Basemark introduced its new GPUScore Sacred Path cross-platform benchmark that lets users comparison wholly antithetic GPU architectures moving connected antithetic operating systems. Our GPU benchmarks hierarchy is constricted to Windows 11 testing, but what if you privation to cognize however the best graphics cards stack up against your smartphone? Now you can.
To springiness a fewer examples, Sacred Path shows however Apple's A16 GPU stacks up against Steam Deck's GPU and Samsung's Galaxy S22+ GPU arsenic good arsenic however show of Nvidia's GeForce RTX 3090 relates to that of Intel's Arc A770 and Apple's M1 Max. Of course, with immoderate benchmark that has to enactment crossed a wide dispersed of hardware capabilities, determination are caveats. But let's spot what the numbers say.
Cross-Platform, Cross-Device Benchmark
There are rather a fewer large hardware graphics platforms for PCs and mobile devices. These see AMD's integrated and discrete GPUs, Apple's PowerVR-derived integrated graphics processors, Arm's Mali graphics solutions, Imagination's PowerVR-based graphics solutions, Intel's standalone Arc Alchemist and integrated Xe-LP GPUs, Nvidia's discrete graphics processors, Qualcomm's Adreno graphics for mobile applications, Samsung RDNA 2-based built-in GPU cores, and a fig of different platforms.
Formally, each of these GPUs are designed to alteration high-performance gaming, though high-end mobile and high-end PC are evidently rather antithetic targets. All the architectures person their ain pros and cons, and successful galore cases they tally nether antithetic operating systems utilizing antithetic APIs (application programming interfaces). That makes it precise hard to find however 1 GPU stacks up against another, since it is astir intolerable to bash an apples-to-apples — oregon Apples-to-desktops — comparison.
Basemark's cross-platform Sacred Path is designed to wide distant immoderate of the confusion, arsenic it performs precisely the aforesaid workload connected each the assorted GPUs (assuming the trial is tally utilizing the authoritative autochthonal solution of 2560x1440). To marque the benchmark much applicable for antithetic devices, it supports assorted show optimization techniques, including adaptable complaint shading, precocious customized shaders, Ground Truth Ambient Occlusion (GTAO), and temporal anti-aliasing.
Repeatable, but With Fluctuations
After playing with the Sacred Path benchmark for a mates of hours connected respective Apple devices, we recovered that portion successful wide the programme produces accordant and repeatable results that correlate with the GPU capabilities of the device, astatine times its results fluctuate importantly — we're talking ~50% oregon much here. Perhaps Apple's mobile platforms accommodate to workloads to supply amended performance, but since smartphones and tablets are not successful our wheelhouse, we'll permission due probe to idiosyncratic else.
Basemark's PowerBoard 4.0 database already has capable submissions astatine the authoritative render solution of 2560x2440 to let comparisons among galore antithetic platforms. To supply a amended perspective, we compared the highest results each graphics processor achieved and did not instrumentality the CPU oregon API into account.
Unfortunately, since these results tin fluctuate rather a bit, we'll person to use due levels of brackish to the PowerBoard 4.0 submissions. Also support successful caput that the results from smartphones, notebooks, and tablets whitethorn alteration greatly owed to antithetic cooling and ambient temperatures — these are not each tally nether trial laboratory conditions by immoderate agelong of the imagination.
High-End Gaming GPUs
Basemark's Sacred Path benchmark doesn't look peculiarly astonishing compared to modern games, but it inactive needs a batch of resources to tally astatine decent speeds. Sacred Path's people represents the mean FPS multiplied by 100 (e.g, 1840 points means 18.4 FPS). Fortunately, adjacent existing GPUs from AMD and Nvidia tin ace this benchmark beauteous easily, delivering 200–280 FPS. That besides means they're perchance not getting anyplace adjacent to 100% GPU utilization, which again is the occupation with trying to make a "universal" GPU benchmark.
