That's basically what you are doing when using any chatbot. ChatGPT is a bunch of server-side GPUs you talk to. lol

Like Me

It is already was with Jensen Huang

just crack an egg on your 4090

2026: RTseX.

@@Blue-bf8lv egg cooling system. Use protein for cooling your GPU

what languages ​​does nvidia support? chat with rtx. arabic

same lol

Same

I thought that there's no way it was the night of the 31st lmao, thought I went into a time jump or something

so how do you run AI and chat with your file local on your computer? There is no program

Visualize possessing a document spanning 1,000 pages.

upload your entire textbook then ask it the homework questions

It can read multiple books in a minutes, can you?

@@user-fe1nr3ws1z I don’t care what it can do if that isn’t useful 💀

Money

It is a weird omission since the 20 series has been getting crazy good 'AI' feature support through drivers to this day (RTX VSR/Video HDR, Ray Reconstruction, DLDSR). Someone on reddit theorized it's because the 20 series tensor cores lack bf16 floating point format support, which is specifically helpful for running LLMs like what Chat with RTX does. That doesn't mean the 20 series can't do Chat with RTX at all, since those cards still have TensorRT support. Nvidia might have to create a different model for them, which may have to be either slower or lower quality. It would be great if this would be cleared up by Nvidia themselves, like how they did for DLSS Frame generation, though this one app isn't as big of a deal as that.

Ai

People in this comment section reeks of clueless and baseless assumptions smh, It's because flash attention 2 doesn't support the Turing architecture. This means that it'd be extremely slow and less accurate. The 20 series also doesn't support CUDA 12.x versions. It's a gpu from 2018 dude what do you expect ?

In the Flash attention 2 requirements it doesn't say anything about Cuda 12 but Cuda 11.6 and they also mention that they are going to add support for Turing.

Same, are you living in the us ? Cause im not and i think thats why

I dont think it will since Turing doesn't support flash attention 2

No, 20 series is old news. My Commodore 64 and Atari 2600 won't run it either mate.

those are junk

Got no problem with mine.​@@Rebe-Caufman

@@brandonejem8620 20 series being old and only second gen tensor might have something to do with it, outdated and slow maybe

@@Fztrmkinda funny as the Nvidia tesla T4 is also a staple in AI and is based off the 2070's

20 series is like 20 years old 💀

lol

> "(and only docx pdf txt are supported)" Yep, and that's why it's useless. I got tons of data in excel files, Outlook, my calendar software, script files, coding files.... all data that this dumb thing wouldn't be able to use.

cmon, you have to pay more to nvidia

Cmon, bro. Spend some money and get lil better like vanilla 4070

Because its an architecture from 2018 plus it doesn't support flash attention 2 for god sake

Necessary since its a 7 Billion parameters

6GB is going to limit you very heavily in regards to AI as everything being released currently are 7 billion parameters and up. Unless you're fine with going at a snails pace, you can use opensource backends with CPU support to use your system RAM.

If you can mount it like a normal drive (SMB drive mapping) it should work.

Cloud ☁️ support enabled

Rtx is a global term for the hardware that enables raytracing, the RT cores and tenser cores

GTX cards dont have AI cores so wouldn't work even if they enabled it tbf.

I mean in this specific case you do need the AI cores and sadly gtx dont have those

No more support for GTX

GTX is obsolete 😂

get a job

Who will you switch to? Everyone will screw you over one way or another

@@Fztrm Yes you are right, but it is not just once or twice, it has been a few times and I am not only talking about marketing, but also about hardware. First a very expensive board with 790i Ultra Chipset, a lot of problems, thinking it was the board I changed it for an Asus Striker Extreme 2, same chipset, same problems, then they took out one more chipset and did not manufacture more. Then a 9800GX2, 600 Euros in its day, when the graphics were cheaper. The worst graphics card I have ever had in my life. Then a Gtx 480, it got to 90Cº with the fan at 100% in a simple game like l4d2. Then came the 970, with its 4Gb of which 512Mb were a joke, and then the fucking marketing with things like Physx that even with 2 graphic cards games like Mafia 2/Sacred 2/ Batman worked like ass. About SLI we don't even talk about it. Now with a red card, its streaming is garbage and RT is behind but when it comes to playing I'm quite happy. Edit: I understand that with such complex hardware there can always be problems, but when they are always the same ones you get tired of it. As you say, surely the same thing happens with the competition, but after so many times it is clear that I just have to change and try. Sorry if my English is not good i am using translator.

@@Nalguita AMD has screwed me over more than anyone, K6, Bulldozer, AM5 (pop goes the 7800X3D) because of lack of information from AMD's side....i still have a SECOND 7800X3D because it is simply the best gaming CPU there is now (And hopefully AMD learned their lesson and i won't have this one pop too) ...Intel screws over people regulary. It is all the same with every single one of them, not once, multiple times. They are all the same, whoever screws you the most is the ones that can afford to do it at the time.

@@Fztrm Yes, my friend, you are right. I speak from my experience, so far with Nvidia. That does not mean that AMD does not have problems, the first Radeon 7xxx with the Vapor Chamber, the first Ryzen 7000 with the bios the 5000 with problems of disconnections in the Usb. The same Intel now with the consumption and before spending years bringing out rehashes. But as I said I've been with Nvidia for years and I'm tired of their "lies" and the problems I've had with them, for me it was time to change. I have a 6800xt and a Ryzen 5000 and so far I am happy, I have no complaints about the graphics card, that does not mean that there are no people with them. About the Ryzen I could tell you that on the Asus website I have more than 22 versions of Bios, but since I bought it late I haven't had any serious problems. Best regards.

@@Nalguita I think everyone is in for a wild ride the next few years on all sides now with all the ray tracing and new features, many with hardware they thought "future proof" will be disappointed

It's because flash attention 2 doesn't support the Turing architecture. This means that it'd be extremely slow and less accurate. The 20 series also doesn't support CUDA 12.x versions. It's a gpu from 2018 dude what do you expect ?

look up why the nvidia stock is so high rn

In the museums?

No lol your 3050 laptop only has 4gb vram

@@Eleganttf2 what's the minimum vram required?

@@sameer52 8GB because the LLM model is worth 7 billion parameter, in laptop it means you must get atleast a laptop with an RTX 3070 or RTX 4060

​@@Eleganttf2what about 3060 12gb varient pc

@@DiluteOxygen its good