I wrote* a modern Linux driver that supports ultrawide resolutions, higher refresh, and faster performance than the original driver(s) on this cheap HDMI only variant server GPU.
I created it because I wanted my Nvidia GPUs to be compute only, and needed a cheap and small display GPU to run X11. I bought this GPU on AliExpress, after learning it had Linux support. It does, but kernel 5.x series and the sm750 driver in modern Linux only supports the older VGA/DVI variants.
Rather than send it back or throw it away, I decided to vibe code a new driver with new features, to make it usable for day to day desktop use. I've a little bit of a background in low level graphics so also added a cool little magic ordered dither (a pretty good one I think!) I devised myself, called "bbdither", and a few hacks to get 2560x1080 at 75hz on a device only officially documented as capable of up to 1920x1200 at 60hz, and PCI-E 1x speed.
Oh nice. I have a little PCIe/VGA one that I use for truenas and it's a bit flaky but I love that it's basically free from a PCIe perspective (just toss it off _any_ root). These are super useful and there are even m.2 variants.
tldr; had Qwen orchestrate, do admin and assist with investigations, builds and debugging when codex was unavailable. Codex had a dedicated test physical machine, and I promoted the driver it created to the actual target (production) machine when it was stable enough. Used syncthing to keep the code synced between the different environments, with Qwen sending me progress updates on Telegram and querying Codex periodically.
Long answer:
1) Ran the concept and investigation past Qwen 3.6, I wanted a card that would fit my spare PCIe 3 X1 socket (faster cards in other sockets). Identified the sm750, and read up on its capabilities.
2) Due to speed and Qwen 3.6's capability, I handed Codex (on high) a whole physical scrap machine / fresh Mint install, with root access and the GPU directly connected and asked it what was needed. Syncthing copied the work folder to my machine desktop for visibility to my main orchestrating Qwen AI. Qwen could also ssh to the Codex owned machine to check progress and send me Telegram messages when any milestones were passed, as well as issues that might need me there.
3) I Identified all the chips on the sm750, whilst it was trying the official older sm750 drivers (attempting to get a HDMI signal) and gave them to it.
4) Used /plan mode to set the first iteration expectations, no DRM/integration into the control panel, modifying a checked out fbdev low level FB to test my magic dither algorithm (this can only be used for test, the integration needs to go elsewhere in the stack typically).
5) Left it a few hours, came back to see colour bars (success!)
6) Asked it to do the basic plumbing for X11, wondered how far I could push the frequencies past the VESA standards, fried the HDMI TV attempting high frequency modes (only realised next day I'd done this after the TV refused to activate that connector)
7) Plugged the card directly into the machine it would be used in, and the actual ultrawide monitor. Resumed the codex session on that machine, when daily limits were hit used Qwen 3.6/3.8 to package up builds, look for solutions to issues I was hitting. At this point the hardware is my "production" system, so effectively I've promoted test to staging.
8) From here I manually iterated with Codex by having it create scripts that would build the module for the current kernel to restart X11 with 1-3 ideas / fixes at a time.
9) In parallel, used Qwen to prep the GitHub project, Codex and Qwen to do research for ideas on how to overclock the chips, feasibility of hacks I wanted investigating
10) Asked Codex and Qwen to identify bugs Vs specs, it found a few in event sequencing, resuming events, wrong bits in packets etc.
11) Switched Codex to Max after the the bugs were fixed because the driver performance was laggy, and recommendations from Qwen and Codex Medium/High for optimisations either caused degradations in some way (e.g. DMA was slower) and asked it to audit and optimise the code. I did this over 3 nights due to credit limits getting hit, again 1 to 3 changes at a time.
12) After I was happy with the quality ran the licencing checks again past Qwen and Codex and set up the project on GitHub, held private until I got workflows working.
It doesn't really do much in the way of processing (https://www.siliconmotion.com/download/3PS/a/SM750_PB_EN_201...); it's pretty much just a dumb framebuffer. You can wire it up to output over an HDMI connector, but with the features it supports you're really just getting a single-link DVI signal.
It is just an old school 2D style card, with some (shape) primitives acceleration (unused), two logical controllers (no idea why), but has DMA (used) and a hardware cursor (also used). There's no onboard audio either.
It's pretty much a GPU for rack mount servers that require occasional log in.
It's exactly what I was looking for desktop work, it runs cinnamon fairly well, and won't clash with my Nvidia setup :)
Yes. Stupid comment of mine tbh, I saw it as pointless for this card, but they wouldn't make a separate version of the controller just for this
One avenue I looked at for a real 2560px width was that the second controller was physically independent and could have different hardware limits of the first controller by virtue of supporting DVI directly (but not being physically connected to another port).
Unfortunately other controller is logical so that was a bust.
I agree but these things are only about $25 USD.
Arguably an older Nvidia GPU is a better investment at this price if you just want basic desktop use. Unfortunately if I did that, the old Nvidia drivers wouldn't support my newer Nvidia Blackwell hardware.
16 * 1024 * 1024 / 1920 / 1080 = 8.0909, so you have enough VRAM for double-buffered 24-bit 1080p, with a bit to spare. For pure software rendering, that's good enough!
