In addition to its revamped CPU subsystem, Apple's M5 Max also boasts a new GPU that is based on a PowerVR-derived microarchitecture developed by Apple. As it turns out, a big integrated GPU and plenty of memory bandwidth can deliver serious GPU compute oomph: the M5 Max scores 232,718 points on the GeekBench 6 GPU compute benchmark when using the Metal API. Apple's previous-generation M4 Max scores up to 204,453 points in the same tests. Evidently, the new GPU is better than the predecessor, but not that significantly.
When compared to non-Apple GPUs, the one inside the M5 Max easily beats the iGPU inside the Ryzen AI Max+ 395, which scores 133,447 points when unconstrained by thermals. When it comes to discrete graphics cards, Apple's flagship iGPU is ahead of Nvidia's GeForce RTX 5070 ( 207,061 points , Vulkan ), but trails the GeForce RTX 5070 Ti ( 253,890 points, Vulkan) and has no chance against the GeForce RTX 5090. Still, building an integrated GPU that delivers compute performance comparable to one of the best graphics cards is a breakthrough.
Follow Tom's Hardware on Google News , or add us as a preferred source , to get our latest news, analysis, & reviews in your feeds.
Anton Shilov is a contributing writer at Tom\u2019s Hardware. Over the past couple of decades, he has covered everything from CPUs and GPUs to supercomputers and from modern process technologies and latest fab tools to high-tech industry trends. ","collapsible":{"enabled":true,"maxHeight":250,"readMoreText":"Read more","readLessText":"Read less"}}), "https://slice.vanilla.futurecdn.net/13-4-18/js/authorBio.js"); } else { console.error('%c FTE ','background: #9306F9; color: #ffffff','no lazy slice hydration function available'); } Anton Shilov Social Links Navigation Contributing Writer Anton Shilov is a contributing writer at Tom’s Hardware. Over the past couple of decades, he has covered everything from CPUs and GPUs to supercomputers and from modern process technologies and latest fab tools to high-tech industry trends.
Gururu Why is there a comparison between a laptop integrated GPU and a 5090? Its faster than 8060s and panther lake (probably), leave it at that and call it the fastest CPU in the world. Reply
andrewzortiz I decided to create this account to just leave this comment. This is the most click bait article I've ever seen from Tom's Hardware. Please do better. Reply
TheyStoppedit There's something we aren't being told. If a 6,000 dollar laptop with an M5 Max can slaughter a $15,000 desktop CPU that has 5x the core count, and 3x the power draw, then there has to be something we aren't being told. I would like to know what that is…. because that doesn't just happen. Is AMD really 10 years behind apple like that? I somehow doubt that. I'd like to know all the details. Reply
Pierce2623 Geekbench heavily favors ARM and it doesn’t actually scale hardly at all with real workloads. If you look at the tests listed in the article, it’s very obviously aimed at testing cellphones. Reply
moon2 It's hidden in the article – geek bench doesn't scale well beyond 32 cores. Presumably that means comparisons beyond 32 core are effectively junk. Maybe they could run four copies of geekbench concurrently and see which CPU as actually scales best – or just run workloads appropriate for a 96 core machine Reply
abufrejoval andrewzortiz said: I decided to create this account to just leave this comment. This is the most click bait article I've ever seen from Tom's Hardware. Please do better. I wouldn't agree, because it's both obviously clickbait and expected performance. Real clickbait plays with the really unexpected and is subtle enough to lead you where you'd never go normally. Could they do better? Zero doubt! Too bad that doesn't pay, but then I block their ads. Reply
PEnns Amazing!!! I am so amazed, I am in amazement heaven….. Now do tell us how that Apple chip bulldozed a comparable IBM chip…or is that reserved only for AMD and Intel is a no-no in such scenarios?? Reply
abufrejoval Having grown up with Lego, I can't help but like the underlying principle and I'd love to see it also available from other vendors for commodity pricing and perhaps more options for the parts (Lego has more than just the 8-notch block). Unified memory and flexible expandability unfortunately don't seem to have economical solutions, let's hope that Apple approach inspires some progress and options elsewhere. In the mean-time I'm just happy that my currently biggest box (RTX 4090, Zen 9 7950X3D, 96GB ECC DDR5-5600, 16TB SSD) still does better for much less anywhere where I need performance, so I'm not even tempted to pay for the potential of the unified address space. I'd still be interested to see how the Max fares against EPYCs with 12-channel DDR5-6400, which should offer similar bandwidth at 600GB/s, although only the M5 iGPU can reach there, while the EPYC CPU cores may still lack the GPGPU ops M5 optimized workloads might be able to take advantage of. E.g. even 96 CPUs may not be be able to beat an M5Max GPU at low-precsion LLM inference, but that's not very likely the principal workload these systems would run. And then CPUs are extending their ISAs towards including huge vectors of low-precision data types to retain or regain bigger chunks of that market. Mostly I'd say that 12 channel external DRAM and on-die DRAM stacks are meeting at a bandwidth point with prices that reflect the technical effort, while the optimum between those two, not entirely exclusive approaches is very application dependent. CPUs with extra HMC and HBM stacks have been made and hybrids remain possible. Hyperscalers can deploy bespoke variants once the scale is big enough, those who want a box on the desk or at their feet, have to chose and pay up front and bet on it paying off over potentially a long time. Reply
thisisaname Gururu said: Why is there a comparison between a laptop integrated GPU and a 5090? Its faster than 8060s and panther lake (probably), leave it at that and call it the fastest CPU in the world. Aye most things are not 5090, if it is that fast that a comparison between it and the 5090 is valid comparison then they have made something rather special. Reply
Key considerations
- Investor positioning can change fast
- Volatility remains possible near catalysts
- Macro rates and liquidity can dominate flows
Reference reading
- https://www.tomshardware.com/pc-components/cpus/SPONSORED_LINK_URL
- https://www.tomshardware.com/pc-components/cpus/apples-18-core-m5-max-destroys-96-core-ryzen-threadripper-pro-9995wx-in-geekbench-gpu-performance-is-much-less-impressive#main
- https://www.tomshardware.com
- Microsoft confirms next-gen Xbox will play PC games — 'Project Helix' teased as more than just a console
- Leading Inference Providers Cut AI Costs by up to 10x With Open Source Models on NVIDIA Blackwell
- From Radiology to Drug Discovery, Survey Reveals AI Is Delivering Clear Return on Investment in Healthcare
- Spiralling memory spot prices could trigger 'industry cycle collapse,' report warns — NAND wafer costs surge 25% in a single month
- Apple's 18-core M5 Max destroys 96-core Ryzen Threadripper Pro 9995WX in Geekbench — GPU performance is much less impressive
Informational only. No financial advice. Do your own research.