The amount of power these use is less than what was used just a couple of years ago, and the next generation will use less power too, so you'll eventually have the computer power required for chatgpt in your desktop.
@@pcmason you already can have it with GPT4ALL. It can run on your cpu or use vulkan, opencl, cuda or the M series of the macbooks and it is pretty fast if you have a gpu with 6 or more gb of ram
@@pcmason While you are definetly not wrong we still have the problem that the usecases for these machines are getting better and better with better performance, leading to a massive growth in usage and thus increase in energy consumption in total. The efficiency increases we have these days are way smaller than the increases in usage overall. Data center energy consumption in projected to double by 2030. The current trend to "daily" AI usage and more compute intense applications being simpler to run is heavily counteracting any improvements we made towards efficiency. Using ChatGPT instead of Google for example uses (up to) 25 times the energy, no matter how efficient the servers are. So tldr, you are both not wrong, but also not right.
1:45 that Nvidia employee forgot she was working with linus. You got to stay on your toes.... And just a few seconds later you realize it's Yvonne, the GOAT of guiding Linus. Here's to Yvonne, saving the company one small catch at a time.
@@gonthyalavishalchandan1193 "is *that* John?" "yes, *it* is" That's a common type of exchange, most people don't wanna be referred to with *it* or *that,* but it's common to use those pronouns in this specific case, I'm sure no disrespect was meant
@@DajovaI mean I would want a case with actual cooking so a radiator case that's just a GPU that I could hook my water loop into the case so maybe we could have a chance to watercolor the RTX 50 cards
@@bluflame5381 This I dream about powerful PC yet I probably would just still stream RU-vid half the time and play Elden Ring/Battlefield and Cyberpunk the other half... Which runs fine on my 4060.
i wonder how powerful it'd be if nvidia could actually build a single gpu which is built out of like 1000 blackwell superchips that would fill up multiple rooms. And i wonder what we'd be able to do with such thing...
I mean, almost? These things are meant for datacenters, so you'd have dozens or hundreds of racks like these, all connected with extremely fast interconnects. That's almost like having a GPU the size of a building.
The amount of power data centers are already consuming is absolutely egregious, and this is going to almost double current usage rates. RIP to the glaciers and Florida💀
We got DX12 and Vulcan...... That either no one bothers trying to use, or no dev bothers doing the "less work than back then" to enable..... Gimme back SLI / Crossfire damit!
@@51Archives And that is what we want, our downfall, after all we all have to die eventually, if we die as the last humans in hisotry then we are special.
@ZaHandle Just use those industrial red power plugs. Youll need at least one of those for the rig if not more. That puppy can supply over 300,000 watts. Psu? Who needs that crap just plug it straight into a nuclear reactor. The rack likes raw unfiltered energy kappa.
I mean,... you would have to probably cover the whole world with 'super' computers of old (Zuse anyone?) and wouldn't have a faction of calculation power of one of these bad boys
bro boutta shutdown an entire house with 1 of those 1k Watts, imagine having multiple of those even your "free energy, solar panels cant take it" exept you have 10 100 Watts ones this shiz is crazy.
I actually work for a hyperscale ODM in QC and testing and we're renovating one of our rooms at the moment for building scale liquid cooling. Am excite.
@@Safetytrousers Sure, and if all you do is play games with the exact same framerate and graphics settings as before then you'll use less power. But people won't do that.
because it didn't. you're either viewing this on a Nokia phone from 2004, or half blind. his hand would even dissapear every time he waved it in front of it
Having it go straight from the GPU, rather than being cloud based, is the real game changer. Programming the AI personality reminded me of how the holodeck is programmed in Star Trek.
@@frankhuurman3955 or better yet, pick nothingness. Maybe now u're not forced to use AI, but sooner or later u would be. At this point someone should amass a "no tech" campaign, since pretty tech nowadays are the same -- making you rely on them.
*Segway* Two-wheeled, self-balancing personal vehicle. A Segway is a two-wheeled, self-balancing personal transporter device invented by Dean Kamen. It is a registered trademark of Segway Inc. It was brought to market in 2001 as the Segway HT, and then subsequently as the Segway PT. HT is an initialism for "human transporter" and PT for "personal transporter." A lesser known fact about Segway is that the company's previous owner died when he fell off a cliff while riding a Segway, kind of ironic. Those things aren't safe, let me tell you. Also, the word you're looking for is 'segue'.
Everyone and their mom knows the 50-series is launching this year. If you spent hundreds of dollars on an upgrade without researching that, it’s on you.
@@trapical Comprehension isn’t your strong suit. You literally took my words and made them mean what you wanted them to so you could argue. Try reading again… slower… and see if you come out with the same result. L
To Nvidia: This admittedly funny marketing stunt won't make me forget that your other GPU's are still overpriced. To Linus: BUY IT AND RUN THINGS ON IT.
He doesn't give a shit about you or any home user. He's got almost a laser focus on the datacenter. Sales of gaming GPUs are just the cherry on top of the cake for him.
8:24 Being a soon to be pharmacist, and looking at this is absolutely crazy to me. Drug development for specific proteins took years to DECADES of tests. And you are now telling me that these can be done in a matter of hours? If they manage to simulate in vivo clinical trials I'll be out of a job before I even begin my career lol.
Simple, MS Winblows used to run like ass on ARM and Nvidia is too dumb for everything else (and burned all goodwill so no one else will do it for 'em).
I mean they have the Shield as a standalone product, along with their Jetson series to some extent. As for selling bare CPUs, there is no standard platform yet on the market to my knowledge. So you end up with proprietary junk on Apple side for consumer Arm, and more proprietary junk on the datacenter space from Nvidia. AMD and Intel are in no hurry to move away from x86 yet. Nvidia and Apple both love money so I don't see how Arm PCs are gonna happen anytime soon. Kind of a shame.
They are going to, in a partnership with MediaTek, it's supposed to be a snapdragon X elite like competitor. Is expected around end of year or 2025 (don't quote me on these numbers, I don't remember it too well)
NVIDIA networking makes me remember when NVIDIA firewall was installed by default with one of my GPUs back in the day that would block absolutely all traffic by default 😂
@@Viewer13128 Neuro will make jokes about whats happening in game, she insults peoples gaming set ups within moments. shes a personal buddy to the creator, and hes just one guy. before chat gbt was even a thing
@@novadestry i was mainly wondering if the ai in this presentation are using more advanced stuff or not. for example the clip where the ai can react to moving game footage. i'm not sure what else the ai does so i was wondering if neuro is indeed doing everything it can do and more. i was just wondering how an entire company with lots of budget still can't make something equal/better than 1 person. that would be quite shocking and they really need to go back to school.
Can wait for the future generations to see this video and be like "Oh yeah, 8 TB/s. That's roughly the transfer speed of those cheaper USB flash drives... What a slow memory those guys head."
Cost is about $70,000 for that CPU + 2 GPU board that plugs into the blade and you need 2 per blade. You are looking at about $3 million per server rack. Do you see why NVIDIA's stock is so high?