Let's figure out how to convert internal MXM connector an external PCI-E x16 box

My AW M17x R4 for sure has no whitelist.

 

Nor does the M17x R2 !

 

Hello
I'm new on this forum, but i was dealing with similar idea for some time.
My idea is to take desktop mini itx board, and put it in laptop case
I will start to make some drawings, when i have more time. For now let the idea develop

 

Hello all! I just joined the forum so that I may post in this thread. I've been lurking for a while now, and I have an idea for powering the eGPU. My personal project involving MXM and a desktop GPU involves enclosing the video card inside the laptop case. I will be laser cutting an acrylic "bottom shell" of my laptop (it will be thick enough to house the card, a universal/low-profile CPU air cooler, a pico PSU, and a controller board for my internal monitor's panel (to avoid the LVDS problem, just connect the controller to panel, and video card output to controller board).

Anyway, I noticed that there was questions regarding powering the GPU and I would like to suggest using a pico psu. They can be found from 90w up to 500w. Hope this helps, and I look forward to helping with this project!

 

I'm really excited to see what you turn up with! The pico PSU is interesting. and plugs into a normal ATX port, so would work on my planned board as well. Does it provide PCIe power? All high end GPUs need one or more 6 or 8 pin plugs... And any eGPU seems like a lot of work when all you end up with is a low/middle-end desktop card (Says the guy with the 560TI...)
Are you using your own wiring diagram or one that WhackingCheese or I cooked up? I'd be interested to see it or the diagram showing which pins you connected, if you're willing to share.
Post lots of pictures! You may consider making your own thread and linking it here, since the MXM part is related but the whole casemod / fanmod / mod-mod is not. Besides which, then you can post even more pictures, making me even happier

Since I'm here, a quick update to the board I'm working on. I'm literally sitting in the airport (having not slept for nearly 21 hours) on the way home. Mysteriously, all my flights have worked out, but there's still one left...
I am going to order my board as soon as possible, though I haven't ordered it yet because the gold plating on the MXM fingers and laser cutting need special care and I wanted to be on the same timezone so I could call the people and explain very carefully what I need.
So, hopefully, I'll have a final answer in the next few weeks!

Until then,
Sompom

 

I haven't built anything just yet, I'm at the drawing board. I just wanted to share my plan. The pico unit I have in mind is from Ali-Express (yes, I know that the site is sketchy, but it's the only place I can find a 500w pico psu), and has a lot of ports! Here are some pictures of it:





As for which wiring I will be using, I'm not sure. I want to go from MXM 2.0 male to PCI-E x16 female (As far as I know, this lines up with WhackingCheese's design). I may use a combination of both yours and Cheese's design. I need to open up the designs on my PC (currently on my phone), and decide from there.

I feel your pain of the low-end GPU problems, my desktop is running a GTS 450 Reference card. Gotta work with what you have, I guess! XD Anyways, the GPU I plan to use for the laptop is a GTX 960 specifically the ASUS Strix edition. I chose the 960 becasue of the fact that it is extremely power-efficient and the Strix cards can be passively cooled under light to medium load (trying to limit heat as much as possible! Less heat = faster computer for the most part! Lol) . I will have the universal cooler installed just in case it ramps up and creates some heat (since the card is enclosed inside the laptop, I am attempting to vent as much heat out as possible).

I hope that my CPU/chipset won't be too much of a bottleneck for the beastly Strix 960! The laptop is an MSI GX723 which has an Intel Core 2 Duo P8600 CPU, an Intel PM45+ICH9M chipset, and an NVIDIA GT 130m (which is MXM 2.0 as far as I know. I'm not sure if it is 2.1 or not, shouldn't matter all that much).

I plan on documenting the entire mod process since I want a record of it as well. I attempt ambitious projects like these to impress future employers with a binder full of working projects; (and to increase my own knowledge, it's a hobby of mine, increase my skills and expertise, etc).

Regards, Nathan "I Hunt Demons"

 

Hey guys, any updates?

I've just bought an eGPU solution and I'm a little disappointed by the bandwidth bottleneck of the mPCI-E with my GTX 960. (Should have expected it to be honest)

The MXM solution, if it works, will definitely trump all other eGPU solutions out there.

I would really love to try it myself, however I lack of any knowledge in this area beyond simple soldering.

Wishing you guys all the best.


