PCI-MATRIX 44UHD

Reviews

1. 

   mrasmus – March 19, 2015

   This device is not a “true” matrix, but rather a partial one. Only one input (#2) can be routed to all four outputs — the other 3 are limited to a pair of outputs each. Input 1 will never display on the higher numbered pair, and inputs 3 and 4 won’t output to the lower ones. This leads to the reality of 32 possible configurations, as limited by the hardware design. This is a dealbreaker for some workflows and applications (it is for me), and not a problem at all for some others, but it’s important to understand it before purchase.

   The EDID manipulation functionality seemed to work well, at least as far as I tested it. I had some trouble with the “furious” functionality, spent about an hour debugging it and then gave up (only to get confirmation of the above limitations, leading me to lose interest in getting it to work).

   Control software works, though a simple visual interface would probably be a nice add-on and simple enough to spin up. The command-line tool is appreciated (the ability to invoke from a batch file is always great), but not great UX for a lot of people. Easy enough fix. Windows-only currently, but there’s nothing preventing them from making Linux or even Mac control software (their interface chip has drivers available for all three platforms, it looks like).

   Didn’t have a 4k HDMI source, nor sink, to test against, though I suspect it works just as well at the higher resolutions. The blanking that occurs on HDMI renegotiation didn’t seem to be happening when the number of sinks paired with a source changed (good), though I did see some second+ blanking times on switching an output (I believe because of differing framerates on sources; I tested mostly with a 60 and a 59.94 source) which I think is more a reflection of the monitor’s re-adjusting to the input’s signal/pixel clock. Not quite instantaneous, but still nicer than most switches I’ve used, and I don’t think it’d be possible to get around that problem without an internal framebuffer/trying to normalize the output FR, which would add lag to the signal itself.

   Actual mechanical design I’m of a couple minds. The PCIe port for power thing is clever; it’d be nice, on a certain level, for it to just enumerate as a PCIe device, so you didn’t have to route and plug in the 5-pin USB connector. That said, the design allows for USB operation controlled by an external computer, potentially, or even an external bay to power it and use as a standalone box (just add 12V1A over the edge connector and you’re good to go; they were talking about maybe making a bay exactly for that purpose, to power multiple of them). I was planning on externalizing mine.

   Installation was a little awkward; putting the card directly above my GPU in the next slot was a no-go because the USB connector was too long (bending it 90° would be a trivial fix with carefully-wielded pliers, or just use a smaller cable connector). The HDMI cables built into the pass-through card were stiff (felt like decent quality) leading to awkward pressure pushing the two apart when installing the daughterboard into an adjacent slot, due to the right angle plug plus the curl of the cable. This may lead to some wobble on the daughterboard mount, as some cases don’t have very tight PCI faceplate anchors, since they’re supposed to anchored laterally by the connector… which the daughterboard doesn’t have. Minor, but I’ve seen some cases that might have issues.

   The cables, once both cards were installed in adjacent slots, curled out to interfere with the space where a full-sized card would go in the adjacent slot. Pushing them down, because of the stiffness, would put undue pressure on the main card, which I wouldn’t be ecstatic to do, but would probably work out alright. The pressure of the design could be relieved by leaving a gap between the two cards (the cables should be long enough); if you have a small card or appropriate empty slot, that’s a solution to consider. Another would be simply not using the daughtercard, and routing the output ports out of your case using cables through a grommet. Different cables on the daughtercard could’ve alleviated this issue, or just making a 2-gang device (though that limits install flexibility and expansion that they seem to be planning, so I’m glad they didn’t go that route).

   The hardware routing limitations are frustrating and spoil it for my use-case, sadly. It otherwise seems like decent hardware, if the routing graph will work for you. I’m going to be returning mine, and hoping they manage something that delivers on the original promise in the next iteration of the product.