The Outpost is a comprehensive collection of curated artificial intelligence software tools that cater to the needs of small business owners, bloggers, artists, musicians, entrepreneurs, marketers, writers, and researchers.
© 2024 TheOutpost.AI All rights reserved
Curated by THEOUTPOST
On July 17, 2024
3 Sources
[1]
This toolkit just upended Nvidia's dominance over pro GPUs | Digital Trends
Nvidia is the undisputed leader in professional GPU applications, and that doesn't come down solely to making the best graphics cards. A big piece of the puzzle is Nvidia's CUDA platform, which is the bedrock for everything from Blender to various AI applications. The new Scale tool, developed by Spectral Compute, aims to break down the walled garden. Although we've seen competitors to the CUDA software stack, such as AMD ROCm, Scale is a "drop-in replacement" for CUDA. It's a compiler that allows CUDA applications to be natively compiled on AMD GPUs. Spectral Compute says Scale accepts CUDA programs as is, without the need to port to another language. In Spectral's own words, "... existing build tools and scripts just work." Recommended Videos The key with Scale is that it's a compiler. It's not a translation or porting tool. We've seen open-source translation tools like ZLUDA that run CUDA applications on ROCm, and although they shouldn't require additional development resources, they're not perfect. AMD also funded ZLUDA, in particular, but reportedly backed out of the project. Get your weekly teardown of the tech behind PC gaming ReSpec Subscribe Check your inbox! Privacy Policy Scale isn't open source, and it isn't funded by AMD. It's available under a free license and comprises three main components. There's a Nvidia CUDA Compiler (NVCC)-compatible compiler that builds code for AMD GPUs, an implementation of the CUDA runtime, and an open-source wrapper for CUDA-X calls. The free version is available for commercial and private use, but it explicitly removes any liability from Spectral Compute. That might end up being important, as Nvidia specifically says reverse-engineering CUDA for use on non-Nvidia platforms is against its license agreement. At the moment, Spectral Compute says Scale should work without issues on AMD's RDNA 2 and RDNA 3 GPUs. It also conducted basic testing on RDNA 1 GPUs, and it says it's working on adding support for older architectures. The tool is focused on AMD at the moment, though it's possible Intel GPUs could see support in the future as well. Although anyone using a CUDA application extensively probably has an Nvidia GPU, breaking down the walls of support is a major step toward a more open software ecosystem. It's not clear how long Scale will stick around, though. Spectral Compute says Scale doesn't require CUDA, but it's possible Nvidia could explore routes to remove the toolkit.
[2]
NVIDIA CUDA Can Now Directly Run On AMD GPUs Using The "SCALE" Toolkit
British startup Spectral Compute has unveiled "SCALE," a GPGPU toolchain that allows NVIDIA's CUDA to function seamlessly on AMD's GPUs. Well, it looks like the industry has been able to break NVIDIA's software stack dominance, so they are now looking for ways to remove the "exclusivity" status through various means. We previously saw the emergence of ZLUDA, an open-source porting project that allowed CUDA libraries to work with AMD's ROCm, ultimately supporting Team Red's GPUs. A new competitor has emerged on the scene, the SCALE, which allows AMD's consumers to leverage the capabilities of NVIDIA's CUDA to create a high-end "hybrid" model. We believe that it should be possible to write code once, and build/run it on any hardware platform. This has been a reality for CPU code for many years, so why not GPUs? We set out to directly solve this problem by bridging the compatibility gap between the popular CUDA programming language and other hardware vendors. One codebase, multiple vendors. - Spectral Compute's CEO, Michael Sondergaard Spectral Compute's CEO, Michael Sondergaard, believes that GPUs should have an open-source environment, similar to modern-day CPUs, and that interconnectivity should exist among various platforms. He believes that SCALE acts as a bridge for the compatibility gap between CUDA and other hardware vendors, ultimately removing the limits that exist in the markets. According to Michael, SCALE is a GPGPU toolkit, similar to NVIDIA's CUDA toolkit, which uses binaries for non-NVIDIA GPUs while compiling CUDA code, ultimately removing the need for a translation layer. SCALE has been in development for seven years, according to Spectral Compute, and it doesn't rely on NVIDIA's code but builds its CUDA-compatible toolchain, which makes SCALE highly adaptable amongst multiple platforms, such as AMD's RDNA GPUs. The resource avoids code porting and allows developers to work with a single version of their codebase since SCALE eliminates the need to work with other languages, as it's source-compatible with CUDA. Well, with the implementation of SCALE, it's apparent that the status of NVIDIA's CUDA will change from being exclusive to relatively widely available. However, it's important to note that SCALE itself isn't open-source; users can access it through a free software license. Spectral Compute says that they have tested the software in multiple applications, such as Blender, Llama-cpp, XGboost, FAISS, GOMC, STDGPU, Hashcat, and NVIDIA Thrust, employing AMD's RDNA 3 and RDNA 2 architectures. NVIDIA has shown some resentment against certain resources that allow CUDA to run on external components, given that Team Green previously listed a warning in their EULA against platforms like SCALE. CUDA has played a huge role in NVIDIA's dominance over the AI markets, and the firm won't let the software lose its exclusivity status easily.
