Geekbench 7 Forces New Chips to Sweat Bigger Tests

Geekbench 7 looks familiar on purpose, but Primate Labs has made the benchmark harder where it counts: bigger datasets, new audio and video workloads, and a redesigned multi-core test.

Primate Labs is releasing the latest generation of its widely used benchmark for computers, phones, and tablets, according to The Verge. The update keeps the Geekbench 6 look and workflow, but changes what the app asks CPUs and GPUs to do.

That matters because Geekbench is a fixture in device reviews. The Verge notes that its short runtime has made it useful for fast comparisons, but also better at measuring quick bursts than longer sustained loads.

Geekbench 7 keeps the old look, then makes the tests heavier

The surface-level surprise is how little Geekbench 7 appears to change. The app “looks and acts just like Geekbench 6,” per The Verge’s description.

The real change sits under the interface. Geekbench 7 adds new video and audio encoding and decoding tests, plus larger datasets for CPU and GPU workloads. In plain terms: modern chips get more to chew through before they earn a score.

Primate Labs frames the update as a realism push.

Geekbench 7 uses “more demanding data sets to more accurately capture the hardware demands of modern applications.”

The new audio and video tests include AV1, Opus, modeled video conferencing, streaming, and content consumption workloads. AV1 and Opus are especially relevant because they’re tied to modern media playback, compression, and communication tasks rather than older, lighter test cases.

Here’s the before-and-after shift:

• Before, Geekbench 6: Familiar quick benchmark runs, with workloads that could be completed very quickly by modern hardware.
• After, Geekbench 7: Larger datasets, new media tests, and revised multi-core behavior designed to push current CPUs and GPUs harder.

For readers comparing devices, the practical point is simple: Geekbench 7 scores are not compatible with Geekbench 6 scores. A Geekbench 7 result starts a new score lane. Mixing the two would blur the comparison.

Redesigned multi-core testing stops treating every task like it scales perfectly

The most important technical change is the redesigned multi-core benchmark. Geekbench 7 workloads will only run multi-threaded when the modeled tasks behave that way in real software, according to The Verge.

That’s a meaningful correction. A benchmark can make a processor look better if every task is split neatly across many cores, even when the real version of that task wouldn’t scale so cleanly. Geekbench 7 appears to narrow that gap by making the multi-core test less automatic and more conditional.

XOOMAR analysis: This is the right place for Primate Labs to focus. Modern processors are increasingly judged not just by peak single-core speed, but by how intelligently they divide work across cores. If the benchmark assigns threads in ways real apps don’t, the score becomes cleaner than the experience it claims to represent.

The Verge also flags the benchmark’s short duration. Geekbench 7 tests still take only a few minutes, so this update doesn’t turn Geekbench into a sustained thermal torture test. It does make the short test more demanding.

That distinction matters for review readers. A quick score can still be useful, but it shouldn’t be mistaken for a full picture of long-run performance under heat, power limits, or extended GPU load. For laptop buyers parsing spec tradeoffs, that same caution applies beyond benchmarks, as seen in our coverage of Less RAM Stings Framework Laptop 13 Pro Preorder Buyers.

Geekbench 7 Pro keeps the $99 license and offline benchmark option

Geekbench 7 Pro keeps the same $99 license price as before. Primate Labs is also offering a launch discount to $79 until August 6th, according to The Verge.

The Pro license still allows offline benchmarking. That means runs don’t have to be uploaded to the Geekbench Browser, the public results database tied to the benchmark.

That offline option is the practical split between casual use and controlled testing. Public uploads are useful for broad comparison, but private runs are useful when someone needs to benchmark locally without posting results through the browser.

The Pro license continues to cover compatible desktop platforms: Windows, macOS, and Linux. The broader Geekbench tool also has a cross-platform role across desktop and mobile systems, including phones and tablets.

For mobile readers, that matters because Geekbench often travels with OS testing and device comparisons. If you’re experimenting with phone software, our guide to Download iOS 27 Free Today Without Wrecking Your iPhone is a separate reminder that benchmark numbers can shift when software changes too.

The score reset is the first real test for Geekbench 7

The immediate breakage is comparability. Geekbench 7 results can’t be stacked directly against Geekbench 6 results, so anyone reading charts needs to check the version before drawing conclusions.

That’s not a flaw. It’s the cost of changing the workload. If Primate Labs wants larger datasets, new media tasks, and multi-core behavior that tracks modeled software more closely, the old scoring scale has to be left behind.

XOOMAR analysis: The next useful signal won’t be one isolated Geekbench 7 score. It’ll be the pattern that forms once enough devices have been tested under the new benchmark. Early results may look noisy simply because the comparison base is new.

The watch item is whether the tougher workloads make Geekbench feel more representative without losing the speed that made it popular. If Geekbench 7 stays quick while asking harder questions of CPUs and GPUs, it could become a cleaner snapshot. If not, reviewers will keep treating it as one fast datapoint among many.

Key Takeaways

• Geekbench 7 may produce more realistic scores for modern phones, tablets, and computers.
• The heavier workloads could make device reviews better reflect current media and multitasking demands.
• Scores from Geekbench 7 should not be directly compared with Geekbench 6 results.