Details on NVIDIA's GeForce GTX 880 have begun to leak, with a report from Tyden.cz claiming that the GeForce GTX 880 will be a Maxwell-based GPU that will dominate the single GPU market. The GeForce GTX 880 will move from the family name of GK110 (the 'K' is for Kepler) to the GM204 part. GM204 is the GK110 equivalent of the Kepler family, but based on the Maxwell architecture.
The leaked specs will see the DirectX 12-capable GPU feature a streaming multiprocessor Maxwell, or SMM, SIMD design that is found in the also Maxwell-based GeForce GTX 750 Ti. The GeForce GTX 880 will just feature more of those SMMs, spread across multiple graphics processing clusters (GPCs) and see some extra performance when compared against the GTX 780 Ti. We should expect something along the lines of these specs:
- 20 nm GM204 silicon
- 7.9 billion transistors
- 3,200 CUDA cores
- 200 TMUs
- 32 ROPs
- 5.7 TFLOP/s single-precision floating-point throughput
- 256-bit wide GDDR5 memory interface
- 4 GB standard memory amount
- 238 GB/s memory bandwidth
- Clock speeds of 900 MHz core, 950 MHz GPU Boost, 7.40 GHz memory
- 230W board power
Comparing the Maxwell-powered GTX 880 against the Kepler-based GTX 780 Ti sees the GTX 880 ahead, but not by much. We have a rumored 3200 CUDA cores versus 2880 on the GTX 780 Ti. Base clocks increased from 875MHz and 928MHz for the Base Clock and Boost Clocks, respectively on the GTX 780 Ti, to 900MHz and 950MHz on the GTX 880.
The memory gets increased from 7.0GHz on the GTX 780 Ti to 7.4GHz on the GTX 880. The big problem here is the memory bandwidth being just 238GB/sec compared to the 336GB/sec on the GTX 780 Ti, which is because of the 256-bit memory bus on the GTX 880, versus the GTX 780 Ti's 384-bit bus. I think that's where we're going to see an issue, but let's hope we see a 512-bit card announced, with a much higher power draw to allow for clocks to be boosted, and the much beefier memory bus.
Or is that what we're seeing as the GTX TITAN Z?