There’s reasons behind this. LPDDR IIRC works most efficiently when it’s closer to the CPU than what dimms would allow for.
Boosts speed and lowers the power requirements.
It also incentivizes people to buy larger SKUs than they originally wanted, which, bluntly, is probably the main driver for going that direction… I’m just saying that there’s technical reasons too
The technical benefits are honestly quite overblown. The M-series didn’t get the massive speed lift because it moved to soldered RAM near the CPU, it got the massive speed lift because it doesn’t have to copy stuff between the CPU and GPU, the proximity to the CPU is a pretty modest improvement. So they could’ve gotten 95% of the benefit while still offering socketed RAM, but they decided not to, probably to drive prices up.
There’s actually an argument that makes the point of driving prices down with soldered RAM.
The individual memory chips and constituent components are cheaper than they would be for the same in a DIMM. We’re talking about a very small difference, and bluntly, OEMs are going to mark it up significantly enough that the end consumer won’t see a reduction for this (but OEMs will see additional profits).
So by making it into unupgradable ewaste, they make an extra buck or two per unit, with the added benefit of our being unupgradable ewaste, so you throw it out and buy a whole new system sooner.
This harkens back to my rant on thin and light phones, where the main point is that they’re racing to the bottom. Same thing here. For thin and light mobile systems, soldered RAM still saves precious space and weight, allowing for it to be thinner and lighter (again, by a very small margin)… That’s the only market segment I kind of understand the practice. For everything else, DIMMs (or the upcoming LPCAMM2)… IMO, I’d rather sacrifice any speed benefit to have the ability to upgrade the RAM.
The one that ticks me off is the underpowered thin/lights that are basically unusable ewaste because they have the equivalent of a Celeron, and barely enough RAM to run the OS they’re designed for. Everything is soldered, and they’re cheap, so people on a tight budget are screwed into buying them.
This is actually a big reason why I’m hoping that the windows-on-ARM thing takes off a bit, because those systems would be far more useful than the budget x86 chips we’ve seen, and far less expensive than anything from Intel or AMD that’s designed for mobile use. People on a tight budget can get a cheap system that’s actually not ewaste.
I don’t think the manufacturing cost is the driver there, the forced upgrade is. The argument isn’t for saving $1-2 per unit or whatever, but forcing customers to pay $100-200 more for that memory upgrade they’re not sure they need, but get because they can’t upgrade later.
thin and light
Yup, and that’s a big reason why I don’t buy those. Saving a little space and size is nice, but not at the cost of upgradability. In fact, that’s why I bought an E-series instead of the more expensive T-series, the E-series didn’t have soldered RAM.
There’s reasons behind this. LPDDR IIRC works most efficiently when it’s closer to the CPU than what dimms would allow for.
Boosts speed and lowers the power requirements.
It also incentivizes people to buy larger SKUs than they originally wanted, which, bluntly, is probably the main driver for going that direction… I’m just saying that there’s technical reasons too
The technical benefits are honestly quite overblown. The M-series didn’t get the massive speed lift because it moved to soldered RAM near the CPU, it got the massive speed lift because it doesn’t have to copy stuff between the CPU and GPU, the proximity to the CPU is a pretty modest improvement. So they could’ve gotten 95% of the benefit while still offering socketed RAM, but they decided not to, probably to drive prices up.
There’s actually an argument that makes the point of driving prices down with soldered RAM.
The individual memory chips and constituent components are cheaper than they would be for the same in a DIMM. We’re talking about a very small difference, and bluntly, OEMs are going to mark it up significantly enough that the end consumer won’t see a reduction for this (but OEMs will see additional profits).
So by making it into unupgradable ewaste, they make an extra buck or two per unit, with the added benefit of our being unupgradable ewaste, so you throw it out and buy a whole new system sooner.
This harkens back to my rant on thin and light phones, where the main point is that they’re racing to the bottom. Same thing here. For thin and light mobile systems, soldered RAM still saves precious space and weight, allowing for it to be thinner and lighter (again, by a very small margin)… That’s the only market segment I kind of understand the practice. For everything else, DIMMs (or the upcoming LPCAMM2)… IMO, I’d rather sacrifice any speed benefit to have the ability to upgrade the RAM.
The one that ticks me off is the underpowered thin/lights that are basically unusable ewaste because they have the equivalent of a Celeron, and barely enough RAM to run the OS they’re designed for. Everything is soldered, and they’re cheap, so people on a tight budget are screwed into buying them. This is actually a big reason why I’m hoping that the windows-on-ARM thing takes off a bit, because those systems would be far more useful than the budget x86 chips we’ve seen, and far less expensive than anything from Intel or AMD that’s designed for mobile use. People on a tight budget can get a cheap system that’s actually not ewaste.
I don’t think the manufacturing cost is the driver there, the forced upgrade is. The argument isn’t for saving $1-2 per unit or whatever, but forcing customers to pay $100-200 more for that memory upgrade they’re not sure they need, but get because they can’t upgrade later.
Yup, and that’s a big reason why I don’t buy those. Saving a little space and size is nice, but not at the cost of upgradability. In fact, that’s why I bought an E-series instead of the more expensive T-series, the E-series didn’t have soldered RAM.
The closest thing to a thin and light that I own is my framework.
I felt like anything less than what framework offers for repairability wouldn’t be sufficient for me.