The strength of Apple Silicon comes from its processing capabilities, not primarily its memory.
The strength of Apple Silicon comes from its processing capabilities, not primarily its memory.
Bandwidth details: 14900k to 90 GB/s on M3. That’s the key spec? Apple must be leveraging LPDDR5 technology effectively.
Only because the CPU's memory controller is capable of handling a certain speed doesn't guarantee the presence of RAM that can actually use it.
400 billion bytes equals about 400 gigabytes, which is roughly four times the standard amount. Storage is integrated directly onto the chip, and being close together helps minimize delays and noise, enabling faster performance. Of course, several other elements influence the chip's overall speed, but unified memory could be the key factor.
Let's break this down clearly. When comparing performance numbers, make sure you're using the same units.
- 14900k vs 90 billion bytes per second: roughly 5x faster.
- M3 Max at 400 billion bytes per second is about 4x faster.
- Memory integration with SoC package allows for high speed and precise timing.
It's similar to how soldered memory in GPUs delivers consistent performance.
The details listed on that page refer to the standard setup. You can adjust faster memory settings. The storage capacity comes with the package, and the emphasis on high memory bandwidth mainly supports the GPU, not the CPU. The CPU does gain from it, but this is likely well beyond the point of diminishing returns.
Sure, I've updated it. Pro specs are now used instead of Max. Pro should still be 150GB/s – see the Wikipedia page for Apple M3 variants.
I'm clearly just playing around with GB to bytes conversion—_- . I was wondering if that's the main reason behind Apple's speed? As YougBlade mentioned again, it might relate to the GPU, but it definitely helps the CPU too. What do you mean by diminishing returns? Programmers give up their wild ideas to create algorithms that work within limits instead of memory constraints. The speed of today's processors is still far beyond what memory can offer, and this has been consistent since the beginning. Keep watching!
Modern GPUs feature extremely fast memory buses. A 4070 delivers around 500mB/s, while a m3 max offers substantial memory bandwidth largely due to the GPU. When comparing a CPU paired with a GPU, you'll notice comparable bandwidth levels.