AMD Big Navi Would Yield 40-50 percent more than the GeForce RTX 2080 Ti
The news about big Navi seems to be endless, as well .. it surely is taking AMD long to release it. Chatter about the product series is getting louder, the dedicated graphics cards based on the “RDNA2” architecture would ultimately yield 40 and 50% more performance than the GeForce RTX 2080 Ti.
There's a new rumor doing rounds originating from Moore's Law is Dead, that the card will have two “Clusters” of 36 computing units each. That's 72 compute units, good for 4608 streaming processors. It has a 384-bit bus, and GDDR6 graphics memory. Based on the bus width, that could mean 12GB of it. The game clock frequency would hover just above the 2 GHz with a maximum boost of 2.15GHz. That would make it significantly faster than a GeForce RTX 2080 Ti, up to 40% faster even.
We know it'll be a DXR compatible card, and thus it can do hardware-assisted attracting. energy consumption would be 310W when using the boost frequency and 285W when using the Game Boost frequency. Big Navi would have twice as many computing units as an RX 5700, which has 36, and would work at a much higher frequency, the RX 5700 has a 1625MHz game clock and a boost of 1725MHz. Adding these factors to the architectural improvements, it probably performs a little more than twice the RX 5700. AMD confirmed that new RDNA2-based GPUs will arrive later this year, ahead of next-gen consoles. According to rumors, they would hit the market during October.
As always, take this with a healthy dose of skepticism and common sense, please.
Senior Member
Posts: 3352
Joined: 2014-10-20
If AMD can produce a similar "wonder" to Ryzen, but this time on the GPU market - this will be a major win for everyone.
MCM is the clear way to go forward.
Senior Member
Posts: 1456
Joined: 2014-10-24
Will certainly be a beast if true. AMD back in the enthusiast level category.
Senior Member
Posts: 746
Joined: 2015-05-19
MCM is the clear way to go forward.
Indeed it would be, but there has been no indication that either AMD or NVIDIA have achieved that yet (both are researching it). It would also be more likely for a full new architecture instead of an "upscale" like "Big Navi".
Senior Member
Posts: 3352
Joined: 2014-10-20
One way to find out.
Member
Posts: 85
Joined: 2007-02-20
Clusters? Is this MCM (multi chip module) or "glued together" with something like infinity fabric? some sort of internal crossfire? wonder how bad the microstutter will be on this