GPU | Score | Device | API | OS |
GeForce RTX 3090 | 28,270 | Desktop, AMD Ryzen 9 5900X | DirectX 12 | Windows 11 Pro |
GeForce RTX 3080 Ti | 25,050 | Desktop, AMD Ryzen 7950X | Vulkan 1.3 | Windows 11 Pro |
Radeon RX 6900 XT | 24,244 | Desktop, Intel Core i5-12600KF | Vulkan 1.3 | Windows 10 Pro |
GeForce RTX 3080 10GB | 22,489 | Desktop, AMD Ryzen 9 5900X | Vulkan 1.3 | Windows 10 Pro |
Radeon RX 6800 XT | 20,569 | Desktop, AMD Ryzen 5800X | Vulkan 1.3 | Windows 10 Pro |
Judging by results demonstrated by high-end gaming graphics cards, we tin reason that Basemark's Sacred Path loves earthy compute performance. The much FP32 teraflops you have, the amended your results volition be. AMD's Radeon RX 6900 XT is precise adjacent to the GeForce RTX 3080 Ti, but the GeForce RTX 3090 is unrivalled.
Sadly, determination are nary RTX 4090 submissions conscionable yet. Jarred ran it but it failed to taxable (perhaps the strategy presently has a whitelist of known GPUs and the 4090 isn't connected it). Manual (custom) investigating utilizing the authoritative settings gave a effect of 48,709 implicit the ~4 2nd run, which astatine slightest seems to beryllium successful the close ballpark. At the aforesaid time, a trial that completes successful seconds means that the GPU isn't adjacent remotely warmed up earlier the trial is over. (Where's my brackish shaker?)
Mid-Range and High-End Mobile Gaming GPUs
GPU | Score | Device | API | OS |
GeForce RTX 3070 Ti | 17980 | Desktop PC, Core i7-12700F | Vulkan 1.3 | Windows 11 Pro |
GeForce RTX 3070 | 15967 | Desktop, Core i7-11700K | Vulkan 1.3 | Windows 11 Pro |
Radeon RX 6700 XT | 15514 | Desktop, Ryzen 7 3700X | Vulkan 1.3 | Windows 10 Enterprise |
GeForce RTX 3060 Ti | 14010 | Desktop, Core i3-10105F | Vulkan 1.3 | Windows 11 Pro |
GeForce RTX 3080 Laptop | 12277 | Acer Nitro AN515-45, Ryzen 9 5900HX | Vulkan 1.3 | Windows 11 Home |
GeForce RTX 3070 Laptop | 12524 | ASUS ROG Strix G533QR, Ryzen 9 5900HX | DirectX 12 | Windows 11 Home |
Midrange and high-end mobile gaming graphics processors don't person arsenic overmuch oomph arsenic high-end desktop cards. Nonetheless, they each — AMD's Navi 22, Intel's ACM-G10, and Nvidia's GA104 — bash rather well. Intel's Arc A770 is 11% down Nvidia's GeForce RTX 3060 Ti, arsenic different constituent of reference.
Meanwhile, assuming that Apple's M1 Ultra scales its graphics show truly good (the M1 Ultra consists of 2 M1 Max SoCs connected together), we would expect its show to beryllium connected par with that of Nvidia's GeForce RTX 3070, which is not atrocious for an integrated graphics processor.
Entry-Level GPUs
GPU | Score | Device | API | OS |
Apple M1 Max | 8939 | MacBook Pro 18.4 | Metal | MacOS 13.0 |
GeForce RTX 3050 | 8117 | Desktop PC, Ryzen 5 5600X | Vulkan 1.3 | Windows 11 Pro |
GeForce RTX 3050 Laptop | 6110 | MechRevo Z3 Air Series GM5TGEO, Intel Core i7-11800H | Vulkan 1.3 | Windows 10 Home |
Intel Arc A380 | 4801 | Desktop PC, Core i5-10400 | DirectX 12 | Windows 11 Pro |
Apple M1 Pro | 4698 | MacBook Pro 14 2021 | Metal | MacOS 12.6 |
Owners of entry-level GPUs person not submitted galore results truthful far, oregon astatine slightest they're not showing up. Fundamentally, the much teraflops a GPU has, the amended it performs successful Sacred Path.
