On a relatively cold Monday afternoon stateside the semiconductor wars escalated when Nvidia said it will offer its first server microprocessors. This is seen as a very aggressive and exciting move by Nvidia with significant negative repercussions for Intel. The server microprocessor market had up until now been Intel’s niche and sweet spot which it controlled and dominated. Now even the server microprocessor space is open war!
Nvidia shares jumped to their all time high ending the day up 6%, they were also helped by their positive comments regarding their first quarter revenue, which they said was above its previous forecasts. Meanwhile Intel shares ended the day down over 4%.
Interestingly enough Nvidia has developed a central processing unit (CPU) based on technology from Arm, a company it has been actively trying to acquire from SoftBank Group. This is Nvidia’s first venture into the CPU space and up until now the company had focused on graphics processing units (GPUs). These have come increasingly into vogue for their ability to power intensive computing tasks and have seen strong demand from cryptocurrency miners over the last few years. Nvidia expanding into the CPU market is seen as a positive for the company as it opens up more revenue opportunities for them.
Nvidia’s Grace chip is named after Grace Hopper, the pioneering computer scientist in the 1950s. This chip has been specifically designed to work closely with Nvidia’s graphics cards to better handle new computing problems that will come from AI training consisting of a trillion data points of information.. Nvidia’s aim when designing this new architecture was to come up with something that would work faster than the current combinations of Nvidia graphics chips and Intel CPUs. The product will be available from early 2023..
Modern AI and data science problems involve training algorithms across millions of data points and up to hundreds of parameters so the Grace chips are designed from the ground up with this in mind. Nvidia claim the new chips reduce this training time from months down to just a few days. Impressive and powerful. One area the new chips will be used is in helping computers to understand human language better, which will ultimately lead to more advanced automated help services. There is also talk of the Grace chip being used to help with complex weather forecasting calculations. In particular, a key area of work here has been a focus on predictions that can be made as much as a whole season in advance.
Nvidia is known to be actively pitching these new CPUs to data centre owners, highlighting that it will allow them to improve their ability to make sense of the flood of data they receive and harness it more effectively. NVIDIA has been actively targeting data centre owners, including Amazon’s AWS and Google in a bid to sell their new CPU.
Nvidia’s decision to use ARM infrastructure in their chips is likely to benefit customers with power constraints given ARM’s high level of expertise with smartphone and mobile technology, where battery life constraints are a key matter of concern.
These chips are going to unleash a lot more power and it is exciting to see what comes of it and how it will benefit us all.
