2 Likes
I’d like to try and impedance match the PCIE lines before I send the board off for a first prototype. This is the last step before I submit it to JLCPCB, but I’m not actually sure how to go about calculating the impedance of whole traces with vias in them. Can anyone help?
Awesome! I’m glad my morning musings look helpful to you. I wish you the best of luck with this project.
Order placed 
. It’ll be a two weeks before it arrives and I’ll need to do a digikey order as well, but at this point I’m just committing and I’ll work with what I’ve got.
7 Likes
Noticed the fan connector models being a different type than what the GPU uses. I.e. check here - the GPU board uses ZIF connectors, while yours seem to be normal “push-in” ones (i saw similar ones in my Dells, where they were made by JST)
1 Like
Hmm. Luckily I had to cancel that order anyway because they were trying to surcharge me for vias that were too small. Now you mentioned it they do look wrong, but I’m fairly certain I copied the exact part number from some reference schematics. I’ll check in a sec.
1 Like
@Joe_Shapiro If I good remember I saw recently comment from @James3 who was proposing someone help with verifying impedance and other design issue. I’m not sure if he can help you, but you may try to reach out to him and check if he will have time to help with it.
I actually already have 
. Saw the same comment, but he’s moved roles and hence doesn’t have access to the tools anymore. I hope it’ll be okay for a first prototype at least
I had to make some changes for it to be manufactured for a reasonable price by JLCPCB, but it has now passed all the pre-checks and is in production.
This board is essentially a V0.3 board with larger vias and updated footprints for the fan connectors. I’ve actually uploaded it to github as well but under the “V0.31” iteration as this is the board that makes most sense for other people who might be experimenting to try. In the end i settled with the SFW4R-2STE1LF from Amphenol, (available on digikey), as a replacement fan connector. @Gmanny correctly identified that the previous footprint I had been using was for a plug type fan, not an FPC cable. I think this must have just been one of the things changed between pre-production of the dual m.2 slot schematic that’s available on github and the one that actually shipped.
6 Likes
Digikey order placed. Parts are likely to arrive long before the PCB, but I’ll at least be able to check the fitment of the MXM card. I’ve ordered some heatpipes along with the rest of the components to prototype the thermal solution, but I’ll just use an old RX580 heatsink for prototyping when it’s all assembled.
5 Likes
This is so sweet! I am down for sure. I want to stress that I respect the RTX 5070 that Framework is now proposing, fair enough.
But I just checked and it would give me 6.7 fps for my use case of a more than 4 year old game (Cyberpunk) in 4k with raytracing - for close to 800 Euros. Hm.
So yes, this thread is all I have. 
Yeah I should probably give an update after the event :).
First of all, huge congratulations to @nrp and the rest of the team at framework - the community has been waiting for both Nvidia and just a general GPU upgrade for ages and this has gone a long way to prove that upgradability is still possible.
That said: this project is very much not stopping. The 4090 mobile still offers a 25% boost over the 5070 mobile, as well as 16gigs of vram instead of 8. Hopefully we’ll even begin to see 50 series mxm cards as well in the future. The 5090 mobile is about 50% faster than the 5070 mobile, for anyone wondering, but obviously at the cost of some extra size, heat and power.
Finally, an update on the PCB. V0.3 is just finishing the inspection phase before it gets shipped out and testing begins. I’m not going to design another one until we’re sure that the most basic design works, (though XenoCow’s ideas will be first in the queue once the V0.3 is validated). Hope this clears up that this thread is certainly continuing, and massive congratulations to the FW team again.
8 Likes
+1, so …
4 Likes
That is unless some Framework staff gives an update soon that they have been working on it and that the 5090 is in fact “fully operational” ;).
PS: But more seriously, awesome thread, and yes, it should not go away any time soon. A 16 inch laptop deserves more than a 5070…
2 Likes
I’m glad to see this project still on track. It’s great to see a first party Framework Nvidia GPU, but I see this project still having a lot of potential even with first party options.
2 Likes
Hi everyone! The MXM gpu has arrived already, Digikey gets here on monday and the PCB has finished production and is waiting for pickup by the carrier.
By the way, for those who have never seen an MXM GPU before, (like me until five minutes ago), they are astonishingly tiny. The picture below is with an RX 5500 board for scale - how someone has managed to put a 4090M onto one of these I’ve got no idea.
I’m afraid there’s not likely to be much more information from me for the next few weeks as I’m both going back to school and waiting for the PCB and dev FW16 to arrive. I’ll put another update here when the Digikey order gets here though, and again as soon as all the other parts arrive. Thanks for all your support again, (including memes…)
9 Likes