Apple’s M3 Max Outpaces RTX 4090 in OpenAI’s Whisper Speech Transcribing Test, But Only with Unoptimized Benchmark
Apple’s M3 Max, the powerful 3nm SoC designed for Mac, has shown exceptional performance in OpenAI’s Whisper speech transcribing test, surpassing even the fastest consumer graphics processor, the RTX 4090 – but there’s a catch. The M3 Max’s triumph is achieved only under specific test conditions using an unoptimized benchmark.
The M3 Max boasts impressive features, including higher multi-core performance compared to the M2 Ultra and a 40-core GPU that can rival the performance of a laptop RTX 4080. However, when it comes to the top-tier RTX 4090, there is still a gap to be bridged. Despite this, the latest speech transcribing test reveals that the M3 Max has the potential to outpace NVIDIA’s renowned graphics card, but only in certain circumstances.
Apple recently unveiled the MLX framework, an open-source solution specifically tailored for their M-series chipsets. To demonstrate the capabilities of this new framework, developer Oliver Wehrens conducted several tests using various Macs and an RTX 4090 GPU. Employing OpenAI’s speech recognition model, Whisper, Wehrens conducted a transcribing test to measure the speed of different chips.
In the initial test results, the M2 Ultra and M3 Max surprisingly outperformed the desktop RTX 4090. However, this outcome stemmed from the utilization of an unoptimized tool for benchmarking, as indicated by the graph displaying the results. Wehrens also conducted tests using an optimized tool, and the outcome aligned more closely with expectations. The RTX 4090 left both the M2 Ultra and M3 Max, along with other SoCs, far behind.
In the subsequent refresh test, the RTX 4090 flawlessly completed the audio transcription in just 8 seconds, while the M2 Ultra required 95 seconds and the M3 Max took slightly longer at 100 seconds. These updated results underline that Apple’s M-series chip lineup still has a considerable journey ahead in terms of raw performance. One area where Apple’s custom silicon shines, however, is power efficiency, consuming notably less energy compared to NVIDIA’s flagship consumer-grade graphics card.
Wehrens appended a disclaimer to his test results, emphasizing that they should not be regarded as scientific measurements. Nonetheless, the data offers valuable insights into the impressive capabilities of Apple’s MLX framework. With the framework specifically designed for Apple Silicon, a MacBook Pro equipped with the M1 Pro chip can achieve RTX 4090 level performance, provided no NVIDIA optimized tools are utilized for testing.
As the tech world eagerly observes the rivalry between Apple’s M-series chips and NVIDIA’s graphics processors, these test results shed light on the potential prowess of Apple’s MLX framework. While the M3 Max exhibits promise, it is worth noting that optimization remains a crucial factor in extracting its full potential. As Apple continues to refine its chipsets, a future where their silicon can truly rival the performance of top-tier graphics cards seems well within reach.
In conclusion, the M3 Max impressively outperformed the RTX 4090 in OpenAI’s Whisper speech transcribing test, albeit under unoptimized benchmarking conditions. Apple’s MLX framework demonstrates its strength as a tailor-made solution for the company’s chipsets, enhancing the performance of Mac devices. While there is still ground to cover to reach the raw performance level of NVIDIA’s flagship graphics card, Apple’s focus on power efficiency sets their custom silicon apart. As technology enthusiasts eagerly await future advancements, Apple’s M-series chips may potentially evolve into fierce competitors to the behemoths dominating the graphics processing landscape.
