framework
Nvidia software engine that assigns parts of an inference workload to the appropriate chip in mixed LPU and GPU deployments.
No evidence-backed relationships are recorded.
Nvidia says DSX MaxLPS power sharing fits up to 40% more GPUs in the same approved power budget, citing an Nscale trial of 192 GPUs against 140. The gain holds only while GPUs rarely peak together, and each operator has to measure that on its own fleet before buying hardware against it.
Reality
Record data center revenue of $75.2 billion and a 25-fold dividend increase sit next to a second-quarter guide implying 11.5 percent sequential growth, against 20 percent just delivered.
Reality
One Dell rollup alone carried 435 CVEs, among them a kernel privilege-escalation bug that had already shipped in two other Dell advisories. That is why a single vendor feed cannot describe a GPU estate.
Reality
MLPerf Inference v6.1 preview submissions put Vera Rubin NVL72 at up to 3.7x GB300 on Qwen3-VL and up to 2.5x on DeepSeek-R1, on two different inference frameworks. The four-rack 99% scaling result is an offline number.
Reality
The one measured number behind NVIDIA's tokens-per-megawatt pitch comes from Lambda's Blackwell cluster, where power reclaimed from static provisioning ran three extra nodes. Amazon's Annapurna Labs and d-Matrix got a line each.
Reality
The kernel was the one place a Rust AI stack still had to change language. NVIDIA's cuda-oxide closes that gap, if you have Linux, a compute capability 8.0 card, CUDA 12.x and the nightly-2026-04-03 toolchain.
Reality
Dynamo can now run the vision encoder as its own worker, and the win is real where encoding dominates a request. The arithmetic behind the headline number tells you which traffic it transfers to, and NVIDIA names the one topology it never beats.
Reality
A CNCF walkthrough puts the money on accelerator utilisation rather than serving throughput, and argues Kubernetes now supplies most of the parts, with the isolation half of the job still sitting on the platform team's desk.
Reality
The first third-party benchmark of the LP30 rack came in at roughly four times the next-fastest public endpoint, measured one request at a time on a model small enough to fit.
Reality
NVIDIA Dynamo keeps a pre-warmed engine on the same GPUs and hands it the resident weights instead of reloading them. The measured recovery window falls to about 2.6% of a cold restart.
Reality
SemiAnalysis has pushed fixed-length serving into maintenance mode. On replayed Claude Code traffic, NVIDIA's GB300 is credited with 15x Hopper, against 40x over H200 on the retired static test.
Reality
NVIDIA says Alibaba's largest open-weight model serves over 4K tokens/sec/GPU and 350 tokens/sec/user in FP8 on a GB300 NVL72. That figure is the self-hosting floor, not a benchmark.
Reality