Sent from my iPhone using Tapatalk

 

Ack. The update is I was busy when I got back home, then I forgot about this... However, the more recent update is I have just started placing the order for the PCBs. (Tried to. It looks like the website of the people I like to order PCBs from is suffering this morning)

If you're brave, all you should need is a steady hand and a soldering iron The bravery part is there is nothing that says these boards won't magically transform your computer into a bunch of poisonous smoke... I have two laptops in various states of already-brokenness to test with, to reduce the bravery needed.

 

Some good news! Today, both PCBs arrived. Unfortunately, I am still missing all the components required to populate them, but the MXM board fits the socket and the pins appear correct.

Figure 1: The MXM board in a broken (short across the power jack) MSI GT-683R motherboard, next to the PCI-e board

 

This is crazy... I wonder how it will work with the system BIOS and internal LCD screen.
For instance I had multiple MXM cards in my laptop but most have some issues with my RGB LED screen. They all work perfectly on external screens.

 

I haven't connected the internal screen on this first version... There is some possibility that could be done in the future (pulled out to an HDMI/Display Port connection, then connected to any GPU), but I wanted something as simple as possible to test against.
I don't know what will happen to the BIOS... I guess that it will show up on the external screen connected to the eGPU, because I pull the correct pins to say that I am the primary display adapter, but maybe it also won't work. There's only one way to find out!

 

A quick update....
All the parts arrived yesterday, but I had one of the ribbon cables twisted... I fixed that and now it's working (YAAY!!) Partially (Awww....)
Linux doesn't seem to report any problems, but as soon as I install the drivers in Windows it goes directly to hell. No display, remote-desktopping in also has nothing. Windows worked before that...
I need to go to bed, so I can't solve the problem right now, but I'll leave you with this picture...

See where I've highlighted "Link Width x16"?

 

Are you testing that on a laptop? What model ?

 

Testing laptop is an MSI GT683r

 

Probably a BIOS incompatibility. You need to investigate...

 

That seems unlikely, given that it boots at all... The BIOS incompatibilities I have had go one of two ways: The BIOS pops up a message telling you it doesn't like the hardware and to go jump off a cliff, or the system just doesn't come on, possibly accompanied by beep codes or flashing lights. Linux works, Windows works until the drivers are installed.
My first thought is that, although it looks like everything is good, there is something miswired which doesn't get poked until the NVidia driver takes over. Could also be that the system reports the internal display, the NVidia driver tries to use it, but since nothing of the sort is connected, it fails.
One more thought is this laptop actually does have an internal Intel GPU on the processor, it is just not connected to anything. It may have somehow come powered on and Nvidia is trying to do Optimus against it, which fails since there is no display connected. I would have thought the display being on the eGPU would get around some problem like this, but maybe not. Setup 1.x may help me.
I'll investigate all these thoughts when I have some time

 

Oh, I thought you didn't pass BIOS.
Yeah, looks good so far.

If you need further testing I am willing to experiment.

 

I have a Alienware 18 with an external Titan card going through the gxp Beast adapter.

I would much rather use a mxm solution so if you need testing tell me and I will pickup any components that are needed.

 

If it's not a driver issue (I'm not sure if there is any micro-controllers, or other logic chips on the adapter), then I would double check every wire (Like you said, the cable was twisted before, which I'm assuming is when it worked in Windows, but then you untwisted it to make it work with Linux). Maybe even check each connection with a multi-meter to ensure that each connection is continuous.

 

Wow, this thread is still going strong.
I'd been thinking of the possibility of using MXM gpu's as lightweight eGPU's for a while now, guess I'll be reading this thread through.

 

Could you possibly give me a link to where I can purchase these PCBs and a parts list please? I would like to start working on my own adapter.

 

Sure,
Here is the parts list:
https://www.dropbox.com/s/dz68vmna4t85rpy/Parts List.doc?dl=0
And here are the gerber files:
MXM board: https://www.dropbox.com/s/ciyrvnvyhhxsl7m/MXM.zip?dl=0
PCI board: https://www.dropbox.com/s/59f2k632g6jpc16/PCI.zip?dl=0

I actually need to make some changes to the PCI board. For instance, if you use an ATX power supply header, it conflicts with the catch on the back of the GPU. "Happily" this is mechanical only and can be cut off (as I did...) For now you should be okay if you just use some other option (SATA power, plug, for instance) to get power to the board.

Also notice that I haven't solved the problem where the screen shows only black after installing NVidia's drivers. I don't have any ATI cards to test with, and only the one NVidia, so maybe it's just a problem with my setup.

