NVIDIA has just announced its new GeForce 800M series.
GeForce GTX 800M
GeForce 800M series will include three achitectures: Kepler, Maxwell and even Fermi. All high-end parts will feature Kepler GK104 GPU variants. More power efficient Maxwell will be added to GeForce 850M/840M/830M. It’s worth noting that there are no GT parts. NVIDIA has dropped all GT branding and left only GTX for high-end segment.
GeForce 800M series | ||||
---|---|---|---|---|
GTX 880M | GTX 870M | GTX 860M #1 | GTX 860M #2 | |
GPU Codename | Kepler GK104 | Kepler GK104 | Kepler GK104 | Maxwell GM107 |
CUDA Cores | 1536 | 1344 | 1152 | 640 |
Video Memory | up to 8GB GDDR5 | up to 6GB GDDR5 | up to 2GB GDDR5 | up to 2GB GDDR5 |
Memory Bus | 256-bit | 192-bit | 128-bit | 128-bit |
Base Clock | 933 MHz +boost | 941 MHz +boost | 797 MHz +boost | 1029 MHz +boost |
Memory Clock | 5000 MHz | 5000 MHz | 5000 MHz | 5000 MHz |
GTX 850M | 840M | 830M | 820M | |
GPU Codename | Maxwell GM107 | Maxwell GM108 | Maxwell GM108 | Fermi |
CUDA Cores | 512 | 384 | ? | ? |
Video Memory | up to 2GB GDDR5 | up to 2GB DDR3 | up to 2GB DDR3 | ? |
Memory Bus | 128-bit | 64-bit | 64-bit | ? |
Base Clock | 876 MHz +boost | 1029 MHz +boost | ? | ? |
Memory Clock | 5000 MHz | 2000 MHz | ? | ? |
NVDIA GeForce GTX 880M/870M — Kepler
The most powerful GPUs are based on Kepler GK104 GPU. The fastest model has 1536 CUDA cores, 256-bit interface and up to 4GB GDDR5 memory. I could not find any confirmation of 8GB variants.
The GTX 870M has narrower interface (192-bit), thus less memory (3GB). It’s using a GK104 variant with 1344 CUDA cores.
NVDIA GeForce GTX 860M/850M — Kepler and Maxwell
There are two GTX 860Ms. One is equipped with GK104 GPU with 1152 CUDA cores, and the latter has GM107 processor with 640 CUDAs. This is quite interesting move, as both cards are using the same model codename.
The GTX 850M is using the same processor as desktop GTX 750 non-Ti. It’s 512 CUDA based GPU with 128-bit interface.
NVDIA GeForce 840M/830M and 820M — Maxwell and Fermi
The GeForce 840M and 830M will utilize new GM108 GPU. The only leak we have says it has 3 SMM. That said the full processor should feature 384 CUDA cores.
The last in the pack is GeForce 820M. There is little information about this card, but according to few sites, it is based on Fermi architecture.
New Features: Battery Boost and ShadowPlay support
Battery Boost is basically a framerate limiter. It doesn’t replace Optimus, it actually works with this technology on hardware level. It is supposed to control power states and overclocking capabilities to improve the smoothness of the gameplay while offering stable framerate.
ShadowPlay and GameStream are now supported with GeForce Experience on notebooks. Current mobile GPUs are powerful enough to allow background video recording.
Performance
I don’t think there’s much to say here. Just refer to the charts below.