ioc.exchange is one of the many independent Mastodon servers you can use to participate in the fediverse.
INDICATORS OF COMPROMISE (IOC) InfoSec Community within the Fediverse. Newbies, experts, gurus - Everyone is Welcome! Instance is supposed to be fast and secure.

Administered by:

Server stats:

1.3K
active users

@niconiconi I'm not sure I believe iFixit's claims; do you?

@whitequark@mastodon.social Of course everything is possible if you play the correct link budget game. If you're willing to use 2.4 mm connectors you can run mmWave through sockets. My read is that I don't believe anyone's going to spend the resource to reengineer a mobile platform (Strix Halo) originally speced without socked memory in mind.

@niconiconi isn't this a bad argument? you could just as well say that nobody is going to engineer smartphones for repairability because it's not profitable to do so

@whitequark@mastodon.social Given the facts that Framework picked up AMD's reference design of an existing platform (Strix Halo) that happens to be the only non-HEDT, non-server x86 platform with the fastest memory. Also given that this platform was not designed without socketed memory in mind, asd since reference designs are usually adopted in whole, major technical changes are rarely (if ever) made - The ODM likely doesn't have the technical capabilities required, and the vendor is unlikely to support such as redesign effort, in spite of the possibility in principle. It's more or less a take-it-and-leave-it situation identical to the status of Apple's M-series ARM desktops and workstations. How should it be called then, "vendor lock-in based on one-of-a-kind design"?

Then what's the alternatives?

1. Should Framework get their own SI engineering team capable of making independent platform-level changes? This would be the best, but it's quite a rarity in the industry, for perhaps 90% of the chipusers, so I don't think it's reasonable to blame the ODM for a platform problem.

2. Or, should Framework take the blame of making a product based on a platform that goes against its philosophy just as matter of principle? But since this platform is a genuinely useful accelerator for things such as my numerical simulations, and that the the alternative of GPUs have worst memory lockdowns, it merely kicks the same problem to other ODMs.

3. Or, as end users, to ignore the platform at all? Get a real workstation or server. Just wait until the regular desktops catch up within 10 years?

So what's the reasonable conclusion here? "Just blame AMD instead, just as how people blame Apple for the same non-upgradability" is the only conclusion I can think of, in additional to "pick your poison".

Andrew Zonenberg

@niconiconi @whitequark I mean, I would love to see Framework hire some actually skilled high speed digital engineers and acquire the lab equipment to do SI workups at DDR5/6/whatever speeds.

It's not going to be cheap though.

@azonenberg @niconiconi yeah, this. can they afford it today? maybe not. should we push them to do that, based on their own principles? absolutely

@whitequark @niconiconi Honestly as someone who actually does from-scratch board designs, it amazes me how much of the EE world is just copying reference designs with tiny changes and not fully understanding anything about what's going on.

Like, I don't want to follow "match traces to within X mils" or something nonsensical like that. I feel much better actually knowing setup/hold timing of the interface, calculating the timing budget, knowing the velocity factor of my stackup, and deriving matching rules based on that.