 

Hello, could you link your project (not sure exactly how to) so that I can edit the CAD files. I can't edit a gerber file and there are a few errors I would like to fix on the gerbers I downloaded from your links.

Thanks

 

Sure thing:
I did the boards with freePCB ( www.freepcb.com). I don't know anything about anything to do with PCB software, but I think that program can export to other file formats if you want to use something else.

Main files:
MXM board: https://dl.dropboxusercontent.com/u/27091889/MXM/MXM_PCI-e.fpc
PCI-e board: https://dl.dropboxusercontent.com/u/27091889/MXM/PCIe_Base.fpc

Library files: (These contain some custom parts like the PCI-e header. I can't remember where they need to be saved to, but I think it was pretty obvious. The computer on which I had FreePCB setup has blown up since I did this >.>)
https://dl.dropboxusercontent.com/u/27091889/MXM/MXM.fpl
https://dl.dropboxusercontent.com/u/27091889/MXM/PCI-e.fpl

If you're planning to fix things, there are a few things I've noticed as well:

The 2x20 pin header holes are a bit too small. I'm sure there's a standard size for pins, and I'm surprised the default FreePCB part wasn't that size, but I don't know what it actually is. I think the holes are 1mm diameter now, and you might try 1.25mm.

The ATX power supply header interferes with the back L-shaped anchor on the GPU. I haven't thought of a good solution, but one possibility is to make the board ~15 mm wider and move the ATX header forward.

Also, it turns out you're not supposed to make 90° turns with signal traces. Who knew? (I guess anyone with basic digital electronic knowledge...) So you might want to re-route all the traces (this could literally be the cause of all my problems). If you're using something like Eagle, it should do it automatically.

I still haven't decided if it's necessary to jumper PWR_GOOD to 3.3V. It probably is. I did before it started working, but I think I had to flip a cable after that.

Could you please tell me what else you've noticed, so I can keep track of what's going on and so I can learn?

Thanks,
Sompom

 

WhackingCheese, did you develop your converter based on the MXM 3.0 connector? I'm in need of a converter for the MXM 2.x connector, and I was hoping that you have schematics already made to save me some time.

 

Hween, are you just fixing some mistakes in Sompom's design (such as connectors, and the other things that he mentioned)? Could you make another adapter (CAD file) that is MXM 2.x to PCI-e 3.0 x16? Sompom's design is an adapter for MXM 3.x and I need 2.x for my current machine. If not, no worries, I may be upgrading to a new laptop soon; however, 2.x is highly preferable since the new machine will be for school/gaming which would leave my 2.x machine open for experimentation, modding, etc.

Thanks!

 

So has anybody a fully working solution yet?

My plan is to use this to connect my late 2009 iMac to an external GPU (was thinking about putting a 980m mxm into it but it would be way to overpriced if the adapter solution works).
Using the little ventilation holes on the bottom for the cables (maybe create a custom connector for them).

 

You'll have to see if Hween comes back with anything, as I'm at the point where I don't know the path forward. One day I'll get around to re-routing the wires to be length-matched and not have 90° turns, which may accidentally fix the problem, but don't rely on it.

 

I never created a pcb but I might try next weekend. Still searching for a pcb manufacturer that doesn't charge 200$ extra for 1.2mm material though...

 

Let me know how it goes! I used IteadStudios: http://www.itead.cc/ , who didn't charge for the 1.2mm boards. You may be able to refer to order number D00000469 and just have them do the same again (except make sure it doesn't ship or bill to me )... It came out to $100 for 5x. Unfortunately, I have broken/used/shared all of my extras at this point :/

 

Thanks for the help and all of your previous work here. I hope it'll work,

 

One late thought for you, which you may already know: You may have trouble with a Mac for the reason that Apple strictly control their hardware. Make sure you get an eGPU which is "Apple Certified" (Or maybe I'm totally wrong; I haven't worked with Mac hardware much)

 

Well, I guess nobody knows what's gonna happen. There is only one way to find out..

 

Sompom is correct. Apple is in fact very closed source, and choose very specific hardware which are then given proprietary software (I'm not sure if it's custom BIOS's, drivers, or both), and then they'll work with OS X. I believe that if you're wanting to run a different OS than OS X, you may be able to get away with using any card. If you do choose OS X, then you may need to do some extra research. I highly recommend that you research the topic so that you don't waste any time or money. One thing that will definitely help you is researching Hackintosh machines. Hackintoshes are computers not from Apple, but still run OS X. The Hackintosh community has tutorials, build guides, Apple Certified hardware and software, and other information that will prove helpful.

