@scaroo.bsky.social
Ok, I read the article, what I should have done before reacting. Lesson learned. Valve was indeed there at the beginning of FEX, and supported nearly from day one its developer.
December 2, 2025 at 8:48 PM
Ok, I read the article, what I should have done before reacting. Lesson learned. Valve was indeed there at the beginning of FEX, and supported nearly from day one its developer.
You can't overstate their impact, their contribs and investment really gave those projects a shot in the arm, sometimes rescuing them from being totally unusable, to be fair. But let's not miscredit anyone or pretend we don't see all the other folks and companies who actually laid the groundwork.
December 2, 2025 at 8:32 PM
You can't overstate their impact, their contribs and investment really gave those projects a shot in the arm, sometimes rescuing them from being totally unusable, to be fair. But let's not miscredit anyone or pretend we don't see all the other folks and companies who actually laid the groundwork.
From what I understand, FEX was already around long before Valve got into the picture with direct contributions or financial backing.
The same goes for Wine, DXVK, VK3D, and RADV.
The same goes for Wine, DXVK, VK3D, and RADV.
December 2, 2025 at 8:32 PM
From what I understand, FEX was already around long before Valve got into the picture with direct contributions or financial backing.
The same goes for Wine, DXVK, VK3D, and RADV.
The same goes for Wine, DXVK, VK3D, and RADV.
Pour que les gens y... viennent.
Ah mince, je ne suis pas sur le chat de CPC, ni ne suis oscar_tilage!
Ah mince, je ne suis pas sur le chat de CPC, ni ne suis oscar_tilage!
December 2, 2025 at 4:06 PM
Pour que les gens y... viennent.
Ah mince, je ne suis pas sur le chat de CPC, ni ne suis oscar_tilage!
Ah mince, je ne suis pas sur le chat de CPC, ni ne suis oscar_tilage!
Something to also remember is that most (all?) of their datacenters deployments for AI, and their own little machine, the Spark, run on Linux.
-- written from my 6900xt (because, yeah, AMD, Red Hat, Valve, Collabora and other companies/individuals are much, much better FLOSS citizen, still) ;)
-- written from my 6900xt (because, yeah, AMD, Red Hat, Valve, Collabora and other companies/individuals are much, much better FLOSS citizen, still) ;)
December 1, 2025 at 7:53 PM
Something to also remember is that most (all?) of their datacenters deployments for AI, and their own little machine, the Spark, run on Linux.
-- written from my 6900xt (because, yeah, AMD, Red Hat, Valve, Collabora and other companies/individuals are much, much better FLOSS citizen, still) ;)
-- written from my 6900xt (because, yeah, AMD, Red Hat, Valve, Collabora and other companies/individuals are much, much better FLOSS citizen, still) ;)
Those kinks are currently being worked out, but the performance debt still surfaces.
That's why, in detailed analyses (like those Gamers Nexus benchmarks), you still occasionally see weird frame-time inconsistencies or unexpected performance drops for Nvidia vs. AMD.
That's why, in detailed analyses (like those Gamers Nexus benchmarks), you still occasionally see weird frame-time inconsistencies or unexpected performance drops for Nvidia vs. AMD.
December 1, 2025 at 7:46 PM
Those kinks are currently being worked out, but the performance debt still surfaces.
That's why, in detailed analyses (like those Gamers Nexus benchmarks), you still occasionally see weird frame-time inconsistencies or unexpected performance drops for Nvidia vs. AMD.
That's why, in detailed analyses (like those Gamers Nexus benchmarks), you still occasionally see weird frame-time inconsistencies or unexpected performance drops for Nvidia vs. AMD.
While the open ecosystem built around shared standards (like GBM and Explicit Sync), Nvidia's display pipeline lagged behind.
This led to inefficiencies and outright incomp that have required huge engineering effort to solve, often by making the open stack accommodate the proprietary one. 😩
This led to inefficiencies and outright incomp that have required huge engineering effort to solve, often by making the open stack accommodate the proprietary one. 😩
December 1, 2025 at 7:45 PM
While the open ecosystem built around shared standards (like GBM and Explicit Sync), Nvidia's display pipeline lagged behind.
This led to inefficiencies and outright incomp that have required huge engineering effort to solve, often by making the open stack accommodate the proprietary one. 😩
This led to inefficiencies and outright incomp that have required huge engineering effort to solve, often by making the open stack accommodate the proprietary one. 😩
Despite the great progress in drivers, Nvidia's historical decision to "do their own thing" in their corner left a lot of proprietary baggage.
The cost of that isolation became painfully clear during the mass transition to Wayland.
The cost of that isolation became painfully clear during the mass transition to Wayland.
December 1, 2025 at 7:44 PM
Despite the great progress in drivers, Nvidia's historical decision to "do their own thing" in their corner left a lot of proprietary baggage.
The cost of that isolation became painfully clear during the mass transition to Wayland.
The cost of that isolation became painfully clear during the mass transition to Wayland.
In 2022, hell froze over. Nvidia released their official Open Kernel Modules.
