The most prestigious law school admissions discussion board in the world.
|
 |
 
|
NVDA is on sale today bc of Chinese panic, just go buy it by the truckload.
Poast new message in this thread
Date: January 27th, 2025 9:54 AM Author: Sepia Skinny Woman Locus jfc no one else is close to them, no reason to panic (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48594243) |
 |
Date: January 27th, 2025 6:35 PM Author: Stimulating drunken pisswyrm potus (dances with raining knives) (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48596046) |
Date: January 27th, 2025 9:56 AM Author: Maniacal arousing kitchen i moved MDB premarket into NVDA and since then MDB is up over 5% while NVDA only .7%
IM 180 |
|
Date: January 27th, 2025 9:57 AM Author: Bull headed becky tattoo Explain why there’s no reason to panic (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48594252) |
|
Date: January 27th, 2025 9:59 AM Author: Lascivious White Laser Beams because Nvidia isn't who would be negatively affected by this development--you still need GPUs to run this shit and a lot of them. it's all the US tech cos pushing closed source models trying to pretend they have a moat who are the ones that may be in trouble (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48594256) |
|
Date: January 27th, 2025 10:02 AM Author: Sepia Skinny Woman Locus Correct. (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48594262) |
|
Date: January 27th, 2025 10:03 AM Author: Canary doctorate (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48594264) |
|
Date: January 27th, 2025 10:10 AM Author: Trip Painfully Honest Heaven Persian interesting (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48594287) |
|
Date: January 27th, 2025 6:39 PM Author: jet-lagged disgusting theater stage (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48596059) |
|
Date: January 27th, 2025 10:13 AM Author: twinkling associate https://x.com/satyanadella/status/1883753899255046301 (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48594298) |
|
Date: January 27th, 2025 1:19 PM Author: Electric bat-shit-crazy theater CR also after this there’s zero chance of Trump getting tricked or cucking on all the export restrictions that help NVIDIA more than any other company. OpenAI might be fucked, though, and it couldn’t have happened to a more deserving Jewish pedophile. (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48594903) |
|
Date: January 27th, 2025 1:30 PM Author: appetizing juggernaut round eye (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48594945) |
|
Date: January 27th, 2025 4:47 PM Author: Glittery reading party cuckoldry (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48595687) |
|
Date: January 27th, 2025 10:15 PM Author: boyish gas station giraffe how is deepseek performing like this if they're chinese though, isn't there a possibility that they DON'T have a ton of nvidia GPUs and figured out a cheaper/simpler way to operate? (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48596700) |
|
Date: January 27th, 2025 11:47 PM Author: aromatic library (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48596836) |
|
Date: January 27th, 2025 10:55 AM Author: Frum crimson turdskin As a big NVDA holder, this is the first time in a long while I am worried about holding the stock
If the DeepSeek architecture is what they say it is, you will need close to an order of magnitude less GPUs. You will still need NVDA, but instead of buying 1000 cards, you will only need like 50 to accomplish the same goal
Needless to say, given the sales growth expectations embedded in NVDA at their current price, this could be a catastrophic re-adjustment
If we see a major US player move to the DeepSeek architecture I would start to freak out |
|
Date: January 27th, 2025 1:16 PM Author: Sepia Skinny Woman Locus lmao okay, and then we have a 20x increase in demand and bam you're back to 1000. (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48594899) |
|
Date: January 27th, 2025 1:20 PM Author: Maniacal arousing kitchen (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48594905) |
|
Date: January 27th, 2025 1:22 PM Author: Electric bat-shit-crazy theater Yes. Also when have the Chinese not been at least 50% full of shit with this kind of big announcement that they are ackshhhually way ahead all of a sudden. But even to the extent that this is the medium-long term it’s like saying I don’t want to invest in cloud-optimized servers in 2012 because the advances in virtualization will eat all the demand, when the much more likely outcome at the time was public cloud growing bigger and faster. (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48594912) |
|
Date: January 27th, 2025 1:31 PM Author: appetizing juggernaut round eye exactly. i would think lower barrier to entry means more players. (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48594950) |
|
Date: January 27th, 2025 1:36 PM Author: exciting mood (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48594965) |
|
Date: January 27th, 2025 2:05 PM Author: Frum crimson turdskin It could be feasible that budgets remain the same, and players use the efficiency savings to simply scale more like you say. However, this could be (for a simple example) going from 1000 GPUs to 500 GPUs and still getting 10x the efficiency as previous architectures. Some players may view that as good enough given manpower limitations etc. There are just not that many companies out there with the capex budgets to be buying chips like this at scale, and those that do exist are already doing so.
