Scale Promises Native Nvidia CUDA Support On AMD Hardware

Expert Verified By

CUDA Support Could Be Huge For AMD!

Story Highlight
  • A toolchain developed by Spectral Compute aims to enable CUDA programs to run natively on AMD GPUs.
  • It handles NVPTX Assembly without manual translation and can replace Nvidia’s nvcc compiler.
  • SCALE has shown success on AMD’s RDNA architecture, enhancing CUDA accessibility beyond Nvidia hardware.

Various efforts have been made to assist in converting CUDA source code to portable C++ code for AMD graphics processors. At the same time, ZLUDA was originally financed by AMD to enable CUDA binaries to operate on AMD.

It looks like a new competitor named SCALE is in town. UK-based company Spectral Compute has been working on SCALE for the past seven years using some open-source LLVM components, helping users run CUDA programs natively on AMD GPUs via the GPGPU toolset.

Jensen Huang Nvidia CEO
CUDA Support Gives Nvidia A Huge Advantage Over AMD

SCALE also has a significant advantage over other alternatives that just help with code translation through manual developer processes. CUDA applications that depend on line NVPTX Assembly can be handled by SCALE, which accepts them just as they are.

Furthermore, the SCALE compiler can be used in place of Nvidia’s nvcc compiler, and it features a runtime that “impersonates” the CUDA Toolkit. While compatibility with Vega is still being worked on, Spectral Compute has begun testing SCALE across RDNA3 and RDNA2 GPUs, in addition to basic testing on RDNA1.

Various tests using Blender, Llama-cpp, XGboost, GOMC, Hashcat, and even Nvidia Thrust have proven successful for SCALE. Nvidia’s CUDA will no longer be exclusive and will instead become broadly available with the adoption of SCALE.

It’s crucial to remember that while SCALE can be accessed by users under a free software license, the program itself is not open-source.

While it’s promising to see CUDA running natively on AMD, Nvidia will probably not lose its exclusivity status so easily, as CUDA has been instrumental in the company’s supremacy over the AI industry.

Was our article helpful? 👨‍💻

Thank you! Please share your positive feedback. 🔋

How could we improve this post? Please Help us. 😔

Gear Up For Latest News

Get exclusive gaming & tech news before it drops. Sign up today!

Join Our Community

Still having issues? Join the Tech4Gamers Forum for expert help and community support!

Latest News

Join Our Community

104,000FansLike
32,122FollowersFollow

Trending

Sony Announces Multiple State of Play Broadcasts For September 3rd

Sony has announced multiple State of Play broadcasts for September 3rd, with updates from its development teams and more.

Star Citizen’s Single-Player Campaign Squadron 42 Delayed to 2027

Cloud Imperium Games CEO Chris Roberts has announced that Squadron 42 has been delayed once again to 2027 due to GTA 6's upcoming launch.

Report: GTA 6 Launch To Cause Over A Billion Dollars In Losses In The US On Day One

An analyst claims GTA 6's launch would be such a big event that it could cause over a billion dollars in losses in the US on day one alone.

Fable Offers 15 to 20 Hours Story Length, Much Shorter Than Most RPGs

Playground has confirmed that the main story in...

Valve Suffers Massive 12TB Leak; Multiple Internal Game Builds Revealed

A major Valve leak appears to include old internal game builds, development files, and rare material tied to older games.