Openspace Captial - Running 850m AUM. Exited Asian Food Network 2013.. Views own. I have a pizza oven :)
singapore
Joined July 2009
- Tweets2.7K
- Following577
- Followers1.6K
- Likes400
hian goh retweeted
MiniMax M3.1 Flash is now on @tenstorrent Cloud -bringing frontier AI to Tensix silicon, Ethernet scale-out, and an open-source stack from the team led by Jim Keller.
Open weights. Diverse architectures. An open, heterogeneous AI future.🫡
Today @MiniMax_AI released the open weights for MiniMax M3.1 Flash – available now on Tenstorrent Cloud.
Explore in the playground or connect to your agent harness. Start building → bit.ly/47vU2zc
hian goh retweeted
We raised $1.5B to deploy Galaxy globally.
Thanks to our lead investors Trace Capital and Innovation X for backing that growth.
More to come: tenstorrent.com/newsroom/ten…
hian goh retweeted
NVIDIA DELETED NVLINK FROM CONSUMER CARDS. JIM KELLER BUILT A COMPANY WHERE THE INTERCONNECT IS THE ENTIRE PRODUCT, OPEN SOURCED THE WHOLE STACK, AND THE CARDS STILL DROPPED OFF THE BUS AFTER TWO DAYS
here is the whole install, unboxing to working:
two Tenstorrent Blackhole P150a, 32GB each, no video out because they are not GPUs -> detected on first boot, then detection vanished -> reseat, swap slots, single card, bios, nothing -> moved one card to a different machine on Fedora and it came up immediately -> LLMs, image gen and speech to text all running on it now.
so the pitch is not "faster than a GPU". the pitch is "your vram can actually be one pool again", and the entire software stack is open source while nvidia's is not.
this is where they are honest and most vendors are not, and it is five things:
the interconnect is the product, not the chip
-> two cards is a dev kit, 32 in a Galaxy Box is the actual machine
-> buying one or two to run a big model is buying the wrong half of the idea
the stack is immature and they say so out loud
-> the documentation targets Ubuntu 22.04, he was on 26, and that mismatch is a real day of your life
-> support answered fast and could not explain the dropout either, which is what a developer platform actually looks like
open source is the long bet
-> you can read every layer of their stack, which is not true of cuda, and that matters more in year three than in week one
-> people underrate this until the first time a driver blob decides their card is obsolete
future architectures are the reason to watch
-> as models move past plain mixture-of-experts into shapes that need constant cross-chip traffic, on-card meshing stops being a nice feature and becomes the requirement
-> right now, for one person who just wants an LLM running tonight, it is not the purchase
nvidia charges $25,000 for a card because pooled memory is the product. tenstorrent put the pooling port on a card you can actually buy and open sourced everything around it.
my take, and it is the uncomfortable one: the company most likely to break nvidia's grip is not shipping a faster chip, it is shipping the connector nvidia deleted, and it will take two more years of driver work before anyone notices.
read the article below. that's the difference between buying a chip and buying a fabric.
hian goh retweeted
Today, we launch JapanFold.
This is the first bio platform built from scratch, as one beautiful stack, optimized from @tenstorrent silicon to @aiand_ serving.
This will make it the highest throughput, lowest cost service for bio models.
Use it with claude, cursor, codex, or web app.
We believe in subsidizing inference for frontier domains. Bio is where we're starting.
JapanFold, built with @tenstorrent, brings the latest open-source drug discovery models — Boltz-2, OpenFold3, ESMFold-2, and more — to everyone: researchers, drug companies, academia, and startups. Industry-leading performance, at no cost to try, running on @aiand_ data centers in Japan.
Read the full announcement here aiand.com/newsroom/japanfold…
hian goh retweeted
I’m reading this on the plane right now, and it genuinely gave me chills. You should all give it a read.
And a huge thank you to @zartbotF for writing such an incredible piece.
zartbot.github.io/blog/arch/…
A Must follow human being.
Ten years ago, we started @cerebras around an approach many believed was impossible.
As a computer architect, it is hard for me to imagine a more exciting time. Model releases are accelerating, and hardware tapeout is compressing from multi-year roadmaps to annual launches.
Hot Chips is my favorite conference, and it’s where I launched Cerebras 7 years ago. This year’s conference was especially exciting, and so much innovation was shared. I am watching the industry recreate itself: SRAM is mainstream, DRAM is moving into the third dimension, networks are being fundamentally redesigned, and AI is helping design and program the chips themselves.