Because the GSP handles the sensitive proprietary work, Nvidia could finally release a GPL-compliant kernel driver. It’s not fully upstream yet, but it allows legal integration with the Linux kernel for the first time.
Because the GSP handles the sensitive proprietary work, Nvidia could finally release a GPL-compliant kernel driver. It’s not fully upstream yet, but it allows legal integration with the Linux kernel for the first time.
December 1, 2025 at 7:40 PM
In 2022, hell froze over. Nvidia released their official Open Kernel Modules.
Because the GSP handles the sensitive proprietary work, Nvidia could finally release a GPL-compliant kernel driver. It’s not fully upstream yet, but it allows legal integration with the Linux kernel for the first time.
Because the GSP handles the sensitive proprietary work, Nvidia could finally release a GPL-compliant kernel driver. It’s not fully upstream yet, but it allows legal integration with the Linux kernel for the first time.
Later, they moved the "secret sauce" (power management, scheduling) out of the driver and into the GSP firmware.
The driver became a "dumb" messenger. Since the logic was now in hardware, the kernel driver could finally be Open Source without leaking IP
The driver became a "dumb" messenger. Since the logic was now in hardware, the kernel driver could finally be Open Source without leaking IP
December 1, 2025 at 7:39 PM
Later, they moved the "secret sauce" (power management, scheduling) out of the driver and into the GSP firmware.
The driver became a "dumb" messenger. Since the logic was now in hardware, the kernel driver could finally be Open Source without leaking IP
The driver became a "dumb" messenger. Since the logic was now in hardware, the kernel driver could finally be Open Source without leaking IP
This isolation is why things like Wayland were a nightmare on Nvidia for so long.
While the ecosystem coalesced around standard allocators (GBM), Nvidia tried to force their own path (EGLStreams) because they weren't part of the shared DRM/KMS infrastructure used by everyone else.
While the ecosystem coalesced around standard allocators (GBM), Nvidia tried to force their own path (EGLStreams) because they weren't part of the shared DRM/KMS infrastructure used by everyone else.
December 1, 2025 at 7:38 PM
This isolation is why things like Wayland were a nightmare on Nvidia for so long.
While the ecosystem coalesced around standard allocators (GBM), Nvidia tried to force their own path (EGLStreams) because they weren't part of the shared DRM/KMS infrastructure used by everyone else.
While the ecosystem coalesced around standard allocators (GBM), Nvidia tried to force their own path (EGLStreams) because they weren't part of the shared DRM/KMS infrastructure used by everyone else.
This is where the friction started. Intel and AMD played ball, contributing to shared GPU subsystems like DRM (Direct Rendering Manager) and KMS (Kernel Mode Setting).
Nvidia? They ignored those standard interfaces because their proprietary blob couldn't legally/technically integrate with them.
Nvidia? They ignored those standard interfaces because their proprietary blob couldn't legally/technically integrate with them.
December 1, 2025 at 7:37 PM
This is where the friction started. Intel and AMD played ball, contributing to shared GPU subsystems like DRM (Direct Rendering Manager) and KMS (Kernel Mode Setting).
Nvidia? They ignored those standard interfaces because their proprietary blob couldn't legally/technically integrate with them.
Nvidia? They ignored those standard interfaces because their proprietary blob couldn't legally/technically integrate with them.
Because the GPL disallows linking the kernel with closed-source code, Nvidia had to use a "shim."
They shipped a massive, proprietary blob that sat outside the kernel, communicating via a thin open-source wrapper. This technically worked, but it made them a legal and structural alien.
They shipped a massive, proprietary blob that sat outside the kernel, communicating via a thin open-source wrapper. This technically worked, but it made them a legal and structural alien.
December 1, 2025 at 7:37 PM
Because the GPL disallows linking the kernel with closed-source code, Nvidia had to use a "shim."
They shipped a massive, proprietary blob that sat outside the kernel, communicating via a thin open-source wrapper. This technically worked, but it made them a legal and structural alien.
They shipped a massive, proprietary blob that sat outside the kernel, communicating via a thin open-source wrapper. This technically worked, but it made them a legal and structural alien.
The issue was never technical incompetence; it was Licensing.
The Linux Kernel is GPL. Ideally, drivers are open-source and part of the upstream kernel. Nvidia refused to open their source to protect their IP (the "secret sauce" of their scheduler and memory management).
The Linux Kernel is GPL. Ideally, drivers are open-source and part of the upstream kernel. Nvidia refused to open their source to protect their IP (the "secret sauce" of their scheduler and memory management).
December 1, 2025 at 7:37 PM
The issue was never technical incompetence; it was Licensing.
The Linux Kernel is GPL. Ideally, drivers are open-source and part of the upstream kernel. Nvidia refused to open their source to protect their IP (the "secret sauce" of their scheduler and memory management).
The Linux Kernel is GPL. Ideally, drivers are open-source and part of the upstream kernel. Nvidia refused to open their source to protect their IP (the "secret sauce" of their scheduler and memory management).
If you needed professional-grade OpenGL or high-FPS gaming on Linux in the 2010s, you bought Nvidia. Period.