And in that case NVDA could miss estimates by 50% down the line. Given it’s valuation multiple they can’t be anything short of perfect to keep rising
Fundamentally, this introduces a lot of uncertainty in trajectories which will cause some people to cash out and wait until it becomes a bit more clear
I am speaking about tactical investing here, and not long term buy and hold |
|
Date: March 6th, 2025 11:23 AM Author: Frum crimson turdskin You could have de-risked like I did, friend. Instead you laughed at me with comically surface level analysis (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48721257) |
|
Date: January 27th, 2025 1:38 PM Author: thriller cruise ship how is deepseek so efficient? (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48594974) |
|
Date: January 27th, 2025 2:10 PM Author: Frum crimson turdskin 8 bit vs 32 bit floating number calculations for a dramatic memory efficiency gain as well as 93% compression on key value indices which eat up VRAM. Also multi-token prediction which effectively doubles inference speed
The chinks basically had a GPU and Memory bottleneck due to sanctions and architected themselves out of it |
|
Date: January 27th, 2025 2:25 PM Author: thriller cruise ship is this computer engineering? (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48595156) |
|
Date: January 27th, 2025 2:44 PM Author: Frum crimson turdskin Software engineering (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48595223) |
|
Date: January 27th, 2025 2:57 PM Author: contagious abnormal kitty Are you an Engineer (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48595275) |
|
Date: January 27th, 2025 2:59 PM Author: Frum crimson turdskin no, I’m a tech investor (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48595283) |
|
Date: January 27th, 2025 10:16 PM Author: boyish gas station giraffe interesting ty (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48596706) |
|
Date: January 27th, 2025 2:52 PM Author: vibrant lilac church alpha What DeepSeek did could be done on the best chips now too. That could be way beyond what DeepSeek is capable of. We’ll see what unfolds. (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48595253) |
|
Date: January 27th, 2025 8:49 PM Author: Green useless brakes spot The marginal benefit you get from that may not be impressive relative to the cost. (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48596473) |
|
Date: January 27th, 2025 11:35 PM Author: self-centered step-uncle's house Maybe but I doubt it. We want multimodal LLMs that can do white collar work from end to end with no human intervention. This is only possible if they train on video and reinforcement learning tasks (games, or have them interact with desktop sessions). Improved training costs could make this feasible. (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48596827) |
|
Date: January 27th, 2025 1:29 PM Author: greedy startling athletic conference orchestra pit (Bitch bois rearranging chairs on the Titanic) (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48594942) |
|
Date: January 27th, 2025 1:34 PM Author: Electric bat-shit-crazy theater (Bidenbot showing off her new openly pro-CCP programming) (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48594960) |
|
Date: January 27th, 2025 1:35 PM Author: avocado halford dilemma yeah i bought $4500 (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48594962) |
|
Date: January 27th, 2025 1:37 PM Author: Maniacal arousing kitchen because of this faggot thread i sold my AMD and put it in NVDA
i also sold MDB and put it in NVDA, since then MDB was up as much as 8% while NVDA shares i got are down 5%
im suing if this thread ends up not accurate |
|
Date: January 27th, 2025 1:43 PM Author: Sepia Skinny Woman Locus Learn to use toilet paper first. (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48595003) |
|
Date: January 27th, 2025 1:54 PM Author: saffron rebellious wagecucks mediation Exactly. (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48595036) |
|
Date: January 27th, 2025 2:12 PM Author: Maniacal arousing kitchen even deepseek cant help me with this (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48595125) |
|
Date: January 27th, 2025 2:45 PM Author: Lascivious White Laser Beams in light of recent developments I've changed course and sold all NVDA -- actually just sold all US stocks. TT is the Jim Cramer of zozol so I have high confidence in this move. (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48595228) |
|
Date: January 27th, 2025 1:50 PM Author: Costumed Lay using a distilled version of R1 locally is like installing a retarded xo poaster.
to run deepseek-r1:671b you need NVIDIA chips - the market wants AI running locally not in China or Silicon Valley. |
|
Date: January 27th, 2025 2:01 PM Author: twinkling associate Agreed for companies, but I'm running the 14b version locally and its really good for its size. (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48595061) |
|
Date: January 27th, 2025 2:37 PM Author: appetizing juggernaut round eye what hardware are you running it on? how much space does it take up? (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48595194) |
|
Date: January 27th, 2025 2:48 PM Author: Lascivious White Laser Beams like 9gb? if you're running it using a GFX card (likely unless you have a couple mac minis laying around) you need at least 12GB of VRAM I think. it's decently fast on my 4080 (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48595239) |
|
Date: January 27th, 2025 2:53 PM Author: appetizing juggernaut round eye ty. i'm surprised that it only takes 9gb of training data to get a reasonable AI running locally. (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48595258) |
|
Date: January 27th, 2025 2:55 PM Author: twinkling associate Nvidia 4070 which has 12GB of VRAM, the 14b model takes up about 9GB (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48595267) |
|
Date: January 27th, 2025 2:45 PM Author: Henna brunch sweet tailpipe Good friend who works for them in this exact area is very bearish today & has been for years but their stock keeps going up so who knows (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48595225) |
|
Date: January 27th, 2025 2:45 PM Author: Supple National sold fuckloads of puts on this POS. that IMPLIED VOLATILITY is through the roof yo (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48595227) |
|
Date: January 27th, 2025 2:57 PM Author: laughsome deranged rehab herrrooo dis uhhh new chat GPT (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48595274) |
|
Date: January 27th, 2025 9:55 PM Author: Glittery reading party cuckoldry (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48596643) |
|
Date: March 6th, 2025 11:54 AM Author: Frum crimson turdskin (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48721352) |
|
Date: January 27th, 2025 3:53 PM Author: Hairraiser home So aren’t many AI models open sourced? If so why is it surprising that China can do it too? (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48595480) |
|
Date: January 27th, 2025 4:21 PM Author: vibrant lilac church alpha None of the major high quality ones are open sourced. Until now.
