![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
If this offered higher gains then why are systems with SIMM socket faster? It's unlikely this will offer the same bandwidth as Apple does, which would be great for the GPU. It does offer lower power consumption because it sits closer to the SoC. Like any great disaster it's always just cheaper for them to manufacture, plus the added bonus to charge a lot more for what would be cheap ram. Granted these chips are meant to be cheap, with a base clock speed of 2.1GHz. If the entire Lunar Lake offering is like this, I don't see a reason not to go AMD for a laptop chip. Especially if AMD adopts CAMM2.
Often have bigger power envelope ?, not sure if SIMM socket system are faster that the big GPU too, all being equal how could SoC memory not be faster exactly (being that it will be closer) ?
SIMM systems are usually accompanied by a discrete GPU with it's own memory. Ram like DDR5 are faster because it has lower latency, because CPU's mostly care about latency over bandwidth. Every iteration of DDR is an exchange of latency over bandwidth, even though DDR5 kept the latency the same as DDR4. To compensate this we see CPU's with more cache which helps mitigate the higher latency of DDR. LPDDR has higher latency because it uses less voltage, but mitigates latency by being really close to the CPU. Plus in Apple's case it uses higher memory bus width like 256-bit on their M1 series, which is meant to feed the GPU. DDR5 isn't all that bad with 64 GB/s vs Apple's M3 base 100 GB/s. According to AMD, dual-channel DDR5-6000 is 96 GB/s. DDR5 uses 1.1V vs LPDDR5's 1.05V or even 0.5V. The main purpose of LPDDR is Lower Power, hence the LP part of the name. It also happens to be cheaper than putting SIMM sockets, plus you can charge more for ram.