I hope that this helps!

 

i want to do it otherway around: using MXM on desktop or convert stander PCIE to MXM slot for smaller build

 

I'm working on a similar idea, I'll be using Mini DisplayPorts instead and running the plugs to were the heat syncs where in my Alienware m18x, the main I'm hoping to make it so while the eGPU PCB isn't connected it will just run the integrated GPU, I'll let you guy's know as I get further in the project

 

Actually, that brings up an interesting question that I had just never thought to ask before. Do laptops with removable GPUs and Optimus work with the GPU removed? I would assume so, provided the BIOS doesn't stop you...

Another interesting thought occured to me while I was chatting with a friend today. Most modern motherboards/processors don't support enough PCI lanes to do 2x16 for SLI, so run in 2x8 instead. Also, PCI lanes are splittable, meaning one could run two x8 connections from a single x16.

No promises, but winter break is coming soon, so hopefully I'll get motivated enough to put together a new version of the MXM converter and fix the various problems I've identified. Maybe also think about a way to split the PCI connection in two, but that starts getting into the problem that SLI/Crossfire capable motherboards need some silly chip or another to tell the cards they are allowed to do SLI. There are software ways around that, but they are unreliable.

 

well from memory m18x ran both in x8 considering that most motherboards do that I would expect laptops to just do that, but it was about a year ago I checked, and it is an "old" laptop now, it seems to have issues frying graphic cards I've gone through 5 of them with that laptop :/, I think it's just a power issue... because it works with a new one for a while before destroying them XD

as far as optimus goes, when my laptop does not detect a gpu when powering up it switches to intel's gpu, I don't think optimus would work with a desktop card, because I don't know if they can send the data back thru the PCI-e lane for the intel gpu to display :/ happy to be wrong though

 

Yikes. That sounds like an expensive laptop to operate!
As I said, high-end, modern desktop processors do not support more than 16 PCI lanes, meaning cards will run at x8. It would not be surprising if laptop cards do as well, since laptop processors usually cut corners compared to desktop processors.

Interestingly, they absolutely can. I can't send a link, because Notebookreview really does not like the other website, but look on Google for "eGPU experience [Version 2]". There, you can find a list of benchmarks and recommendations, and notice that laptops with an integrated Intel which is able to do Optimus with an NVidia card has noticeably higher performance than those which can't.
This is actually the easiest way to get the eGPU working on the laptop's internal screen. For reference to that, you can see a picture of a laptop/eGPU I setup awhile ago: https://scontent-ord1-1.xx.fbcdn.ne...=b3cfe5939ae101007c11618a43f3ef87&oe=56E98ACD (Let me know if that link doesn't work. It was just easiest to link directly from my Facebook upload )
The setup there is to remove the WiFi card and connect the eGPU in its place, which is only an x1 connection, which lets the GPU operate at about 80% performance.

Interestingly, the reverse does not seem possible. I have not seen a laptop with a desktop processor and Optimus vs. a laptop GPU, nor have I heard of Optimus being used in a desktop to allow the motherboard connectors to be used for high-performance graphics output.

 

Thats really cool, there'd still be a bottle neck with using optimus but and it's only really using the eGPU for possessing and then gets filtered threw the iGPU as far as I know (it has been a while since I've researched this), while I know it's not much for laptop cards, but it is still noticeable, I don't know what about it slows it down to know if it would be worse with eGPU, but the easiest way to prove the difference with optimus and without is running a bench mark on just your laptop display then HDMI with laptop display disabled, you'll get a better score.
It's cool that, that works though I didn't think it would. It'd still be cool to have the option of running the GPU display back through the laptop, the only downside would be you'd need to turn off and back on again your laptop to switch between iGPU and eGPU, but that lil bit extra never hurts I suppose I don't need to worry about that until after I get it working now
I've just finished fixing up my Inspiron 1720 so it can play Final Fantasy ARR so this is my next project.
Merry Christmas people

 

i wonder how this might work with an alienware 17 R1?

EDIT: is someone printing the parts and making a kit?

