I do, yes, but to make a long story short I have the wrong component for it. This board does have the right footprint for the correct fan headers, (something I fixed on the V2 boards), but I’ve been trying to hold off buying the correct components until I’ve got a larger digikey order, as i reused all of the components from V1 on V2 . Im just now soldering on the mxm connector to test the PCIE link, so once I figure out what else is broken I’ll put in the order.
3 Likes
You might want to order some thin enamel-insulated copper “wire‐wrap” wire. About 0.2mm (30 awg) is a good size.
It is a little easier to work with than the black wire you were using to fix up traces.
Great news you got the EEPROM working. I guess we will never know why the old board failed.
2 Likes
Will do.
Here’s the update then:
Nothing blew up when I plugged it in and switched it on. The eeprom is being read in the bios, which is great, and the GTX 740M MXM card is getting noticeable toasty thanks to my woefully inadequate cooling solution. Its detecting a PCIE translator card, (or something along those lines) in device manager, but no display adapter. I have neither Linux nor a USB stick on hand to check whether Linux works. For the moment I’m just going to continue checking continuity in case anything is missing. I’ve unfortunately also just realised that there aren’t any drivers available for windows 11 and the 740M, (though if anyone can show them to me that would be amazing.)
Please throw ideas at me and I’ll happily try them!
3 Likes
Did it get the PCI ids correct for that card?
If it has those correct, at least the I2C configuration interface for the card is working.
I don’t know what the equivalent of “lspci -vvv” is in windows.
To see what speed PCIe links it has negotiated, and how many lanes.
Downloading NVIDIA drivers:
GTX 740M MXM, no drivers for Windows 11, but has drivers for Windows 10 and drivers for Linux.
3 Likes
I’d try win10 drivers on win11, there’s likely some backcompat which will allow you to get some sort of confirmation that the card is working
3 Likes
I’ll check once it boots on.
Ive tried the windows 10 drivers already I’m afraid, but the installer won’t let me. If it’s possible to install them without the Nvidia exe do let me know.
After checking, (and fixing) various continuity lines, the computer seems to be booting but not displaying anything, which could actually mean it’s working and there’s just an issue with my display port connections. I’ll keep on trying though.
Is there any way to boot it with the GPU attached but using the IGPU?
2 Likes
Hmm I’m confused. I’ve checked the continuity of every trace and fixed all the issues, but I’m not getting any display from the framework with the MXM card installed. Is the BIOS normally run from the iGPU or dGPU?
Aside from this, I’ve been experiencing extreme screen flashing with no identifiable cause while the screen and main board are out of the chassis.
1 Like
It might just be a driver problem causing the screen flashing etc.
With a lack of suitable drivers for Windows 11, I would suggest maybe continue to test using Linux. Linux has the DRI_PRIME=0 environmental variable to force it to use the iGPU.
I don’t know which the BIOS prefers.
Another possible cause of the screen flash might be RF interference.
When the mainboard is in the laptop chassis, it is surrounded by metal shielding.
It also looks like you are using an eDP extension cable. That might contribute to the RF noise.
I don’t know how well the FW 16 does DP pre-emphasis and swing negotiation.
1 Like
It’s probably RF, just annoying that it’s happening. Im using an external monitor now to force output via the iGPU.
I think it knows that there’s something connected, but nothing apart from a generic “x8 PCIE" link is showing up:
4 Likes
I’m going to try one more total reflow of the whole board in the oven just to fix up any potential loose connections.
Okay stuff now fully isn’t working. I’m not even getting voltage now on the main line of the card.
Here’s the plan then:
-
I’ll put in the Digikey order with more components now. This includes:
- The correct fan headers
- 0 ohm resistors
- Extra EEPROM chips
- A new MXM connector, (this one is warped)
-
I might also try a new MXM card. I’m just not sure if this one actually works, and windows 11 drivers would certainly be useful.
Aside from that, it’s now back to the waiting game until the parts arrive. Thanks again for all of your help so far.