VGA displays are something of a dying breed, and HDMI is the single likeliest interface you might find on a random newish display, so there is a market for absolute cheapest, most basic HDMI output device you can make. If you're going to build that interface, 1080p is the resolution you need to target. My point was simply that 16MB is just barely enough to build that interface, and barely enough is still enough.
datacenter guys didn't like digitals. they embrace blurry SVGA pictures coming through excessively long cables and sometimes mechanical switches, and to them DVI/HDMI felt to be an all-or-nothing risk factor
Possibly, but in order to be safe for everyone, reviewers may want me to remove features that allow the driver to achieve the non-standard performance. There's a lot of hoops to jump through :)
How did they react to it being vibe coded? Were they pragmatic about it, as Linus recently suggested
I have like 4 drivers to upstream myself, plus a couple of patches here and there. Quality is good and they're tested, and while I understand the code and how things work low level, I wouldn't be able to write them myself.
I've never tried it, I am put off by the amount of work for a few weekends on what I considered a personal project, plus I don't understand the code base well enough (problem domain is ok, I grew up with 8 and 16 Bit machines, and pushing graphics driver boundaries in Amiga and VGA monitor era machines). This is why my readme asks for help with issues :)
The dev on this driver used a combination of local Qwen 3.6 and 3.8 27B for admin and basic work, and extensive Codex 5.6 Sol Max to catch the issues I found and do last stage optimisations / isolate the bugs.
I'll freely admit I relied heavily on AI, my expertise and day job is in other software stacks.
I wrote* a modern Linux driver that supports ultrawide resolutions, higher refresh, and faster performance than the original driver(s) on this cheap HDMI only variant server GPU.
I created it because I wanted my Nvidia GPUs to be compute only, and needed a cheap and small display GPU to run X11. I bought this GPU on AliExpress, after learning it had Linux support. It does, but kernel 5.x series and the sm750 driver in modern Linux only supports the older VGA/DVI variants.
Rather than send it back or throw it away, I decided to vibe code a new driver with new features, to make it usable for day to day desktop use. I've a little bit of a background in low level graphics so also added a cool little magic ordered dither (a pretty good one I think!) I devised myself, called "bbdither", and a few hacks to get 2560x1080 at 75hz on a device only officially documented as capable of up to 1920x1200 at 60hz, and PCI-E 1x speed.
Screenshots here: https://github.com/KodeMunkie/sm750hdmifb#screenshots
If you try it out, be careful - read the project disclaimer :)
*Vibe coded, as mentioned in Hackernews comments here https://news.ycombinator.com/item?id=49415282
Oh nice. I have a little PCIe/VGA one that I use for truenas and it's a bit flaky but I love that it's basically free from a PCIe perspective (just toss it off _any_ root). These are super useful and there are even m.2 variants.
Ah you mean this https://www.pcworld.com/article/393815/the-asrock-rack-m2-vg...
That one is VGA so I think the original upstream driver should support it.
looks cool, and thanks for being upfront about the LLM-assisted part. It seems like a genuine project with a lot of time put into it!
Title fix proposal: Why vibecoding rocks
Nice. Is your vibecoded driver from scratch or does it bring the old driver forward ?
Hey, how did you iterate on it? Quite interested in your process here considering you say it’s vibe coded. The iteration loop is crucial for that.
Initially I just needed a driver for the device.
tldr; had Qwen orchestrate, do admin and assist with investigations, builds and debugging when codex was unavailable. Codex had a dedicated test physical machine, and I promoted the driver it created to the actual target (production) machine when it was stable enough. Used syncthing to keep the code synced between the different environments, with Qwen sending me progress updates on Telegram and querying Codex periodically.
Long answer:
1) Ran the concept and investigation past Qwen 3.6, I wanted a card that would fit my spare PCIe 3 X1 socket (faster cards in other sockets). Identified the sm750, and read up on its capabilities.
2) Due to speed and Qwen 3.6's capability, I handed Codex (on high) a whole physical scrap machine / fresh Mint install, with root access and the GPU directly connected and asked it what was needed. Syncthing copied the work folder to my machine desktop for visibility to my main orchestrating Qwen AI. Qwen could also ssh to the Codex owned machine to check progress and send me Telegram messages when any milestones were passed, as well as issues that might need me there.
3) I Identified all the chips on the sm750, whilst it was trying the official older sm750 drivers (attempting to get a HDMI signal) and gave them to it.
4) Used /plan mode to set the first iteration expectations, no DRM/integration into the control panel, modifying a checked out fbdev low level FB to test my magic dither algorithm (this can only be used for test, the integration needs to go elsewhere in the stack typically).
5) Left it a few hours, came back to see colour bars (success!)
6) Asked it to do the basic plumbing for X11, wondered how far I could push the frequencies past the VESA standards, fried the HDMI TV attempting high frequency modes (only realised next day I'd done this after the TV refused to activate that connector)
7) Plugged the card directly into the machine it would be used in, and the actual ultrawide monitor. Resumed the codex session on that machine, when daily limits were hit used Qwen 3.6/3.8 to package up builds, look for solutions to issues I was hitting. At this point the hardware is my "production" system, so effectively I've promoted test to staging.
