Damn
>>719284785That'd be great, if the TDP can be held at low levels, too. Like a 125 watt TDP, with 3D cache and enough RAM dedicated to it. RAM would likely be on the expensive side, with upward of 64GB and tight timings. But the savings on the GPU, power, and therefore cooling, would be huge
pricetag?what's the catch?
>>719285138won't support nvidia features needed to run some 2d indie games
>>719284785Medusa family isn't growing the IGP>>719285038Its not happening and MLID is a faggot
>>719285294>>Its not happening and MLID is a faggotWhat's MLID?
>>719285575"Moore's law is dead" a youtube/twitter fag who has a track record of being wrong more often than not.Even if this clickbait headline was true, and its not, it would be the replacement for Strix Halo. That is a chip exclusively sold in $1700+ laptops and mini PCs, the higher end of the market, and even the lower binned SKUs are not much cheaper.
>>719285754Oh, that guy. Last time I heard him talk about it, it was supposed to be 9060 non-XT levels, or 5060 levels. What made it jump to 5070?Also, doesn't AMD sell APUs anymore?I mean, an APU with 3D cache, paired up with high capacity, low timings RAM would be turn the GPU market upside down. Like, it'd still be in the $300US+ range, but you wouldn't need to spend $700US+ to get a GPU.
>>719285138APUs are always heavily bottlenecked by having to use system RAM instead of dedicated VRAM. Probably won't be able to throw it in a desktop motherboard like a normal processor either.
>>719286105>Also, doesn't AMD sell APUs anymore?Yeah, tons of them, its one of their main staples. An APU with 3D V-Cache probably isn't happening though. The added cost of this advanced packaging doesn't suit itself to what is fundamentally a laptop chip.The market competition is tight here. Nvidia has great performing parts, but their power consumption sucks hot dog shit. AMD can sell a system that pulls 120w max under full load CPU+GPU, but an Nvidia GPU might pull that on its own while performing the same or slightly better. As far as the average NPC consumer cares the Nvidia system is better.AMD is still selling tons of units and gaining market share though.
>>719286463AMD certainly sells a lot of chips to handheld manufacturers.
>>719284785I don't care, PS6 will be more powerful
>>719285196What Nvidia specific features do 2d indie games need?
>>719286591Not that many, they sell way more to random OEMs for laptops. The handheld/UMPC market is super small compared to some random Lenovo laptop.
>>719286370Heh
>>719284785non-engineer retard hereWhy are GPUs so big and bulky when iGPUs can get nearly the same performance on a tiny die?Why are GPUs so long and thin instead of square shaped like being concentrated in a tiny chip that can have a massive cooler attached to it like CPUs do?
>>719285138It won't be sold as a socketed APU, it's a halo laptop chip that needs special integrated RAM.
>>719286730The Jews.
>>719286730An IGP cannot get "nearly the same performance" unless you're comparing it against a GPU many generations older. There is no magic anywhere in this world. Xtors are xtors. Cache is cache. You compare different arch on ISO process and the differences totally evaporate into nothingness.
>>719286730The long and short of it is that GPUs are that small, and what you think of a "GPU" is a whole board with its own power components ram, processor cooler etc.
>>719286730>iGPUs can get nearly the same performancethey can't
>>719286730>Why are GPUs so big and bulky when iGPUs can get nearly the same performance on a tiny die?>Why are GPUs so long and thin instead of square shaped like being concentrated in a tiny chip that can have a massive cooler attached to it like CPUs do?"GPU"s as they're sold now are the chip, board, VRAM, cooler, and fans all sold as a single package. It's all a holdover from the fact that GPUs are still a "secondary" processor and are designed as add-in boards for a main board that's housing a CPU.Pic related is what the GPU die, RAM, and VRMs by themselves look like without the cooler on top.
>>719286697I see plenty of AMD laptops too. I'm just using that as proof of AMD being desirable for the low power draw gaming CPU market.
>>719284785Sounds like BS. The current IGPUs are only as powerful as 10 year old above midrange cards.
>>719286730I'm just wondering why they don't sell those iGPUs as low power dedicated cards for people looking to upgrade old systems anymore. I love things like the gt 1030 and r7 250 because of how small and low power draw they are while supporting modern (for the time) APIs and having a decent amount of VRAM (for the time).
>>719285138no memetracing and dlss smudgefilter
>>719286730It's mostly power and memory bandwidth.Imagine if you had to triple the power that has to be delivered to your CPU.It would be quite horrifying don't it?And on memory bandwidth, GPUs are surrounded by memory chips because every memory chip gets it's own exclusive lanes to the GPU, while on a CPU that is not as bandwidth hungry you can get away with shared vias between the memory sticks, and each memory stick share the lanes between every chip in em.
