Hello all,
So I recently got it into my head after seeing NVIDIA open up GPU passthrough for GeForce cards that I wanted to try something a little unusual, though perhaps not so much on this forum.
I'm a 3d artist and do a fair bit of rendering but am also an avid tinkerer and like messing around with combining odd bits and ends and seeing what I can make of it. As a result, I came up with this idea to build an external PCIE expansion box on my own, but not only that, I wanted it to use the Threadripper platform and be supported by bifurcation so I could maximize my slot density. The Asus Zenith Extreme motherboard has 2 x16 slots that can be split to 4 each (supposedly, I did put a hyper m.2 card in one so I know that much works) 2 x8 slots that can also be split, a x4 lot and an x1 slot. So of these, it looks like I can possibly get 13 x4 slots in on my board.
In a nutshell, I bought some of these intriguing risers from a website called maxcloudon, you can find them here. They claimed their risers support bifurcation some x8x8 and some x4x4x4x4 and a few in between. I went ahead and ordered some from them and much to my surprise I was able to get two old quadros I had lying around to be recognized on the bottom most slot of my Zenith Extreme motherboard (x8 or x4+u.2 or x4x4 aka pcie raid as Asus calls it). Great, easy enough.
These risers appear to use sff-8087 instead of the USB connectors that you would typically see on something like a mining riser. So it got me thinking, why not get a sff-8087 to sff-8088 adapter card and hook them up to see if I could relocate the cards instead to a separate enclosure where I could better control my thermals and possibly set it up for water-cooling.
So I ordered a pair of each the sff-8088 cables and 8088-8087 adapters from Amazon. They arrived and I went to connect up a single known working riser to the new hardware. Lo and behold, no signal from the card and it's no longer recognized in the device manager. However, I did note that my boot sequence was longer than normal as if the board realized something was different and needed to reconfigure.
So my question to those of you that have experience in one or more of the technologies used here is where did I mess up or where should I start looking? I already tried every combination of new cabling with new adapters to rule that out (though if I have more than one piece of DOA hardware I really don't have any other way to test) and I did plug the riser back up in it's default configuration to make sure I didn't fry something somewhere during my testing and the card is once again recognized when connected directly. I'm assuming it's not due to signal loss, as I've seen before a ridiculous number of daisy chained pcie risers courtesy of Linus Tech Tips.
Thanks for taking the time to look through this!
So I recently got it into my head after seeing NVIDIA open up GPU passthrough for GeForce cards that I wanted to try something a little unusual, though perhaps not so much on this forum.
I'm a 3d artist and do a fair bit of rendering but am also an avid tinkerer and like messing around with combining odd bits and ends and seeing what I can make of it. As a result, I came up with this idea to build an external PCIE expansion box on my own, but not only that, I wanted it to use the Threadripper platform and be supported by bifurcation so I could maximize my slot density. The Asus Zenith Extreme motherboard has 2 x16 slots that can be split to 4 each (supposedly, I did put a hyper m.2 card in one so I know that much works) 2 x8 slots that can also be split, a x4 lot and an x1 slot. So of these, it looks like I can possibly get 13 x4 slots in on my board.
In a nutshell, I bought some of these intriguing risers from a website called maxcloudon, you can find them here. They claimed their risers support bifurcation some x8x8 and some x4x4x4x4 and a few in between. I went ahead and ordered some from them and much to my surprise I was able to get two old quadros I had lying around to be recognized on the bottom most slot of my Zenith Extreme motherboard (x8 or x4+u.2 or x4x4 aka pcie raid as Asus calls it). Great, easy enough.
These risers appear to use sff-8087 instead of the USB connectors that you would typically see on something like a mining riser. So it got me thinking, why not get a sff-8087 to sff-8088 adapter card and hook them up to see if I could relocate the cards instead to a separate enclosure where I could better control my thermals and possibly set it up for water-cooling.
So I ordered a pair of each the sff-8088 cables and 8088-8087 adapters from Amazon. They arrived and I went to connect up a single known working riser to the new hardware. Lo and behold, no signal from the card and it's no longer recognized in the device manager. However, I did note that my boot sequence was longer than normal as if the board realized something was different and needed to reconfigure.
So my question to those of you that have experience in one or more of the technologies used here is where did I mess up or where should I start looking? I already tried every combination of new cabling with new adapters to rule that out (though if I have more than one piece of DOA hardware I really don't have any other way to test) and I did plug the riser back up in it's default configuration to make sure I didn't fry something somewhere during my testing and the card is once again recognized when connected directly. I'm assuming it's not due to signal loss, as I've seen before a ridiculous number of daisy chained pcie risers courtesy of Linus Tech Tips.
Thanks for taking the time to look through this!
Last edited: