Dropless moe training just went 10x faster on gpus. nvidia's transformer engine work, out september…
dropless moe training just went 10x faster on gpus. nvidia's transformer engine work, out september 14, processes every token assigned to an expert without dropping under load imbalance, hitting 97% scaling across 1,024 gpus.
the bottleneck is now the network, not the math.
Context
NVIDIA's technical blog of 14 September 2026, Accelerating Dropless MoE Training in JAX with NVIDIA Transformer Engine, reports 10.4x throughput on GB200 for DeepSeek-V3 (103 to 1,068 TFLOPS per GPU) and 97 percent scaling efficiency at 1,024 GPUs on GB300 NVL72 training DeepSeek-V3 671B. Dropless MoE processes every token without dropping or padding. The post says communication took 84 percent of accumulated kernel time in the unoptimized baseline.
Both figures are first-party and vendor-reported. They are different configurations: the 10x is on GB200 against an unoptimized baseline of the blog's own, and the 97 percent is on GB300 NVL72 at 1,024 GPUs; the note combines them. The bottleneck is now the network is the author's take; the 84 percent describes the baseline and not a current bottleneck.
Watch next
- Third-party reproduction.
Sources
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 20 September 2026 at 22:04 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →