Good thing nobody gives a shit. If I wanna translate CUDA then I’ll do it.
Nvidia can simply make faster products with more VRAM if they really care.
GanzNa
I don’t know what this means
Cross_Whales
Is this something related to ZLUDA?
CyberWeirdo420
No idea what this means
Alternative-Doubt452
Edit: the below reply is my analysis after working in virtual compute, high performance compute at the enterprise level, NOT gpt’s response. I asked got for background understanding at the hardware level to make my estimation on its use cases for consumer vs enterprise systems use cases.
But thanks for saying I sound dumb, good to know I’m still paid for my knowledge at architect level.
Since i don’t typically dive into the memory allocation theory I asked gpt about it and based on my understanding of infrastructure architecture and the current enterprise system shift i would say they did this for a reason retail/gaming wouldn’t understand or meet the reason to allow them. IE they did it to improve performance on memory uniform allocation and response time, but only for that reason potentially. Translation layers are beneficial for complex data not necessarily for gaming performance use cases. Which if that’s the case makes sense if they disable it for users at the consumer level to ensure a uniform behavior for all gamers. Whatever the reason, it’s most likely a corporate decision to prevent companies from using consumer systems for enterprise use cases forcing enterprise users to use the GPUs that don’t have these features locked out at a higher price point running different flashed firmware for the specific GPU they need. That’s my back of napkin understanding of the situation.
an_0w1
For anyone wondering what this means, NVIDIA says you cant run CUDA software on non NVIDIA GPUs.
Its a good thing I don’t need to accept a EULA to do it anyway.
SoDrunkRightNow2
NVIDIA is doing everything it possibly can to monopolize the industry.
Remember ray tracing? We were promised it was some kind of new tech for making better graphics. It turned out to basically be bullshit. Instead of actually useful technology, it was a way for NVIDIA to attempt to force consumers to use its hardware in order to play certain games.
This is the same type of thing. NVIDIA is saying certain software can only be used on its devices. They don’t want you to buy an Intel or AMD GPU.
thicc_toe
we should party like its 2023👹🦾
hodak2
Strange. If I buy an AMD card (I currently have one). Then I have no need or reason to accept their EULA….odd as it seems other than being unhappy there is little to no actual recourse for folks who likely have not agreed to their terms anyway. 😂
B3taWats0n
CUDA cores are vital for scientific research. This is why nvidia is the worst
MonteCrysto31
Are ROCm and ZLUDA safe or is Nshitia just shitting their pants here?
Zoltar-Wizdom
Did they use AI for his voice? It sounds like wish Keanu
-Retro-Kinetic-
I have to use a Nvidia for some of my production work, this feels more like by force than by choice. Trying to prevent competitors from using “translations layers” is just out right anti-competitive.
zanziTHEhero
This was a weird exchange. V, it’s not capitalism I’m at war with, but with the system of exploitation for the profit of corporations???
*insert they’re the same picture meme*
wsippel
It doesn’t even matter anymore. Zluda was the Plan B, in case developers wouldn’t adopt HIP. But pretty much every major project eventually did, and the bleeding edge stuff is starting to target MLIR, which is vendor-agnostic anyway. This decision only makes Nvidia look bad, and does nothing to solidify their fading monopoly. If anything, it makes project owners more likely to look into cross-API and MLIR.
amberoze
>Oh no!
>Anyway.
GermaneRiposte101
This is weird. Whilst walking the dog prior to seeing this post I was wondering if there were any CUDA translation layers. Never crossed my mind otherwise.
Now I guess the question is moot.
Thial92
Meanwhile AMD made FSR3 compatible with Nvidia cards. It was such a good company but over the last 6-7 years it has been a disaster. Gonna enjoy AMD products soon.
Philocrastination
This is one of the hardest monologues in recent gaming for me
jhankuP
Nvidia is anti competitive.
