I'm glad that the $250 price point has been borne out. This makes it reasonably attractive, and with Intel releasing this before the end of the holiday season there's a chance it will end up under some Christmas trees.
I will wait for reviews, but based on the specs it looks like an attractive card. We'll have to wait and see what AMD brings to the table.
I want to know die size. If the early rumors of 400mm2 on 4nm are true, this seems like a disaster. I just couldn't believe that.
I would hope for their sakes that it's under 280mm2, because Nvidia at this point could get us a 200-240mm2 die at this performance level no problem.
It matters to Intel, because it determines profitability. If they can't make a profit, and can't afford to lower prices, eventually they'll give up on GPUs. I think they recently even said something about only doing integrated graphics in the future. I don't think they are doing dedicated GPUs anymore for much longer, it seems from their recent comments.
Because if Intel goes under, AMD has no competition on CPUs. With the CEO being kicked out yesterday, I do wonder if they could go back to dedicated GPUs after bettering their drivers on integrated first. If they already build the architecture for integrated, maybe it's not too late to go back if things go well. Mostly I just don't want Intel to fail I want the to be good competition.
>Both the Arc B580 and B570 are based on the "BMG-G21" a new monolithic silicon built on the TSMC 5 nm EUV process node. The silicon has a die-area of 272 mm², and a transistor count of 19.6 billion.
From TPU. Solid but not fantastic. 4060 is \~40% smaller on the same node, 7600XT is 25% smaller on a larger node, but this is probably a hint quicker than those. Technically though the 4060 ti is faster still at 188 mm2.
Damn, I totally forgot the 4060ti was under 200mm2. I thought it was the 4060 that was around 188mm2. The question is how small Nvidia's next generation is, because I'd expect only improvements in that regard. This thing might be competing with the 20 sm GB207 in the RTX 5050ti in a few months.
What would you do with a 5050/even 5060 series card with 16gb besides very niche scenarios? 5050 for sure will not be powerful enough to even utilize 16gb. With the incredible drop off of the 5090-5080 I wouldn’t be surprised if Nvidia does what it always does with 60 series and less cards (besides the 3060ti), and makes it absolutely trash. I hope that 60 series cards running 2k with RT are powerful enough for gaming, but I severely doubt it. I’d bet on the 5070 being the absolute minimum for gaming on 2k (on modern games) with higher settings, as that’s the precedent set by Nvidia already. On another note, the leaks suggest the same style of lineup as the 40 series-meaning 5070ti garbage, 5080 being too expensive, and the 5090 being ridiculous (and ridiculously priced-even more so with the inclusion of gddr7). I have 0 hope for any performance/value outside of the xx70 series.
The answer is longevity. VRAM usage only rises with time, cards with more VRAM last longer. A 4060 with 12GB today could handle all 1080p ultra settings and can even play reasonably well at 1440p. 8gb limits what the 4060 can do more than it's compute power.
Honestly? 0 reason. And from a business perspective the 3060 12GB was a mistake. Nvidia should have stuck with the original 6GB plan and it still would have sold out pandemic and all. Had that been the case the 4060 would have been better in every way and a welcomed upgrade over the base 1060, 2060 and 3060.
There's rumors that Nvidia will increase the xx60 class VRAM later down the line when higher capacity chips are available. I still wouldn't do it unless it's flat out cheaper.
Yes and no. I understand what you’re saying-but with a 4060 ultra settings I doubt it would be powerful enough to utilize the ram. It really is dependent on texture settings at 8gb-but for example a 3070 could run it, while a 4060 is just too weak to be able to utilize the ram.
There’s no theoretical minimum to the speed required to run AI models (you can inference in CPUs if you’d like), but you do need a certain amount of RAM. Nvidia can sell a 16GB 5050 as an AI card; it’ll be the slowest one they sell, but it’ll beat a higher end 8GB/12GB card if the model you’re trying to run won’t fit in VRAM.
I mean people are buying the 7600xt which can't even get playable ray tracing frame rates, and yet has 16gb. RT does take an additional 1-2gb. Frame generation also takes around 1gb. No idea what other features Nvidia will bring next. 12gb certainly seems like a more suitable option for a GPU of this class. But if it's 5% faster than a 4060, that's around 20%-25% faster than a 3060 which has 12gb. 16gb would be a better option than 8gb at least.
