*GeekBench* 6.7 has made its move! The moment IBOT is enabled, the result is immediately deemed invalid.

On April 9, Primate Labs officially launched GeekBench 6.7. At first glance, this update looks like a routine version iteration, but in reality it brings one of the most sensitive topics in the hardware world directly into the spotlight: should Intel’s newly introduced IBOT binary optimization be considered a genuine performance improvement, or just a variable in benchmark scores? For users who regularly follow processor reviews, device comparisons, or even use benchmark results as a buying reference, this is not a minor tweak, but a recalibration at the rules level.

The core change in GeekBench 6.7 is very clear: it now includes precise detection of IBOT status. Previously, older versions of GeekBench could not accurately determine whether IBOT was enabled and could only issue a warning. Now the logic is more direct—once the system is found to have this binary optimization technology enabled, the score will be marked as “Invalid.” The system will further indicate that “This result may be invalid because the system is using a binary modification tool.” The message conveyed by this approach is very clear: you are free to optimize, but the resulting score is no longer suitable for public cross-platform or cross-device comparisons.

Why has this triggered such intense discussion? The key is that IBOT itself is not an ordinary driver update. Based on currently available information, Intel introduced this technology alongside the upgraded Core Ultra 200 Plus series. Through lossless low-level binary optimization, it can improve gaming performance by an average of 8%, with gains reaching as high as 22%. From the perspective of actual player experience alone, that kind of boost is certainly appealing, especially for those who want to squeeze out a few more frames without changing hardware—it feels a bit like a free power-up pack. But once it enters the benchmarking scenario, the issue shifts to a different dimension: are you measuring the chip’s native capability, or the result after additional tools have stepped in?

In my view, GeekBench’s decision this time is clearly to side with “comparability.” It is a tough stance strategically, and one that aligns well with its role as a benchmarking tool. Developer Primate Labs has already stated that this move is intended to ensure that test results can be fairly compared across platforms. Put simply, if people on the same ranking list are using low-level binary optimization while others are not, then that list loses its value as a reference for consumers, media, and manufacturers alike. That is also why some netizens jokingly call GeekBench “AppleBench”—but at its core, it is really the same controversy: how should a cross-platform tool define “fairness”?

From a broader trend perspective, this update is not aimed only at Intel. GeekBench 6.7 also improves Android SoC detection accuracy, RISC-V processor recognition, and Linux Arm system stability, while scores from the new version remain directly comparable with versions 6.3 through 6.6. This shows that what Primate Labs wants to do is not simply block a specific company’s technology, but bring the testing environment back to a unified standard. The practical advice for gamers and hardware users is also simple: if you care about “how fast games can run in daily use,” technologies like IBOT are worth paying attention to; if you care about “whose chip is inherently stronger,” then prioritize GeekBench scores that have not been deemed invalid. Benchmark scores have never been the game itself, but they do influence how players spend their money—and that is what truly makes this 6.7 update worth discussing.