←back to thread

317 points diwank | 1 comments | | HN request time: 0.216s | source
Show context
throwaway48476 ◴[] No.41850185[source]
It is interesting that one of their examples is a "community repair fair", they want to market a sheen of social responsibility without actually taking part themselves.
replies(6): >>41850328 #>>41850406 #>>41850430 #>>41853373 #>>41856540 #>>41857113 #
giancarlostoro ◴[] No.41850430[source]
Lowkey wish their laptops would be as they used to be. Being able to swap RAM or hard drives is so basic but so useful.
replies(4): >>41852852 #>>41853334 #>>41853505 #>>41858086 #
talldayo ◴[] No.41852852[source]
Drives in particular. Let them solder the memory if they absolutely have to, but exposing even an empty NVMe slot should be standard for laptops. Unfortunately, Apple makes a pretty penny off the storage surcharge so I wouldn't really anticipate that anytime soon.
replies(3): >>41853139 #>>41853202 #>>41853315 #
loopdoend ◴[] No.41853139[source]
I will accept the trade off for the performance boost tbh.
replies(1): >>41853581 #
lucb1e ◴[] No.41853581[source]
What performance boost? As in, same software running for comparison on the hardware of interest, one soldered and the other not. I never heard that soldering your SSD on makes it faster...
replies(2): >>41853731 #>>41856273 #
loopdoend ◴[] No.41856273[source]
Sorry I was referring to the boost you get from having ram integrated into the chip vis-à-vis apple’s M-line of processors.

Having replaceable ram is not really a marketable feature these days.

replies(2): >>41858557 #>>41882737 #
angoragoats ◴[] No.41858557[source]
This might be nitpicking but 1) the ram is not integrated into the chip, per se, it's still discrete and soldered on a PCB right next to the CPU, and 2) the increased speed comes from additional memory channels built into the M-series CPUs, not necessarily the fact that the memory is closer to the CPU.

It is true that it's not currently feasible to have socketed memory in a laptop offering 8+ memory channels to enable 200GB/sec+ bandwidth, but you can absolutely get the same (or greater!) memory speeds as an M-series CPU from an x86 desktop workstation.

If Intel/AMD wanted to prioritize memory bandwidth, they could probably work with JEDEC or another industry org to develop a new standard for socketed memory with multiple channels per socket, to enable the kind of speeds that Apple offers. The fact that they haven't (to my knowledge) indicates to me that they don't see it as a big enough priority or benefit.

replies(1): >>41882764 #
lucb1e ◴[] No.41882764[source]
> 1) the ram is not integrated into the chip, per se, it's still discrete and soldered on a PCB right next to the CPU, and 2) the increased speed comes from additional memory channels built into the M-series CPUs

Thanks, I did not know this! I would have honestly have bought into Apple's marketing that the soldering is what allows them to make it more integrated and faster

replies(1): >>41882974 #
1. angoragoats ◴[] No.41882974[source]
You're welcome! In fairness to Apple, having shorter traces between the CPU and main memory does in fact decrease latency and power requirements. It's just not the only, or even the best, way to get more performance out of memory chips.

The CAMM2/LPCAMM2 standard is a new way of having replaceable memory which takes up less physical space and is faster, if you're interested. There are a few laptops (and desktops) out there using it already. It still only supports dual-channel memory, though.

As I said originally, my suspicion is that "200GB/s+ memory bandwidth!" might be good marketing copy and make for good synthetic benchmark results, but just isn't actually that beneficial for the average computer user in the real world. This could be why you don't see other computer manufacturers pursuing it, at least not in laptops.