 

I have my R1 to test when I get my prototype working, I'll let you know how it goes
I just bought 2 Graphics Amplifier for Alienware from an auction I'm interested to see if I can just use them as the PCI side of things with my MXM card, I haven't looked at how Alienware connects with the amplifier other than it uses it's own plug, I just came across them and put a low bid on them not thinking anything of it but it does have it's own 470watt PSU so that will save some stuffing around hopefully

EDIT: got the Amp, it only uses a x4 pci link, the PSU looks like it will do the job nicely though, I'm going to sell this one, but i think I might see how much replacement PSU's are for them because it's nice and tidy

 

Some of you may still remember me. I’m WhackingCheese. I got interested in external GPUs after my laptop started lagging behind in terms of performance and I found out that I couldn’t replace the GPU with a newer one. I found this forum and saw people using MiniPCI-Express for external GPUs but I quickly realized that those cards were only capable of doing 1x PCI-E which is enough for what MPCI-E cards are meant to do (wifi, sim, bluetooth, etcetera), but it’s not nearly enough for a full on graphics card. So I started looking for other things and on the very same forum I found this thread “Let's figure out how to convert internal MXM connector an external PCI-E x16 box”. It had been dead for over a year at that point with everyone who originally got interrested in the idea when the thread was originally posted back in 2009 having left. I was unsure if my laptop even had this connector, so I went for a nice short google and yes, yes it did! It had this amazing thing called MXM or mobile PCI-E module. An interconnect designed by MXM-SIG, whom are controlled by Nvidia. This interconnect was specifically designed for graphics cards for mobile or small form factor systems. Mine happened to be version 3. And this thread was started when version 2 was the current version. I posted that I’d be trying to create my own connector, and the thread got rolling again. I started posting regularly, updates my findings, documents and ideas. And it was great. I loved it and was very excited to be working on this. We had made plenty of progress but than something happened and I stopped posting.

Last time I posted was on the fist of January 2015. I had just found the Electromechanical spec for PCI-Express, so I had all of the documents required to make this thing happen (excluding the software specs). I had the MXM Electromechanical spec, PCI-Express Electromechanical spec, my laptops schematics and I whole bunch of other documents most of which I cant even remember anymore. I decided to go straight ahead and encorporate everything within all of those document in to my design to achieve optimal performance and system stability. I had length matched all the PCI-E lanes. Made the traces follow to specs entirely. I even found low profile right-angle PCI-E connectors which I was going to put on my MXM board to transfer all of the signals with. It was optimal. As many of you may know PCI-E is fragile, if something goes wrong with it when the system is running, if it gets disconnected or interrupted, it may and probably will crash the system. Which Is why I decided to use things that were designed to handle PCI-E signals for optimal stability and such. I even had planned to use a PCI-E riser to move the signal from the MXM board to the PCI-Express board. I had placed pins on the board for every signal that I was unsure of needing. And that’s just the MXM board. The PCI-Express board was a whole another set of things. It had a cardedge PCI-E connector, like the on at the bottom of a graphics card to move the signal without having it go through too many connectors. The expansion card would plug into the PCI-E slot, which would be powered by a 24pin Molex connector, so I wouldn’t have to use any jumpers or anything to start it. Plus it should be enough for the slot anyway, and the card would be powered normally, just like in you’r standard PC. And I did all of this in the free version of Eagle which took me about a 120 hours of trace routing which was a ****ing NIGHTMARE, but i'm still fairly impressed with myself concidering that I had no previous experience in PCB design and trace routing.

So I had pretty much finished my design when the unthinkable happened. Something that I had never even concidered. I was moving files around, cleaning my desktop, just the usual, when suddently my computer froze completely. I had to hard reset it, which I did and it wouldn’t boot. I went into my bios and found that the my main storage drive wasn’t being detected anymore. My hard drive had died. The one that had all the project files ready, ready to be sent out to a PCB manufacturer. I tried connecting it to different PSU’s, different SATA ports on my motherboard and I even plugged it up to my secondary laptop. Needless to say It didn’t work. Then I got pretty pissed. That **** took a lot of time. Length matching differential pairs in the free version of eagle was a ****ing nightmare, a NIGHTMARE. I was stupid enough to never back it up anywhere. No backups. Which Is why I stopped posting things on here. I gave up on the project, I put too much time into something that just disapeared before my very own eyes. I got sad, angry and quit.

