---
format: "aidr-story-markdown/v1"
id: "3005162da20f4ccbd78451d93a63a755223affb90d66faa1b8a445540c65602d"
canonical_url: "https://aidr.today/3005162d?lang=en"
title: "Nvidia Delivers Up to 42 Times AMD’s DeepSeek Performance per Dollar, SemiAnalysis Says"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-09-12T08:40:12.000Z"
category: "Chips"
topics: ["nvidia","amd","deepseek","chips","inference"]
source_urls: ["https://huggingnews.com/ai/update-nvidia-delivers-up-to-42-times-amds-deepseek-performance-per-doll-46674fd3","https://x.com/SemiAnalysis_/status/2098618867035557984","https://x.com/DesignArena/status/2098489852027375784","https://x.com/SemiAnalysis_/status/2098501866401218813","https://x.com/SemiAnalysis_/status/2098501868406137138"]
summary: "AMD publicly released a software image for running DeepSeek V4.1 Flash two days after Nvidia’s CUDA version of vLLM supported the model, SemiAnalysis said. The image works out of the box, but Nvidia’s B200 and B300 deliver up to 42 times its performance per dollar, while the H200 delivers up to 14.8 times, the research firm said. SemiAnalysis had previously reported that the image specified in AMD’s vLLM documentation remained unavailable 23 hours after the model’s release, while Nvidia’s implementation worked across six chip models at launch. The vLLM project, however, announced on Sept. 10 that its launch support was “verified on NVIDIA and AMD GPUs!” SemiAnalysis attributed Nvidia’s advantage to CUDA’s software ecosystem and collaboration with a community of six million developers."
---

# Nvidia Delivers Up to 42 Times AMD’s DeepSeek Performance per Dollar, SemiAnalysis Says

> [Open the canonical story](<https://aidr.today/3005162d?lang=en>)

**Published:** 2026-09-12T08:40:12.000Z
**Category:** Chips
**Topics:** nvidia, amd, deepseek, chips, inference

## Summary

AMD publicly released a software image for running DeepSeek V4\.1 Flash two days after Nvidia’s CUDA version of vLLM supported the model, SemiAnalysis said\. The image works out of the box, but Nvidia’s B200 and B300 deliver up to 42 times its performance per dollar, while the H200 delivers up to 14\.8 times, the research firm said\. SemiAnalysis had previously reported that the image specified in AMD’s vLLM documentation remained unavailable 23 hours after the model’s release, while Nvidia’s implementation worked across six chip models at launch\. The vLLM project, however, announced on Sept\. 10 that its launch support was “verified on NVIDIA and AMD GPUs\!” SemiAnalysis attributed Nvidia’s advantage to CUDA’s software ecosystem and collaboration with a community of six million developers\.

## Sources

- [Story source](<https://huggingnews.com/ai/update-nvidia-delivers-up-to-42-times-amds-deepseek-performance-per-doll-46674fd3>)
- [Supporting source](<https://x.com/SemiAnalysis_/status/2098618867035557984>)
- [Story source](<https://x.com/DesignArena/status/2098489852027375784>)
- [Supporting source](<https://x.com/SemiAnalysis_/status/2098501866401218813>)
- [Supporting source](<https://x.com/SemiAnalysis_/status/2098501868406137138>)