Their proprietary driver was parity-matched with Windows. They were often the first to implement new Vulkan extensions and pushed the API forward massively. The silicon was never the problem.
Their proprietary driver was parity-matched with Windows. They were often the first to implement new Vulkan extensions and pushed the API forward massively. The silicon was never the problem.
December 1, 2025 at 7:37 PM
If you needed professional-grade OpenGL or high-FPS gaming on Linux in the 2010s, you bought Nvidia. Period.
Their proprietary driver was parity-matched with Windows. They were often the first to implement new Vulkan extensions and pushed the API forward massively. The silicon was never the problem.
Their proprietary driver was parity-matched with Windows. They were often the first to implement new Vulkan extensions and pushed the API forward massively. The silicon was never the problem.
@dachsjaeger.bsky.social @digitalfoundry.bsky.social Hot take: Historically, Nvidia didn't have "bad" Linux support. In fact, for a long time, they were the only option for serious perf.
The narrative that Nvidia failed Linux is a misunderstanding of a legal standoff masquerading as a technical one.
The narrative that Nvidia failed Linux is a misunderstanding of a legal standoff masquerading as a technical one.
December 1, 2025 at 7:36 PM
@dachsjaeger.bsky.social @digitalfoundry.bsky.social Hot take: Historically, Nvidia didn't have "bad" Linux support. In fact, for a long time, they were the only option for serious perf.
The narrative that Nvidia failed Linux is a misunderstanding of a legal standoff masquerading as a technical one.
The narrative that Nvidia failed Linux is a misunderstanding of a legal standoff masquerading as a technical one.
Ah oui : Achetez sur GOG, vos jeux vous y appartiennent vraiment ;)
November 25, 2025 at 4:25 PM
Ah oui : Achetez sur GOG, vos jeux vous y appartiennent vraiment ;)
Et pour le joueur console ? Une PS5 est moins chère et offre le catalogue des gros jeux multi du moment. À la fin, c'est l'analyse produit/prix/besoin qui doit l'emporter, pas la sympathie pour le logo.
November 25, 2025 at 4:17 PM
Et pour le joueur console ? Une PS5 est moins chère et offre le catalogue des gros jeux multi du moment. À la fin, c'est l'analyse produit/prix/besoin qui doit l'emporter, pas la sympathie pour le logo.
Pour un joueur PC, mieux vaut se monter sa propre machine plus performante, le GPU et la VRAM y sont anémiques. Elle sera évolutive, et le coût d'usage réel (en revendant les pièces) pourrait être similaire, voire plus bas que le prix de la Machine.
November 25, 2025 at 4:17 PM
Pour un joueur PC, mieux vaut se monter sa propre machine plus performante, le GPU et la VRAM y sont anémiques. Elle sera évolutive, et le coût d'usage réel (en revendant les pièces) pourrait être similaire, voire plus bas que le prix de la Machine.
Et concrètement, est-ce que la Steam Machine (> 550€) est une si bonne affaire ? J'ai des doutes sur le rapport qualité-prix en général.
November 25, 2025 at 4:16 PM
Et concrètement, est-ce que la Steam Machine (> 550€) est une si bonne affaire ? J'ai des doutes sur le rapport qualité-prix en général.
Moralité : aucune boîte ne veut votre bien, ni votre mal d'ailleurs. Juger un produit selon l'image de "bon" ou "mauvais" samaritain de son fabricant, c'est se tirer une balle dans le pied.
November 25, 2025 at 4:14 PM
Moralité : aucune boîte ne veut votre bien, ni votre mal d'ailleurs. Juger un produit selon l'image de "bon" ou "mauvais" samaritain de son fabricant, c'est se tirer une balle dans le pied.
Ironie : c'est la politique de non-fermeture de l'écosystème Windows par MS pendant 40 ans qui a permis à Steam d'exister. Et rappel, Gabe Newell a fait sa première fortune chez... MS. Le monde est petit.
November 25, 2025 at 4:14 PM
Ironie : c'est la politique de non-fermeture de l'écosystème Windows par MS pendant 40 ans qui a permis à Steam d'exister. Et rappel, Gabe Newell a fait sa première fortune chez... MS. Le monde est petit.
Leur image de "boîte cool" (non cotée, organisation horizontale, grands jeux comme HL/Portal) crée un contraste avec le "béhémoth capitaliste" Microsoft. Ça aide à faire passer la pilule en surface.
November 25, 2025 at 4:14 PM
Leur image de "boîte cool" (non cotée, organisation horizontale, grands jeux comme HL/Portal) crée un contraste avec le "béhémoth capitaliste" Microsoft. Ça aide à faire passer la pilule en surface.
Je ne critique pas Valve, qui reste un très bon citoyen FLOSS par ses contributions concrètes. Mais je suis étonné par la couverture médiatique si complaisante. On dirait que tout le monde oublie le contexte business.
November 25, 2025 at 4:14 PM
Je ne critique pas Valve, qui reste un très bon citoyen FLOSS par ses contributions concrètes. Mais je suis étonné par la couverture médiatique si complaisante. On dirait que tout le monde oublie le contexte business.