Features of nvidia's new low-end gpu: gk208

Last month Nvidia introduced its new Geforce GT 640 and GT 630 v2 GPUs, its new inexpensive GPUs based on the also new Kepler GK208 graphics core, currently the only chip based on Nvidia's third-generation Kepler graphics architecture.

The Kepler GK110 architecture introduced many new features over the previous Kepler graphic architecture, one of the most celebrated being its greater orientation towards games, increasing the number of its integer processing units (ALUs or shader processors) but sacrificing its performance in intensive GPU-accelerated (FP64) applications.

The original Kepler architecture has undergone two evolutions aimed at offering a greater number of features and greater performance at the weak points of the original Kepler (also known as the first-generation Kepler).

Among the changes and improvements introduced by Kepler evolutions we can mention:

Kepler second generation "GK11x Series"

Nvidia added the ability to execute merged FP16 instructions and duplicated the circuitry necessary to execute integer swap operations; improvements with which the GPU achieves a performance between 100 to 200% (at best) superior to that of the first generation Kepler in GPU-accelerated applications.

At the moment the only second-generation Kepler-based Nvidia graphics core is GK110, the core on which the Geforce GTX Titan and Geforce GTX 780 GPUs are based.

Kepler of the third generation “GK20x Series”

It includes all the improvements introduced in the second-generation Kepler, to which is added the hardware support to the Microsoft DirectX 11.1 API, officially launched in October last year with the Microsoft Windows 8 operating system.

At the moment the only graphics core based on the third-generation Kepler graphics architecture is GK208, the core on which the GeForce GT 640 V2 and GeForce GT 630 V2 GPUs are based.

GK208

Graphic core oriented towards the low range, which is why Nvidia has tried to reduce it to its minimum expression in several aspects as you can see in its specifications:

• Compatible with API DirectX 11.1.1270 million transistors. 384 shader processors organized in 2 SMX (192 shaders each), which make up a GPC. 2 tessellation engines (Polymorph Engines 2.0).32 texture units "TMUs" (16 for each SMX).8 “ROPs” rendering units (4 for each memory controller).64-bit DDR3 / GDDR5 memory controller (2 dual-channel 32-bit memory controllers). PCI Express (PCIe) controller 2.0 with 8 lines (PCIe 2.0 8X).

Although in terms of specifications, GK208 seems somewhat lower than GK107, but it compensates with higher operating frequencies, higher performance in GPU-accelerated computing, lower consumption and full compatibility with the hardware acceleration functions that the DirectX 11.1 API incorporates..

The GeForce GT 640 V2 and GeForce GT 630 V2 GPUs

The first GPUs based on the GK208 graphics core are the new Nvidia GeForce GT 640 V2 and GeForce GT 630 V2 GPUs, GPUs that replace their predecessors (same trade name "without V2") based on the old Kepler GK107 graphics cores (GeForce GT 640) and Fermi GF108 (GeForce GT 630 or renamed GeForce GT 440).

These two new Nvidia GPUs, although oriented to the low range, offer a good number of features not present in any of its older brothers from previous generations.

WE RECOMMEND YOU 2K Games withdraws its games from GeForce NOW

Here is a table with its specifications:

Here are some benchmarks made to these new GPUs:

At last Nvidia adopts the Microsoft DirectX 11.1 API, but curiously only for its low range, because at the moment it is unknown if it plans to launch other graphics cores based on its third-generation Kepler architecture apart from GK208.

This decision, as in the case of GT21x GPUs (GeForce 300, 240, 230 220, 210 Series) based on the third-generation Tesla graphics architecture, is presumed to respond to certification issues for the impending Microsoft Windows operating system. 8.1, the one that launches the new API DirectX 11.2 API that has several functions retro-compatible by hardware with DirectX 11.1.

So far the third-generation Kepler graphics architecture has performed wonderfully well in tests, managing to outperform its first-generation Kepler and Fermi-based predecessors, despite using only a 64-bit memory bus and having half the texture and rendering units.

These new GPUs become the most attractive low-end chips for price and specifications, which is good news for users of low-end products, and will be even more so for notebook and ultrabook users, as they are also expected. GeForce 700M Series GPUs based on GK208.