jrk

8 ms·

31.
by jrk 2y ago
SRAM is scaling significantly more slowly than logic in recent process nodes.
32.
by jrk 2y ago
It is a coincidence. They are unrelated hardware blocks and very different architectures.
33.
by jrk 2y ago
The have built and operated a growing number of their own data centers for years. Presumably this will go into those.
34.
by jrk 2y ago
If you annotate PDFs in Preview it is very common to find that they disappear when the PDF is shared with someone else (e.g. when giving feedback on a document), or even when it’s reopened in Preview in the future. Frequent inexplicable dat
35.
by jrk 2y ago
Within the market of discrete gaming GPUs, there has been a hierarchy from "low-" to "high-end" for decades -- since before integrated GPUs even existed. For almost 20 years, things with 192- or 256-bit memory busses h
36.
by jrk 3y ago
I was curious to follow your development, but the lack of RSS feed for your blog means I basically can't.
37.
by jrk 3y ago
A warning to others: This is far more contested, and more nuanced, than this post makes it out to be. I believe this narrative as stated would not fly without lots of debate and pushback in any mainstream economics department today. Among m
38.
by jrk 3y ago
Actually each of those x86 CPUs probably has at least two AVX FMA units, and can issue 16xFP32 FMAs per cycle – it’s at least “64 AVX cores”! :)
39.
by jrk 3y ago
As far as I can tell, the article notably never defines what "slices" are or what "multi-slice" means.
40.
by jrk 3y ago
Chapel has at least several full-time developers at Cray/HPE and (I think) the US national labs, and has had some for almost two decades. That's much more than $100k. Chapel is also just one of many other projects broadly intere
41.
by jrk 3y ago
My usual approach with valuable iCloud Drive folders is to have an automatic process rsync them somewhere else and auto-version them there (Git, Dropbox, etc.). This wouldn't be great for very large data, but most of what I care to agg
42.
by jrk 3y ago
Note that the shortage now is different than the past few years: it is not in fabricating any of the primary chips, it is entirely attributed to insufficient capacity in TSMC’s CoWoS packaging. This is only used for the highest-end accelera
43.
by jrk 3y ago
Nearly every time single old houses in in-demand areas are bought for renovation or redevelopment into denser duplexes, they're "bought by investors, including LLCs and other corporate entities." This doesn't mean 20% of
44.
by jrk 3y ago
That's a wonderfully evocative image, and I am a total non-expert so this sort of depth may be common, but I do note that that map pin is in what looks like harbor landfill. It may require an unusually deep foundation to reach bedrock
45.
by jrk 3y ago
SIM swapping lets you take over a phone number . It does not let you clone the keys stored in your physical phone’s secure storage / enclave / key storage. So SIM swapping makes SMS verification vulnerable (that just depends on c
46.
by jrk 3y ago
I think you may have triggered some concern or misunderstanding. What do you mean by "pretend it's private"?
47.
by jrk 3y ago
This is just a raw result of many frames strung together from a view synthesis technique, with an arbitrary, programmer-designed camera path. Neither of these issues is fundamental to the technique: - The camera path could be anything, and
48.
by jrk 3y ago
It's not an increase in capacity of per-GPU "GPU memory" (the HBM directly connected to the H100 here is up to 96GB, where the previous generation was 80GB), but rather reflects the product of two things: 1. Each node here is
49.
by jrk 3y ago
I am not the original poster, but I assumed they meant as an embodied entity, interacting in the physical world.
50.
by jrk 3y ago
No, starting with Ice Lake these are on the first actually-new core micro architecture after Skylake (it’s the “Cove” family, starting with Sunny Cove in Ice Lake). They have Lake in the name but these are actually a new micro architecture,
51.
by jrk 4y ago
https://www.sqlite.org/json1.html ?
52.
by jrk 4y ago
I think you missed the implied wink. This post is not arguing for throwing away all performance and safety, it is showing how “ugly syntax” is actually mostly inherent complexity from trying to be safe and fast.
53.
by jrk 4y ago
Indeed, almost the entire pitch reads like a recapitulation of the pitch for Terra 10 years ago. There are a few different design choices, but discussion of Terra is a striking omission from the FAQ.
54.
by jrk 4y ago
No, they don't. In spite of the confusing wording in the post you cite, its petabytes/year claim is not derived from the 500m tweets/day claim – it must include metadata and/or multimedia. This was all already derived (c
55.
by jrk 4y ago
This has been true for decades and has nothing to do with “AI,” it’s just classic computer graphics.
56.
by jrk 4y ago
I have always wondered exactly this. Market segmentation on what platforms/controllers support ECC: fine, whatever. But market segmentation of what is an “ECC DIMM” vs. a “regular DIMM”? It makes no sense that the commodity memory manu
57.
by jrk 4y ago
Yes. The calls to the platform graphics APIs and underlying drivers are identical (and literally hit the same code), whether from x86/Rosetta or native/ARM64.
58.
by jrk 4y ago
This is just https://www.wired.com/story/facebooks-listening-smartphone-m... in a different medium.
59.
by jrk 4y ago
"Energy crops" are burned as fuel. They aspire to be carbon-neutral (and are often still fed a lot of petroleum-based fertilizer), not carbon-negative.
60.
by jrk 4y ago
It’s still early days, but in addition to the paper the Exo language project lives here: https://exo-lang.dev You can think of this as an attempt to take our idea of a “user-schedulable language” from Halide much farther, both i

More ›