NVIDIA grabs the headlines as tech’s New Big Thing
With the explosion on the scene of Artificial Intelligence platforms such as ChatGPT tech writers, podcasters, and Internet Performers have a lot to talk about. Separating the wheat from the chaff proves difficult. Slate’s “What Next TBD” presents a solid look at graphics processor manufacturer NVIDIA in their pod this week. TBD host Emily Peck explains why a company that’s been around for years leapt over Alphabet, Meta, and other tech giants to join the trillion-dollar club.
Graphics for Gamers
My personal experience with NVIDIA traveled through three phases. First was the, Oh, these graphics cards look neat. Were I a gamer, I’d probably sink some money beyond the SVGA/HDMI that comes with your basic computer. When my son graduated from Georgia Tech, he wanted us to build a gaming workstation for him. (Joining the US Navy changed that, as desktop computers don’t fit well on submarines.) So, fancy graphics cards were items on the mail-order websites.
Switching to Linux as a regular desktop environment carried me into NVIDIA Phase 2. Many laptop manufacturers integrated the company’s graphics cards into their computers. While this was OK for running whatever version of Windows came with the computer, Linux distributions demanded hardware with open-source drivers. NVIDIA held back on some of their code. That meant some distros only supported their hardware as “third party add-on,” if at all. In the long run, the company gave in, releasing all of the code for basic laptop graphics. This wasn’t a huge sacrifice, of course, since the proprietary goodness powered the better, more expensive, cards.
My third NVIDIA phase developed in 2019. That’s when Hitachi introduced its DS7000 series of servers. While the company’s Compute Blade, CB500 product line enjoyed commercial success, the product’s design hit a dead end with the release by Intel of a new generation of processors. Hitachi needed a new design. They established an OEM relationship with Atos, and sold that company’s servers. They were branded as the DS7000s.
The DS7000 series consisted of the 2U DS7020, 4U DS7040, and 8U DS7080. Essentially, the DS7020 served as a building block for the larger configurations. The DS7020 consisted of a 1U base module. This section contained the motherboard, with CPU, RAM, and peripheral support. A second 1U module sat on top of the base. That module could be either what we called a “nothing” module that filled in the space, a “disk” module offering expansion using SAS, SSD, or NVMe drives. The third 1U module, the GPU Module, was where NVIDIA came in.
Those GPU modules enabled the user to install two NVIDIA graphics processor cards in each DS7020. So, if you purchased a DS7080 unit, Hitachi field service folks racked in an 8U computer for you that contained eight Intel Cascade Lake CPUs, a ton of RAM, lots of internal disk support, and eight NVIDIA GPUs.
When I regularly taught the DS7000 series install/config class for Hitachi Vantara Global Learning, we ran through all the possible configurations. Most of our customers here in the Americas, however, purchased the systems as server virtualization platforms. So, the field service folks didn’t work all that much with the GPU modules. They regarded DS7000s set up in that configuration as Bitcoin-mining devices.
The AI boom
Of course, ChatGPT and its competition didn’t emerge from a vacuum. AI development started years ago, and servers like the DS7000 series functioned as the hardware platforms.