You can tell Nvidia Ampere (opens in new tab) cards for us PC gamers are getting close because of the increasing cadence of leaky leaks, and I've got some bad news for you… it's looking like the Founders Edition tax is back. Over the weekend we had reports from Chinese forum, Chiphell, showing off a potential (fugly) design for the Founders Edition (FE) version of the GeForce RTX 3080 card (opens in new tab), and now we're getting some more supposed specs confirmation for three launch day Ampere cards.
Igor's Lab (opens in new tab) has published some more details about both the leaked RTX 3080 shroud designs as well as the memory configurations of the different cards coming at launch. Those are supposedly going to be the RTX 3090 Ti, RTX 3080 Ti, and RTX 3080. That immediately sounds frickin' weird; a pair of different Ti cards at launch and a whole new top naming scheme.
That looks like a trio of ultra-expensive, enthusiast-class graphics cards to me, and it feels like we might end up waiting a while for the more affordable options to surface. Which might also tell you a little about what's happening with the competition.
The AMD Big Navi (opens in new tab), RDNA 2 GPUs are sounding genuinely competitive with Nvidia's high-end cards and so I wouldn't be surprised to see Nvidia rolling out some shock and awe tactics with the aim of battering AMD into submission from the off. But AMD could be on for a big launch this time around.
Both companies will be aiming to have the top halo GPU of the new generation of graphics cards arriving around September. Ish. And they're going to be pulling out all the stops to ensure that their card sits at the top of the mountain. I'm genuinely excited about both the technology coming later this year and the inevitable pricing and configuration shenanigans both AMD and Nvidia are likely to engage in around launch.
But yes, Igor Wallossek is suggesting the leaked images from the weekend are likely now confirmed as the first pictures of the Nvidia Ampere Founders Edition shrouds. That's because his own sources have told him an internal investigation has been launched to discover the source of the leaked cooler shots. The focus is apparently on the contract manufacturers of the FE cards: Foxconn and BYD.
He also says that the design is now likely to change because of the leaks, with Nvidia aiming to put the lie to the released images by making some changes to the shroud over the next three months until launch.
But it's his other information about the FE cards that has my interest piqued and has me a little concerned for the wallet-crushing nature of these GPUs. With a whole new architecture, and new production process, we were expecting a bit of a pricing bump over Turing, but Wallossek says the production price for the Founders Edition shroud alone is around $150.
So yeah, you can bet the FE cards are going to be far more expensive than other versions, with that cooler adding another layer of premium pricing to the equation. It's also reportedly using a completely different PCB to the boards going out to Nvidia's graphics card partners, such as MSI, Asus, and Gigabyte. The Founders Edition PCB is set to be used across all three launch day cards, and is set to have a cut out to accommodate the cooler design. The boards going out to manufacturers are reportedly a more traditional rectangular board design.
|Header Cell - Column 0||RTX 3090 Ti||RTX 3080 Ti||RTX 3080|
|Memory||24GB GDDR6X||11GB GDDR6X||10GB GDDR6X|
|Total Board Power (TBP)||350W||320W||320W|
But whatever the PCB being used, each of the three new cards is supposed to be based on the same GA 102 GPU. That's a marked departure from the Turing launch, where three separate slices of graphics silicon was used for the RTX 2080 Ti, RTX 2080, and RTX 2070. Interestingly, Wallossek is also claiming Nvidia will be using the as-yet-still-unannounced GDDR6X video memory in different configurations.
All of this has to be taken with the requisite level of sodium chloride, of course. Especially as, at the time of writing, neither Samsung, SK Hynix, or Micron have announced a subsequent version of their GDDR6 memory, so it will be interesting what the 'X' might end up meaning. I would guess at 18Gbps, but that was expected to be part of the standard GDDR6 spec anyways.
The GA102 is rumoured to have a full spec of 7,552 CUDA cores, though that's unlikely to be seen in anything other than a professional card or a Titan AMP/Ampere/RTAmpX. But with TBP levels of 350W and 320W respectively they're still going to be vast, powerful GPUs inside the RTX 3090 Ti et al.
But at the moment we're still guessing on specs until we actually see something more concrete instead of a few unnamed sources. And as I alluded to before, everything will probably be quite fluid up to the launch of both AMD and Nvidia's next-gen GPUs. The competition is going to be fierce this year, and that's surely going to make for some fireworks.