English
Eric Quinnell
582 posts
Eric Quinnell
@divBy_zero
Founding Chip Architect, Stealth Startup. Fmr AWS Tranium, Tesla Dojo, CPUs (x86+ARM). PhD Computer Arithmetic
Katılım Aralık 2022
391 Takip Edilen2.4K Takipçiler
@divBy_zero @corsix Lol so what everyone is teaching me is that there's no hope for a much saner world if I switch career to silicon design.
English
As a SW guy what puzzles me most in chip design is that it's a total monopoly of one programming language: Verilog. Yeah there's some higher-level stuff, some variants, but all ultimately translated to Verilog.
Where's the fun without a hundred different PLs to confuse you
English
@opinali Verilog already comes with the confusion level of a hundred different PLs
English
@Leik0w0 It’s usually better for hardware systolic arrays. Hw doesn’t execute matmuls in the same physical directions as paper and pencil matrices - the row and col weights come from the same physical place, so it has to be row+row or col+col. Someone is doing the xpose somewhere
English
@insane_analyst He who controls the SPICE controls the universe
English
It's been over an hour and I still can't import external SPICE models correctly.
WTF am I doing wrong I made a spice directive like the tutorials said.
English
To be fair, at low enough FP quantization it does indeed return to associative behavior.
Bc it’s INT
SemiAnalysis@SemiAnalysis_
Floating point math is not associative! And many of the highest performance kernels split the workload among SMs and accumulate partial results in a nondeterministic order. Many AI labs just accept this, or pay a huge performance penalty for determinism. DeepSeek decided to do neither. (1/4) 🧵
English
Eric Quinnell retweetledi
i believe it was @divBy_zero that wisely said that, contrary to common sense, it is always easier to fix performance problems in silicon than change an entrenched sw stack
English
why isn't there a startup doing a RISC-V extension for the Python Virtual Machine
put a bunch of PyCores on a chip and you win the whole "Agentic CPU" market
English
@yacineMTB Confirmed anecdotal data point.
(Mid 40s, but close enough. This vibe coding stuff is legit)
English
Callout to the many leads and engineers who worked this over the years, esp @rawat_ritvik @aaronsrogers and Pete.
This will be a massive (and needed) upgrade to all cars and bots.
English
Congrats AI5 team, I know it was a rocky road
Elon Musk@elonmusk
Congrats to the @Tesla_AI chip design team on taping out AI5! AI6, Dojo3 & other exciting chips in work.
English
Eric Quinnell retweetledi
Hello, Moon. It’s great to be back.
Here’s a taste of what the Artemis II astronauts photographed during their flight around the Moon. Check out more photos from the mission: nasa.gov/artemis-ii-mul…
English
@LottoLabs @ptremblay Pedantically you are correct, yes. Way less loss than current quantization, and that detail would derail non technical folks ever further, so I didn’t split the hairs
English
@ptremblay @divBy_zero Fair its no loss in accuracy but not lossless in the traditional sense of recovering bit for bit
English
@CliffLattner Yes, exactly. If doing inference, the compute will hide under the dram loads, even at high batch. For training, it is extra compute to pay.
English
@divBy_zero IMO its a no brainer that if you are willing to spend more cycles on quantization/dequantization, and forgo the savings you get from computing on the quantized data, you can do better than ordinary abs-max. I doubt that it is anywhere close to 8x over say fp8.
English
Two days of weird takes, so I must:
“8x perf” is 32-bit baseline vs 4-bit compressed
“KV cache” is merely a use case and hard to capitalize full perf. And yes, many already compress here.
But it’s lossless. The others aren’t.
We should have been using this all along.
Google Research@GoogleResearch
Introducing TurboQuant: Our new compression algorithm that reduces LLM key-value cache memory by at least 6x and delivers up to 8x speedup, all with zero accuracy loss, redefining AI efficiency. Read the blog to learn how it achieves these results: goo.gle/4bsq2qI
English
My flight lands, I exit the plane, look for the Baggage Claim sign, and then there it was:
A 448Gpbs PAM4 Keysight waveform analyzer advertisement, on an LED backlight mega screen.
Thank you SJC
English
@GoogleResearch Classic compressor, seems obvious in retrospect, we should have all done this earlier.
Well done Google Research, truly
English
Introducing TurboQuant: Our new compression algorithm that reduces LLM key-value cache memory by at least 6x and delivers up to 8x speedup, all with zero accuracy loss, redefining AI efficiency. Read the blog to learn how it achieves these results: goo.gle/4bsq2qI
GIF
English
Eric Quinnell retweetledi
Levels should end asking you “is it dark outside? No looking at your watch”, after cabling and manually PXE booting for 16 hours in a row
P.M@p_misirov
there is a game called "data center" on steam which let's you build and manage your own data center. this is lowkey genius, the best way to educate people on a new trait. hyperscalers should learn a thing or two from "edutainment".
English