4 ms·

They were soldering RAM on well before there were performance advantages - actually there were often performance disadvantages as the soldered on RAM was often

They were soldering RAM on well before there were performance advantages - actually there were often performance disadvantages as the soldered on RAM was often a slower speed tier than what you could get cheaply just a few years later. So the "lets solve that before we can swap-able RAM" argument doesn't really fly. There are also Performance advantages to soldering the ram directly to the CPU die. So make the whole CPU with ram on it swappable then?

rfoo 3y ago
The industry does not accept regression. It's okay back then cause what you described is not a regression when they happened - "a few years later" doesn't count, most people don't buy laptop for upgradability. For the last two DDR generations, we have: - During the early days of DDR4: (soldered) LPDDR4-3200 vs (socketed) SO-DIMM DDR4-2400 - During the late days of DDR4: (soldered) LPDDR4-4266 vs (socketed) SO-DIMM DDR4-3200 - During the early days of DDR5 (now): (soldered) LPDDR5-7500 vs (socketed) SO-DIMM DDR5-5600 Good luck explaining "ok uh our latest laptop is slower than our last generation but hey you can change the RAM after a few years and it will be faster!" to ordinary users, then.
- ClumsyPilot 3y ago
  > The industry does not accept regression In that caae I want my microphone jack returned please, and laptops with A-sized USB ports again.
  - blkhawk 3y ago
    that is not regression, that is just brave.
- blkhawk 3y ago
  yes, but how much of that was more that the laptop vendors for those models simply chose better RAM chips while the slower binned part got sold as mainstream modules? What I mean is that your example might just be caused by something else not "we don't know how we can make fast ram on modules"
  - Brian_K_White 3y ago
    None. The problem is the traces and the kind of signal that has to travel on the traces, not the chips.
- opan 3y ago
  It's not just about upgrading, there's also the ability to replace a bad DIMM. Although upgrading is nice for the secondhand market since most laptop configurations are not the most high-end ones, and they're more likely to become e-waste if you can't make them the best they could've been.

rfoo 3y ago

The industry does not accept regression. It's okay back then cause what you described is not a regression when they happened - "a few years later" doesn't count, most people don't buy laptop for upgradability. For the last two DDR generations, we have: - During the early days of DDR4: (soldered) LPDDR4-3200 vs (socketed) SO-DIMM DDR4-2400 - During the late days of DDR4: (soldered) LPDDR4-4266 vs (socketed) SO-DIMM DDR4-3200 - During the early days of DDR5 (now): (soldered) LPDDR5-7500 vs (socketed) SO-DIMM DDR5-5600 Good luck explaining "ok uh our latest laptop is slower than our last generation but hey you can change the RAM after a few years and it will be faster!" to ordinary users, then.

ClumsyPilot 3y ago
> The industry does not accept regression In that caae I want my microphone jack returned please, and laptops with A-sized USB ports again.
- blkhawk 3y ago
  that is not regression, that is just brave.
blkhawk 3y ago
yes, but how much of that was more that the laptop vendors for those models simply chose better RAM chips while the slower binned part got sold as mainstream modules? What I mean is that your example might just be caused by something else not "we don't know how we can make fast ram on modules"
- Brian_K_White 3y ago
  None. The problem is the traces and the kind of signal that has to travel on the traces, not the chips.
opan 3y ago
It's not just about upgrading, there's also the ability to replace a bad DIMM. Although upgrading is nice for the secondhand market since most laptop configurations are not the most high-end ones, and they're more likely to become e-waste if you can't make them the best they could've been.