- Joined
- Nov 14, 2012
A patron going by as "K" has donated $6000 on the condition that it be put towards hardware. His original idea was to buy a NVIDIA RTX 6000 ADA, which is roughly $6000 online, but such a device would not fit in our existing servers nor do we have a PSU that can support it.
My ultimate plan for the forum is a first step towards the XenForo software rewrite. I want to write a media processor that converts user uploads into playlists. Modern platforms like YouTube do not deliver .mp4 files or byte ranges of .mp4 files for various reasons, what they do is use playlists and a JavaScript player to send 2s chunks of videos. They also often include multiple formats at different resolutions. (To be clear, I do not want to get rid of downloading source files, but anyone who uses the forum knows how bad trying to scan through long videos is.)
After deliberating on what formats to use, I am tempted to go to 1080p in AV1 + 720/480/360p in MPEG. This means excluding a lot of cards which do not support that AV1 codec released in 2019.
I have a usual ensemble of people I trust on hardware decisions. I've gotten a diversity of input. I'm not sure what to do.
Guy 1. He suggests getting an Intel GPU, probably an Intel Arc.
Guy 2. Suggests a GPU server with an NVIDIA Tesla.
Guy 3. Suggests just using an Intel CPU because they have QuickSync and are very good at conversions and support AV1. Suggested an Intel Arc to support it.
Some other guy suggested a Xilinx U30 but those don't support AV1. With Av1 as a requirement he suggested an Intel Arc.
Claude also suggested Arcs on an AM5 board.
One of my pet memes has been a media processing card like from NETINT but I cannot find those surplus. I'm not sure if I want to actually hassle the company to get two of them. They run like $1500 each but can parallel process 32 1080p videos each.
There's also the thought that I could put it towards something that also supports AI for my security tools I asked about before. The original $6000 card he wanted to get could do this, but oddly despite AI originally being ran on GPUs, media processing and tensor cores are at odds with each other these days. I don't think it's possible to build a server that does both media and AI without just getting big boy NVIDIAs, spending a ton of money, and plugging it into a 250W outlet.
My ultimate plan for the forum is a first step towards the XenForo software rewrite. I want to write a media processor that converts user uploads into playlists. Modern platforms like YouTube do not deliver .mp4 files or byte ranges of .mp4 files for various reasons, what they do is use playlists and a JavaScript player to send 2s chunks of videos. They also often include multiple formats at different resolutions. (To be clear, I do not want to get rid of downloading source files, but anyone who uses the forum knows how bad trying to scan through long videos is.)
After deliberating on what formats to use, I am tempted to go to 1080p in AV1 + 720/480/360p in MPEG. This means excluding a lot of cards which do not support that AV1 codec released in 2019.
I have a usual ensemble of people I trust on hardware decisions. I've gotten a diversity of input. I'm not sure what to do.
Guy 1. He suggests getting an Intel GPU, probably an Intel Arc.
Guy 2. Suggests a GPU server with an NVIDIA Tesla.
Guy 3. Suggests just using an Intel CPU because they have QuickSync and are very good at conversions and support AV1. Suggested an Intel Arc to support it.
Some other guy suggested a Xilinx U30 but those don't support AV1. With Av1 as a requirement he suggested an Intel Arc.
Claude also suggested Arcs on an AM5 board.
One of my pet memes has been a media processing card like from NETINT but I cannot find those surplus. I'm not sure if I want to actually hassle the company to get two of them. They run like $1500 each but can parallel process 32 1080p videos each.
There's also the thought that I could put it towards something that also supports AI for my security tools I asked about before. The original $6000 card he wanted to get could do this, but oddly despite AI originally being ran on GPUs, media processing and tensor cores are at odds with each other these days. I don't think it's possible to build a server that does both media and AI without just getting big boy NVIDIAs, spending a ton of money, and plugging it into a 250W outlet.