JJ
Edit: Order placed 
5 Likes
Digikey delivers quickly these days… about to bake an entirely fresh board.
Improvements:
Using 0 ohm resistors now instead of jumpers
Fan headers!
Capacitors on the 12V fan lines
And most crucially a brand new Mxm connector to be applied with solder paste, not hand soldered on.
9 Likes
Some definite improvements here. The Mxm connector isn’t even slightly warped, and the capacitors look straight out of a factory. A few more passives to hand install and some fixes to make, along with sorting out this abomination of a bridging catastrophe on the connector. Might have to try more viscous solder paste next time.
9 Likes
Well… not a roaring success.
Good news: main board still works
Bad news: smelled magic smoke from adapter board.
I’m starting to feel as though the actual problems here are stemming from the MXM connector. I continuity checked everything for shorts, but clearly something still slipped under the radar.
Any ideas for the best way to get the connector on there? I keep getting loads of bridged connections with a solder stencil which I then have to clear up by hand.
Different solder paste maybe?
2 Likes
Not a way to get the connector on there (I generally just hand solder personal project stuff like that) but concerning the “smelled smoke” thing:
If you have a meter on-hand, it’s a great idea to check that all of the shorts have been removed with something other than eyeballs. You very well may have already done that of course, but it’s also worth taking a few minutes to ensure there aren’t any hidden shorts between pins that didn’t look like they were initially shorted.
Normally when I’m soldering connectors, I’ll add solder to one or two pads on an edge, situate he connector by hand, flow the solder-containing pad to the associated pin, zoom out and check that the connector is sitting flush, then solder down a pin lateral to the first soldered pin. Double-check the connector orientation, fix any problems, and away we go.
A microscope is so handy, too. Unsure if you mentioned having one or not, but if you don’t AND you think it’s a reasonable investment (i.e. it’s something you will find at least semi regular use for), it’s great for ensuring quality soldering!
1 Like
This is what I tried though, checking the major voltage lines for continuity with ground. A microscope might also I’ve be useful in the future. Your comment did remind me though that I failed to check for shorts between different voltage lines - maybe this was the problem.
2 Likes
This can happen for a number of reasons.
- too much solder. Make the solder mask slightly smaller than the pad resulting in less solder and less migration.
- in the reflow oven, change the profile. Warm it up first to just below solder melt temp, then peak it above for as short time as you can, to make a good solder, but limit solder migration. There is a delicate balance here, because the connector can warp. The connector datasheet should have reflow guidelines.
- experiment with different viscosity solder.
1 Like
Oh and not just between voltage lines! If you have suspicions about a shorting connector, I’d suggest measuring continuity between pin 1 and 2, then 2 and 3, 3, and 4, et al. There will obviously be plenty of pins that should have a short between them (gnd, etc.) that can be skipped.
There may be some value in testing the board alone with a current limited power supply first (having test points to easily inject voltage can massively help there) and checking for known current draw and if it’s higher checking for hot-spots with a thermal camera or alcohol or something.
My bet is on less solder paste primarily
That is a good point but man there are a lot of pins XD.
1 Like
For testing to avoid smoke.
On a production line one would have a test MXM board, largish test points on the board, so an automated jig can connect and a test board and to connect via interposer.
One then tests first for shorts, then gradually raise voltages, making sure the expected current is drawn, and then lastly, test patterns in sequence to test other parts. Depending on the path, ac or dc test volts are used.
For a one off board like you have, you can’t do all the above, because the test equipment itself would be too complex.
You might be able to do a MXM board that just has a small microcontroller that switches a low voltage,current limited source, switching the source across all the Mxm pins, that then cycles a detector across all the pins making current, volt measurements.
Or, even simpler, make a test MXM board that simply breaks out all the pins, so you can fit your voltmeter probes on more easily. If you arrange the breakout pads well, you can simply swipe the probes down a line of paired pads, completing all the short tests.
Maybe do similar breakout for the interposer.
3 Likes