That is the greatness of it. It reminds me of democracy: The less powerful join together to give everyone an equal vote, rather than having one vote per dollar.
I mean, Cornelis Networks is trying with the resurrected OmniPath. I hope they pull it off, but I'm not holding my breath.
HPC cluster builds are complex enough due to the presence of multiple networks (2x moderate-scale infiniband chassis and 2x ethernet chassis as a minimum) without introducing unknown vendors. At that point, if you're doing IB, why not just go Mellanox since you will almost certainly buy 200GE Connectix NICs and not the 100G NICs from these guys.
UEC will - like most standards - _eventually_ work. In the mean time, Nvidia pods are the obvious choice for anyone who really cares about performance, and other vendors (Cisco, Arista) if they don't.
Modern HPC mostly means GPU compute and tons of data shuffling, but point taken. CPU bound jobs aren't going to stress the i/o, so you probably could have done 100GE for less. I'm curious what you did for storage but I'm guessing with CPU-bound that is again much less of an issue.
One of my greatest joys learning to program is reading old (well written) code, and piecing apart the timeless from the legacy.
Freesync wasn’t really market-ready until at least 2015 and the early products were mostly junk, it’s not until the gsync compatible program that any vendors really cared about LFC or flickering issues.
Generally speaking you get standards consortiums when there is a clear winner that is mopping up the space.
Here's an example that's happening right now: Nvidia-NvLink-Infiniband.
Nvidia owns the highspeed interconnect inside the chassis (HGX), the NICs (Mellanox), the inter-host interconnect (Infiniband), the high performance inter-host interconnect (switched NvLink), the and the ethernet network (Mellanox has the same 52.1Tbps switch performance that everyone else has now). GPU training is RDMA heavy and this is a place where both NvLink and Infiniband shine, ethernet much less so. Retransmissions are very bad, in global-performance-terms, for ROCEv2 transfers. Right now Nvidia is just crushing it and there's zero chance anyone is going to catch up by introducing new Infiniband ASICs.
So what happens? You have a consortium spun up by all of the companies in the Ethernet space - Ultra Ethernet Consortium - to try and use "standards" to push back on customers who don't want to make big investments in "non-standard." UEC is pretty vague but seems to be promising Broadcom-style cellized fabrics, the whole point of which is to have an ethernet-like standard that avoids ECMP-induced flow collisions and retransmissions - that is, get Ethernet into the same territory as Infiniband.
If you look back in time in the tech industry, you see this over and over and over and over. Standards are great, they make certain kinds of multi-sided markets and markets that need broad participation to be viable possible - but they are also routinely about the losers joining together to compete.