I built myself a desktop mid last year and I have been using it for a while now. I recently found the hard drive I from that computer, I decided to plug it in just for a cheap laugh, and It was detected in the BIOS. I froze, I was surprised as ****. I booted into windows and I saw that there was no actual data showing up on the hard drive, almost as if it was formatted. Maybe the lookup table got wrecked or something. I am concidering trying to recover what was once on it, but I don’t really know if the hassle is worth it. So many things have happened in the tech industry since I stopped posting. A whole set of new things that could be utilized for external GPUs: M.2, thunderbolt becoming more widespread, USB 3.1 and most intriguingly USB 3.1 Type C. Some manufacturers even started making laptops designed to be connected up to external GPUs, weather via Thunderbolt or a proprietary connectors like MSI did for their whatever laptop. And now at CES there was plenty of external graphics solutions for laptops showing up. Manufacturers are starting to get into a market which really never existed outside of niche DIY projects like this one. So the DIY scene for external graphics might soon be non-existant. In the future I would expect laptops to have USB 3.1 Type C carrying Thunderbolt (PCI-E). allowing anyone to go out buy a GPU enclosure and connect it right up to the laptop using that USB port. So I think that the DIY scene of external GPUs might soon no longer exist.

So here I stand, after writing all these words, asking you people if you are still even interested in the project. I noticed that Sompom and a couple of others did go and get one of some made and manufactured and it sort of worked and I am noticing new people coming in and asking for kits and stuff. But I really am not sure if this project is even worth continuing after seeing the stuff at CES. Thoughts? I’d be willing to get back on the wagon just to make a more powerful LAN machine out of my old laptop though.

- WhackingCheese (Still alive)

 

Until thunderbolt/usb/any other non proprietary port doesn't support full pci 3.0*16 and isn't provided to all mxm laptops. This project has scope, though I agree it has diminished.

Sent from my Micromax AQ4501 using Tapatalk

 

@WhackingCheese and anyone else interested here's some food for thought. Mxm to 2*pcie 3.0*8 orpcie 3.0*8+2*pcie 3.0*4, i.e, 2/3 way sli/crossfire from a single mxm slot!

Sent from my Micromax AQ4501 using Tapatalk

 

@god1729
Theoretically possible, but I don't see this as a feasible option. Putting two gpus in Sli should be possible but mobile cpus are the limit here, there are only so many different cpus any given laptop can take and most of them are more than likely not powerful enough to support two high end cards in sli. I think that using 8 lanes for a GPU and rest for other expansion cards would be a better option. But first we need to focus on making a design that actually works. I'm currently in school but once I get home for the weekend I will attempt data recovery in the read only mode just so I don't accidentally **** it up.
Also I disagree that those egpu solutions need 16x support. All cards are fine with as low as 4x.

 

@WhackingCheese
1. CPU bottleneck depends on the mobo. I'm pretty sure the zm and new dm series won't struggle with this(6700k !). Even the hk CPU in the new MSI titan handles 2 gtx 980s(desktop).
2. The gtx Titan x looses approx 1% performance over PCIe 3.0*8 and 5% over PCIe 3.0*4, this trend is exponential as lanes reduce(read this a while ago, can't find the source). 5% might not seem like much for a top of the line card but this will increase for next gen cards.

Sent from my Micromax AQ4501 using Tapatalk

 

Most laptops don't have the ability to house desktop class cpus. And use laptop cpus which are unfortunately inferior. I am currently unaware how many laptops on the market right now have mxm slots and which ones do. But I see this as a thing for people how don't want to buy a whole new laptop, but rather wanting to upgrade the one they have or modify it to suite their needs.
Also new cards won't use more pcie bandwidth, the same amount of stuff will have to travel through the pcie bus. It's the applications that will require more stuff do be moved over to an expansion card. Remember all the processing happens on the card, and once something is no longer needed it is wiped. The pcie connection is mostly for receiving and giving instructions to the CPU and loading files meant for the graphics card. But yes the requirement will go up as the industry progresses and it will become a bigger hindrance. But not too the point that 4x won't be enough. We also have to keep in mind material costs, board complexity and whatnot and saving on board space with a smaller connector might be a thing to consider.

 

If I end up being unable to recover my project I will start over. This time in circuit maker which is basically a free version of the multi thousand dollar licenced version of allium designer. It was released while I was gone . I will have to learn how to use it, but it will be hella easier routing those pcie signals with that rather than with Eagle.

 

Clevo DMs and the MSI titans and most of the MSI dominator models are the only ones left, so yea the current target group is small. But even if they eventually kill MXM the future, the people owning MXM systems,even the oldest ones can stretch the gaming lifespan of their systems with this solution as long as CPU/PCIe bottlenecks are tolerable

Sent from my Micromax AQ4501 using Tapatalk