Apple M2 Ultra GPU Outpaces RTX 4070 Ti in Early Compute Benchmarks
Apple M2 Ultra GPU benchmarks have leaked online.Today we spotted sharing various social media posts Geekbench GPU Compute Score and GFXBench Aztec Ruins Score. On the surface, these results seem to indicate that the new M2 Ultra’s GPU is very powerful, offering GPU computing performance between the GeForce RTX 4070 Ti and RTX 4080. But I have to step back and explain exactly what it is. we are watching
We reported on the announcement of the new Apple M2 Ultra SoC at the company’s annual WWDC earlier this week. The chip comes in an updated Mac Studio and Mac Pro design and is expected to ship next week. However, with the large amount of benchmarks published online these days, it seems that someone is testing these new Mac computers/processors before consumers get their hands on them.
Yesterday, we handpicked the Apple M2 Ultra benchmarks, focusing on the performance of the SoC’s CPU cores. Today we’ll take a closer look at the potential computing power of the M2 Ultra’s GPU thanks to the Geekbench and GFXbench leaks mentioned at the beginning.
M2 Ultra Geekbench 6 Compute Benchmark
A few pre-release M2 Ultra Apple Mac system users appear to be running Geekbench 6 Metal and OpenCL GPU benchmarks. Apple’s Metal API is a proprietary graphics API developed by the company for fast “direct-to-the-metal” hardware addressing, similar to Apple’s optimizations alongside Microsoft’s DirectX.
These are very interesting synthetic computing benchmarks, but soon there will be thorough real-world testing of the new Apple Macs with M2 Ultra processors, especially in the gaming sector, which has been a thorn in Apple’s chips. I’m looking forward to it.
Note that GPU computing tends to scale much better with multi-chip approaches than GPU graphics. Remember Ethereum mining back then? When you connect 8 (or more) GPUs to a single modest CPU via PCIe x1 connections, they are all basically running at 100% of maximum performance.Not all compute workloads scale or Well, it’s still very different from the scaling traditionally seen in real-time graphics used in games.
The new M2 Ultra chip’s GPU features up to 76 integrated graphics cores, which looks impressive when compared to the previous generation M1 Ultra which had up to 64 graphics cores. In a head-to-head comparison using Geekbench 6, the new generation M2 Ultra scores around 220,000, while the average M1 Ultra scores around 155,000.
According to this metric, the M2 Ultra is over 40% faster than its predecessor in GPU computing tasks. The number of graphics cores is only about 10% more, so Apple has done well to achieve an improvement of this magnitude.
Turning to another comparison, the new Apple M2 Ultra’s 220,000 Geekbench 6 Compute Score (Metal) sits between the GeForce RTX 4070 Ti (208,340 OpenCL) and the RTX 4080 (245,706 OpenCL). In a direct Geekbench 6 OpenCL comparison, the Apple M2 Ultra Open CL score of about 155,000 is much closer to PC GPUs such as the Nvidia RTX A5000 and AMD Radeon RX 6800 XT.
Again, it should be noted that Geekbench 6 OpenCL’s compute performance typically does not translate well to real-world gaming workloads. Synthetic Geekbench 6 performance is very different from real world graphics.
M2 Ultra GFXBench Benchmark
Another user ran an Apple Mac with an M2 Ultra through the GFXBench test suite. In a 4K Aztec ruins off-screen test, the new Apple SoC Over 55% faster than its predecessor. This seems to be an even better result than the Geekbench 6 GPU compute comparison. In theory, since this is a graphics workload, not a compute one, it should more accurately represent 3D acceleration and improved gaming performance.
However, GFXBench is not an ideal desktop GPU comparison tool as it usually cannot keep up with the highest end graphics hardware. The Aztec Ruins benchmark is better suited for smartphone SoCs than GPUs like the RTX 4070 Ti. Additionally, the focus should be on GFXBench’s “off-screen” tests, which evaluate the GPU’s rendering capabilities without worrying about the display resolution in use. Given that the entire benchmark suite is less than 1GB in size, it’s definitely not very representative of modern games that total more than 100GB.
The screenshot above shows that the Apple M2 Ultra can achieve almost 315 FPS in the 4K Aztec Ruins high-rise off-screen test (Metal API). Above we said that this score is over 55% faster than the M1 Ultra. Comparable scores on the PC platform would still be between GeForce RTX 4070 Ti and GeForce RTX 4070 Ti. RTX4080 (DX12 API).
Here you need a tablespoon of salt. Ideally, you should be looking at the performance of games and other graphics and computing applications rather than synthetic workloads that may not represent anything really useful. We’re looking forward to intensive real-world testing of the new Apple Macs with M2 Ultra processors due to be released next week.