I started out playing around with code generation using Ollama/open-webui and qwen 2.5 coder 14b on a 3060 12GB, but ended up on a winding journey with an ex datacenter card called the AMD V620. Its roughly equivalent to an RX 6800XT, but with double the VRAM.
At this point i’ve really done nothing productive with it but learned a lot about bios settings, GPU/ROCm drivers, and custom fan solutions/PWM controls trying to get it setup and optimized haha.
It’s pretty sick though, that amount of VRAM with 512GB/s bandwidth can run Qwen 3.6 27B dense with 100k context window at 20 tokens/sec in LM studio. Draws 300 watts at the wall on my ITX chassis (idling about 30w).
I’ve been dabbling in building an aviation weather and field condition report application using this, but my next step is to rebuild my VS Code environment into a new machine. I’m kinda enjoying just fucking around with building the hardware too though
If you are having trouble getting the 6800xt to work with pi.dev I’d be surprised if the V620 would be any different, but I haven’t tried that tool. I can attempt it and get back to ya in a couple days if you’d like.
I ended up getting it purely as it seemed like the cheapest option for 32GB VRAM that didnt have discontinued driver support. Around Jan/Feb 2026 the MI60’s had recently blown up in price but the V620 still seemed niche/slept on partially because AMD hasn’t released an SR-IOV driver for this. Servethehome forums had a big thread about how these aren’t particularly useful for home server/virtual machines as a result. I think it’s still possible to pass it through to docker containers but I haven’t tried it yet.
This guy accepted a $350 offer for mine: www.ebay.com/itm/157133307609
Then you’ll need a shroud: www.ebay.com/itm/286347509481
The optional included fan works well, pushes 60CFM but is LOUD. I ended up replacing it with an Arctic P8 Max which is much quieter but only pushes 40CFM, but cools it fine with -100mV undervolt in LACT.
I started out playing around with code generation using Ollama/open-webui and qwen 2.5 coder 14b on a 3060 12GB, but ended up on a winding journey with an ex datacenter card called the AMD V620. Its roughly equivalent to an RX 6800XT, but with double the VRAM. At this point i’ve really done nothing productive with it but learned a lot about bios settings, GPU/ROCm drivers, and custom fan solutions/PWM controls trying to get it setup and optimized haha.
It’s pretty sick though, that amount of VRAM with 512GB/s bandwidth can run Qwen 3.6 27B dense with 100k context window at 20 tokens/sec in LM studio. Draws 300 watts at the wall on my ITX chassis (idling about 30w).
I’ve been dabbling in building an aviation weather and field condition report application using this, but my next step is to rebuild my VS Code environment into a new machine. I’m kinda enjoying just fucking around with building the hardware too though
Oh…i recognise this sickness :)
I went down the same rabbit hole. I have a 6800xt however but have issues getting it to perform outside of llm chats into using tools like pi.dev
Is it worth getting a v620?
If you are having trouble getting the 6800xt to work with pi.dev I’d be surprised if the V620 would be any different, but I haven’t tried that tool. I can attempt it and get back to ya in a couple days if you’d like.
I ended up getting it purely as it seemed like the cheapest option for 32GB VRAM that didnt have discontinued driver support. Around Jan/Feb 2026 the MI60’s had recently blown up in price but the V620 still seemed niche/slept on partially because AMD hasn’t released an SR-IOV driver for this. Servethehome forums had a big thread about how these aren’t particularly useful for home server/virtual machines as a result. I think it’s still possible to pass it through to docker containers but I haven’t tried it yet.
This guy accepted a $350 offer for mine:
www.ebay.com/itm/157133307609
Then you’ll need a shroud:
www.ebay.com/itm/286347509481
The optional included fan works well, pushes 60CFM but is LOUD. I ended up replacing it with an Arctic P8 Max which is much quieter but only pushes 40CFM, but cools it fine with -100mV undervolt in LACT.