Apple Unveils M1 Ultra SOC: CPU Faster Than Intel 12900K at 100W Less Power, GPU On Par With NVIDIA RTX 3090 at 200W Less Power

Hassan Mujtaba

Apple has just unveiled its flagship M1 SOC, the M1 Ultra, and provided some impressive metrics in regards to its CPU & GPU performance.

Apple M1 Ultra SOC Offers Faster CPU & GPU Performance Than Flagship Desktop Intel Alder Lake & NVIDIA RTX 3090 Chips While Consuming Lower Power

The Apple M1 Ultra SOC is the combination of two M1 Max chips using the brand new Apple UltraFusion architecture. The Ultra SOC still utilizes the 5nm process node but the doubling provides up to 114 Billion transistors on the same package. The new UltraFusion architecture delivers up to 2.5 TB/s of low-latency interprocessor bandwidth and 800 GB/s of fast memory bandwidth. The interconnect architecture features more than 10,000 signals between the two chips.

Related Story Apple Silently Changed The Design of The M3 Max Chip Which Could Have A Huge Impact on The M3 Ultra Chip

A low-level view of the Apple M1 Ultra SOC reveals a 20-core CPU with 16 high-performance and 4 high-efficiency cores. Like the previous M1 chips, the high-performance cores are based on an ultra wide execution architecture with 192KB instruction cache, 128 KB data cache & 48 MB of L2 cache while the efficiency cores are based on a Wide Execution architecture with 128KB instruction cache, 64KB data cache and 8 MB of L2 cache. There are also 32 cores as a part of the M1 Neural Engine which provide up to 22 trillion operations per second.

2022-03-08_23-30-10
2022-03-08_23-30-17
2022-03-08_23-30-18
2022-03-08_23-30-24
2022-03-08_23-30-33
2022-03-08_23-30-42
2022-03-08_23-29-01
2022-03-08_23-29-29
2022-03-08_23-29-54
2022-03-08_23-30-04
2022-03-08_23-30-29

The GPU also gets a huge bump with a 64-core unit that offers 8192 execution units for up to 21 TFLOPs of single-precision horsepower, 660 GTexels/s, and 330 GPixels/s. The media engine on the Apple M1 Ultra is composed of the latest Hardware-accelerated H.264, HEVC, ProRes & ProRes with 2 video decode engines, 4 videos encode engines and 4 ProRes encode/decode engines. The GPU is getting access to 128 GB of unified memory.

Apple M1 Ultra SOC CPU Performance (vs Intel Alder Lake):

2022-03-08_23-31-22
2022-03-08_23-31-28
fnwsob9vqaehexl

Coming straight to the performance numbers on display, Apple has stated that the M1 Ultra can deliver up to 90% more performance than the flagship Intel Alder Lake Desktop CPU, the Core i9-12900K at the same wattage and similar performance while sipping in 100W less power. It is also stated that the chip offers 90% more performance than an Intel Xeon 16-core chip and 2.5x faster performance than a 10 core Intel Core i9 desktop CPU. The specific models for both chips weren't mentioned.

Apple M1 Ultra SOC GPU Performance (vs NVIDIA/AMD GPUs):

2022-03-08_23-31-37
2022-03-08_23-31-45
fnwsoctvcae1k-d

The same is true for the GPU which offers the same performance as the NVIDIA GeForce RTX 3060 Ti at 1/3rd the power and similar performance to NVIDIA's flagship GeForce RTX 3090 graphics card while sipping in 200W less power. Apple also showed comparative numbers against the Radeon Pro W6900X on their existing Mac Pro solutions. The Mac Studio with M1 Ultra SOC offers up to 80% faster performance than the Radeon Pro solution.

2022-03-08_23-29-50
2022-03-08_23-29-15
2022-03-08_23-29-38

Mac Studio Powered by M1 Ultra Enables:

  • Up to 3.8x faster CPU performance than the fastest 27-inch iMac with the 10-core processor.
  • Up to 90 percent faster CPU performance than Mac Pro with 16-core Xeon processor.
  • Up to 60 percent faster CPU performance than 28-core Mac Pro.
  • Up to 4.5x faster graphics performance than the 27-inch iMac, and up to 80 percent faster than the fastest Mac graphics card available today.
  • Up to 12x faster than the 27-inch iMac, and up to 5.6x faster than 28-core Mac Pro when transcoding video.

Apple M1 Ultra SOC UltraFusion Chip-To-Chip Interconnect:

Apple M1 Ultra SOC Benchmark Leaks Out, Slower Than Intel Alder Lake In Single-Thread But Faster Than 32 Core AMD Threadripper In Multi-Thread
apple-m1-ultra-soc

Footnotes:

  1. Testing was conducted by Apple in February 2022 using preproduction Mac Studio systems with Apple M1 Ultra, 20-core CPU, 64-core GPU, 128GB of RAM, and 8TB SSD, as well as production 2.5GHz 28-core Intel Xeon W-based Mac Pro systems with 384GB of RAM and AMD Radeon Pro W6900X graphics with 32GB of GDDR6, configured with Afterburner and 4TB SSD. Prerelease Compressor 4.6.1 tested using a three-minute clip with 5K Apple ProRes RAW media, at 5760x3240 resolution and 24 frames per second, transcoded to Apple ProRes 422. Performance tests are conducted using specific computer systems and reflect the approximate performance of Mac Studio and Mac Pro.
  2. Testing was conducted by Apple in February 2022 using preproduction Mac Studio systems with Apple M1 Max, 10-core CPU, and 32-core GPU, and preproduction Mac Studio systems with Apple M1 Ultra, 20-core CPU, and 64-core GPU. Performance measured using select industry‑standard benchmarks. 10-core PC desktop CPU performance data tested from Core i5-12600K and DDR5 memory. 16-core PC desktop CPU performance data tested from Core i9-12900K and DDR5 memory. Performance tests are conducted using specific computer systems and reflect the approximate performance of Mac Studio.
  3. Testing was conducted by Apple in February 2022 using preproduction Mac Studio systems with Apple M1 Max, 10-core CPU, and 32-core GPU, and preproduction Mac Studio systems with Apple M1 Ultra, 20-core CPU, and 64-core GPU. Performance was measured using select industry‑standard benchmarks. Popular discrete GPU performance data tested from Core i9-12900K with DDR5 memory and GeForce RTX 3060 Ti. Highest-end discrete GPU performance data tested from Core i9-12900K with DDR5 memory and GeForce RTX 3090. Performance tests are conducted using specific computer systems and reflect the approximate performance of Mac Studio.
  4. Testing was conducted by Apple in February 2022 using preproduction Mac Studio systems with Apple M1 Ultra, 20-core CPU and 64-core GPU, and 128GB of RAM, and configured with 8TB SSD. Prerelease Final Cut Pro 10.6.2 was tested using a one-minute picture-in-picture project with 18 streams of Apple ProRes 422 video at 8192x4320 resolution and 30 frames per second, as well as a one-minute picture-in-picture project with nine streams of Apple ProRes 422 video at 8192x4320 resolution and 30 frames per second. Performance tests are conducted using specific computer systems and reflect the approximate performance of Mac Studio.
  5. Testing was conducted by Apple in February 2022 using preproduction Mac Studio systems with Apple M1 Ultra with 20-core CPU and 64-core GPU. Power was measured using a representative workload in a commercial application. High-end PC desktop data was acquired from testing Alienware Aurora R13 with Core i9-12900KF and GeForce RTX 3090. Performance tests are conducted using specific computer systems and reflect the approximate performance of Mac Studio.
Share this story

Deal of the Day

Comments