5 ms·
Note that Cinebench R23 (the main CPU benchmark we use) runs for 10 minutes, so it’s a measure of sustained performance. Boost (PL2) is typically 30 seconds or
by nrp 4y ago
Note that Cinebench R23 (the main CPU benchmark we use) runs for 10 minutes, so it’s a measure of sustained performance. Boost (PL2) is typically 30 seconds or less with a long period after that of staying at the PL1 limit.
- senttoschool 4y agoAlso note that Cinebench R23 is a terrible general purpose CPU benchmark. It uses Intel Embree engine which is hand optimized for x86. It heavily favors CPUs with many slow cores even though most people will benefit from CPUs with fewer, faster cores. Cinebench is a great benchmark if you use Cinema4D, which I asumme 99.99% of the people buying these laptops won't use. Cinema4D is a niche of a niche. Geekbench is far more representative of what kind of performance you can expect from a CPU. https://reddit.com/r/hardware/comments/pitid6/eli5_why_does_it_seem_like_cinebench_is_now_the/ https://reddit.com/r/hardware/comments/pitid6/eli5_why_does_...
- senttoschool 4y agoTo clarify, Cinebench correlates poorly with gaming, office software, web browsing, and video editing. Those are what the vast majority of people buying laptops will use it for. For people that code, it also correlates poorly with parallel code compilation.
- 2OEH8eoCRo0 4y agoAnandtech benchmark puts Intel 12900K on top for both RISCV toolchain compilation and Cinebench.
- senttoschool 4y ago12900K is the fastest x86 DIY CPU available. It'll be tops in nearly all benchmarks and applications. You can say the same about gaming. It's tops in Cinebench and most games. Therefore Cinebench and gaming must have a high correlation? No.
- nabakin 4y agoThanks for your input, this isn't said enough. So many CPU benchmarks aren't effective at evaluating the general use case and yet are held up as this golden standard
- senttoschool 4y agoIt's because Cinebench was disproportionately favoring the Zen architecture so AMD pushed it in marketing.
- kllrnohj 4y ago> Geekbench is far more representative of what kind of performance you can expect from a CPU. Geekbench is an even worse benchmark than cinebench. It's extremely heavily influenced by the underlying OS rather than the hardware being compared. Cinebench 1T and mT at least actually benchmark the hardware.
- sgerenser 4y agoBenchmark the hardware doing 3D rendering. Which is a pretty niche use case for most people that doesn’t correlate well with more common cpu-intensive tasks like gaming or video editing.
- kllrnohj 4y agoVideo editing is just as niche as 3d rendering. But really neither are actually that niche, see for example https://www.blender.org/news/blender-by-the-numbers-2020/ https://www.blender.org/news/blender-by-the-numbers-2020/ And to put those blender numbers in comparison https://prodesigntools.com/number-of-creative-cloud-subscribers.html https://prodesigntools.com/number-of-creative-cloud-subscrib...
- saagarjha 4y ago> It's extremely heavily influenced by the underlying OS rather than the hardware being compared. How so?
- SpacePortKnight 4y agoI agree that 12th gen can sustain good performance given decent cooling. However, the issue is still heat. It could be fun to have a `lap benchmark`. That is can you keep a laptop on your lap while it runs Cinebench R23 for 10 minutes? With TB disabled, I can.