OpenAI Reports Up to 14% PPA Gain on AMD Flagship GPU IP From Model-Rewritten RTL
OpenAI reported on November 18, 2026 that its models achieved up to a 14% improvement in power, performance and area (PPA) on AMD flagship GPU intellectual property, measured against heavily optimized baselines. Some of the circuit rewrites were included in a design sent for manufacturing.
OpenAI reported on November 18, 2026 that its models achieved up to a 14% improvement in power, performance and area (PPA) relative to heavily optimized baselines on AMD flagship GPU intellectual property (IP). According to OpenAI, some of the circuit rewrites were included in a design sent for manufacturing. Unite.AI covered the report.
How the gains were produced
The improvements came from rewriting SystemVerilog RTL, the code that describes a chip's circuits. The model used Synopsys tools for two tasks:
- Verifying that each rewrite preserved the behavior of the block it modified.
- Assessing PPA trade-offs for each change.
The 14% figure is stated as an upper bound ("up to") and is measured against baselines that OpenAI describes as heavily optimized.
Key figures
- Improvement: up to 14% in PPA
- Comparison point: heavily optimized baselines
- Target: AMD flagship GPU IP
- Method: model-driven rewrites of SystemVerilog RTL
- Verification tooling: Synopsys
- Manufacturing status: some rewrites included in a design sent for manufacturing
OpenAI has not named a specific model in the available summary, and no pricing, context window or benchmark scores apply to this report.
What this means
The notable detail is that the rewrites are not confined to an experiment. Some are included in a design sent for manufacturing, so model-generated RTL changes have reached a production hardware flow. The use of Synopsys tools to check that each rewrite preserves a block's behavior, and to assess PPA trade-offs, shows the workflow pairs model output with established verification steps rather than relying on the model alone.
The 14% figure is OpenAI's own claim and a maximum, so it should not be read as an average across the AMD IP. The baselines are described as heavily optimized, which makes the stated gain a comparison against already-tuned designs rather than unrefined ones.
Related Articles
OpenAI models optimized circuits used in Jalapeño silicon and improved RISC-V core designs
OpenAI reports that its models optimized circuits used in Jalapeño silicon and improved RISC-V core designs. The company presents the work as progress in AI-assisted chip design with GPT-6 Astra.
OpenAI Python SDK v3.53.0 adds GPT-6.2 Astra, Astra Law and Sol identifiers
OpenAI Python SDK v3.53.0, dated 2026-11-17, adds model identifiers for GPT-6.2 Astra, GPT-6.2 Astra Law and GPT-6.2 Sol. It also adds 2-hour and 8-hour Responses background windows.
OpenAI's GPT-6.2 Astra ranks second on Artificial Analysis index at 62, behind Claude Fable 5.2 (65)
OpenAI released GPT-6.2 Astra and GPT-6.2 Sol. Artificial Analysis ranks Astra second in its Intelligence Index with 62 points, behind Claude Fable 5.2 at 65, and finds coding performance flat versus the previous generation. Astra's API price is $8 per 1M input tokens and $40 per 1M output tokens.
OpenAI's GPT-6.2 Astra and GPT-6.2 Sol are generally available in GitHub Copilot
OpenAI's GPT-6.2 Astra and GPT-6.2 Sol are now generally available in GitHub Copilot, according to the GitHub changelog. GitHub says both models are designed for agentic coding and code review, with stronger coding judgment than earlier GPT-6 models.
Comments
Loading...