Woefully slow is an interesting way to put it. Yes, there are some losses to the iGPU (roughly 10% from early metrics), however, that’s the price we pay for upgradeable RAM.
Soldered RAM was a bad idea that just increased obsolescence and I’m seeing a worrying amount of people vying for it on this forum…
@Shiroudan I see a pattern of people creating an account with a very short username just to post one unique comment about how LPDDR is better than DDR.
Maybe people in the laptop industry afraid of the raising Framework startup and trying to defend their business using troll farm tactics?
I mean it is better than ddr in pretty much every way, you just can’t socket it which makes it unusable for a framework.
I do kinda hope that new memory form-factor dell started takes off, that could work with the lpddr options if I understood that correctly. I am not married to sodimms but I want swappable memory that is made by more than one manufacturer.
@Adrian_Joachim What is the form factor used by Dell? If I understand correctly, the very fact that there is a socket (with connectors) is responsible for the slower speed…
Another road would be to expect that Framework caters to the crowd equipped with the necessary to (re)solder LPDDR, like hot-air guns and special solder pastes… but that would mean they would shrink their market a lot instead of developing it… (not even mentioning soldering accidents, both for the board components and the human components…)
CAMM, it’s not just dell though. As far as it sounds the traces are much shorter and can be closer to the socket which in theory would allow lpddr to be used but it’s still in a mostly draft phase so who knows.
That is a very small crowd, bga soldering isn’t the hardest core stuff there is but it is up there.
Kinda depends on what you take as limiting factor, it works with solwer speeds and I can see pepole using even slower speed for better battery life but faster ram DOES perform better, not by a huge ammount but measurably, especially in the igpu department.
Even at the same speed lpddr is more power efficient which is better in my book.
If there was a way to get swappability (without a rework station) and lpddr, that would be best but if I have to choose between the 2 I choose swappability.
However if CAMM actually works out I’d want that (and you can still use sodimms with an adapter if there is enough space in the case).
Probably nice to have, but it’s very far in my list of priorities.
I’m quite happy with the current DDR system, and before begging for a replacement of this tech I’d like to see many other feature requests (and fixes) addressed.
Supposedly, this video from GPD shows off the performance of the Radeon 780M and 760M iGPUs, using the 680M for comparison. Not 100% sure if accurate, but I found the performance of the 760M particularly interesting.
Worth noticing the computer with the 680M has LPDDR5-6400 memory and the 780M and 760M computers have LPDDR5-7600 memory, so while useful the comparison is not apples to apples.
Unfortunately AMD doesn’t actively engage with us over Linux laptop testing and the major OEMs/ODMs aren’t actively engaging at large with the relatively small Linux niche especially from the PR side.
The Windows benchmarking should still give a relative approximation, you then want to pick from the OEMs who offer that chip in a combination you can accept with Linux support adequate for the bits you want.
Luxury of a computer shop, if you happen to work in one, is trying this stuff in the process of selling it to someone else; you could assemble a build for a client, then on your own time that evening put Linux on it with a spare drive and test, swapping the customer drive back in before sale. Every job has it’s perks.