>>719286730When people say similar performance to a 5070 or whatever, they're usually referring to the M version. Also it's usually the lower wattage version.
>>719287254the "i" part is what makes them cheap. turning an iGPU into a dedicated card would make it as expensive as shitty low end GPUs. there's no free lunch.
>>719286730>Why are GPUs so big and bulky when iGPUs can get nearly the same performance on a tiny die?Only if you compare to midrange GPUs from 4-5 generations ago. And due to advancements in lithography and new tech such as DDR5 (which is still slow as fuck for VRAM, also shared, making it even slower)
>>719286370I'll take your entire stock
>>719287470They don't sell shitty low end GPUs anymore. The last low end card Nvidia made is the the GT 1630 and the last low end card AMD made is the RX 6400.
>>719284785Onboard video is actually good now which is nice but I like AI porn too much to give it up. I could've probably spent like $500 less on my current machine if I was just accounting for my gaming needs but being able to run image gen and LLMs is too fucking juicy
>>719286463I read that as >hot dog shit
>>719285138It’s going to be for the new steam home console
>>719287691Intel is moving into that market.
>>719285196>nVidia featuresYou mean like the AI bullshit you will never use, but convinced yourself you "need" it somehow?
>>7192876915050 is basically an overpriced 5030
>>719284785Let me guess AM5 MOBO, so I would need to build a new system completely
>>719287920And oversized and too power hungry. I want something I can slap in a single width half height slot in a ten year old office PC and be set.
>>719287847They still haven't made a proper low end card. They don't have anything properly small and low power draw. Only 750 TI equivalent stuff.
>>719287691AMD is making an RX 7400 with only 45W power draw but it's looking like it'll be an OEM only card. Sucks since it looks like it would be the perfect GT 1030 follow-up.
>>719288239I don't mind OEM only cards since they always end up flooding the used market anyways. As long as its actually card, and not just a laptop GPU. What's the RAM size on it?
>>719288078There are some low profile versions out there and I bet it's not that power hungry if you power limit, modern cards are only so power hungry because they are clocked way above their efficiency range (ever noticed how OC essentially died? They are essentially factory OC'd)
>>719287964Not even, it's a laptop chip (Medusa Point) that'll only be sold to OEMs. Best you might be able to get is if Framework makes a Desktop 2 so you can buy it as an ITX board.
>>719286370
>>719288427I want something that doesn't need any external power connectors since OEM PSUs never have them, and they usually aren't well built enough to handle the load of a power hungry GPU even if they're supposedly rated for it.
>>719288450Might as well buy a AsRock AMD BC-250 (mining board with a PS5 APU)
>>7192883158GB. Not great but it's not like you can get any better unless you step up to the 7700XT anyways.
>>719288575There are SATA to 8 pin adapters. And you are fine if you keep it under 80W, which is basically what the PCIe slot delivers anyway. Just power limit it
>>719288720All I need is the bare minimum. I held off on the 6400 because while it is RT capable, it lacks the VRAM to start and of the RT exclusive games. >>719288834I'm not chancing anything more that 50 Watts. Also many OEM PSUs don't have spare SATA power connectors either. They're made to order so its normal to see them have exactly enough for one HDD and one disc drive.
>>719284785No, it won't. You can't blast enough power through there, and you don't have fast enough memory access.
>>719289794with magic it will, so kys
>>719284785APU will never beat real dedicated cards unless they figure out vram
>>719285138It's not true
>>719290710Except it is. My grandfather worked for AND back in the 1940's and they were already planning this out from back then.
>>719286463you are a retard, like half this board, who spreads misinformation. the 5080 is better and uses less power than the 9070xt: https://www.techpowerup.com/review/sapphire-radeon-rx-9070-xt-pulse/41.htmlIn general, Nvidia is more efficient than AMD except for the premium 90 series cards which are top of the line powerhouses that AMD doesn't even compete against anymore.
>>719284785havent they been saying this since A8?
Steam deck 2? I'm very much looking forward to easily max out 7th gen games and a 4 hour in game battery life.
>>719289794>>719290657Strix Halo solves the iGPU VRAM issue by going the Apple route and using high speed integrated memory that's shared between the CPU/GPU sections. Both processors are connected to the same 256-bit memory bus that allows them to access the same memory pool at 200gbps. (For reference, the RTX 5060 mobile runs at 18gbps)
>>719285138In practice it won't beat a 4050 because it's severely bottlenecked by the anemic memory bandwidth, making the thing a waste of silicon.
>>719291102The current gen version is between the 4050 and 4060, it's surprisingly performant for being an APU.
