Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
zzzoom
searching Neon…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
14 ms
·
1.
▲
by
zzzoom
22d ago
No way to avoid the memory cartel, even if CXMT catches up.
2.
▲
by
zzzoom
1mo ago
Each minute that a GPU isn't running is money evaporating
3.
▲
by
zzzoom
1mo ago
Wow, if only the Chinese knew that! So easy!
4.
▲
by
zzzoom
2mo ago
Matrix multiplication is one of the few operations that isn't regularly limited by memory bandwidth. BLAS implementations come with several heavily optimized, architecture-specific versions of sgemm.
5.
▲
by
zzzoom
3mo ago
Isn't the delta fed to an arithmetic coder?
6.
▲
by
zzzoom
3mo ago
Not as many as the ones that the US snatched in operation paperclip
7.
▲
by
zzzoom
3mo ago
I'd bet on the inverse: China scaling DRAM production until the price crashes, and the whole US stock market that is propped on top of that scarcity going down with it.
8.
▲
by
zzzoom
4mo ago
Prefill (GEMM) is compute bound, decode (GEMV) is memory bound.
9.
▲
by
zzzoom
4mo ago
EPYC 9965: 614GBps of 12-channel DDR5-6400 A100: 1935GBps of HBM2e Most of those FLOPS are constrained by memory bandwidth.
10.
▲
by
zzzoom
4mo ago
A social safety net
11.
▲
by
zzzoom
4mo ago
And definitely not more evil than the workers at current Meta.
12.
▲
by
zzzoom
4mo ago
Is it? Spack has only one package that depends on chapel.
13.
▲
by
zzzoom
4mo ago
And state actor LLMs probably don't need convincing
14.
▲
by
zzzoom
4mo ago
Copy Fail was found using AI and made our last week miserable (thanks RedHat). Anyone with access to a SOTA model can point it at any package in your stack and go fish. You're underreacting.
15.
▲
by
zzzoom
5mo ago
> GabeN also had the wrong take in that it's a "supply problem" or whatever nonsense he said. It is a supply problem. Steam regional pricing and game passes have demolished piracy in countries where people wouldn't ha
16.
▲
by
zzzoom
5mo ago
If the image contains information required to bring up the device, why isn't that data shipped in firmware?
17.
▲
by
zzzoom
5mo ago
At some point SBCs that require a custom linux image will become unacceptable, right? Right?
18.
▲
by
zzzoom
5mo ago
It's predatory pricing.
19.
▲
by
zzzoom
5mo ago
If any of those addresses sold a single sat the price would crash hard .
20.
▲
by
zzzoom
6mo ago
That's a genius idea, keep adding broken stuff into the standard until there's no choice but to break compatibility to fix it.
21.
▲
by
zzzoom
6mo ago
Imagine considering some bird poop staining the paint dangerous instead of the air pollution that's slowly killing you.
22.
▲
by
zzzoom
6mo ago
I can't even play the decent free games I got because I can't find them in the UI. It doesn't have sort by rating (or any other popularity metric) so you have to wade through the junk. Imagine paying for that experience...
23.
▲
by
zzzoom
6mo ago
They did try to pull a Valve and use their successful game to fund a game store that prints money.
24.
▲
by
zzzoom
6mo ago
The 9800X3D has wider everything . Decoder, execution ports, vectors, cache, memory bandwidth...
25.
▲
by
zzzoom
8mo ago
If you get the "You're absolutely right!" response from an LLM that screwed up on a field you're familiar with and still let them play with your health, you're...courageous to say the least.
26.
▲
by
zzzoom
8mo ago
You'd have to be quite tall to average an ~80cm step. 193cm (6'4") according to a quick search.
27.
▲
by
zzzoom
9mo ago
Because most ARM SBCs are still limited to whatever linux distro they added support to. Intel SBCs might underperform but you can be sure it will run anything built for x86-64.
28.
▲
by
zzzoom
9mo ago
Apple's A4 was launched 4 years before RISC-V.
29.
▲
by
zzzoom
9mo ago
Why would the market jump from one proprietary ISA to another proprietary ISA?
30.
▲
by
zzzoom
10mo ago
> They could make a systolic array TPU and software, perhaps. But it would mean abandoning 18 years of CUDA. Tensor cores are specialized and have CUDA support.
More ›