I understand people buy these lower gpus-my counter argument is that if you’re willing to pay $200-300 for a gpu-why not save up for something that is vastly better? Unless you’re making an emulator rig, and have no interest in playing anything higher than 1080p, that persons patience will be highly rewarded. The drop off from both brands is crazy
I don't know. Some people I guess just save a long time just so they can be afford an RTX 4060, and a 4070 it's way too far away. The 4060ti seems like a worse deal than anything. A used 3080 is a risk that you might not be willing to take, and maybe your power supply can't take a 320w GPU either.
B580 192bit 12GB 10% faster on average vs 4060 128bit 8GB at 1440P ultra settings.
What about 1080P? Slower than 4060? 4060 is probably running out of vram in some games dragging 4060
yeah, it seems impressive but not as impressive when you realize the 4060 is due for an update soon and this would most likely get easily get overshadowed by that.
They're a few months earlier tho, this would probably still be a good buy for the people wanting to build a PC now
The [rumours](https://videocardz.com/newz/rumors-suggest-nvidia-could-launch-rtx-5070-in-february-rtx-5060-series-already-in-march) say that most of the RTX 50 cards will be available by March.
>RTX 5090/5090D/5080: January 2025
>RTX 5070/5070Ti: February 2025
>RTX 5060/5060Ti: March 2025
It's the first time since Maxwell that there is no new xx80 card release every two years.
GTX 980 - 2014
GTX 1080 - 2016
RTX 2080 - 2018
RTX 3080 - 2020
RTX 4080 - 2022
**RTX 5080 - 2025**
If the 5090/5080 were released in Q4 2024 the 5060/Ti in Q1 2025 would seem pretty normal.
For comparison:
B580: 272 mm2, 19.6B transistors
4060: 159 mm2, 18.9B transistors
4060 Ti: 188 mm2, 22.9B transistors
The B580 does have a 192-bit bus, which takes some extra die space compared to the 128-bit bus for the GeForce cards. Still, this indicates that Intel is behind in performance per mm2. The Intel chip costs significantly more to make than the 4060 Ti and yet will sell for significantly less.
The chip IS the GPU. Unless you mean the actual graphics board.
At scale the size of the chip dictates a big chunk of the cost of the final chip from both manufacturing; because larger dies require more expensive packaging, for example. As well as design; larger dies may require more validation effort.
So although some other factors may indeed affect the overall cost structure, where a smaller die ends up being a more expensive SKU for the manufacturer. It is very unlikely that at this point of the cycle NVIDIA is having a more expensive overall package (in terms of their gross costs) than the newer (larger) intel GPU competing at the same bracket.
If intel requires a much larger die to compete with a smaller previous gen NVIDIA GPU, they are in trouble from a margins perspective.
Okay, thanks. Makes sense, and it's not something we could really know.
I'd expect yields are mostly affected by chip size.
As for the GPU cost, I'd expect that for chips that aren't too power-hungry RAM cost will have the largest effect on price. Cooling and VRM are affected by power use. I'm not sure how much of a difference they make here between the 190W B580 and 160W 4060 Ti, but they probably save a little.
48 Comments
2080TiPULLZ450watts@reddit
No_Attitude_7779@reddit
johnshonz@reddit
1sh0t1b33r@reddit
ET3D@reddit
bubblesort33@reddit
Severe_Line_4723@reddit
vhailorx@reddit
bubblesort33@reddit
Severe_Line_4723@reddit
bubblesort33@reddit
Toojara@reddit
bubblesort33@reddit
Vb_33@reddit
bubblesort33@reddit
Subject_Gene2@reddit
Vb_33@reddit
slvrsmth@reddit
Vb_33@reddit
Subject_Gene2@reddit
ThankGodImBipolar@reddit
Strazdas1@reddit
Subject_Gene2@reddit
bubblesort33@reddit
Subject_Gene2@reddit
bubblesort33@reddit
Vb_33@reddit
excaliflop@reddit
NeroClaudius199907@reddit
salcedoge@reddit
InconspicuousRadish@reddit
dab1@reddit
InconspicuousRadish@reddit
zerinho6@reddit
dj_antares@reddit
NeroClaudius199907@reddit
conquer69@reddit
NeroClaudius199907@reddit
ET3D@reddit
Frexxia@reddit
ET3D@reddit
Frexxia@reddit
Adromedae@reddit
Frexxia@reddit
ET3D@reddit
bubblesort33@reddit
imaginary_num6er@reddit
logosuwu@reddit (OP)