bout to get a 5090what can i use it for?
You literally can’t. Unless if you’re doing real nigga shit you’ll just be stuck playing games goy. If you don’t know how to code why even waste the money on the hardware.
>>59654176minecraft (with mods)
I'll try to get one on the 30th when they go for sale but not sure I'll be quick enough to snag one online.
>>59654176Play Doom.
>>59654176Should I try scalping this shit?
>>59654176keeping your coffee mug hot
>>59654176generating AI furry porn locally on your machine
>>59654285The (credible) reviewers will say it's a waste of money so most of the gaymen crowd will not be as thirsty for these as they were for the 4090.
>>59654176If you have to ask it's not for you
>>59654176GB 10 makes RTX completely obsolete.For just 3k you get 128 GB ram, 1 petaflops of compute and extra 4 TB disk + 20 core cpu5090 only has 32 GB RAM. You've been scammed.
>>59654394That's 128 GB slow-ass shared system memory vs 32 GB bitchin' fast GDDR7.
>>59654394Vram is not regular RAM you retard.
>>59654434Nope it's not. You need as much RAM as possible for training and infering in batches.
ima wait for the 7090
>>59654176PDAI
>>59654434Lmao yeah that retard probably thinks using integrated graphics is better than an actual dedicated GPU cause his system has 64 or 128gb ram. Then wonders why whatever games his playing is always very slow and he's still on 720p in 2025.
>>59654462Stop spouting idiotic nonsense.
>>59654480I'm not referring to games. The GB 10 has 128 GB unified memory, which might be slower, but it's still much better to have more, because you can't even fit large models on 32 GB.
>>59654176are there any other places besides newegg that sells hardware for crypto
>>59654394what the fuck is gb10?
>>59654481for some workloads he's right tho, for ai inference people have been using mac minis or macbooks just because of the huge integrated memory, yes it's slow but you can fit larger models in your memory
>>59654176we need to know how much faster it is at stable diffusion porn generation. it's the only metric that matters.
>>59654753It would take ten minutes to get a simple generation from the largest LLM you can fit in there. That is abysmal and completely useless. These things are, at best, suited for educational purposes to teach children about AI.
>>59654953I think you're talking out of your ass.
>>59654964No it would unironically take ten minutes to spit out a 200 token output (text). Educate yourself.
>>59654176Generating fake frames, AI waifu shit.
>>59655001The whole selling point of project digits / GB10 is that it can run and train large models. Ask yourself the question why nvidia would want the RTX to outperform it in AI tasks. We don't know about the bandwidth yet, but the chip has more tflops.
>>59654176keep warm in winter
Counter strike 1.6
>>59654176pretending your graphics card is actually good when in reality its using fancy upscaling tricks and AI for frame interpolation instead of actually increasing performance.
>>59654176I'm getting one to heat my house. Should be nice and comfy, 500 watts of spaceheater bitches.>>59654704You haven't heard? Newegg is run by the Chinese now, it's all scams and broken shit. Seriously. If you pay with BTC, you can't do a chargeback, and don't ever expect to get a refund.
