if they had made iOS devices friendly to third-party OSs. Maybe a nice homebrew culture would have emerged and LineageOS would have a nice set of images to install and all the geeks would hav
111.
▲
by
pedrocr
4y ago
I do think billionares are undertaxed, but mostly because they can easily get around the tax system itself. The personal wealth comparison seems flawed though because it's treating wealth as if it's just bank accounts of cash and
112.
▲
by
pedrocr
4y ago
In theory it should allow fine grained control over raid levels at the file/directory/subvolume level. So that depending on the specific needs you can set direct/raid1/raid10/raid5/raid6 without having to setup
113.
▲
by
pedrocr
4y ago
There are box trucks in the market and in service already. They're definitely further along than semis. Tesla is really good at pushing the public perception of EVs and yet doesn't make one, so it's possible it's just a
114.
▲
by
pedrocr
4y ago
We're still short and always will be of the vision and endless possibilities of zombo.com
115.
▲
by
pedrocr
4y ago
Makes sense, thanks
116.
▲
by
pedrocr
4y ago
For the first point, why isn't that debt just cancelled?
117.
▲
by
pedrocr
4y ago
Linux raid does that already. You can also add disks to grow an array or change it from raid5 to raid6. The gradual phase-in is what's missing. It makes sense as having what is effectively a raid0 of a 4 drive raid6 and a 6 drive raid6
118.
▲
by
pedrocr
4y ago
Does anyone know if any of these next generation filesystems can do raid6 with more flexibility than normal linux software raid? Say you do a 6x4TB raid 6 array in Linux and get a 16TB virtual drive. Now if you update 4 of the drives to 8TB
119.
▲
by
pedrocr
4y ago
If those languages achieve their memory safety at runtime, like C# does with GC, then that becomes relevant to the point of the compile time performance. The C# compiler has to do less as that complexity has been pushed elsewhere with diffe
120.
▲
by
pedrocr
4y ago
While it's true that memory safety has been done in a lot of languages for a long time this overstates the point. Doing memory safety at compile time is an important change that allows replacing C/C++ in a lot more situations. In
More ›