With Nvidia damage control in full swing, lets take a look at why the Denver core is having problems. If you understand the underlying tech, some of the problems are obvious.
At CES this year, Nvidia finally gave some hints about the Denver core, something SemiAccurate detailed to the pipelines more than two years ago. Nvidia was previously telling analysts that it would be out in Q4/2012 or Q1/2013 at the latest, we had a different analysis. We think, in retrospect, that our analysis was accurate as we don’t seem to see production silicon including the Denver core as of March 2014. They claimed that Project Denver would be based on the Maxwell GPU, something that wasn’t due to exist until a year after Denver was supposedly scheduled. In short Nvidia’s scheduling did not make sense.
This year at GTC, the Parker SoC that was to be the first Denver release vanished from the roadmap without mention much less explanation. Once SemiAccurate pointed that out, the Nvidia spin machine went in to overdrive and those not willing to question the official line were handed an explanation to reprint. The explanations all roughly say, “The cores Nvidia was showing at CES were pushed out a year. The new Erista SoC will come out before Parker but don’t read anything in to that”. No actual reasoning was given.
If this sounds like damage control by Nvidia PR, you would be right. If you have silicon in hand for a product, you are months from product release. Nvidia claimed they had functional Denver/Parker silicon at CES 2014 to a room full of analysts less than three months ago. They claimed it would be out this year. Two days ago, it vanished without a trace, only “explained” after SemiAccurate called them on this “oversight”.
Lets take a look why we think Nvidia’s story holds no water. It helps if you understand what Denver is, technically speaking. The details and the benchmarks explain why delay means death at this point regardless of what PR operatives hand to sites afraid of losing ad revenue. You may notice that no one questioned the obvious holes in the official story, it wasn’t by chance.
Note: The following is for professional and student level subscribers.
Disclosures: Charlie Demerjian and Stone Arch Networking Services, Inc. have no consulting relationships, investment relationships, or hold any investment positions with any of the companies mentioned in this report.
Latest posts by Charlie Demerjian (see all)
- HyperX ships it’s 60 millionth enthusiast memory module - Oct 15, 2018
- Bittware/Nallatech water cools 300W of Xilinx FPGA - Oct 12, 2018
- More on Intel’s 10nm process problems - Sep 17, 2018
- Intel puts out another 14nm 2020 server platform - Sep 11, 2018
- Why Can’t Intel Supply Enough 14nm Xeons? - Sep 10, 2018