--- title: "MTIA 300: Meta’s First Training Chip with Built-in NICs and Communication-Offloading Engines" url: https://daily.dev/posts/mtia-300-meta-s-first-training-chip-with-built-in-nics-and-communication-offloading-engines-jpxe0edsp source_url: https://engineering.fb.com/2026/08/24/networking-traffic/mtia-300-meta-training-chip-built-in-nics type: article source: "Facebook Engineering\n" published: 2026-08-24T17:50:21.320Z updated: 2026-08-24T17:50:46.116Z tags: ["machine-learning", "pytorch"] reading_time: 5 upvotes: 0 comments: 0 language: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # MTIA 300: Meta’s First Training Chip with Built-in NICs and Communication-Offloading Engines **[Facebook Engineering ](https://daily.dev/sources/facebook_code)** · 5 min read · 0 upvotes · 0 comments ## Summary Meta introduces MTIA 300, its first training-optimized chip built specifically for recommendation and ranking models, featuring built-in NIC chiplets and dedicated communication-offloading engines. Unlike GPUs, where collective communication operations compete with compute for shared resources, MTIA 300 integrates 12 custom 800 Gbps RDMA NICs directly on-chip and 16 dedicated message engines that handle AllReduce, AllToAll, and AllGather collectives independently, achieving less than 0.5% compute degradation versus over 20% on GPUs. Co-designed with the HCCL communication library, which compiles collectives into autonomous work-queue subgraphs executed without host involvement, the system delivers up to 940 GB/s bandwidth per rack and 3.9x faster communication than equivalent GPU clusters on a 150-billion-parameter production model across 40 accelerators. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What makes Meta's MTIA 300 chip different from GPUs for training recommendation models? MTIA 300 integrates 12 custom 800 Gbps RDMA NICs directly into the chip package, providing 1.2 TB/s of I/O bandwidth without crossing a PCIe bus, and includes 16 dedicated message engines that handle AllReduce, AllToAll, and AllGather collectives independently of the compute grid. This isolation limits compute degradation to under 0.5% during concurrent collectives, versus over 20% degradation seen on general-purpose GPUs where NCCL kernels compete with training kernels for streaming multiprocessors. _Teams weighing custom AI silicon against GPUs for training can track chip architecture news like this on daily.dev._ ### How much faster is MTIA 300 than GPUs for communication in production recommendation model training? On a 150-billion-parameter production recommendation model running across 40 accelerators, MTIA 300's total communication time is 3.9 times faster than an equivalent GPU cluster. HCCL, the co-designed communication library, achieves up to 940 GB/s of communication bandwidth within a single rack by compiling collectives into autonomous work-queue subgraphs executed without host involvement. _Engineers benchmarking accelerator communication performance can follow hardware deep dives like this via daily.dev._ ### How does HCCL integrate with PyTorch for collective communication on MTIA hardware? HCCL integrates with PyTorch's c10d and torchcomms interfaces, allowing collectives traced through torch.compile to be compiled into a single graph alongside compute operators. HCCL then selects topology-aware algorithms that exploit the asymmetric bandwidth between scale-up (within-rack, up to 1 TB/s) and scale-out (cross-rack, 200 GB/s) communication, minimizing cross-rack traffic where bandwidth is constrained. _Developers integrating custom accelerators with PyTorch workflows can find architecture writeups like this on daily.dev._ ## Similar posts on daily.dev - [Four MTIA Chips in Two Years: Scaling AI Experiences for Billions](https://daily.dev/posts/four-mtia-chips-in-two-years-scaling-ai-experiences-for-billions-kg6utbveo) · AI at Meta · 0 upvotes · 0 comments - [Meta to put its own AI chip into production in September, aiming to double computing capacity](https://daily.dev/posts/meta-to-put-its-own-ai-chip-into-production-in-september-aiming-to-double-computing-capacity-vr3c7tjss) · The Next Web · 0 upvotes · 0 comments - [Meta reveals custom AI chips it says beat Nvidia](https://daily.dev/posts/meta-reveals-custom-ai-chips-it-says-beat-nvidia-ei3upkpvd) · The Register · 1 upvotes · 0 comments --- Tags: [#machine-learning](https://daily.dev/tags/machine-learning), [#pytorch](https://daily.dev/tags/pytorch) [View this post on daily.dev](https://daily.dev/posts/mtia-300-meta-s-first-training-chip-with-built-in-nics-and-communication-offloading-engines-jpxe0edsp)