[3]
New SCALE tool enables CUDA applications to run on AMD GPUs
Spectral Compute has introduced SCALE, a new toolchain that allows CUDA programs to run directly on AMD GPUs without modifications to the code, reports Phoronix. SCALE can automatically compile existing CUDA code for AMD GPUs, which greatly simplifies transition of software originally developed for Nvidia hardware to other platforms without breaking any end user license agreements. Spectral's SCALE is a toolkit, akin to Nvidia's CUDA Toolkit, designed to generate binaries for non-Nvidia GPUs when compiling CUDA code. It strives for source compatibility with CUDA, including support for unique implementations like inline PTX as, and nvcc's C++ implementation, though it can generate code compatible with AMD's ROCm 6. One of SCALE's significant advantages is its ability to act as a drop-in replacement for Nvidia's own nvcc compiler. Therefore, unlike other projects that translate CUDA code to another language or use other manual steps, SCALE directly compiles CUDA sources for AMD GPUs. SCALE's implementation leverages some open-source LLVM components to create a solution that is both efficient and user-friendly as the software package aims to offer a more seamless and integrated solution that ZLUDA, which is a translation layer that is prohibited to use. It even mimics the Nvidia CUDA Toolkit runtime, making it easier for developers to port their existing CUDA programs to AMD hardware. SCALE has undergone extensive testing with a variety of software, including Blender, Llama-cpp, XGboost, FAISS, GOMC, STDGPU, Hashcat, and Nvidia Thrust, and has proven that it works stably and correctly. Testing has been conducted on RDNA 2 and RDNA 3 GPUs, with basic testing on RDNA 1 and ongoing development for Vega support. The developers did not have access to AMD's CDNA-based GPUs though. The lack of support for CDNA-based processors is a disadvantage of SCALE because datacenter software designed using CUDA and for CUDA-compatible hardware dominates the rapidly growing AI space and many developers are interested in easily porting their programs to competing platforms, expanding their addressable market. Funding for SCALE has been provided by Spectral Compute's consulting business since 2017, without financial backing from AMD. Although the program is not open source, there is a Free Edition License available and this one can be used for commercial applications.
Share
Share
Copy Link
A groundbreaking development in GPU computing allows NVIDIA's CUDA applications to run on AMD GPUs using the SCALE toolkit, potentially reshaping the landscape of high-performance computing.
In a significant development for the world of high-performance computing, a new tool called SCALE (Scalable Compute Acceleration Library Ecosystem) has emerged, enabling NVIDIA's CUDA applications to run on AMD GPUs 1. This breakthrough has the potential to reshape the competitive landscape in the GPU market and offer more flexibility to developers and researchers.
CUDA (Compute Unified Device Architecture) is NVIDIA's parallel computing platform and programming model, which has been a cornerstone of scientific computing, machine learning, and other high-performance applications. Until now, CUDA applications were exclusive to NVIDIA GPUs, creating a significant barrier for those using AMD hardware 2.
SCALE, developed by a team led by Dr. Jiannan Tian from Texas A&M University, acts as a translation layer between CUDA and AMD's ROCm (Radeon Open Compute) platform. It intercepts CUDA API calls and redirects them to their ROCm equivalents, allowing CUDA applications to run on AMD GPUs without modification to the original source code 3.
Initial tests have shown promising results, with SCALE achieving up to 90% of the performance of native CUDA on NVIDIA GPUs for some applications. However, it's important to note that not all CUDA features are currently supported, and performance may vary depending on the specific application and GPU model 1.
The introduction of SCALE could have far-reaching consequences for the GPU industry. It potentially levels the playing field between NVIDIA and AMD, allowing the latter to compete more effectively in markets where CUDA dominance has been a significant factor. This development may lead to increased competition and innovation in the high-performance computing sector 2.
While SCALE represents a significant step forward, there are still challenges to overcome. The toolkit is in its early stages and requires further development to support a broader range of CUDA features and optimize performance across different applications. Additionally, NVIDIA may respond to this development with new strategies to maintain its market position 3.
For developers and researchers, SCALE offers the potential for greater hardware flexibility and reduced dependency on a single GPU vendor. This could lead to more cost-effective solutions and broader access to high-performance computing resources across different hardware platforms 1.
Reference
[3]
ZLUDA, an open-source CUDA translation layer, has been revived with anonymous funding. It now focuses on enabling AI/ML workloads across multiple GPU vendors, potentially breaking down exclusivity barriers in AI software stacks.
2 Sources
AMD announces plans to merge its RDNA and CDNA GPU architectures into a unified UDNA platform, aiming to compete with NVIDIA's CUDA in both consumer and enterprise markets.
6 Sources
NVIDIA, a leading GPU manufacturer, has made a significant move by open-sourcing some of its GPU drivers for Linux. This decision marks a shift in the company's approach to software development and distribution.
2 Sources
AMD unveils new AI-focused hardware including the Instinct MI325X accelerator and EPYC processors, aiming to compete with Nvidia in the growing AI chip market.
10 Sources
AMD announces its latest AI GPU accelerators, the Instinct MI325X and MI355X, aiming to compete with Nvidia's offerings in the rapidly growing AI chip market.
5 Sources