Most of the higher idle in dGPUs are due to the dedicated VRAM. Everything that is responsible for transferring data and traffic needs to have a high-on state, so it has to be the last one that can be put to idle.
| Device | Median score | Number of benchmarks |
|---|---|---|
| Apple M5 | 277.02 | 19 |
| Intel Core Ultra 9 285H | 229.82 | 37 |
| Apple M4 | 221.71 | 176 |
| AMD Ryzen AI 7 350 w/ Radeon 860M | 185.67 | 24 |
| Snapdragon X Plus (8 core) @ 3.30 GHz | 178.18 | 1 |
| Snapdragon X Plus — X1P42100 — Qualcomm Oryon CPU | 177.93 | 4 |
Not only is it misleading but there is almost no world where the owners of these SoCs will use the CPU to render instead of the GPU. GPU results would be far more relevant here. Apple Silicon pulls even further ahead of Intel, AMD, Qualcomm if we use GPU rendering.
Goes the same way. Qualcomm's CPUs aren't as bad in the official benchmark which suggests that Blender 4.0 and above have more Arm Windows optimizations.
Classroom is just a scene to render, you can use it with whatever blender version you want. You need the tester to say which blender version they used but I highly doubt it was that old.
As Hitman said it's just a scene - nothing about it is new or old, it's just what might be classed as a standardised workload by which you can judge the performance of any given hardware, assuming a pixel accurate 1:1 implementation with no bugs or hacks for each backend.
I was referring to the fact that graph is more clear and notes what blender version is used.
Blender 4.2 is not good. Use 4.3 and higher. 4.3 is when Arm got experimental native support on Windows. 4.4 and 4.5 had further CPU optimizations.
browser.geekbench.com
browser.geekbench.com
I like USB A. USB C over the long term are actually not that reliable. Apple has it better with the Lightning connector.
Yea this is not early 2000's where every tech upgrade was almost necessary. 90% of my devices are USB-A, and will continue to be so. Also from the reliability perspective, USB-C power is a downgrade. The negotiation required means more parts required, means more failures, and harder to repair on the circuitry side. If it was instead a dedicated power port, like a 20V laptop power jack, it will always be at 20V and only need filters to use that thing.
Yes, but there should always be justification in having systems that you can upgrade, which gets you the trifecta of benefits: lower long term cost for you, REAL environmental benefits, and better serviceability.
It’s a mini PC. It idles at 0.5 watts for the whole system. If it was a tower PC, it would make sense to have proper upgradability.
I'm getting 0.9-1.1W on my Pentium G6400, GTX 1080 desktop through Kill-a-watt. Margin of error, that's about equal to that Apple. That's why I am thinking of buying a power pack so I can sleep this system without fear of data loss when power goes out.
Facebook
Twitter