The industry has never moved faster and some of the hardest architectural questions are still wide open.
2
304
Skeptics will say Cerebras does not work because they only served small models but their latest efforts to serve gpt 5.6 sol ultrafast has gotten me to change their mind. Maybe they really have managed to build something different? 🤔🤔@insane_analyst @IanCutress @bubbleboi
1
298
hian goh retweeted
average chip architect role interview
interviewer: why do you want to join this company?
interviewee: to make the compiler engineers lives as difficult as possible.
hired!
hian goh retweeted
🌶️ @OpenAI
hian goh retweeted
Hello you fine Internet folks,
Continuing our @hotchipsorg 2026 coverage is talking about NVIDIA porting CUDA to RISC-V and how they are going about that.
Hope y'all enjoy!
chipsandcheese.com/p/hot-chi…
hian goh retweeted
Finally got to try the new deepseek v4 flash model this weekend and it's been very excited about the use cases it unlocked for me. Here are the experiments I've ran so far.
hian goh retweeted
🚨 6 NVIDIA competitors all found the same weakness in CUDA
@gpuemi wrote deep dives on 6 nvidia competitors. different teams, architectures, and even countries, but all 6 independently arrived at similar conclusions about what's wrong with CUDA:
every single one killed hardware caches.
tenstorrent, cerebras, trainium, TPU, sambanova, furiosa all use software-managed SRAM with zero hardware caches. NVIDIA is the only major AI chip maker still running L1/L2. the bet is that AI workloads have predictable enough access patterns that speculative caching is wasted silicon. you never pay for cache misses if every byte is explicitly placed.
every single one abandoned threads. CUDA's thread/warp/block model exists because GPUs were originally built for graphics, where you don't know which pixels need more work. AI workloads are regular. matmul, attention, MoE routing, the access patterns are completely predictable.
tenstorrent uses producer-consumer pipelines. cerebras uses dataflow with wavelets. trainium writes sequential code and lets the compiler extract parallelism. TPU executes sequentially on a wide vector unit. sambanova maps spatially onto silicon. furiosa compiles tensor contractions directly. six solutions all agreeing threads are the wrong abstraction for AI.
most of them eliminated synchronization entirely. __syncthreads(), atomics, barriers, race conditions, which is arguably the hardest part of GPU programming.
cerebras doesn't need synchronization because dataflow is inherently ordered.
TPU doesn't need it because execution is sequential.
sambanova doesn't need it because the pipeline is spatial.
tenstorrent replaced it with circular buffers.
all six keep data on-chip between layers. on a GPU, every transformer layer reads from HBM and writes back to HBM. on these chips, data stays in SRAM and flows directly to the next operation.
there's also a clear spectrum forming. tenstorrent gives you a full C++ kernel SDK. trainium and TPU give you python DSLs. sambanova and furiosa currently give you no kernel API at all, but furiosa does plan to offer a low-level programming API (kernel-level) and a higher-level API (DSL-level) to the developer community in the coming months.
generally, more control = higher ceiling but harder to program. less control = at the mercy of the compiler.
hian goh retweeted
Tenstorrent Blackhole p150a is live.
Fired up tt-toplike to monitor it. Really cool arcade visualizer by @episod
The case for glm5.2
I cannot believe how good GLM 5.2 is. Several weeks in now and it's mostly all I have been using. It doesn't have the alignment issue of cloud AI, it's much more clear what it can do and can't because it doesn't care about you taking another $$$ spin at the token slot machine.
3
435
hian goh retweeted
Thank you Tokyo! 🇯🇵 Here’s everything we announced at TT-Deploy Japan:
Faster AI Inference
• Kimi K2.6 900 t/s/u, 3x faster than GPUs
• LTX 2.3 Fast 6 sec video gen in ~6 sec, 144 frames, 1080p, 4x faster than GPUs
• DeepSeek-R1-0528 671B 400+ t/s/u
TT-Ascalon S Available Today
• A licensable RISC-V CPU built for the next generation of agentic AI applications
Heterogenous or Stand Alone
• Easily deploy Tenstorrent Galaxy alongside existing infrastructure or standalone
• @aiand_'s sovereign heterogenous inference platform with Tenstorrent Galaxy™ superclusters
