Search on ebay for the full height bracket for whatever cards those are. Something like;
https://www.ebay.com/sch/i.html?_nkw=Intel+DC+P-series+full+height+bracket
With no ports on the back, you can definitely McGyver something with a generic full height slot cover, but really the easiest thing is to just track down an original as a spare part.
I bet you have a lot of mew friends.
Oh nice I didn’t even think about that! That’s a good approach. I may have to return the extension risers I just ordered on eBay, this method would be cleaner if I can find the right ones but like you said I’ll rig something up somehow! LOL about the friends, how do you think I funded this entire build 
not a penny of my own cash, I was a tad mind blown that my boss was at the datacenter a little while back and give to a sudo friend (acquaintance really) some old stuff we decommed that had 80K worth of RAM in it. I almost threw up when he told me.
I scolded him I was like DUDE…DUDE!!! WHAT ABOUT ME!!!
Anyways, he says he knows of another “stash” of tons of em, I made it crystal clear to him don’t you dare give it to anyone except myself and him! I think he gets it now. I was literally depressed for a few days knowing I could have had 80K in ebay sales in the blink of an eye, that was when RAM was really peaking too. Its cooling down a bit now. Formatting some 2.5" SSD Sata drives to get them ready as part of the next “test phase”…also awaiting on some fans I screwed up I though one part of my case held 3x 140’s, nope it is 2x 140’s OR 3x 120’s. I learned that when I went to put on the third fan and it wouldn’t fit LOL. I was so angry at myself because I had stared at the case diagram over and over. Sigh. So I ordered up 3x 120 Noctura’s and also a fan controller module from them as I have 10 case fans plus the CPU fan (thinking about adding a 2nd 140mm Noctura fan to the Noctura cooler, cause well why not!. I may even hide a raspberry pi in there to collect environmental data at different points in the case and send that back to a grafana dashboard or something. Dunno. I haven’t seen the case actually kick in and do PWM but since its wide open it probably doesn’t need anything the idle speed on the cpu cooler is enough. I need to format that 2nd m2 NVMe that’s going onto the MB in slot m2_1 (pins moved to 2-3).
Anyone know or ever have used all the other connectors the MB has? It can do all sorts of stuff but I guess it needs the right device to talk back to it. I did order up a ASROCK branded TPM2.0 module from their list of QVLs. I’m scared of implosion of the system honestly with installing that. Never done that even in my sysadmin world. I saw it has some I2C and SPI connections..curious what some of them to. I’m a microcontroller nerd so I have all the lab grade equipment to speak I2C and SPI. But I’m also scared to break stuff so yeah probably best not to fuss with it. It’s like the bios. It’s at 3.8, I know there are newer versions and my sys admin brain says of course always keep your bios/firmware up to date, but when the Asrock website says if it’s not broke dont fix it. I take them at their word.
Always open to thoughts suggestions. Remember this is going to be XCP-NG hypervisor hosting a small batch of VM’s that will be controlled as much as possible by AI, as we have a lab at work on XCP-NG to try and prove to the leaders that be that we can get off VMware’s insane cost and model now. I love VMware don’t get me wrong (even worked there for 6 years right till the BC takeover when I bowed out on my own). So if we can automate and simulate some of the VMware stuff with AI, we will look like rock stars so that’s why I built this lab in the first place for work. Though its “mine” I’m sure I’ll end up buying a domain, setting up email and dns and ldap (no active directory / Windows allowed in this house!) and who knows what else. I’m hoping to grab another 1-2 TB of RAM in 64GB sticks to hold on to for spares ect.. if there is an insane amount well then yeah some might find their way to ebay. Having all this space in the MB almost makes me want to get a decent gfx card to run very small local models crazy fast and push the heavy work down to my spark. The case has room for 3.5" HDDs too. Thoughts on those? Are there some seriously fast enterprise NVMe drives I could hook up? I think I’m out of NVMe capability with the m2_1 getting another 980 pro. I don’t know about oculink, I’ve heard it can handle NVMe. I haven’t put research into it yet. Once I got the box access to my network I had my AI Hermes agent on my mac do all the work on those drives for me. Hermes is tied to the Spark for its engines. I have never used a “paid” AI model in my life. I live having privacy with my local models that are plenty big for all tasks except something really insane. Sure its not as good as Claude but still 120B models are sweet. Back to prepping sata drives!