http://www.extremetech.com/gaming/2...otype-allegedly-powered-by-maxwell-not-pascal
When Nvidia’s CEO, Jen-Hsun Huang took the stage at CES last week he unveiled the company’s
next-generation self-driving car platform, the Drive PX 2. According to Nvidia, its Drive PX 2 platform packs the same amount of compute power as six Titan X boards, in just two GPUs. During the show, Jen-Hsun displayed the new system — but what he showed from stage almost certainly wasn’t Pascal.
As Anandtech
readers noted, the hardware Jen-Hsun showed was nearly identical to the GTX 980 in an MXM configuration. The new Drive PX 2 is shown above, the GTX 980 MXM is shown below. The hardware isn’t just similar — the chips appear to be identical. Some readers have also claimed they can read the date code on the die as 1503A1 — which would mean the GPUs were produced in the third week of 2015.
If Nvidia actually used a GTX 980 MXM board for their mockup, it would explain why the Drive PX 2 looks as though it only uses GDDR5. While Nvidia could still be tapping that memory standard for its next-generation driving platform, this kind of specialized automotive system is going to be anything but cheap. We’ve said before that we expect GDDR5 and HBM to split the upcoming generation, but we expect that split in consumer hardware with relatively low amounts of GPU memory (2-4GB) and small memory busses. The Drive PX 2 platform sports four Denver CPU cores, eight Cortex-A57 CPUs, 8 TFLOPS worth of single-precision floating point, and a total power consumption of 250W. Nvidia has already said that they’ll be water-cooling the module in electric vehicles and offering a radiator block for conventional cars. Any way you slice it, this is no tiny embedded product serving as a digital entertainment front-end.
Then again, it is still possible that the compute-heavy workloads the Drive PX 2 will perform don’t require HBM. It seems unlikely, but it’s possible.
Wood screws 2.0?
These issues with Pascal and the Drive PX 2 echo the Fermi “wood screw” even of 2009. Back then, Jen-Hsun held up a Fermi board that was nothing but a mock-up, proclaimed the chip was in full production, and would launch before the end of the year. In reality, NV was having major problems with GF100 and the GPU only launched in late March, 2010.
The good news is, we’ve seen no sign that Nvidia is having the same types of problems that delayed Fermi’s launch and hurt the final product. As far as we know, both AMD and Nvidia are on-track to launch new architectural revisions in 2016.
What’s more perplexing is why Nvidia engages in these kind of overreaches in the first place. Claiming the first public demo of a 16nm FinFET GPU may be a decent PR win, but claiming it in the context of automotive market isn’t going to ignite the coals of a GPU fanboy’s heart. Nvidia’s push into self-driving cars and deep learning is a long-term game. Launching a new Shield or game GPU might push some fans to upgrade immediately; precious few people are going to have the luxury of planning a new car purchase on the basis of Nvidia’s PX 2, no matter how powerful it might be.
The entire point of holding up a product from stage is to demonstrate to the audience that the hardware actually exists. If it’s later shown that the hardware in question wasn’t what it was claimed to be, it undercuts the original point. Worse, it invites the audience to question
why the company is playing fast and loose with the truth. There’s no reason to think Pascal is suffering an unusual delay — but these kind of antics invite speculation to the contrary.
Facebook
Twitter