This week NVIDIA announced the launch of the RTX SPARK super chip for personal AI computers. The processor part of the chip is customized and developed by NVIDIA and MediaTek based on the ARM architecture. The GPU part is the NVIDIA GB10 chip. NVIDIA's super chip can provide 1 peta FLOP of computing power and can locally run artificial intelligence models with parameter scales up to 120B.

The main players in the field of desktop processors are Intel and AMD. For this reason, Tom's Hardware interviewed Nish Niralojanan, senior director of product management of Intel's Client Computing Division, about Nvidia's launch of desktop chips. He wanted to hear Intel's views on Nvidia's entry into the desktop processor market.

Intel is cautious about this:

Nira Lojanan said that Nvidia's new RTX SPARK super chip is indeed great. Nvidia is proficient in game development and good at handling various application scenarios, but Intel is always cautious about this because Intel is confident in its products.

From a compatibility perspective, when new ARM CPUs enter the market, they are bound to face a large number of compatibility, DRM (digital copyright protection mechanisms, but Nvidia has emphasized that its chips will be compatible with all mainstream DRM) and backward compatibility issues, while Intel provides customers with the right combination of CPU and GPU, whether it is gaming or has the best products in the field of AI inference workloads.

Niralogjanan also mentioned that Nvidia is also a very good partner for Intel. Intel will continue to cooperate with Nvidia. For example, the two parties will jointly expand different parts of the product roadmap, cooperate in certain areas, and of course compete in certain areas. This is a good thing for the entire market.

What’s interesting is that Nila Lojanan also mentioned the price issue of RTX SPARK. There is no doubt about this: NVIDIA super chips are mainly targeted at high-end AI equipment, and the final product price will definitely be very high, so Nila Lojanan believes that price is also another important factor in whether the RTX SPARK chip can gain market share.

Compatibility is indeed an issue worth paying attention to:

The cooperation between Qualcomm and Microsoft has brought the ARM architecture into the consumer desktop market. Unfortunately, Qualcomm and Microsoft have not succeeded in realizing the prosperity of the Windows on Arm ecosystem. Therefore, the RTX SPARK compatibility mentioned by Intel is indeed an issue worthy of attention, because there are still a large number of software that do not provide native Windows on Arm versions.

As far as the current actual situation is concerned, the method adopted by Microsoft is compatibility layer conversion, which is to convert traditional x86 software through the compatibility layer into a version that can run on ARM chips. However, compatibility layer conversion will lose performance. This situation has lasted for many years, at least the current Windows on Arm ecosystem has not seen too obvious progress.

For example: Adobe's creative productivity tools Photoshop and Premiere Pro do not have native Windows on Arm versions, and the running performance will be much worse through compatibility layer conversion. However, NVIDIA revealed that Adobe is developing a native Arm version to support RTX SPARK. After all, this super chip provides higher AI performance and can cooperate with Adobe's AI tools to help creators achieve more enhancements locally.

In the long run, the Windows on Arm ecosystem can still develop. In the Qualcomm era, this development speed was relatively slow but it is also developing slowly. After Nvidia joins, the construction of this ecosystem may be accelerated, and ultimately achieve achievements similar to Apple's in the Arm ecological field (most popular macOS software now provides native versions of Apple M chips, rather than Intel x64 versions).