Monday, June 26th 2023
More Pictures of NVIDIA's Cinder Block-sized RTX 4090 Ti Cooler Surface
Back in January, we got our first look at the cinder block-like 4-slot cooling solution of NVIDIA's upcoming flagship graphics card (called either the RTX 4090 Ti, or the TITAN (Ada). "ExperteVallah" on Twitter scored additional pictures of the cooler. Its design sees the heat dissipation surface pushed to the entire thickness of the cooler, and ventilated the entire length.
The card's PCB isn't conventional—not perpendicular to the plane of the motherboard like any other add-in card—but is rather along the plane of the motherboard, with additional breakaway daughter cards interfacing with the sole 12VHPWR power connector, and the PCIe slot. This slender, ruler-shaped PCB spans the entire length of the card, without coming in the way of its heat dissipation surfaces. The length is used for the large AD102 ASIC that's probably maxed out (with all its 144 SM enabled), twelve GDDR6X (possibly faster 23 Gbps), and a mammoth VRM that nearly maxes out the 600 W continuous power delivery design limit of the 12VHPWR.
Sources:
ExperteVallah (Twitter), Hassan Mujtaba (Twitter), VideoCardz
The card's PCB isn't conventional—not perpendicular to the plane of the motherboard like any other add-in card—but is rather along the plane of the motherboard, with additional breakaway daughter cards interfacing with the sole 12VHPWR power connector, and the PCIe slot. This slender, ruler-shaped PCB spans the entire length of the card, without coming in the way of its heat dissipation surfaces. The length is used for the large AD102 ASIC that's probably maxed out (with all its 144 SM enabled), twelve GDDR6X (possibly faster 23 Gbps), and a mammoth VRM that nearly maxes out the 600 W continuous power delivery design limit of the 12VHPWR.
145 Comments on More Pictures of NVIDIA's Cinder Block-sized RTX 4090 Ti Cooler Surface
One slot is 2cm, 4 slots just wasted.
That seems like an awfully small margin for error/excursions.
As I recall, both previous PCIe power plugs routinely went well above their spec'd rating.
<10% overhead is absolutely unacceptable in any AC Mains receptacle/plug, why is it okay here?
Sure, the power pins aren't an issue but the signaling pins might be, and may require a non-standard solution, especially if the creators of the standard were shortsighted enough and did't think that any GPU's consumption, ever, could exceed 0.6 kW.
The PC community has literally always been chasing the performance dragon. Remember clock doubling? The AMD 5x86? Bus OCing? It sounds like you are fundamentally unhappy. :mad: you still have 200w GPUs, in the form of cards like the 3060ti. You have the 3050. You can easily buy one of these cards, undervolt it by 5%, and have the uber efficient GPU of your dreams for not that much money.
If they took a 3060, and made it the 3090ti, would that make you happy? Or would you immediately be complaining that nvidia was sandbagging performance for the next generation? Something tells me you would be fundamentally unhappy with nvidia no matter what they did, unless they gave you 3090ti with a 3050 power envelope for $100, and even then, you'd be unhappy that it wasnt a 4090. Funny, people said the same thing with the 8800 ultra. Almost like innovation is scary. Why? The air cooler is relatively quiet, keeps temps in check, and is much easier to ship and install then an AIO based card. Most people dont want to deal with water cooling, so if you can manage with an air cooler, why even bother?
Also, this argument is totally stilted anyway. People always whine and cry about the power use, like you did with "Making something that eats 2x more power but is 4x faster is different from making something that's 2x more power hungry but is only 10% faster", but the GPUs we are talking about are the likes of the 4090/ti, you know, the GPU that is not only significantly faster then a 3090ti but also uses less power? Makes the argument fall on its face.
If we're being serious here, the entire 4000 series is much more efficient then the 3000 series. Your complaint holds no water. The 4060ti pulls less power then the 3060, or the 2060, while being as fast as a 3060ti or 3070. Same for the rest of the gen. As defenders of the 7600 love to point out, it draws 90w compared to 135w for the 6650xt.
It reinforces my argument that the people whining about power use are just that, whiners. They will never be happy, they have been given a generation that is significantly more efficient and all they can do is whine about how it still uses too much power for them, why isnt the 4090 a 150w GPU, blah blah blah, while simultaniously ALSO complaining that the 4000s dont provide enough generational uplift. Like WTF do you people want? The 4000s dont need undervolted like 3000s do, and yet, all we hear about is the 4000s pull too much power.
It's all so tiresome.
And the answer is simple: the delta between idle and load power consumption wasn't anywhere near it is today. Don't get me wrong, being efficient at idle is a good thing. All I'm saying is, we (at least in my area) were rocking noname 350-400 W office PSUs back then, while a 600 W quality unit is the minimum when you think about building an even slightly gaming-capable PC today. I don't know what people want. What I want is midrange cards to stay where they are/were in power consumption. I remember the 1060 being extremely efficient with its 120 W TDP, then the 2060 increased on it by a lot, and then the 3060 by even more. Of course the 40-series is efficient because it has the performance to compensate for power needs in the high end, and it seems to be improving on power consumption while not giving a lot more in performance in the lower segments, just like the 7600 doesn't. I definitely appreciate the move towards better power vs better performance, but I'm not sure it convinces everyone, and that's probably what the crying is about. The lower end doesn't improve on last gen in performance, while the higher end consumes enormous amounts of power.
The 2060 introduced RT cores and clocked noticeably higher. The 3060 offered a huge relative performance boost over the 2060, along with better RT, and was stuck on samsung's inferior node. The 4060ti offers nearly double the performance of a 2060 while pulling less power then a 2060. The high end has ALWAYS consumed enormous amounts of power, the limit was always the node, not the GPU. For some reason, people want those high end GPUs to be 150 w power sippers, against any and all reason.
It's lost that "apple" appeal.. ;)
They picked Samsung to save money, it was a choice.
Then you abandon the pure 60 class comparison to go for the 4060ti, a class higher and obviously also priced higher and using to compare power draw when we know the 4060 ti is a joke that had no evolution and barely beats the 3060ti and sometimes even loses to it. Using it as argument to discuss power draw evolution is absurd, it's a card that brings zero evolution
Oh and just wanted to add on here - it's funny you mentioned the 8800 Ultra...it was after that card (I had the 8800 GTX with a heavily modified cooler that did 8800 Ultra clocks but took up like 3 slots) that I started moving to dedicated water for both the CPU & GPU. I was tired of the insanity with the air cooler size, hah. See above. I understand your point and I'm not saying the cooler is poorly designed for what it does in terms of handling the heat + noise, only that when you have the cooler taking up 3-4 slots preventing usage of the southern hemisphere (lol) of your mainboard I'd say it wouldn't necessarily be the worst thing to look at other solutions because things are getting a bit out of hand.
I believe the 3090 FE cooler was 3-slots wide and as you noted, it's not a poorly designed device considering the job it does, but then I look at my own 3090 which is an ROG Strix with the front and back EK waterblock, and it doesn't even take up the full 2nd slot that is reserved by the block's mounting bracket. So yes, while some folks don't want to deal with watercooling, it could offer some nice advantages in that area that may offer some incentives - better cooling, even less noise, and will allow cards to not consume as much space.
IIRC, stuff like this is/was called 'Halo Tier' (as in: 'above the head of its class')
Why don't they just make an AIO Water Cooler for the 4090 Ti FE? Like with how wide that cooler is, i would actually be concerned about the GPU being heavy enough for the PCIE Slot on the motherboard to get torn off and fall