Apple's M1 Max (an iGPU) and Nvidia's GA106 successful the signifier of desktop GeForce RTX 3050 pb the battalion here. Meanwhile, Intel's standalone Arc A380 tin hardly support up with Apple's M1 Pro integrated GPU. Even Nvidia's GeForce RTX 3050 for laptops is importantly faster than the Intel Arc A380.
Handhelds, Ultra-Portables, Integrated
GPU | Score | Device | API | OS |
Apple M2 | 3124 | MacBook Air 2022 | Metal | MacOS 13 |
Apple M1 | 2477 | Mac Mini 2020 | Metal | MacOS 12.5.1 |
Apple M1 | 2423 | iPad Pro 11 2021 | Metal | iPadOS 15.6.1 |
Apple A16 | 1837 | iPhone 14 Pro | Metal | iOS 16.0.3 |
AMD Van Gogh | 1828 | Steam Deck | Vulkan 1.2 | SteamOS |
AMD Ryzen 9 5900HX | 1744 | Acer Nitro AN515-45 | Vulkan 1.3 | Windows 11 |
Apple A15 | 1595 | iPhone 13 Pro Max | Metal | iOS 16.0.2 |
Qualcomm Adreno 730 (Snapdragon 8 Gen 1 — SM8450) | 1365 | Oppo Find X5 Pro | Vulkan 1.1 | Android 12 |
Samsung Xclipse 920 | 1265 | Galaxy S22+ | Vulkan 1.1 | Android 12 |
Arm Mali-G710 MC10 (MediaTek Dimensity 9000) | 899 | Xiaomi Redmi K50 Pro | Vulkan 1.1 | Android 12 |
When it comes to cross-platform benchmarking, handhelds, ultra-portables, and iGPUs are arguably the astir absorbing categories arsenic they includes aggregate hardware and bundle architectures arsenic good arsenic antithetic form-factors.
Since Apple uses its M1 and M2 system-on-chips for its tablets, it's tenable to see them into this examination adjacent though they tin powerfulness full-fledged PCs similar MacBook Air/MacBook Pro. After all, AMD's Ryzen SoCs with higher-end GPUs are besides designed for this precise purpose. Meanwhile, Apple's M1 and M2 iGPUs are faster than AMD's Cezanne iGPU by 40%–80%.
As for SoCs for handheld devices, Apple's A16 tin bushed adjacent the RDNA 2-based AMD Van Gogh iGPU utilized successful the Steam Deck crippled console, but the iPhone 14 Pro needs to tally the trial for a mates of times to deed thing similar 2200 points. Qualcomm's Adreno 730 utilized successful the Snapdragon 8 Gen 1 (SM8450) is simply a bully performer, but it's inactive down Apple's previous-generation A15 SoC. Meanwhile, RDNA 2-based Samsung Xclipse 920 looks somewhat disappointing erstwhile compared to rivals.
Summary
Basemark's Sacred Path benchmark tin so beryllium utilized to trial each kinds of graphics processors, from humble GPUs aimed astatine smartphones and each the mode to monstrous graphics cards that tin beryllium utilized not conscionable for games but besides arsenic abstraction heaters oregon AI supercomputers.
The bundle produces accordant and repeatable results astir of the time, and it heavy depends connected the compute show of modern GPUs. While that mightiness beryllium utile connected immoderate level arsenic a show metric, bash support successful caput that a 1.15GB benchmark that tin implicit successful seconds connected the fastest GPUs volition ne'er adequately amusement the differences betwixt GPUs that mightiness person anyplace from 1GB to 24GB (or more) VRAM. In this test, earthy teraflops wins the game, adjacent though existent crippled titles besides beryllium connected graphics representation capacity, representation bandwidth, and different factors.
The benchmark besides doesn't usage ray tracing successful a bid to enactment connected each graphics processors from the past decennary oregon more, truthful it cannot show each the benefits that leading-edge GPUs for PCs support. There are besides immoderate fluctuation-related hiccups with immoderate platforms, but that's capable for this look astatine the vagaries of cross-platform "universal" GPU testing.