Haiku ❤ Nvidia (porting Nvidia driver for Turing+ GPUs)

Bro I am still confused how do you even install it

Actually, you can’t install it; it’s a prototype in development. Even when finished it may not appear in Haiku before a while.

2 Likes

bruh so is there any release date??

1 Like

Well, the port barely started, it’s a bit premature to talk about release dates and such, best arm yourself with some patience as it will take a while.

Best case scenario, once a good amount of issues are ironed out on his side X512 might need testers to try the driver on compatible cards.
In that case you could get the driver early, but still, that would take some time.

6 Likes

I still think GLTeapot could be used as a benchmark. But that number could have been faked (AI and Photoshop and whatnot, yaknow), so unless we see it in an actual video clip… :grin:

Oh, wait…AI can fake videos, now, too! Ok… um… how we gonna prove this is legit? :grin:

For GPUs, it can be only used as various components overhead and latency benchmark, not rendering performance benchmark. For actual rendering performance benchmark, more heavy and complex 3D scene is needed.

It is much easier, no Photoshop or AI are needed and video clip do not help with anything. Just adjust value to be displayed as FPS in source code (for example to 10005000) and you are done.

8 Likes

AH-HA! You admit it… it’s a fake number! Conjured up, just to make it seem like you were doing something extra special for the rest of us, when the actual number is probably something in the tens of digits… slower than a snail doing the backstroke, through chilled molassass! :rofl:

6 posts were split to a new topic: Kdenlive on Haiku

Ref: Mesa - latest version - on Haiku x86 (32 bit) and x86_64 (64 bit) - #14 by cocobean

I had a similar benchmark awhile ago… fastest with GLTeapot was about 2801 FPS.

More about the accuracy and quality of the rendering than just the speed…

Take note of the GPU encoder/decoder components too…

1 Like

Some 20 years ago, GeForce 4mx on BeOS, on an early driver version of mine. Indeed, a lot slower :wink:

19 Likes

You should have edited out the decimal point :wink:

2 Likes

That was great !!!. At some extent on which nvidia cards this work? I have gt200 gt520 and gt630 and and a gtx 780 and gtx 960 and even an agp on older system that i know it maybe works geforce 6200.

1 Like

All of those cards are too old for the new stuff X512 is working on, and too new for the old BeOS era stuff.

New - Turing architecture up so 16xx, 20xx and newer
Old - TNT, TNT2, Geforce 1 and Geforce 2 based cards (which includes some Geforce 4s)

Why my post is being edited? Topic title represents contrast with Linux, where Linus Torvalds showed complaints against Nvidia multiple times.

UPDATE: fixed and added Turing+ mention.

3 Likes

(post deleted by author)

It is original authors (me) name, please do not edit it without prior agreement.

7 Likes

Nvidia RTX A2000 (Ampere)

27 Likes

Multiple GPUs at the same time:

34 Likes

X512 is amazing :grin:

2 Likes

There is nothing inapropriate about kitsuneprefectures post. Please don’t report stuff just because you don’t agree with it. (Especially since you kept the edit and only slightly reworded it…)

2 Likes