_zoltan_

10 ms·

211.
by _zoltan_ 10mo ago
which is exactly what a business should do. it's not like Nvidia doesn't invest a ton into R&D, but hey, they have the cash, why not use it? like a good business.
212.
by _zoltan_ 10mo ago
is this an ad? do you work there? I'm following the chip industry on a daily basis and never heard of them...
213.
by _zoltan_ 10mo ago
in several countries it's fully legal to download anything for personal use, so I don't care :P
214.
by _zoltan_ 10mo ago
the world is moving towards this, so all the naysayers will be dinosaurs :-)
215.
by _zoltan_ 10mo ago
Get used to it. Nobody will write readmes by hand. I've always hated it and now I'll just let the LLM write it.
216.
by _zoltan_ 10mo ago
that seems like a lot of work compared to click click watch that one can achieve with *arr stack.
217.
by _zoltan_ 10mo ago
could you list 3 things that you think are more important than the internet? (I know the internet is going to be fine; I just want to understand what you think ranks higher globally...)
218.
by _zoltan_ 10mo ago
I've meant in usec, sorry if that wasn't clear, given that the discussion that I've replied was about rpc latency.
219.
by _zoltan_ 10mo ago
just as an NVL72 rack today has 72 71 links ( 18 probably) in the rack connecting all those GPUs together.
220.
by _zoltan_ 10mo ago
what do you think "ethernet's overhead" is?
221.
by _zoltan_ 10mo ago
anybody writing single core code in 2025 professionally isn't very professional.. the core count doesn't matter. a top of the line Turin system has less than 1.4TB/s memory for the whole dual CPU system. A 2020 era A100 has 2
222.
by _zoltan_ 10mo ago
there are uses cases where those cores are used for aux processing. there is more to these boxes than AI :-)
223.
by _zoltan_ 10mo ago
go on reddit and you can see a million of these vibe coded codebases. is that not good enough?
224.
by _zoltan_ 10mo ago
yet it works? where have you been for the last 2 years? calling this snake oil is like when the horse carriage riders were against cars.
225.
by _zoltan_ 10mo ago
what progress has been made on understanding Teichmüller theory in the past decade? not a lot...
226.
by _zoltan_ 10mo ago
fully agree! with MGX and CX8 we see PCIe root moving to the NIC, which is very exciting.
227.
by _zoltan_ 10mo ago
the GH/GB compute has LPDDR5X - a single or dual GPU shares 480GB, depending if it's GH or GB, in addition to the HBM memory, with NVLink C2C - it's not bad!
228.
by _zoltan_ 11mo ago
what's the opposite problem statement?
229.
by _zoltan_ 11mo ago
Scanning Tunneling Microscope, high-temperature superconductivity - 2 Nobel prize right there. Then laser eye surgery, magnetic storage, relational databases, UPC barcodes, DES, FFT, RISC, ... yeah, almost nothing. /s disclaimer: I wor
230.
by _zoltan_ 11mo ago
nsys and nvtx are awesome. many don't know but you can use them without GPUs :)
231.
by _zoltan_ 11mo ago
I've never heard of jj before this thread...
232.
by _zoltan_ 11mo ago
my VW Multivan's gearbox needed a replacement. 17000 CHF, I kid you not. Luckily VW Germany paid for the whole ordeal, but I wouldn't have been happy to pay that...
233.
by _zoltan_ 11mo ago
you mixed up "is dead" with "is vital" :-)
234.
by _zoltan_ 11mo ago
I can't wait for BYD to enter the European market with a true minibus like VW's ID Buzz. There is a rumor that the M9 is coming in 2026.
235.
by _zoltan_ 11mo ago
how much?
236.
by _zoltan_ 11mo ago
tens of millions? I think you're severely underestimating it.
237.
by _zoltan_ 11mo ago
It's a coding model and you're not coding with it with that instruction.
238.
by _zoltan_ 11mo ago
into? bonds?
239.
by _zoltan_ 11mo ago
what is there to manage on a single instance, single VM...
240.
by _zoltan_ 11mo ago
the two statements can be true at the same time: they can still view software developers as second class while having a great hardware vision. what OP is saying is that after a point it doesn't matter how good your HW is if your SW sta

More ›