A week ago, I temporarily turned positive on Nvidia (NVDA). I cited as a reason the encouraging Tegra K1 benchmarks being leaked in several websites. This was particularly relevant due to the bad performance of the Tegra line, which clearly needed a boost. It also left the idea that Nvidia might end up being able to sell IP on its mobile GPU technology.
Already in the comments to that article I explained that my position would be very transitory. I find that Nvidia is being exposed to an incredible amount of competition in the mobile space at the same time the PC side of things faces two major threats:
- An overall decline due to cannibalization by the mobile space (smartphones, tablets);
- And the risk that Intel (INTC) and Advanced Micro Devices (AMD) will make CPU+GPU integration on-die the standard.
Both of these trends work against Nvidia in the PC space while massive competition works against it in the mobile space. With these trends in place I cannot stay bullish Nvidia for long, and so my bullish stance was always going to be very temporary.
Now it ended
With that in mind, my bullish stance ended. Not because of the negative underlying trends, though, but because of two other factors:
- First, there are significant doubts regarding the Tegra K1 benchmarks. These doubts revolve around the power consumption at the performance levels which Nvidia boasted about; or alternatively at the performance given a reasonable power envelope. I already had doubts regarding this which I wrote about in my previous article, but SemiAccurate made these doubts much more serious for me;
- Second, I believe AMD's Mantle might be a major problem for Nvidia. Mantle makes it possible to sidestep limitations in Directx which crimp performance significantly. Games making use of Mantle will be able to improve performance 50%-200% on the same hardware.
A bit more on the implications of Mantle
Directx introduces significant overhead. This was already known by game developers when developing for the older generation Xbox 360 and PS3. Due to this overhead it was possible to make many more draw calls on the aging gaming hardware than on much more powerful and recent PCs.
Mantle brings the lower overhead of consoles to the PC, thus allowing the same hardware to perform much more efficiently. But at this point, it will do so only for AMD hardware. Combine this with the fact that new generation gaming consoles run on AMD hardware and it becomes possible that developers will adopt Mantle as a way to further optimize their titles. Electronic Arts is already promising to do this with Battlefield 4, and others will certainly follow.
While initiatives such as "The way it's meant to be played" by Nvidia already called for a level of customization toward a supplier's hardware, Mantle goes much further because it hits hard on the overall performance of the game by a very large factor.
If Mantle spreads significantly, for a gamer the decision on which hardware to buy becomes a no-brainer, since at the same hardware power and price, games will run 50-200% faster on AMD hardware due to Mantle.
The risk stemming from this is thus simply too large for one to hold a bullish opinion on Nvidia at this point, even though Mantle is just taking off right now.
My bullish opinion on Nvidia was always going to be very temporary due to the underlying trends affecting it. However, I again turned neutral even sooner than I expected, both due to the doubts regarding the Tegra K1 power consumption, and more importantly, due to the major risk implied for Nvidia from AMD's Mantle initiative.
I must also add that if Mantle is successfully adopted by enough developers, it will also constitute a risk for Intel and a positive for AMD.