https://videocardz.com/newz/asrock-unveils-radeon-rx-9050-with-1024-cores-and-8gb-4gb-memory-optionsWhat great times we live in
>>109399798yay my rx 550 is relevant again
waste of sand, unironically
>>109399798let me guess, its more than the $200 mrsp of the 10 year old rx 470, which had 4gb
Good sign the AI bubble is popping
>8gb in today's day and age>there's also a 4gb that won't be able to play anything made in the last 11 yearsthe poors are holding us back yet again...
>>109400129It's a fucking entry level garbage card anyway.The GPU is more likely to hold it back than the VRAM.
>>109399798...why not get a cpu with decent igpu at this point?
Still rocking my rx 570 8GB, AI can go fuck itself
>>109400280That has pretty much been the case with "entry level" GPUs for a long time.Even with using system RAM, the iGPUs beat the crappiest of them.
>>109399798Recession indicator edition
>>109399798 I mean I'm playing the new grim dawn expansion on 1440P with ultrasettings and it's 2.1 GB of vram so I mean that's not that bad actually you just gotta play Older games
>>109400347>lower specs is fine just play older gamesWhile there are still plenty of older games, it is a poor excuse for paying more for less.
>>109400280I don't think an igpu is better. Don't get me wrong they're incredibly powerful now, I used to play l4d2 at 720p60 for a few months and it was fine, but even a shitty (at the time) 2015 gpu is miles ahead.
>>109400589Rather than get an ewaste 4gb vram gpu it's honestly better to stick to an igpu for a while after putting your build together until you can afford an actual gpu, it's just a waste of money to get something like this because even if it outperforms your igpu you're going to want to replace it sooner than later
>>109400597Oh you meant it that way, sorry. Then yeah I completely agree. But for OP wouldn't it be better to get a cpu without igpu? To get the absolute most compute per watt per dollar? I imagine the difference must not be very significative.
>>1093997984gb vram just like my gtx 980 12 years ago
>>109399798You don't need more. Anything above 1080p is slop anyways. Anything above 60Hz is probably fake gens unless it's 20 yo cybersport multiplayer game.>>109400321Not true. Although both are binned garbage, you cannot compete with descrete GPU cards, they are decades ahead of mobile and integrated nonsense.Unless you say you don't need more and either play with 30fps like people did for decades before samsung and lg started shilling 240Hz blurry nonsense. Or you just don't do gayming at all. Then it makes sense AND beats integrated trash.
>>109400280Because they're not only worse, they also cost more.
>>109399798I'd take if it goes below 180$
>>109400597It's for kids who'll have their parents buy them a "gaming computer" to play f2p unrealslop or trans indie games.>>109400156
>>109399798>I just want a new basic GPU to jerk off to anime on my coding machine>here you go only $275
>>109400761It won't.
An iGPU+16GB of ram will beat this trash.
>>109399798>64bit 144GB/swtf
>>109400993Not likely, unless you're talking about something like the Z2 Extreme or Strix Halo. Or BC-250.
>>109400280>zen6 removes igpu for muh AI
>>109401023AMD is due for new generation of iGPUs soon
>9060 XT 16GB
>>109401046From what I heard they're scrapping iGPUs entirely and making them into NPUs. For muh copilot AI.
>>109401073this
>>109399798still way better than the rx 6400 shitfest they launched or are they skipping the encoders here too lmao
>>109399798If it has 65W TDP, it would be a good buy.
>>109399798I just ordered a decade old low end card with 4GB of RAM to use on the host while I pass my beefy GPU to a VM.
>>109401075That seems kinda retardedigpus are useful for troubleshooting and shitI've never seen NPUs used for anything useful
>>109401190>I've never seen NPUs used for anything usefulMy Orange Pi has 6. I intend to use it to run a small LLM while freeing my CPU to other tasks
>>109401207Run a small LLM for what? The tiny 3B parameter models you can run on these SBC NPUs are absolutely useless.
>>109399798You VVILL pay $889 and you VVILL be happy with what you're getting goy
>>109401236It’s to help copilot and other AI apps installed on your PC
>>109401245>8GB DDR4People were making poverty builds with double this amount a decade ago.
>>109401245I think 8GB of ram is worse then 4GB of vram
>>109399798They are doing this because the rest of the 90 series is having a 20% price increase
>>109401257How? You mean use it like with VSCode's BYOM feature? That shit is a complete waste of time and power with any model below 27B parameters, and even then it's not worth it. You get nothing out of it.
>>109401236>Run a small LLM for what?Small tasks like summaries, some fun daily newsletter for a Telegram group I'm part of with online friends where response can be lazy. It works nicely. The problem is the setup, which if proves to be too much of a hassle, I'll use DeepSeek V4 since it's fairly cheap. >>109401257 is not me.
NPUs are an alternative to data centers. If every PC has one that OpenAI doesn’t need to make data centers because they can leverage your NPU
>>109401283It doesn't matter to me because my thirdie country is charging that extra 20% anyway, RAM crisis or not.
>>109399798I hope this gets revealed to actually be 3.5GB just to close the circle kek
>>109401491now you're getting an another 20% on top of that extra 20%
