Intel shows us its dedicated GPU prototype: for the moment efficiency reigns, not power

A few weeks ago we learned how Intel was preparing a dedicated GPU with which theoretically would compete in that segment with the absolute dominating, AMD and NVIDIA.
Those expectations that we all had with this project are now a bit down: the first data we have finally these dedidac GPUs point to some graphics that bet on efficiency and not power .
A lot of way to go
Intel's proposal seems to go beyond its traditional GPUs integrated into its SoCs with the traditional microphones of the Intel Core family, but it certainly does not seem to pose any kind of battle to the more powerful NVIDIA or AMD dedicated graphics.
At the ISSCC event in San Francisco, Intel showed what is still a job with a lot of margin to change things . Even so, the data provided by those responsible for Intel point to a very modest GPU in power.
That prototype - which could well stay in that and never appear on the market - has a chip manufactured in 14 nm technology with 1.5 billion transistors . Most NVIDIA Pascal GPUs use a 16 nm manufacturing process. The GTX 1080, for example, has 7,200 million transistors and operates at a base frequency of 1.5 GHz that can approach 1.7 GHz according to NVIDIA specifications.
In this prototype of Intel the native frequency of work would be of only 50 MHz of native form, although it could arrive at the 400 MHz . Of course, the voltage of these two modes of work is really low: 0.51 and 1.2 volts respectively.
It seems clear that at least in this prototype of which Intel spoke the focus is on efficiency and not power . It is of course soon to talk about these GPUs that do not even exist as a physical product, so we will have to be patient and let Intel work quietly in this segment. If there is something that Intel has, of course, it is with resources for the development and manufacture of solutions in this area.
Via | Liliputing

No comments