Nvidia-patch: removes restriction simultaneous video encoding sessions | Hacker News Reader

Nvidia-patch: removes restriction simultaneous video encoding sessions

github.com

86 pointsjordigh12 comments

Pretty clean shell script: https://github.com/keylase/nvidia-patch/blob/master/patch.sh

Surprised there are no protection/obfuscation methods against patching the `.so` in this fashion. No checksums, signatures, etc.

Give it time, the Bogeyman for this will be attestation and verified boot/data integrity+safety

Took them a while to sort out BIOS mods/signing firmware. The irons are in the other fires.

Not an endorsement, I got into computing because of the openness. I'm with Torvalds RE: nvidia

That's because the target market isn't going to risk running this kind of patch.

From my experience working on enterprise software, you just need something basic to curb the "oh, intern installed it and then we copied it to production" type behaviour. As soon as you have something that gets flagged to legal/managers you're good. Not worth going beyond that.

I see the limitation used to be 3, but was raised over a year to 8 for most modern GPUs:

https://videocardz.com/newz/nvdia-geforce-gpus-now-support-u...

Anyone know what sort of a difference "unlimited" NVENC streams means in practice for a 4090?

Found this reference to show the potential capabilities of each card:

https://www.elpamsoft.com/?p=Plex-Hardware-Transcoding

Seems a 4090 is capable of 22 concurrent h264 / av1 / hevc streams, quite a bit less than the default unlocked 8.

Is decoding limited then?

That link above talks about only 1x nvdec task. What does this mean in practice? Clearly it doesn't mean just decoding a single h264 stream??

So I think if I'm reading it right, it means there's 1x NVDEC chips. Each chip is capable of handling multiple encodings / decodings, as each of the transcodings list means concurrent streams of NVDEC -> NVENC.

What is the purpose of restricting it in the first place? Are there enough people out there streaming dozens of videos for Nvidia to care to milk them?

Upselling the few commercial users.

They want to push people who would use more than that to a more commercial grade GPU, which costs thousands more for the hardware and have a $450 driver.

There's a big jump in price at the very low end of their enterprise line for a not-very-large difference in performance from their high end gaming/consumer GPU.

So they do stuff like this, and explicitly blacklisting vGPU support (even if you pay the $450 price tag for their enterprise driver, you also need the hardware).

As soon as you start to scale GPU-intensive activities, you either have to overpay cloud services or tack on overhead for enterprise capability as an Nvidia tax.

In a normal society, nvidia would be sued

But we live in clowntopia, so we deserve this

There’s no artificial limit on AMD and Intel I think.

Proves that competition fixes these greedy nvidia artificially restricted capabilities.

If it’s up to 8 then you gotta wonder why bother have any limit at all.

Reply on news.ycombinator.com