Why is NVIDIA a chaser? Yes, NVIDIA actually is the King of the Hill in the GPU market, but it is the Intel’s chaser in the Enterprise market, and now I want to explain why.
As we have seen, Intel gets the 99% of Enterprise market, so it’s the King of the Hill. However, the Enterprise market is quickly changing thanks to Cloud Computing and HPC new sub-markets. Intel has just to defend its positions. On the other hand, there are a lot of virgin territories to conquer, and NVIDIA is interested in.
“Leaders should play defensive, not offensive, warfare. Offensive warfare is a game for the No. 2 or No. 3 company in a given field. This is a company strong enough to mount a sustained offensive against the leader”. NVIDIA has to consider the main Intel’s weak points in order to master a good attack plan: “Find a weakness in the leader’s strength and attack at that point”.
Intel has the x86 ISA, NVIDIA not. Intel can easily attack Cloud Computing market, NVIDIA not. But Intel doesn’t have a good portfolio products to attack HPC market, NVIDIA has its own GPUs and CUDA. But Intel is the leader, so NVIDIA has to do a good attack. How?
“Launch the attack on as narrow a front as possible. Preferably with a single product. The “Full line” is a luxury only leaders can afford. Offensive warfare should be waged with narrow lines, as close to single products as possible. […] Only when a breakthrough was achieved did the attacking forces expand laterally to occupy territory”. Intel will try the same strategy in HPC market like it did in x86 market, when AMD showed innovative products and technologies. Intel will try to take advantage of its partnerships (SuperMicro, AsRock Rack, etc), of its brand, and of its software compatibility (Xeon Phi is a HPC chip that use x86 cores, and x86 code gets 99% of Enterprise market!).
NVIDIA has to commercialize as soon as possible its best HPC GPU, in order to counter attack Xeon Phi and gain a lot of market share. Intel, on the other hand, has a lot of money, and it can do a friction war against NVIDIA: Intel can give for free its Xeon Phi, in order to gain market share. For this reason NVIDIA is now partner of IBM in this market: The enemy of my enemy is my friend! When Intel deposed DEC as best CPU manufacturer, Intel joined IBM to create the PC revolution. Now IBM with its Power ISA (About 15% Server market share, and about 30% Infrastructure and Middleware Server market share) is an enemy of Intel.
Turning now to different matters, in the GPU consumer market NVIDIA is the leader: it gets about 80 percent of the market. But the GPU market is not like x86 CPU market, because every uArch is a whole other story. There isn’t a x86 ISA to follow. We have seen during these years different uArchs: Fermi, Kepler, Maxwell, VILW4, VILW5, GCN, etc.
But, like Intel, NVIDIA has some strong point: people thinks that NVIDIA drivers are the best, that NVIDIA cards offer the better compatibility, and so on. Even if AMD show better cards, NVIDIA clients will buy green cards (like during Pentium 4 vs Athlon64 war). NVIDIA is challenging against itself. NVIDIA has just to persuade people to change older NVIDIA card/cards: “Because of leadership position, the defender own a strong point in the mind of the prospect. The best way to improve your position is by constantly attacking it. In other words, you strengthen your position by introducing new products or services that obsolete your existing ones”. G-Sync monitors fit perfectly in this strategy: they require at least a Kepler card (No AMD cards supported!). Also, if you are a G-Sync owner, you have to buy a NVIDIA card in the future to use it, even if AMD cards could be better. Another example, Physx: every so often, NVIDIA drops the support for old cards, and you have to buy a new card to use Physx in your video games. NVIDIA has to create standards, where there are none.
Thanks to this strategy, NVIDIA can have a safe market (gaming GPUs), like Intel (x86 CPUs), in order to attack a virgin territory (HPC servers).