Geekbench 7 is out, and the biggest change is not the length of the test list. It is how the scores are produced. Primate Labs released Geekbench 7 on July 23 for Android, iOS, Windows, macOS, and Linux, rebuilding the multi-core benchmark, adding media and machine-learning workloads, and moving the scoring baseline to an AMD Ryzen 7 7700 desktop. The practical result: the numbers you see on spec sheets and review videos are about to look different, and they cannot be compared with Geekbench 6 results.
What is new in the Geekbench 7 CPU benchmark
Geekbench is a cross-platform benchmark - a test app that gives a device a score so you can compare processors across phones, tablets, laptops, and desktops. Version 7 adds a set of media workloads that model what people actually do on a video call or while watching content:
- Video Encoder - encodes screen-sharing video with the AV1 codec (a modern video compression format), modelling the screen-share feature in conferencing apps.
- Audio Encoder - compresses music and spoken word with the Opus codec, modelling voice memo and podcast apps.
- Video Decoder - decodes video and audio while generating live captions using the Whisper speech recognition model, modelling playback with automatic subtitles switched on.
On top of that, there is a new Game Physics workload built on the Jolt physics engine used in shipping games, an expanded Photo Editor test, and a Photo Library workload that now imports and processes modern image formats such as JPEG XL and DNG. Primate Labs also enlarged the data sets behind the File Compression and PDF Viewer tests so they push harder on the hardware.
Why the multi-core score works differently now
This is the change that will move numbers the most. In older versions, every workload was run across multiple CPU cores. Geekbench 7 only runs a workload in multi-threaded mode if the real application it models is genuinely multi-threaded.
The HTML5 Browser test is the clearest example: it is excluded from the multi-core suite because web browsers in practice are single-threaded or only lightly threaded. In Primate Labs' words, "Not every task in the real world is multi-threaded, and pretending otherwise distorts scores without telling you anything useful about your device."
The GPU benchmark leans into machine learning, and CUDA is back
The GPU side has been refocused on machine learning and content creation. New workloads track faces and apply real-time filter effects to video, upscale images with machine learning, and blur backgrounds in video-conferencing streams. There are also new image editing and synthesis tests: RAW image processing, LUT-based video colour grading, path tracing, and fluid simulation.
For desktop PC owners, the notable addition is CUDA. NVIDIA's CUDA API now sits alongside OpenCL, Vulkan, and Metal, so an NVIDIA GPU can be measured using the interface most of its demanding software actually uses.
Geekbench 7 vs Geekbench 6: what early testing shows
Geekbench 7 is recalibrated against a new baseline of 2,500 points, now pegged to an AMD Ryzen 7 7700 desktop instead of the Intel Core i7-12700 used by Geekbench 6. Combined with the heavier data sets, that means the same hardware scores lower on single-core than before.
The analyst firm Signal65 ran both versions back to back on four current laptops. Signal65's testing used Windows Balanced power mode (or the macOS equivalent) on AC power, averaged over three runs:
| Laptop and chip | Single-core change | Multi-core change |
|---|
| Lenovo Yoga Slim 7X (Snapdragon X2 Elite) | about -15% | about +16% (19,764 to 22,871) |
| Apple MacBook Air 15-inch (M5) | about -14% | about +2% |
| HP OmniBook Ultra 14-inch (Intel Core Ultra X9 388H) | about -10% | about -2% |
| HP OmniBook 5 14-inch (AMD Ryzen AI 9 465) | about -9% | about +4% |
The pattern is consistent with the redesign: single-core falls across the board, while multi-core moves in different directions depending on how many cores a chip has and how the new threading rules treat them. The 12-core Snapdragon part gains the most; Intel's chip is the only one that slips.
Price, platforms, and the launch discount
Geekbench 7 is free for personal use, and Primate Labs says it will stay that way. It is available now on Android, iOS, Windows, macOS, and Linux.
The paid Geekbench 7 Pro licence, aimed at professional and commercial use, is listed at $99 (around ₱6,100). Primate Labs is running a launch promotion of 20% off until August 6, which brings it to roughly $79 (around ₱4,900). MacRumors reported the same launch pricing and the same set of workload changes on release day.
What this means if you shop by benchmark scores in the Philippines
Geekbench numbers are everywhere in Philippine phone and laptop buying - in review videos, in Lazada and Shopee listings, and in the spec arguments people have before upgrading. Three practical points follow from this update:
- Do not mix versions. A Geekbench 7 score can only be compared with another Geekbench 7 score. If a listing quotes a big number without saying which version produced it, treat it as unusable.
- Expect lower single-core figures. A phone or laptop that looked strong under Geekbench 6 has not become slower. The scale moved.
- The app itself costs nothing here. Geekbench 7 is a free download on Android and iOS, so you can measure a unit yourself before or after buying. Only the Pro licence is paid, and it is priced in US dollars with no separate Philippine SRP.
If you are cross-shopping recent releases such as the OnePlus Pad 4 in the Philippines, or comparing laptops in a market where Acer has led Philippine sales for two decades, it is worth waiting for a fresh batch of Geekbench 7 results rather than reading across from old figures.