8) From here I manually iterated with Codex by having it create scripts that would build the module for the current kernel to restart X11 with 1-3 ideas / fixes at a time.
9) In parallel, used Qwen to prep the GitHub project, Codex and Qwen to do research for ideas on how to overclock the chips, feasibility of hacks I wanted investigating
10) Asked Codex and Qwen to identify bugs Vs specs, it found a few in event sequencing, resuming events, wrong bits in packets etc.
11) Switched Codex to Max after the the bugs were fixed because the driver performance was laggy, and recommendations from Qwen and Codex Medium/High for optimisations either caused degradations in some way (e.g. DMA was slower) and asked it to audit and optimise the code. I did this over 3 nights due to credit limits getting hit, again 1 to 3 changes at a time.
12) After I was happy with the quality ran the licencing checks again past Qwen and Codex and set up the project on GitHub, held private until I got workflows working.
Excellent setup and I think your write up would be pretty good.
A GPU with HDMI and 16MB of VRAM is a very strange combination.
It doesn't really do much in the way of processing (https://www.siliconmotion.com/download/3PS/a/SM750_PB_EN_201...); it's pretty much just a dumb framebuffer. You can wire it up to output over an HDMI connector, but with the features it supports you're really just getting a single-link DVI signal.
It is just an old school 2D style card, with some (shape) primitives acceleration (unused), two logical controllers (no idea why), but has DMA (used) and a hardware cursor (also used). There's no onboard audio either. It's pretty much a GPU for rack mount servers that require occasional log in. It's exactly what I was looking for desktop work, it runs cinnamon fairly well, and won't clash with my Nvidia setup :)
two logical controllers (no idea why)
For dual-monitor versions?
Yes. Stupid comment of mine tbh, I saw it as pointless for this card, but they wouldn't make a separate version of the controller just for this
One avenue I looked at for a real 2560px width was that the second controller was physically independent and could have different hardware limits of the first controller by virtue of supporting DVI directly (but not being physically connected to another port).
Unfortunately other controller is logical so that was a bust.
I agree but these things are only about $25 USD. Arguably an older Nvidia GPU is a better investment at this price if you just want basic desktop use. Unfortunately if I did that, the old Nvidia drivers wouldn't support my newer Nvidia Blackwell hardware.
old nvidia cards are pretty terrible actually on linux due to the proprietary drivers not working on mainline kernel after ~6.6
Maybe an LLM can also assist with modifying the Nvidia driver appropriately.
16 * 1024 * 1024 / 1920 / 1080 = 8.0909, so you have enough VRAM for double-buffered 24-bit 1080p, with a bit to spare. For pure software rendering, that's good enough!
It's the HDMI output that's the oddity; mid-90s cards with that much (little?) VRAM could already do 2048x1536x32, but usually had a VGA connector.
Yeah, I could've been clearer.
VGA displays are something of a dying breed, and HDMI is the single likeliest interface you might find on a random newish display, so there is a market for absolute cheapest, most basic HDMI output device you can make. If you're going to build that interface, 1080p is the resolution you need to target. My point was simply that 16MB is just barely enough to build that interface, and barely enough is still enough.
datacenter guys didn't like digitals. they embrace blurry SVGA pictures coming through excessively long cables and sometimes mechanical switches, and to them DVI/HDMI felt to be an all-or-nothing risk factor
3840 x 2160[px] x 24[bit/px] / 8[bit/byte] = 24883200[bytes]
^ / 1024^2[bytes/MB] = 23.73[MB]
Proprietary drivers for obscure hardware were the bane of my existence. Glad to see this SM750 driver open-sourced.
Hmmm... How about adding it to the Free Software Foundation's "respects your freedom" hardware list?
What kind of motherboard has no onboard iGPU these days?
Another piece of silicon saved from the landfill! Open source keeps hardware relevant way past its commercial lifecycle.
Will you upstream it?
Possibly, but in order to be safe for everyone, reviewers may want me to remove features that allow the driver to achieve the non-standard performance. There's a lot of hoops to jump through :)
How did they react to it being vibe coded? Were they pragmatic about it, as Linus recently suggested
I have like 4 drivers to upstream myself, plus a couple of patches here and there. Quality is good and they're tested, and while I understand the code and how things work low level, I wouldn't be able to write them myself.
I've never tried it, I am put off by the amount of work for a few weekends on what I considered a personal project, plus I don't understand the code base well enough (problem domain is ok, I grew up with 8 and 16 Bit machines, and pushing graphics driver boundaries in Amiga and VGA monitor era machines). This is why my readme asks for help with issues :)
The dev on this driver used a combination of local Qwen 3.6 and 3.8 27B for admin and basic work, and extensive Codex 5.6 Sol Max to catch the issues I found and do last stage optimisations / isolate the bugs.
I'll freely admit I relied heavily on AI, my expertise and day job is in other software stacks.
I stand on the back of giants :)