Arctic Islands GPU Will Double Present Technology’s Efficiency Per Watt!
Arctic Islands GPU: As a result of shift to a decrease node, it seems that 2016’s GPUs might be a major enchancment over the present technology.
At AMD’s earnings name, Lisa Su stated that the following technology of AMD graphic playing cards will double the efficiency per watt provided by the present technology (Fiji) lineup, whereas earlier stories stated that Nvidia’s Pascal has 17 Billion transistors (greater than twice as a lot because the Maxwell flasghip).
Arctic Islands processors will ‘Twice the efficiency per watt’ of Fiji
Though the massive efficiency enhance has been hinted at beforehand, this is likely one of the uncommon events when the company has made the elevated official. On the newest earnings name for the 2015 fiscal yr’s third quarter outcomes, Lisa Su stated:
Based on AMD, the flagship of the Arctic Islands is thought by the codename “Greenland,” and if that is correct (and there’s no purpose to consider it isn’t), then we must always count on efficiency to develop by an element of 100 in comparison with Fiji.
However what does that quantity to, numerically talking? Our present understanding of the FinFET process permits us to offer a tough estimate. In distinction to AMD, whose participation in TSMC’s 16nm FinFET+ node has not but been confirmed, Nvidia is already there.
Extra dependable claims have surfaced, nevertheless, suggesting that AMD has moved on to the 16FF+ node relatively than the 14nm FinFET.
Assuming this to be the case, we will count on AMD’s upcoming process to comply with the identical pointers as Nvidia’s subsequent. Right here’s the official rundown of TSMC’s 16FF+ node:
The Fiji GPU’s die measurement is 596mm2 and it accommodates 8.9 billion transistors and 4096 stream processors. To get this sort of pace enhance, the Greenland (Arctic Islands) processor will doubtless must have roughly 18 billion transistors.
Assuming an preliminary die measurement of 500mm2, you’ll be able to put as many as 15 billion transistors on it. Contemplating the efficiency boosts from architectural upgrades, the 500mm assumption can see a rise of 66%+ (in actuality, something from 70% to 80%).
The accuracy might not be 100%, but it surely’s adequate for our functions. The precise variety of cores can be not possible to guess with out figuring out the scale of the stream processor, which we clearly don’t have.
ICYMI: AMD to launch next-gen ‘Arctic Islands’ GPU in mid-2016 on 14/16nm https://t.co/RmFVIZBUhJ pic.twitter.com/MacXg0WW4V
— TweakTown (@TweakTown) December 4, 2015
An affordable estimate places the determine round 8192 for a 600 mm die and 6784 for a 500 mm die (minus a couple of multiples of 128). Nonetheless, the GM200 solely has 8 billion transistors, subsequently doubling the density would lead to roughly 16 billion transistors for a 600mm chip based on TSMC specs, presuming that the transistor traits would stay unchanged.
Assuming a 500mm2 gadget for the primary 16nm FinFET+ run yields a transistor rely of about 14 billion. Nvidia’s next-generation Pascal GPUs could have a brand new transistor-to-core ratio and design if the story of 17 billion transistors being delivered to the desk by Nvidia seems to be real (inconceivable).
Whereas TSMC’s 20nm back-end expertise lends credence to rumors of a 600mm2 die, I’ve bother believing that these companies will debut with such huge dies.
I feel the probability of an preliminary 500mm2 die is way larger than the likelihood that competitors will compel them to extend to that measurement.
In any case, the primary true, 4K @60fps monsters will arrive within the DirectX 12 battlefield in 2016, and the competitors might be intense. Being a PC gamer proper now could be unquestionably a optimistic expertise.
Please maintain visiting Techballad.com for updates. Preserve our web site bookmarked so you’ll be able to rapidly return to examine for brand new content material. Like Elden, Ring Gamer Transforms Their Avatar
