Would you be interested in NVIDIA N1 / N1X mainboards?
If so, which are you most interested in?
Note: This poll is simply measuring interest; I’m a consumer who is not in any position to directly influence future products.
Desktop with N1X and high TDP would be a ML dev monster machine.
A FW16 with N1X would likely never happen… Strix-Halo wasn’t a practical fit for the '16 because of its discrete CPU/GPU modular architecture, N1/N1X is the same.
A FW13P with N1X on the other hand… could be a true Mac killer.
I train big ML models for a living, and while the cluster we have at work is the go-to for training, a powerful laptop to do quick experiments when around is a paradise. Even if the 5070 in the FW16 works for small stuff, having those 128GB of unified memory makes me drool.
I wish the Desktop was more impressive at AI workloads. It’s decent, but so are a lot of computers these days.
Other than the N1X being nVidia (which, admittedly, is a big boost, you shoulda paid more attention to your software stack AMD)… other than being nVidia, does it bring much to the table? It’s also 256bit LPDDR5X, so real close to the Desktop’s RAM bandwidth. So prefill in half the time, then generate tokens slightly faster than the Desktop?
Gotta admit, personally I’m hoping for more.
It doesn’t make sense to have a separate item for 13 and 13 Pro. They both have compatible motherboards.
Your poll is missing the option for “No interest”, therefore it is 100% biased in one direction.
A thread for this but with the intel G3/G3 ex processors would be interesting.
A lean/fast processor like the handheld processors could be pretty awesome for a framework 12 mainboard option. Yes the G3 series is nominally for handhelds, but a gaming handheld processor designed for a 25-45 tdp would make for a killer mini-laptop option since something that can run well in a tiny tablet with high performance would be a cool chip for a elite but tiny laptop.
An interesting and still useful conversation piece for when it has the power to interface with a high performance monitor just to confound people who think small = weak.
The wildcat lake chips are likely inevitable for the smaller frameworks, but a powerful variant of the 12 is hypothetically an option with how quickly dedicated handheld chips are gaining power relative to regular laptop mobile chipsets, and that could rock once prices stabilize.
Fair point. I don’t see a way to update the options, but I suspect those not interested are reflected in the amount of (or lack thereof) interaction and by the content within comments.
In all honesty, I think it would be nice to have one of these in a FW12.
Right, a powerful sleeper configuration or three or the 12 would be kinda fun because why else are we making processors whose power draw is nil and whose compute speed is the speed of sound in a tungsten cube if it isn’t to have this tiny little thing with the horsepower of an old prestige pc.
Those are already out as the dgx spark (taped to a really fancy nic).
It’s probably not even beating pantherlake in a fw13 power envelope while very likely costing quite a bit more than even m5 pro max ultra turbo super configs.
Those are just somewhat differently configured pantherlake chips like the one the 13 is already getting, just like the z1 extreme was pretty much just a confused 7840u.
I mean chiplet designs have made most cpu’s different configurations of another cpu in their series, just with different cores enabled or disabled.
The big thing is the cheaper/lower end models are often made from the silicon that didn’t turn out as nice, needed to have bugged cores disabled, etc etc.
So a two pcore processor with a restricted IO die with fewer ports I’d expect to be cheaper than a 4-6 p core configuration of the same processor simply because its able to use silicon that wasn’t able to be shipped in a fully ready flagship processor.
Course, the G series having a ton of working E-Cores is what makes me doubtful that the chips are actually cheap. Still, I’d like to imagine the P core issue is something that is a big selling point on whether or not a die is considered valuable.
I kind of doubt anything in the pantherlake family is anything resembling cheap but it looks like intel is determined to actually compete again which is nice. This feels like when amd came out with the first rdna based igpus and leapffrogged everything.
The z1 extremes didn’t seem to be binned down from the 7840u all that much except for the disabled npu.
Still it doesn’t make much sense for framework to introduce another sku for a same just slightly different chip that may or may not be a little cheaper while the ram for it makes up the bulk of the effective cost. But I’d assume it’s on the same bga socket and platform so would probably not be a huge ammount of work to implement, cut down io may be an issue though as framework is probably one of the only laptop manufacturers that actually uses all 4 thunderbolt ports.
Honestly 75% of cool projects depend on ram and SDD prices returning to being sane to like, actually be worth it.
Cool stuff, worth it, and cost actually depend on things actually meeting up with each other.
I would say my interest is extremely…meh. At least right now. Part of that could just be due to what I use a laptop for. I don’t do any LLM/ML type stuff.
Don’t get me wrong, I’m still always interested in seeing leaps in power and efficiency. But right now, N1/N1X have a lot of question marks around them. Until they actually launch and we can see what kind of real-world performance they have in a laptop power envelope, I can’t really get too excited. Yes, they might be amazing. But they also might result in similar performance per watt to other options we already have, or be only a slight improvement. They also might be super expensive. They might be both of those things.
They also are pretty much guaranteed to require soldered RAM. For certain use cases, that might be fine or make sense. But again, for my laptop usage, I prefer socketed RAM where possible.
Time will tell. I can’t say for sure what machine I might or might not want them in right now, if any. If your poll had a “No interest” option, that would probably be the most accurate thing I could pick right now.
If they release and it turns out they offer great performance per watt at a reasonable price, maybe I’ll find myself interested.
Performance per watt maybe, at least at higher loads but I am extremely doubtful about the reasonable price bit, the estimated prices from ces coverage made high end strix halo laptop prices sound like chromebooks.
Ha! Yeah, that’s a big reason why I’m not really interested right now. I mean, soldered RAM, efficiency uncertainty, and just NVIDIA in general are other reasons. But the near certainty that it will be super expensive makes me pretty uninterested in it for a Framework laptop at the moment.
It’s also not even a new platform, for peak performance you can just look at the dgx spark and those have been out of a while.
i believe their stage demonstrators had 2024 markings on the actual chip covers as well.
This is just old back catalogue projects for the consumer space that they finally got the power solution to be stable for as far as I know, they are trying to be an infrastructure provider here and the consumer products more a coincidence that’s been a long time in the making.
Support is going to be dubious since if we don’t meet their fantasy requirements to compete with their infrastructure business they will likely just put the money that would go into support back into investing in the people buying their infrastructure again, since a diverse product portfolio hasn’t been what NVIDIA has been pushing for lately.
The prices that are estimated for their products are far beyond what we’ve been seeing of other products that people are hesitant to buy. We would need events that would shatter our faith in humanity for this to succeed on a level that pulls them back from more server infrastructure.
That or terrible things to happen with the infrastructure business, in which case then there will be a time lag on their next project and they’ll have to force this chip to succeed.
Its a uncertain on whether or not these get held in stock for so long they go for clearance at a good price, but it would take really unsettling economic events for the price to fall to sane levels AND the chip to get supported. Its just. I’d like it to be affordable but all the scary prerequisites make me dizzy.
They were allready that vintage when they came out in the dgx spark in 2025.
I am guessing windows support definitely not as bad as the arm surfaces and probably worse than qualqom.
Linux tbd but spark apparently works pretty decently.
I am a human, I don’t need a laptop for robots.