PerceptionQueasy3540
Makes me glad to have an AMD card. Nvidia has moved into shit territory
jefflukey123
Weird. I had this interaction after placide. Where is this? (I haven’t finished, no spoilers I beg of you)
FrostyFootLong
Does this mean if Im an nvidia fanboy I don’t have anything to worry about?
23 Comments
Good thing nobody gives a shit. If I wanna translate CUDA then I’ll do it.
Nvidia can simply make faster products with more VRAM if they really care.
I don’t know what this means
Is this something related to ZLUDA?
No idea what this means
Edit: the below reply is my analysis after working in virtual compute, high performance compute at the enterprise level, NOT gpt’s response. I asked got for background understanding at the hardware level to make my estimation on its use cases for consumer vs enterprise systems use cases.
But thanks for saying I sound dumb, good to know I’m still paid for my knowledge at architect level.
Since i don’t typically dive into the memory allocation theory I asked gpt about it and based on my understanding of infrastructure architecture and the current enterprise system shift i would say they did this for a reason retail/gaming wouldn’t understand or meet the reason to allow them. IE they did it to improve performance on memory uniform allocation and response time, but only for that reason potentially. Translation layers are beneficial for complex data not necessarily for gaming performance use cases. Which if that’s the case makes sense if they disable it for users at the consumer level to ensure a uniform behavior for all gamers. Whatever the reason, it’s most likely a corporate decision to prevent companies from using consumer systems for enterprise use cases forcing enterprise users to use the GPUs that don’t have these features locked out at a higher price point running different flashed firmware for the specific GPU they need. That’s my back of napkin understanding of the situation.
For anyone wondering what this means, NVIDIA says you cant run CUDA software on non NVIDIA GPUs.
Its a good thing I don’t need to accept a EULA to do it anyway.
NVIDIA is doing everything it possibly can to monopolize the industry.
Remember ray tracing? We were promised it was some kind of new tech for making better graphics. It turned out to basically be bullshit. Instead of actually useful technology, it was a way for NVIDIA to attempt to force consumers to use its hardware in order to play certain games.
This is the same type of thing. NVIDIA is saying certain software can only be used on its devices. They don’t want you to buy an Intel or AMD GPU.
we should party like its 2023👹🦾
Strange. If I buy an AMD card (I currently have one). Then I have no need or reason to accept their EULA….odd as it seems other than being unhappy there is little to no actual recourse for folks who likely have not agreed to their terms anyway. 😂
CUDA cores are vital for scientific research. This is why nvidia is the worst
Are ROCm and ZLUDA safe or is Nshitia just shitting their pants here?
Did they use AI for his voice? It sounds like wish Keanu
I have to use a Nvidia for some of my production work, this feels more like by force than by choice. Trying to prevent competitors from using “translations layers” is just out right anti-competitive.
This was a weird exchange. V, it’s not capitalism I’m at war with, but with the system of exploitation for the profit of corporations???
*insert they’re the same picture meme*
It doesn’t even matter anymore. Zluda was the Plan B, in case developers wouldn’t adopt HIP. But pretty much every major project eventually did, and the bleeding edge stuff is starting to target MLIR, which is vendor-agnostic anyway. This decision only makes Nvidia look bad, and does nothing to solidify their fading monopoly. If anything, it makes project owners more likely to look into cross-API and MLIR.
>Oh no!
>Anyway.
This is weird. Whilst walking the dog prior to seeing this post I was wondering if there were any CUDA translation layers. Never crossed my mind otherwise.
Now I guess the question is moot.
Meanwhile AMD made FSR3 compatible with Nvidia cards. It was such a good company but over the last 6-7 years it has been a disaster. Gonna enjoy AMD products soon.
This is one of the hardest monologues in recent gaming for me
Nvidia is anti competitive.
Makes me glad to have an AMD card. Nvidia has moved into shit territory
Weird. I had this interaction after placide. Where is this? (I haven’t finished, no spoilers I beg of you)
Does this mean if Im an nvidia fanboy I don’t have anything to worry about?