2023-12-22 13:35:01
“Would have, would have bicycle chain” seems to be the prevailing leadership principle at Intel. Because instead of realistically dealing with the company’s economic future and the massive competitive pressure from Nvidia and other manufacturers, Intel CEO Pat Gelsinger is crying over a project that was doomed to failure from the start and for which he himself was responsible: Larrabee. This doesn’t bode well for Intel’s future development.
Job market
Knappschaft Kliniken Solution GmbH, Bochum
ekom21 – KGRZ Hesse, Darmstadt, Giessen, Kassel, Fulda
This was shown in one Question and answer session with Gelsinger at the Massachusetts Institute of Technology (MIT). When asked by Daniela Rus, professor of electrical engineering and computer science and head of MIT’s renowned Computer Science and Artificial Intelligence Laboratory (CSAIL), what Intel is doing to develop AI hardware, Gelsinger initially evades the question.
Transfigured view of Intel and Nvidia
Instead of answering the specific question, Gelsinger laments not only his career at Intel up until his expulsion, but also the failed Larrabee GPU project. Gelsinger says: “When I was kicked out of Intel 13 years ago, the project that would have fundamentally changed AI was canceled.”
Gelsinger’s assessment of the current AI hardware dominator Nvidia under the decades-long leadership of Jensen Huang is just as unrealistic as his previous time at Intel. According to Gelsinger, this initially only focused on computing with a high data throughput, especially for graphics cards, “and then he was extremely lucky”.
Larrabee was doomed
However, everything that can be wrong about Gelsinger’s supposedly critical analysis is actually wrong. Under Gelsinger as Intel Vice President, Intel correctly analyzed the emergence of Nvidia’s Cuda and the associated possibility of using graphics cards as GPGPU accelerators in supercomputers as a danger in 2007.
But the answer that had been prepared since 2006 was not a real GPU with massive parallelizability, but a reconfigurable many-core CPU system that was only partially reminiscent of GPUs.
This was followed by rumors about a dedicated Larrabee GPU, a secrecy and salami tactic about Larrabee’s capabilities by Gelsinger himself – and ultimately a repeatedly postponed introduction of Larrabee, for which Gelsinger was also responsible.
However, Larrabee’s actual capabilities were so disappointing even at the time of introduction that Intel canceled graphics chips with the architecture shortly afterwards and Intel ultimately completely buried the entire project parallel to Gelsinger’s departure.
However, contrary to what Gelsinger claims, Intel continued to pursue the Larrabee idea more or less unsuccessfully for almost ten years afterwards. However, the concept could hardly score points, especially in the AI environment. Gelsinger’s analysis with its comparison to Nvidia is also misguided, probably because Gelsinger wants to cover up a wrong decision on his part. Because Intel would have had practical competition.
1703252995
#hardware #Intels #CEO #lives #dream #world