Howdy, Stranger!
It looks like you're new here. If you want to get involved, click one of these buttons!
Quick Links
Nvidia shares fall after Morgan Stanley says the performance of its new gaming card is disappointing
Nvidia shares are down 2.6 percent Thursday.
Moore noted that Nvidia's new RTX 2080 card performed only 3 percent better than the previous generation's 1080Ti card at 4K resolutions.
"We are surprised that the 2080 is only slightly better than the 1080ti, which has been available for over a year and is slightly less expensive," he said. "With higher clock speeds, higher core count, and 40% higher memory bandwidth, we had expected a bigger boost."
Comments
Had they paired the release with some actual Raytracing-enabled shipping software (by waiting a few weeks for Windows to catch up, for example), you could have been looking at an entirely different narrative. Without raytracing (and DLSS to a smaller extent) to showcase the difference in Turing, all anyone had to look at was the increase in price for .... the promise of something to come in the future.
(not saying I buy into the raytracing thing as it has been presented by nVidia so far, but with nothing available now it's a moot point ... )
I don't think many companies will intentionally suppress their stock just so they can set up for the rebound later on.. I wouldn't say never, but I'm sure it's against some SEC regulation (otherwise, companies would do it all the time for stock buybacks, then rebound back later on to inflate their worth). It would be a big gamble even if it were a tactic - there's the strong chance you don't bounce back enough to overtake the initial loss.
I still feel like Turing RTX was more a play for stock to buy time for their 7nm generation - they knew on rasterizing it wasn't significantly better than Pascal, so everything hinged on the hype of the added functions. The snafu was in the timing, those added functions weren't ready in time, the hardware beat them out so the hardware doesn't have anything to showcase.
Amazon did it for years. Tesla is doing it now.
But eventually, Wall Street will tell you to show them the money. You can't baffle them with bullshit forever.
For comparison, Nvidia stock is up about 3% since announcing Turing, while the S&P 500 as a whole is also up about 3% in that time. That means that Turing has been a net was wash as far as stock price is concerned.
I think it's entirely appropriate for the reviews to mean Nvidia stock is down as a result of the Turing reviews, though the Morgan Stanley guy doesn't seem to know why. There are two reasons.
One reason is that performance per mm^2 went down. You can ordinarily increase performance by adding die space, but it adds to the cost. That's the difference between higher and lower end cards of a given generation. It's the reason why Pascal is a better gaming architecture than Vega: a GTX 1080 and an RX Vega 64 are competitive in performance, but the latter takes a lot more die space to do it.
When I've brought this up in the past, some people have been unconvinced that gamers should care about die size. But if you're hoping to make a profit by investing in Nvidia, expenses are every bit as important as revenue.
Performance per mm^2 going down is in spite of moving to a new and theoretically better process node, which makes it even worse. If the GeForce RTX 2080 die had been of the Pascal architecture rather than Turing, but exactly the same die size on the same process node, and with GDDR6 instead of GDDR5X, it likely would have been about 25% faster than a GeForce GTX 1080 Ti, and for exactly the same cost of production as an RTX 2080.
The RTX 2080 didn't get there. While it will likely look somewhat better with driver improvements, that's probably not going to be enough to catch what Pascal could have offered.
The second reason is that you don't get a new architecture very often. This is Nvidia's first since Maxwell in 2014. That its efficiency went down doesn't just mean that the RTX 2080 series isn't as efficient as expected. It means that the GeForce 3000 and likely 4000 series are probably also going to be less efficient than you'd have expected two months ago.
Again, I think the real problem is that Nvidia wasted too much die space on stupid stuff that doesn't matter. I'm not sure how much space the ray tracing cores take, but I'd bet on the tensor cores taking quite a bit. Suppose that you were going to buy one of the new cards and had the choice between:
1) The option to run DLSS on the cards as they are,
2) A 10% lower price tag, but no DLSS, or
3) A 10% performance increase across the board, but no DLSS
Nvidia basically had that choice and went with option (1). I doubt that that is what very many gamers would have chosen. The 10% figure is something pulled out of thin air, but there is some percentage figure, and it could as easily be larger than that as smaller.
Maybe Nvidia can redo Turing without the tensor cores on 7 nm and get the die space savings back. But historically, they haven't been able to get major mid-architecture efficiency gains like that, apart from the Fermi respin to fix yields.
If you meant to compare GTX 1080 Ti to RTX 2080, then they trade blows with each other, but overall RTX 2080 is still a bit faster. GTX 1080 is a lot better price/performance ratio.
Monitor resolution barely affects memory requirements anymore. It used to, when a single 1024x768 frame buffer could require a considerable fraction of the 16 MB that a video card had. But since then, memory capacities have grown massively faster than monitor resolutions, and today, some tens of MB for a frame buffer is nearly a rounding error as compared to the several GB of memory that a video card has. For graphics today, memory capacity requirements are mostly driven by the need to store textures in GPU memory, and higher resolution textures can inflate memory requirements in a hurry.