With the R515 driver, NVIDIA released a set of Linux GPU kernel modules in May 2022 as open source with dual GPL and MIT licensing. The initial release targeted datacenter compute GPUs…
It’s more that back then was a better time for price to performance value. The 3000 and 4000 series cards were basically linear upgrades in terms of price to performance.
It’s an indicator that there haven’t been major innovations in the GPU space, besides perhaps the addition of the AI and Raytracing stuff, if you want to count those as upgrades.
RTX 3050 (which got a new 6 gb version less than a year ago) is similar to 1070 Ti in terms of performance and 1080s are of course even better. Definitely a ton of staying power, even in 2024.
I bought a secondhand 1080 a couple years ago when the crypto bubble burst finally and it’s still serving my needs just fine. It could play Baldur’s Gate 3 just fine on release last year, which was the last “new” game I played on it. Seems like it’ll still be good for a few years to come so yeah.
That was mostly because the 20 series was so bad. Expensive, didn’t perform lightyears better to justify the price, raytracing wasn’t used in any games (until recently).
The 30 series was supposed to be more of a return to form, then covid + mining ruined things.
I got a 2060 super and i must say i’m very happy, i do 3d stuff so the ray tracing was plenty useful and despite it getting a bit it fairs pretty great in most games and the price was okay at the time (500 €still a bit high since it was during the bitcoin mining madness =-=")
It’s not really planned obsolescence, they changed the way their drivers work with the 16xx/20xx series. Up till the 10xx series, they did a lot of algorithms and processing in the software. Then they switched to doing most of that on the GPU in the form of firmware. The 10 series GPUs can’t do that.
Like most hardware vendors, Nvidia doesn’t want (and probably isn’t allowed to) publish all of their special sauce source code. They can open source the driver and load a binary blob like, most hardware does, but only on the newer cards.
The older cards will have to keep the special sauce in software on the CPU, so those devices will need to stick to the proprietary driver.
I doubt very much it’s about whether they are allowed too or not. They’re the ones at the top of the hardware supply chain, designing their own chips and having them fabricated. It’s them telling other companies, like Gigabyte and EVGA, what they are allowed or not allowed to do.
Nvidia also buys and licenses code from other companies. These days they’re on top of the chain, but they used to be a lot smaller. Maybe they rewrote their drivers to remove the external code, but I wouldn’t be surprised if they still have old external code in their drivers.
AMD tried to open their source code for a display technique (VRR I think? Not sure what it was) but was prevented from doing so by the standards authority, presumably because they used licensed reference code. I don’t think this applies to the older 10xx series of cards, but these factors are difficult to work around.
Hmdi 2.1 and the hdmi consortium prevented them from releasing code. It wasn’t even proprietary, just based on a licensed implementation from what I understood.
So 20-series onwards.
My ol’ 1070 doesn’t make the cut hey… ;-;
Maybe it’s just because I’m older and more jaded, but that really feels like the last truly good era for GPUs.
Those 10 series cards had a ton of staying power, and the 480/580 were such damn good value cards.
It’s more that back then was a better time for price to performance value. The 3000 and 4000 series cards were basically linear upgrades in terms of price to performance.
It’s an indicator that there haven’t been major innovations in the GPU space, besides perhaps the addition of the AI and Raytracing stuff, if you want to count those as upgrades.
It feels like the crypto mining goldrush really changed the way GPU manufacturers view the market.
I feel like AI has changed the game. Why sell retail when people are paying you billions to run LLMs in the cloud.
RTX 3050 (which got a new 6 gb version less than a year ago) is similar to 1070 Ti in terms of performance and 1080s are of course even better. Definitely a ton of staying power, even in 2024.
I bought a secondhand 1080 a couple years ago when the crypto bubble burst finally and it’s still serving my needs just fine. It could play Baldur’s Gate 3 just fine on release last year, which was the last “new” game I played on it. Seems like it’ll still be good for a few years to come so yeah.
That was mostly because the 20 series was so bad. Expensive, didn’t perform lightyears better to justify the price, raytracing wasn’t used in any games (until recently).
The 30 series was supposed to be more of a return to form, then covid + mining ruined things.
I got a 2060 super and i must say i’m very happy, i do 3d stuff so the ray tracing was plenty useful and despite it getting a bit it fairs pretty great in most games and the price was okay at the time (500 €still a bit high since it was during the bitcoin mining madness =-=")
Still have a beautifully running 1070. 👌
Comrade. (☞ ͡° ͜ʖ ͡°)☞
I think it works but the performance might not be ideal. Keep on the proprietary module.
And 16xx
Yep! My pre-built 1660 super i got years ago is still chugging along amazingly as a streaming device for my steam deck.
Yes. Everything older is unsupported in terms of the new Linux stuff anymore. Planned obsolescence yk?
It’s not really planned obsolescence, they changed the way their drivers work with the 16xx/20xx series. Up till the 10xx series, they did a lot of algorithms and processing in the software. Then they switched to doing most of that on the GPU in the form of firmware. The 10 series GPUs can’t do that.
Like most hardware vendors, Nvidia doesn’t want (and probably isn’t allowed to) publish all of their special sauce source code. They can open source the driver and load a binary blob like, most hardware does, but only on the newer cards.
The older cards will have to keep the special sauce in software on the CPU, so those devices will need to stick to the proprietary driver.
I doubt very much it’s about whether they are allowed too or not. They’re the ones at the top of the hardware supply chain, designing their own chips and having them fabricated. It’s them telling other companies, like Gigabyte and EVGA, what they are allowed or not allowed to do.
Nvidia also buys and licenses code from other companies. These days they’re on top of the chain, but they used to be a lot smaller. Maybe they rewrote their drivers to remove the external code, but I wouldn’t be surprised if they still have old external code in their drivers.
AMD tried to open their source code for a display technique (VRR I think? Not sure what it was) but was prevented from doing so by the standards authority, presumably because they used licensed reference code. I don’t think this applies to the older 10xx series of cards, but these factors are difficult to work around.
Hmdi 2.1 and the hdmi consortium prevented them from releasing code. It wasn’t even proprietary, just based on a licensed implementation from what I understood.
Well their proprietary driver works fine for older hardware.