GeForce 700 series
Nvidia's 2013 GPU series spanning Kepler, Fermi, and Maxwell.
The GeForce 700 series is a lineup of graphics processing units from Nvidia. Most of these cards are a refresh of the Kepler microarchitecture (chips with GK codenames), though some earlier models use Fermi (GF) and later ones use Maxwell (GM). The series launched in 2013, beginning with the GeForce GTX Titan on February 19, 2013, and followed by the GeForce GTX 780 on May 23, 2013. Mobile versions of the GeForce 700 series first appeared in April 2013.
The GK110 chip, which powers the higher-end cards, was built with computational performance as a priority. It contains 7.1 billion transistors and aims for energy efficiency by running as many tasks in parallel as its streaming processors allow. Compared to earlier models, GK110 offers more memory space and bandwidth for both the register file and the L2 cache. At the SMX level, the register file has grown to 256KB (made up of 64K 32-bit registers), up from Fermi's 128KB (32K 32-bit registers). The L2 cache has increased to up to 1.5MB—twice the size of the GF110—and both the L2 cache and register file have doubled their bandwidth.
Performance in situations where registers are scarce has improved, as each thread now has access to more registers. The total number of registers a thread can address has jumped from 63 to 255 with GK110. Nvidia also reworked the GPU texture cache for compute tasks. It is 48KB in size and, when used for compute, becomes a read-only cache that specializes in handling unaligned memory access workloads. Error detection capabilities were added to make it safer for workloads that rely on ECC. Later Nvidia drivers added Dynamic Super Resolution (DSR) support to Kepler GPUs.
The GeForce 700 series combines features from both GK104 and GK110. Kepler-based members of the series include standard features derived from GK104: a PCI Express 3.0 interface, DisplayPort 1.2, HDMI 1.4a with 4K x 2K video output, PureVideo VP5 hardware video acceleration (up to 4K x 2K H.264 decode), a hardware H.264 encoding block (NVENC), support for up to four independent 2D displays or three stereoscopic/3D displays (NV Surround), Bindless Textures, GPU Boost, TXAA, and manufacturing by TSMC on a 28 nm process.
Quick Facts
- Codename
- GK110 · GK208 · GM108 · GM107
- Architecture
- Fermi · Kepler · Maxwell
- Created
- February 19, 2013
- Model
- GeForce series
- Fab
- TSMC 40 nm · TSMC 28 nm
Facts from the source article.
Lore & Background
The GeForce 700 series includes both GK104-derived and GK110-derived features. GK110 was designed and marketed with computational performance in mind, containing 7.1 billion transistors and aiming to maximize energy efficiency through parallel execution. Increases in memory space and bandwidth for both the register file and the L2 cache over previous models are seen. At the SMX level, GK110's register file space increased to 256KB composed of 64K 32-bit registers, compared to Fermi's 128 KB. The L2 cache space increased by up to 1.5MB, twice as big as GF110, and both L2 cache and register file bandwidth doubled. Performance in register-starved scenarios improved as more registers became available per thread, with the total number of registers each thread can address moving from 63 to 255.
New features from GK110 include Compute Focus SMX Improvement, CUDA Compute Capability 3.5, new shuffle instructions, Dynamic Parallelism, Hyper-Q, a Grid Management Unit, Nvidia GPUDirect, and GPU-Boost 2.0. Hyper-Q expands hardware work queues from 1 to 32, allowing higher utilization by putting different task streams on otherwise idle SMXs. Dynamic parallelism enables kernels to dispatch other kernels, saving time by not having to go back to the CPU. The texture cache was reworked for compute, with 48KB becoming a read-only cache specializing in unaligned memory access workloads. Error detection capabilities were added for ECC workloads. Dynamic Super Resolution (DSR) was added to Kepler GPUs with the latest Nvidia drivers.
Reader's Guide
The GeForce 700 series represents a transitional period for Nvidia, spanning multiple microarchitectures. Its significance lies in the introduction of GK110, a chip focused on computational performance with 7.1 billion transistors and substantial increases in register file and L2 cache capacity. The series brought features such as Hyper-Q and Dynamic Parallelism, which improved GPU utilization and allowed kernels to dispatch child kernels without CPU intervention. The series also saw the addition of shuffle instructions for faster data sharing within warps and overhauled atomic operations. Support for DirectX 12.0 with feature level 11_0 was included across all cards. Nvidia later transitioned Kepler desktop GPUs to legacy support from September 2021, with critical security updates through September 2024. Kepler notebook GPUs moved to legacy support in April 2019 and stopped receiving security updates in April 2020. In December 2025, Nvidia discontinued Game Ready Driver support for Maxwell-based models of the GeForce 700 series, with security updates continuing until October 2028. The last driver to fully support CUDA with 64-bit Compute Capability 3.5 for Kepler in Windows 7 and Windows 8.1 64-bit was 388.71.
Did You Know?
- GK110 contains 7.1 billion transistors.
- GK110's register file space increased to 256KB, compared to Fermi's 128 KB.
- Hyper-Q expands GK110 hardware work queues from 1 to 32.
- The GeForce 700 series includes cards using Fermi (GF) and later Maxwell (GM) in addition to Kepler.
More in Nvidia and AMD Graphics Processors 1-24
Spotted an error? Know more?
Reader corrections go straight into our review queue. Suggest an edit · How this site is sourced