Update 9/1/20: It’s official. Our original story follows.
As the launch date for Ampere approaches, the details on the next-generation platform are firming up and the rumors are theoretically becoming more accurate. The first is a given, with details on Ampere arriving this week from Nvidia, while the second is… look, details on Ampere arrive this week from Nvidia, and I’m tired of talking about salt. Stop eating so much salt. Take your rumors with something healthy, like a giant spoonful of high fructose corn syrup.
If these new, tasty, tasty rumors are accurate, the highest-end Ampere will cram up to 24GB of RAM onboard, while the RTX 3090 and RTX 3070 will feature 10GB and 8GB, respectively. This seems… kind of low, honestly — but it may be that looks are deceiving here.
On the one hand, this would be the third generation in a row in which Nvidia mostly held the line on 8GB cards. While the RTX 3080 would move up to 10GB, the RTX 3070 at 8GB means that effectively, 8GB is going to be the RAM target for high-end hardware. No one is going to build games that require 10-24GB of VRAM if only a few percent of the market can play them.
On the other hand, however, we’re finally seeing games tapping the power of SSDs this generation rather than continuing to rely on ever-large amounts of console RAM, and it’s already been stated that PCs are expected to share in that bounty. Thus, we should be able to expect the same kinds of graphics updates courtesy of leveraging NVMe and solid-state storage on the platform side of things, without the need for larger memory pools.
Here’s what VideoCardz expects as far as speeds and feeds:
Assume, for the sake of argument, that Ampere and Turing offer identical performance-per-clock (they probably don’t, but it makes the math easier). The RTX 3090 is 1.21x wider than the old RTX 2080 Ti, while the new RTX 3080 is the size of the RTX 2080 Ti. Clocks would be modestly higher than what we saw in the last generation, with the RTX 3090 picking up about 1.1x clock compared with the RTX 2080 Ti. The RTX 3080 would be almost 1.5x wider than the old RTX 2080 (non-Super), making it a significant upgrade in the same price bracket and an unknown value until we know more about how Nvidia will price these new cards.
Looking at these cards, the big question is price. If Nvidia holds the RTX 3080 steady at the RTX 2080 Super’s pricing, it would be a tremendous upgrade. If it raises prices — and if I’m being honest, the sharp increase in VRAM and core counts could both point in that direction — then Ampere might offer performance benefits in-line with its architectural improvements but not dramatically exceeding them, in terms of performance-per-dollar. It’s a little harder to predict Nvidia’s pricing here than in the past because the company raised prices last time it launched, then cut them under pressure from AMD once its own RDNA architecture debuted. Nvidia’s overall share of the graphics market has increased to 80 percent of the discrete space, which probably hasn’t signaled to the company that it ought to consider a price cut.
Then again, if Nvidia comes in at markedly higher prices than AMD plans to target, we’ll see costs come down when AMD launches its own RDNA2 GPUs, an event my crystal ball tells me…to expect. This year, if things are still on schedule, but we don’t know more than that.
It’s been a bit of an odd year in gaming. With a number of Xbox Series X and PlayStation launch titles pushed back, there’s been less to talk about, and for every game that’s formally announced a delay, there are more still floating in limbo. One reason it’s hard to predict how Nvidia will price ray tracing with Ampere is that there hasn’t been as much of a high-profile push around next-generation gaming titles.
Assuming VideoCardz specs are accurate, a lot of how the 3090 – 3070 compare against their predecessors is going to come down to price.
Now Read: