Possible GPU upgrade for Precision M4600

That may be true in most laptops (ie. with switchable graphics, but integral to the mainboard, soldered on, not via MXM).

I think i might've posted this here before (although i could be wrong), but here's the "block diagram" of the M4800 LVDS / QHD+ version (LA-9771P) of the mainboard:

http://laptopserviz.bg/media/catalo...aq10_pt_la-9771p_rev_0.2_x01_r0.2_0115_02.jpg

Since the LA-9772P is the eDP version, i can only assume that does away with the eDP-to-LVDS converter for the IGP, and has some eDP multiplexer for the IGP/MXM switching for the internal display.

 

Fair enough -I'm no ee, and my ability to comprehend the intricacies of "PS8331 – 2:1 DP to DP Mux and TS3DV20812 Differential SWITCH 8-Bit, 1: 2 MUX/DEMUX With 3-Side Band Signal" datasheets is utterly and completely nonexistent.

You did, indeed post that block diagram previously, as well as one for the M4600 -I think they all look great -Very pretty, and I especially like all of the little boxes (very modern aesthetic).

I'll just have to trust you on the whole "LVDS MUX is basically a switch, so no display for you" thing. At least I should be able to use this with hdmi out on the m4800. When it arrives, I'll try it out in the m4600 first, just for fun.

In the end, I guess it's cheaper than a new game console, so there's that.

 

https://en.wikipedia.org/wiki/Multiplexer

And for details of how they work in the M4800, just follow the signal flow (lines) from their respective sources, to the outputs (ie. video connectors), while keeping in mind only one input / signal source can be selected at any one time.

 

Thanks for that KhronX, nothing like a little eli5 wiki to help clarify everything. /s

I originally assumed that igp framebuffer access bypassed that entire chain, and the switching hardware was used tor discrete gpu access bypassing the igp when switchable graphics are disabled via sbios.

 

At least in the M4800, that's unlikely. For example, my Intel GPU is driving my internal display most of the time (ie. unless i'm gaming), while my 4k TV is permanently driven by my M5100 (ie. the TV's the only "display" that shows up in the Radeon Settings, and my internal display only, shows up in the Intel control panel).

Unfortunately i can't recall whether that's the case simply because the IGP can't drive 4k; i don't remember if the display "association" was the same with my previous (fullHD) TV.

That being said, i've seen a few other laptop "block diagrams" like those, with similar-generation AMD GPUs and Intel CPUs, where all the display outputs came out of the CPU or PCH; the discrete GPU was only connected via PCIe to the CPU:

http://laptopserviz.bg/media/catalo...rghini_y40_la-b131p_r10_0303_r1.0_0304_02.jpg

 

So, glancing at the Y40 specs, there is no way to disable the igp via BIOS. This makes sense for that laptop, as it's a 'gaming laptop'. In the case of the precision series, Dell had to include the option to disable the igp completely, as a large portion of the corporate customers who purchased these machines utilized them for high-end modeling/rendering/engineering/whatever, and needed direct access to the workstation-class discrete gpu.

Every piece of documentation I've looked at for powerxpress/dynamic switchable graphics/enduro indicates that no muxing has been utilized since powerexpress 4.0 in 2010. There are TONS of articles describing how optimus/enduro works, and every single one describes accessing the frame buffer over pcie.

Once again, I'm no ee, but I don't see how your component diagram indicates that the m4800 isn't compatible with powerxpress/dynamic switchable graphics/enduro. Dell's marketing material says that it supports both optimus and enduro, depending on which discrete gpu is installed.

Do you have some documentation/information that I don't that leads you to believe that your m4800 isn't functioning in the manner described above ie "igp handles the frame buffer and outputs to the laptop screen, and the discrete gpu fills the back buffer"?

 

OK... So Gecube sent me this:

https://nofile.io/f/xTaWUZRfnGx/SCHEMATIC1+_+18.pdf

Along with the (poorly translated by me) message:

Hello

RX560 does not support EDP and LVDS.

We need an attached file with your motherboard video output circuit diagram so that we can configure the DP / HDMI output according to the circuit diagram.

@KhronX Any chance you have access to the full schematics for this board?

 

So, I sent back a copy of the block diagram and the message:

DP_A = Displayport 1.2
DP_B = Dock Connector
DP_C = HDMI 1.4
DP_D = LVDS Laptop Screen

This doesn't seem to match up with other mxm 3.0 display output schematics that I was able to dig up. On some other schematics, they have HDMI on DP_A and Ex Displayport on DP_C

 

OK, so now I have to wait until the 26th for their engineers to return from vacation and 'check inventory'. Got a little more info from gecube -Apparently the rx560 chip they're using is what doesn't support lvds or edp. I confirmed that they're running a desktop chip in the gtx1050, maybe it's the same with the rx560.

Time is ticking, as my friend will be returning from China in a couple of weeks. Looks like I might have bought a m4800 without getting the graphics card it was intended for. We'll see...

 

Which "frame buffer"? To me, that's basically the VRAM.

I half-recall having typed it out before, but on my previous laptop (Lenovo Thinkpad E520 - Sandy Bridge + Radeon 6630M) i needed specific drivers, either straight from Lenovo, and later from Leshcat, to be able to have the whole muxless switching thing working "properly" (ie. the Radeon outputting the video data back to the CPU / IGP via PCIe).

On the M4800, the switching seems to be done "physically" / in hardware - there would be zero point in having those multiplexer (ie. input selector) and converter chips otherwise, between both GPUs and the video outputs, wouldn't there? Plus i've installed the "normal" Radeon drivers on several occasions, with no loss of any functionality.

PS: The IGP's "frame buffer" can only possibly be (part of) the system RAM, which is dog-slow compared to the GDDR5 present on the MXM boards, so... that kinda-sorta doesn't really hold water (for me) either.

At least in the case of the M4800, it kinda doesn't matter which DisplayPort (hence the "DP_x") goes where - each output seems to have its own "thing". As long as they're enabled, it's all good.

As far as i can tell, the only question mark left regarding the eDP support, is the data that gets communicated back and forth, on the AUX channel. The four pairs of data lines are same between eDP and the "full fat" one, the HPD (hot-plug detect) signal is also part of the standard... So physically and electrically, the GPU / video card shouldn't care what device it outputs the video data, on any specific output - the DisplayPort protocol's the same. That's part of the push to "universal-ize" it, so you wouldn't need to tinker with settings for every specific output for ever specific application. But as i mentioned, that AUX data is the "unknown" that's left...

I'm really not buying the "no eDP support" thing. LVDS i can understand, eDP's here to replace it, but i really don't believe the no support thing.

Is that supposed to be the "pinout" for the MXM connector of the RX560 board? No, it can't possibly be. See all those capacitors to the right of the MXM slot? Those are on the mainboard, so they sent you a page out of the schematics of some... most likely, embedded mainboard or something like that, which only has a VGA and a DVI output. That chip on the top-right is a level-shifter, and its outputs have the "DVI" prefix, and the only DisplayPort used is the "C" which goes to that chip - you do the math

Well, unless they sent you that only to confirm the "names" of the DisplayPort outputs...

Either way, here's the link to a PDF with the MXM pinout of the connector in the LVDS version of the M4800 mainboard.
(I wanted to attach it, but even zipped it was over the 2MB limit...)

https://www.dropbox.com/s/1mq63ez5ddy34sx/M4800_MXM.pdf?dl=0