Also, it’s technically not open source, but the weights are available and the weights are like the recipe. |
|
Date: January 27th, 2025 4:46 PM Author: Lascivious White Laser Beams it isn't surprising at all. that's what's so crazy about this.
people are just flipping out because the team that happened to do it was in china, although this could have just as well happened in france or russia or meta's open source models or whatever |
|
Date: January 27th, 2025 4:56 PM Author: Frum crimson turdskin US AI models are closed source as the firms creating them want to retain the value of said models (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48595718) |
|
Date: January 27th, 2025 4:57 PM Author: Hairraiser home I thought Facebook was doing it open source (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48595721) |
|
Date: January 27th, 2025 4:59 PM Author: Glittery reading party cuckoldry they are (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48595725) |
|
Date: January 27th, 2025 5:35 PM Author: appetizing juggernaut round eye facebook and twitter are open source, i think. the rest aren't.
|
|
Date: January 27th, 2025 8:39 PM Author: Green useless brakes spot I don't think the problem is "closed source" so much as its the policing that's being done to make the LLMs "safe." Some of it seems reasonable. You don't want the LLM to give people instructions on how to make pipe bombs or cook meth. (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48596449) |
|
Date: January 27th, 2025 6:34 PM Author: self-centered step-uncle's house the 180 part is that DeepSeek wants to open source AGI. everyone will be able to have superhuman intelligent AGIs running on their own computers, with competing goals to acquire more resources and make money for their owners. (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48596043) |
|
Date: January 27th, 2025 6:38 PM Author: Trip Painfully Honest Heaven Persian AI seems pretty fucking stupid from what I’ve seen (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48596054) |
|
Date: January 27th, 2025 6:43 PM Author: jet-lagged disgusting theater stage I've trained it to write adequate responses to FROG 17.1, which is all I need it for. (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48596071) |
|
Date: January 28th, 2025 6:00 PM Author: Hairraiser home how does it know the basis for your denial (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48599141) |
|
Date: January 27th, 2025 8:33 PM Author: Costumed Lay May release…
NVIDIA Project DIGITS
A Grace Blackwell AI Supercomputer on your desk. Powered by the NVIDIA GB10 Grace Blackwell Superchip, Project DIGITS delivers a petaflop of AI performance in a power-efficient, compact form factor. With the NVIDIA AI software stack preinstalled and 128GB of memory, developers can prototype, fine-tune, and inference large AI models of up to 200B parameters locally, and seamlessly deploy to the data center or cloud.
$3000 each
$9k to run DeepSeek locally in each of SMBs and WFH mos who want to replace themselves
|
|
Date: January 27th, 2025 8:42 PM Author: Green useless brakes spot In theory you should be able to do this on just about any modern GPU. China probably used something more like NPUs. Nvidia's major advantage is the developer support it offers, not the GPU designs. (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48596456) |
|
Date: January 27th, 2025 8:56 PM Author: embarrassed to the bone provocative hominid crackhouse The real question is why would you ever believe China are telling the truth about this? (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48596486) |
|
Date: January 28th, 2025 8:15 PM Author: Odious regret (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48599467) |
|
Date: January 28th, 2025 12:25 AM Author: Spectacular center azn Would nvda be a buy at 115? (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48596872) |
|
Date: March 6th, 2025 11:27 AM Author: underhanded newt i just bought it all. there's no shares left. now we short squeeze. (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48721273) |
|
Date: February 26th, 2025 4:38 PM Author: Sepia Skinny Woman Locus https://www.wsj.com/business/earnings/nvidia-earnings-report-4q-2024-nvda-9de1a075 You're welcome. |
|
Date: March 6th, 2025 12:29 PM Author: Spectacular center azn So is it a buy at 110? (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48721442) |
|
Date: April 5th, 2025 5:30 PM Author: Floppy confused son of senegal (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#48819061) |
|
Date: August 6th, 2025 4:28 PM Author: lfo It's now 180 AND $180, brothers, I hope you weren't faggots like lots of the pumos ITT (http://www.autoadmit.com/thread.php?thread_id=5671126&forum_id=2:#49162701) |