Which is BETTER? ... M4 Max OR M3 Ultra ... Mac Studio Test

Comparison of Mac Studio M3 Ultra vs. M4 Max

Overview of the Devices

• The video compares two models of Mac Studio: one with the M3 Ultra and another with the M4 Max, focusing on performance, fan noise, heat, and power consumption.

• The entry-level M3 Ultra features 28 CPU cores and 60 GPU cores; a full version has additional cores at a higher price point ($1,500).

• The recommended configuration for most users is the full M4 Max with 16 CPU cores and 40 GPU cores, offering a balance of performance and value.

Performance Insights

• In single-core performance tests like Geekbench 6, the M4 Max shows a significant uplift (25%) over the M3 Ultra.

• Real-world applications may not show noticeable differences due to the already competent performance of the M3 in many tasks.

• Multi-core performance comparisons can be misleading; Cinebench results indicate that while both chips perform well, the M3 Ultra outperforms in certain scenarios.

Power Consumption and Heat Management

• During Cinebench testing, power draw was measured: M4 Max averaged about 95W while the M3 Ultra drew around 128W under load.

• Despite peak power capabilities being higher for the Ultra chip, it often consumes less power during low-load situations compared to the M4 Max.

• Both machines maintained similar temperatures under light loads (41.4°C), but during stress tests, they showed slight variations in temperature rise.

Cooling Solutions and Noise Levels

• The cooling solutions differ: aluminum heat sink for M4 Max versus copper for M3 Ultra; however, both managed efficient cooling without audible fan noise during testing.

GPU Performance Comparison

• In GPU benchmarks like Geekbench 6 Open CL test, results show that while both GPUs are capable, the older generation (M3 Ultra) performed slightly better in some areas (11% more).

• When using ray tracing capabilities in Cinebench tests for rendering tasks, results favored the newer architecture (M4 Max), though overall graphics performance still leaned towards the older model depending on specific workloads.

Creative Workflows Analysis

• For creative workflows involving mixed CPU/GPU tasks such as photo editing or graphic design software like Affinity Photo 2:

• Overall scores suggest that while there are advantages to each model based on specific tasks (single core vs multi-core), real-world differences may be minimal between them.

Exporting Photos: Performance Comparison

Photo Export Times

• The M4 Max finishes exporting 12 edited 40-megapixel Fujifilm RAW files to JPEG in 39 seconds, while the M3 Ultra takes 42 seconds.

• For professional photographers dealing with hundreds of images, even a small time difference can accumulate significantly.

AI Performance Analysis

Neural Engine Comparison

• The M4 generation neural engine shows significant improvement over the M3, but claims of performance may be exaggerated due to different precision levels used for TOPS figures (M3: 18 trillion vs. M4: 38 trillion).

• The Geekbench AI test does not fully utilize the additional cores of the M3 Ultra, leading to misleading performance comparisons.

Real World Application Insights

• In single precision tests, the M4 Max is about 3.5% more performant than the M3 Ultra; however, this is negligible for real-world applications where GPU usage is preferred.

• At half precision, the performance gap widens to 20%, indicating that while theoretically capable, the M3 Ultra's extra cores are underutilized in practice.

Large Language Models (LLMs) Testing

Model Performance on Mac Machines

• Both machines were tested using the Gemma 3 model with 12 billion parameters, revealing no substantial performance differences between them for this specific task.

• The advantage of allocating unified memory primarily benefits users running multiple models or larger models simultaneously on machines like the M3 Ultra with up to 512 GB of memory.

Video Editing Benchmarking

Da Vinci Resolve Testing

• Puget Bench for Da Vinci Resolve evaluates various video resolutions and codecs, providing insights into CPU, GPU, and media engine performance across both machines.

• Detailed results show that while generally outperforming in most tests, the M4 Max excels in long GOP process codecs and spatial noise reduction tasks due to its architecture advantages.

Overall Performance Summary

• The basic score shows that the M4 Max scores 12,390, whereas the M3 Ultra leads at 14,066, reflecting a 14% uplift for the latter overall across various tests.

Comparison of Video Editing Machines

Performance Analysis of Resolve Studio vs. Premiere Pro

• The speaker compares Resolve Studio and Premiere Pro, noting that both are optimized for specific chips, with Resolve traditionally performing better.

• It is suggested to seek various opinions from different reviewers on YouTube to make an informed decision regarding video editing software.

Rendering Times and Machine Comparison

• A test was conducted using a short timeline in Resolve with various video codecs, effects, and color grades; the M4 Max rendered in 38 seconds while the M3 Ultra took 28 seconds.

• For larger projects, the difference in rendering times can significantly impact productivity; thus, the M3 Ultra is recommended for those prioritizing speed.

Cost-Benefit Analysis of M3 Ultra vs. M4 Max

• While the M3 Ultra offers advantages like Thunderbolt 5 ports for enhanced connectivity (up to eight displays), its overall performance does not justify its higher cost in many scenarios.

• The recommendation leans towards the M4 Max with 64GB unified memory and 1TB storage as it provides excellent value for most users.

Conclusion and User Engagement