GT215/GT216/GT218, GeForce 200 series
GT215/GT216/GT218 specifications
- Code names: GT215/GT216/GT218
- Process technology: 40nm
- 727/486/260 million transistors
- Unified architecture with an array of unified processors for handling vertices and pixels, as well as other data.
- Hardware support for DirectX 10.1, including Shader Model 4.1
- 128/64-bit memory bus
- Core clock: 550-625 MHz
- More than doubled ALU frequency: 1360-1402 MHz
- 96/48/16 scalar floating-point ALUs (integer and floating-point formats, support for FP32 according to IEEE 754
- 32/16/8 texture address and filtering units, support for FP16 and FP32 components in textures
- Two or one wide ROP (8/4 pixels) supporting antialiasing with up to 16 samples per pixel, including FP16 or FP32 frame buffer. Each unit consists of an array of flexibly configurable ALUs and is responsible for Z generation and comparison, MSAA, blending. Peak performance of this subsystem is up to 16-32 MSAA samples (+ 16-32 Z) per cycle, in Z only mode - 32-64 samples per cycle
- Multiple render targets (up to 8 buffers)
- Interfaces (2 x RAMDAC, Dual Link DVI, HDMI, DisplayPort) are integrated into the chip
GeForce GT 240 reference specifications
- Core clock: 550 MHz
- Unified processors frequency: 1360 MHz
- Unified processors: 96
- 32 texture units, 8 blending units
- Effective memory frequency: 1000(2000) or 850(3400) MHz
- Memory type: GDDR3 or GDDR5
- Memory: 512/1024 MB
- Memory bandwidth: 32.0-54.4 GB/s
- Maximum theoretical fillrate: 4.4 gigapixel per second
- Theoretical texture sampling rate: 17.6 gigatexel per second
- PCI Express 2.0
- Support for DVI, HDMI, DisplayPort
- Power consumption: up to 69 W
- Single-slot design
- Recommended initial price: $99
GeForce GT 220 reference specifications
- Core clock: 625 MHz
- Unified processors frequency: 1360 MHz
- Unified processors: 48
- 16 texture units, 8 blending units
- Effective memory frequency: 800(1600) MHz
- Memory type: DDR3
- Memory: 512/1024 MB
- Memory bandwidth: 25.3 GB/s
- Maximum theoretical fillrate: 5.0 gigapixel per second
- Theoretical texture sampling rate: 10.0 gigatexel per second
- PCI Express 2.0
- Support for DVI, HDMI, DisplayPort
- Power consumption: up to 58 W
- Single-slot design
- Recommended initial price: $69
GeForce 210 reference specifications
- Core clock: 589 MHz
- Unified processors frequency: 1402 MHz
- Unified processors: 16
- 8 texture units, 4 blending units
- Effective memory frequency: 500(1000) MHz
- Memory type: DDR2
- Memory: 512 MB
- Memory bandwidth: 8.0 GB/s
- Maximum theoretical fillrate: 2.4 gigapixel per second
- Theoretical texture sampling rate: 4.7 gigatexel per second
- PCI Express 2.0
- Support for DVI, HDMI, DisplayPort
- Power consumption: up to 31 W
- Single-slot design
- Recommended initial price: $49
With a long delay NVIDIA launched the overhauled Low-End solutions with price tags below $100. Their official names are: GeForce GT 240, GeForce GT 220, and GeForce 210 in the order of performance. These graphics cards are based on the new GPU family: GT215, GT216, and GT218 correspondingly.
These GPUs are based on the modified architecture GT200. They are manufactured by the 40nm process technology at TSMC factories. These cores are very small and consume very little power. However, power consumption is relatively low, because these solutions are not very powerful.
The new GPUs copy characteristics of some previous G9x GPUs (for example, GT 240 - GeForce 9600 GSO), but only as far as the number of stream processors are concerned (96). The number of texture units and ROPs was decreased. Besides, memory controllers were modified, these chips support a 64 and 128 bit bus, but together with GDDR5 memory, which is installed only in the top solution. It's also possible to install other memory types.
Let's talk about differences from the parent architecture GT200. GeForce GTX 285 has ten clusters (TPC - texture-processing cluster), each one consisting of 24 stream processors and eight texture units. Besides, GT200 has eight wide ROPs, executing four operations per cycle. As a result, GeForce GTX 285 is considered to incorporate 240 stream processors, 80 texture units, and 32 ROPs plugged to the 512 bit bus.
The new GPUs differ much from the ancestor, as they are designed for Low-End. For example, GT215 (GeForce GT 240) has four TPC with 24 stream processors (96 in total) and eight TMUs (32 all in all). And the number of wide ROPs has been decreased to two (8 ROPs in total). Two 64 bit memory controllers provide the 128 bit memory interface.
GT216 (GeForce GT 220) has even fewer execution units - only two TPCs, 24 ALUs (48 in total) and 16 TMUs each. The memory bus remains the same as in the top solution (if we may say so). Two 64 bit controllers give us the 128 bit bus.
GT218 has been stripped to minimum, and this GPU has only one TPC with 16 stream processors and eight TMUs. There are twice as few ROPs and memory controllers. 64 bit memory and only four ROPs.
NVIDIA needed to upgrade to the 40nm process technology to cut down manufacturing costs and raise income. But the 40nm process technology is not the only new feature of the latest Low-End GPUs. They all support DirectX 10.1 and Shader Model 4.1. Yes, the DX 10.1, which has been supported by competing products for so long. Unfortunately, it was impossible to launch them together with DirectX 11 solutions, which will be delayed even further. But now NVIDIA has Low-End GPUs that offer better functionality than top GPUs.
Another interesting novelty in the GT215 is support for the new memory type. GeForce 210 and GeForce GT 220 come equipped only with DDR2, DDR3, and GDDR3, and GeForce GT 240 may come with DDR3 and GDDR5 (they have finally caught up with AMD). It's really important for performance, as GDDR5 memory provides twice as high bandwidth. So memory bandwidth of GeForce GT 240 almost matches that of the 256 bit GeForce 9600 GT with GDDR3 memory.
Interestingly, GeForce GT 240 is not only the first graphics card from NVIDIA to support GDDR5, it's the first graphics card with this memory type in the Low-End segment. Unfortunately, only models with GDDR3 memory (reduced memory bandwidth) will be available at first, as NVIDIA partners had some problems with GDDR5 models. So these cards will appear a bit later.
The new cards are not only energy-efficient, but they also possess new features: support for DirectX 10.1, CUDA, PhysX, and 3D Vision. And they are equipped with the integrated audio codec, which allows to output audio to HDMI devices without extra cables inside the PC enclosure. The latter is very important for graphics cards of this price level, unlike PhysX and DirectX 10.1.
But the most important thing about the Low-End GeForce 200 series is the fact that discontinued solutions like GeForce 9600 GT have been replaced with modern solutions. Unfortunately, they do not support DirectX 11, and they are based on the relatively old architecture GT200. When new RV8xx-based solutions are out, it will be very difficult for the new cards from NVIDIA to compete with them, especially with the current level of prices.
Write a comment below. No registration needed!