This one is a short one, because I wanted to chronicle it and found it interesting.
Over the course of this week, a series of photos purporting to be leaked from manufacturing partners for Nvidia claim to show the reference (or Founder’s Edition) design of the upcoming RTX 3080 GPUs.
These cards are pretty highly anticipated, because the virtual GTC event last month showed a non-raytracing professional big data version of the GPU with drastically higher rated performance from a small-ish increase in overall compute units.
A lot of people and rumors expected the cards to look like the RTX 20 series founder’s edition cards with 3 fans instead of the 20-series 2 fans. The same basic, attempted elegant design, but longer with more cooling capacity.
The reference design from a GPU manufacturer is an important factor for the launch of a new graphics card. While add-in board partners (for Nvidia, companies like EVGA, Asus, MSI, Gigabyte, etc) will often design custom circuit boards with higher-quality components, those designs usually come later after launch, so at launch, most of these partners sell Nvidia’s reference circuit board design and simply test for higher quality cards that can be sold at a markup with some factory overclocking applied. With that differentiation created, they then bolt on their own designs for coolers, usually with more gamer-y aesthetics, thicker heatsinks, triple fan and triple expansion slot designs, and RGB LEDs for customized lighting.
Given that, it’s typical that Nvidia does not push the boundaries of the reference design too much. Going to two-fan designs last generation was a big shift in that it forced board partners to make thicker and three-fan heatsinks more common as a selling point for their cards, but the circuit board at the heart of the design was still a standard design with a rectangular PCB and a fairly common component layout.
All this preamble is to wind up the unexpected part of the rumor and leaked images – the supposed RTX 3080 is a bizarre card with a V-shaped cutout and what is likely to be wires or a second PCB for the power delivery hardware and PCI-Express power connectors. So without further ado, here is the photo!
This design is…odd, to say the least. Polarizing, but with a large number of negative responses is probably a better way to put it. When I first saw it, I wasn’t sure what to make of it – it looks, well, weird.
As I thought about it, though, I actually started to find myself liking it. Firstly, for my subjective tastes, the design of the card is cool – I like the industrial nature of it, with the angular heatsinks and multiple compartments. That left the cooling design and efficiency. Nvidia’s stock designs are often okay, but not great. Even the dual-fan design of the RTX 2080 Ti wasn’t the best because it was an open card that had a shroud that obstructed most of the airflow out of the card, and had a small heatsink to make room for, ironically, the extra fan.
So this design is a bit weird, because it has two fans, but one on each side of the card in both the left-to-right and front-to-back meaning of the term. Given the shape of the blades and what we would expect, it appears that the fans are both pushing air into the heatsink, with the fan on the back of the card pushing cleanly through the heatsink and out of the shroud, and the front fan pushing air against the PCB and that air then pushing out in all directions – the back of the card, then the side, heading towards the middle heatsink segments and the one towards the front of the case, being pushed further by the second fan out of the card.
This is a design that seems weird in a lot of ways at first, because it seems inefficient and like it may be hotter – the fan on the back pulls air from near the CPU, the hottest area of the PC. Blowing a majority of the hot air back out the front of the system will cause the fan on the front of the card to pull heated air back in, which is not very good! It also means that in ITX small form factor systems, where the GPU is often run via a PCI-E extension cable and sandwiched behind or on top of the motherboard, the back fan is likely to suffocate.
However, the more I look at it, the more I like it. The heatsink is much larger than the RTX 20 series cards reference design, and is far more open and breathable compared to that design. When you see the heatsink without the shroud, it looks a lot more interesting because there is obviously thought given to it.
Prior Nvidia reference coolers used a vapor chamber, which was fine, but often kept the heat to radiate into the heatsink directly above, keeping it local to the heat-generating components. This new design uses large, mostly-straight heatpipes, which will move the heat away from the GPU and memory thanks to capillary action, taking it to the front-of-case heatsink section, which has airflow right through the card out of the GPU and into the case. With the other fan directly over the GPU and memory blowing some amount of heat out of the case, it should do a reasonably okay job removing heat from the system, although a fair amount will still recirculate through the case and GPU on a few loops before it is removed, so long gaming sessions will slowly heat-soak the card and eventually cause the turbo to back off.
So, in that regard, it is a stock Nvidia card in almost every way – visually designed to appeal, with a cooling solution that has at least been thought about, but the function is unlikely to win out over board partner designs with thicker heatsinks, larger and more fans, and a more conventional design with regards to fan placement.
But outside of spending a ton of text nerding out over the cooling implications of the design, let’s talk about the second part – is it a real leak?
Well, a few things to evaluate without even discussing events that have happened since the initial leak. Firstly, leaked designs that end up being false are usually “safe” in some way – a conventional, predictable heatsink design, standard fans and card length, usually with one sort of oddball element like an extra fan, weird layout, RGB gaudiness, or the like.
This one is, in a way, almost too weird to be an attempted convincing fake. The design fits the aesthetic from the RTX 20-series cards and matches the Nvidia branding guidelines they passed to partners during the ill-fated Geforce Partner Program era, but then takes a weird turn with a 4-chamber heatsink, heatpipes instead of Nvidia’s love affair with vapor chambers, and then there’s the fact that the power connectors for the card are not visible in any of the photos leaked. However, on the flip side, the heatsink only photo lends a lot of credence to the design being legit – having clear photos of a fully-assembled card and the standalone heatsink creates a far more convincing narrative.
Then, there is the rumor confirmation coming out of tech site Igor’s Lab – that the design is near-final if not already final and that many Nvidia staff, including project managers on the team working on the card, hadn’t seen production hardware yet before the leak. The idea of Nvidia setting out investigators on the leak is a little bit funny, given that it has given them a lot of press in advance, but also understandable given corporate contracts on this type of manufacturing. The news about specs and the cost of the cooler are a little harder to confirm and odd, so rather than discuss those, I’ll leave them be for now.
Either way, the fall is going to be fascinating for gaming hardware, with new high end graphics cards from AMD and Nvidia, new CPUs from AMD, and the likely launch of both next-gen consoles bringing tons of new toys for developers to take advantage of, many of which will require this new hardware or even further-out launches for things like storage, and that is pretty cool!