[HN Gopher] OpenTPU: Open-Source Reimplementation of Google Tens...
       ___________________________________________________________________
        
       OpenTPU: Open-Source Reimplementation of Google Tensor Processing
       Unit (TPU)
        
       Author : walterbell
       Score  : 152 points
       Date   : 2025-05-27 23:10 UTC (23 hours ago)
        
 (HTM) web link (github.com)
 (TXT) w3m dump (github.com)
        
       | mdaniel wrote:
       | Yeowzers that FAQ is filled with watch-outs
       | 
       | The /forks contained https://github.com/csirlin/OpenTGPTPU which
       | had a commit 3 hours ago but it seems they have not yet updated
       | the FAQ for their version. Anyway, the fact it has commits
       | greater than 8 years ago makes it seem like a more reasonable
       | submission
        
       | walterbell wrote:
       | Google TPU engineers used open-source Chisel for ASIC design
       | (2018), https://youtube.com/watch?v=x85342Cny8c
       | 
       |  _" Google Edge TPU devices"_, 100 comments (2019),
       | https://news.ycombinator.com/item?id=19130896 &
       | https://news.ycombinator.com/item?id=19313813
       | 
       |  _" Coral Edge TPU review"_, 100 comments (2020),
       | https://news.ycombinator.com/item?id=24808755
       | 
       |  _" TPU transformation: 10 years of our AI-specialized chips"_,
       | 60 comments (2024), https://news.ycombinator.com/item?id=41148532
        
         | dekhn wrote:
         | The site confuses the inference engine in the Edge TPU with the
         | datacenter TPU. They are two unrelated projects. Based on the
         | paper they're borrowing from, I think they are trying to go for
         | a much older datacenter inference-only TPU, or only
         | implementing the inference capabilities of the datacenter TPU.
        
           | walterbell wrote:
           | Are there recent papers on datacenter TPU?
        
             | dekhn wrote:
             | Yes.
        
               | walterbell wrote:
               | David Patterson overview (2023),
               | https://www.cs.ucla.edu/wp-
               | content/uploads/cs/PATTERSON-10-L...
               | 
               | TPU v4 (2023), https://arxiv.org/abs/2304.01433
        
       | flakiness wrote:
       | [2017] (https://arxiv.org/abs/1704.04760)
        
         | walterbell wrote:
         | [May 2025]
         | (https://github.com/csirlin/OpenTGPTPU/commits/master)
        
           | flakiness wrote:
           | Wow they have kept working on this! Thanks for pointing this!
           | very impressive.
        
       | whimsicalism wrote:
       | > The TPU is Google's custom ASIC for accelerating the inference
       | phase of neural network computations.
       | 
       | this seems hopelessly out of date/confused
        
         | walterbell wrote:
         | Additional text from Google's 2017 paper abstract says:
         | This paper evaluates a custom ASIC---called a Tensor Processing
         | Unit (TPU)---deployed in datacenters since 2015 that
         | accelerates the inference phase of neural networks (NN). The
         | heart of the TPU is a 65,536 8-bit MAC matrix multiply unit
         | that offers a peak throughput of 92 TeraOps/second (TOPS) and a
         | large (28 MiB) software-managed on-chip memory.             The
         | TPU's deterministic execution model is a better match to the
         | 99th-percentile response-time requirement of our NN
         | applications than are the time-varying optimizations of CPUs
         | and GPUs (caches, out-of-order execution, multithreading,
         | multiprocessing, prefetching, ...) that help average throughput
         | more than guaranteed latency.             The lack of such
         | features helps explain why, despite having myriad MACs and a
         | big memory, the TPU is relatively small and low power. We
         | compare the TPU to a server-class Intel Haswell CPU and an
         | Nvidia K80 GPU, which are contemporaries deployed in the same
         | datacenters.
        
           | whimsicalism wrote:
           | hence the out of date part of my comment
        
             | walterbell wrote:
             | Recent (2024) description by Google,
             | https://cloud.google.com/blog/transform/ai-specialized-
             | chips...                 TPUs were purpose-built
             | specifically for AI. TPUs are an application-specific
             | integrated circuit (ASIC), a chip designed for a single,
             | specific purpose: running the unique matrix and vector-
             | based mathematics that's needed for building and running AI
             | models..            TPU v2.. built an interconnected
             | machine -- our first TPU pod -- with 256 TPU chips
             | connected with a very high-bandwidth, custom interconnect..
             | liquid cooling was added with TPU v3 to help address
             | efficiency needs, while TPU v4 introduced optical circuit
             | switches to allow the chips in pods to communicate even
             | faster and more reliably.             TPUs also underpin
             | Google DeepMind's cutting-edge foundation models, including
             | the newly unveiled Gemini 1.5 Flash, Imagen 3, and Gemma 2,
             | propelling advancements in AI.. Forget about a single chip,
             | or a single TPU pod -- we're building a global network of
             | data centers filled with TPUs.
        
           | throwawaymaths wrote:
           | what's the memory bandwidth? IIRC that is the limiting factor
           | in LLM hardware today
        
             | walterbell wrote:
             | Slide 21, https://files.futurememorystorage.com/proceedings
             | /2024/20240...                           TPUv3     TPUv4
             | HBM2 BW   900 GB/s  1200 GB/s
        
         | surfmike wrote:
         | How would you describe it instead? Curious and learning
        
           | imtringued wrote:
           | Google does everything, both inference and training, on their
           | TPUs.
           | 
           | Inference is easier, since the person deploying a model knows
           | the architecture ahead of time and therefore can write custom
           | code for their particular model.
           | 
           | When training you want to be as flexible as possible. The
           | framework and hardware should not impose any particular
           | architecture. This means lots of kernels and combinations of
           | kernels. Miss one and you're out.
        
             | throwawaymaths wrote:
             | > Miss one and you're out.
             | 
             | well these days since everything is transformer, your pool
             | of choices is less daunting and theres only about four or
             | five places that someone might get clever.
        
         | dgacmu wrote:
         | They're not confused at all, this is just a (correct)
         | description of TPU v1. The repository is 8 years old.
        
       | andutu wrote:
       | There is an excellent paper and talk on how Google's TPU cluster
       | is managed:
       | https://www.usenix.org/conference/nsdi24/presentation/zu.
        
       | westurner wrote:
       | Can [OpenTPU] TPUs be fabricated out of graphene, with
       | nanoimprinting or a more efficient approach?
       | 
       | From https://news.ycombinator.com/item?id=42314333 :
       | 
       | >> _From "A carbon-nanotube-based tensor processing unit" (2024)
       | https://www.nature.com/articles/s41928-024-01211-2 :_
       | 
       | >>> _Using system-level simulations, we estimate that an 8 bit
       | TPU made with nanotube transistors at a 180 nm technology node
       | could reach a main frequency of 850 MHz and an energy efficiency
       | of 1 tera-operations per second per watt._
        
         | westurner wrote:
         | What about QPUs though?
         | 
         | Can QPUs (Quantum Processing Units) built on with electrons in
         | superconducting graphene ever be faster than photons in
         | integrated nanophotonics?
         | 
         | There are integrated parametric single-photon emitters and
         | detectors.
         | 
         | Is there a lower cost integrated nanophotonic coherent light
         | source for [quantum] computing than a thin metal wire?
         | 
         | "Electrons turn piece of wire into laser-like light source"
         | (2022) https://news.ycombinator.com/item?id=33493885
        
       ___________________________________________________________________
       (page generated 2025-05-28 23:01 UTC)