NVIDIA introduces today a new a new graphic card range: the GeForce 6200 TurboCache. This technology allows the graphics processing unit (GPU) to share the capacity and bandwidth of dedicated video memory and dynamically available system memory. ATI is also working on a similar technology named HyperMemory.
To enable this technology, some innovations were necessary. First the combination of local memory and additional system memory allocated for graphics usage based on application need. One main advantage: except for 3D the graphic card doesn’t use the system memory. So memory is still 100% available for other computer uses whether if it is for quantity or speed.
The other innovation is the ability to render more rapidly system memory. Thanks to an optimum use of the PCI Express x16 interface, NVIDIA is able to reach bidirectional bandwidth of PCI Express of up to 8 GBps (4 GBps in both directions).
However it is important to notice that according to NVIDIA some of the chipsets might not be able to reach such performances. Intel’s chipsets for example are restricted to 3 GBps in the chipset to graphic card direction and 1 GBps in the other direction. So in practice, NVIDIA announces that a GeForce 6200 TurboCache might be 20% faster with a nForce4 chipset than with an Intel i915 chipset.
The GeForce 6200 « TC », known as NV44, includes most of the standard GeForce 6200 characteristics (which is a castrated 6600). This graphic card has 4 pixel pipelines, 3 vertex engines and is clocked at 350 MHz. The GeForce 6 range functionalities are of course included, like the Shader Model 3.0 and the PureVideo video engine, which should be (finally!) activated soon. Several GeForce 6200 TC are available :
- With 16 MB of DDR-SDRAM in 32 bits and clocked at 350 MHz (2.8 GBps)
- With 32 MB of DDR-SDRAM in 64 bits clocked at 350 MHz (5.6 GBps)
- With 64 MB of DDR-SDRAM in 64 bits clocked at 350 MHz (5.6 GBps)
What about performances? The TurboCache seems to provide good performances as the 16 and 32 MB version are respectively comparable to the X300 SE and X300 128 MB. There is however sometime an important performance gap between the 6200 TC 32 MB and the X300. ATI also indicates similar performances between a Radeon X300 SE 32 MB HyperMemory and a X300 SE 128 MB (see this news ).
Of course only one of our test could confirm this announcement, but as we don’t have anything lese we will have to trust them for the moment. Performances aren’t however revolutionary and we feel that if you want to play with a comfortable resolution and framerate (1024*768) it is better not to save too much money on your graphic card. With this type of graphic cards only 800*600 resolutions are accessible, and it isn’t much especially if you are using a LCD screen (most of them have mediocre interpolations).
From a pragmatic point of view this technology advantage is to reduce memory cost for entry level products. The advantage for chip manufacturers is obvious as they will be able to use a more expensive chip for the same products which will have in end the same price. That means more income for chips manufacturers and not memory manufacturers. For the user, a more expensive chip means more efficiency and functionality. This is also a plus starting from the fact that HyperMemory and TurboCache technologies are efficient.
Actually the real issue is the product name. A GeForce 6200 TurboCache with 16 MB of DDR-SDRAM in 32 bits and clocked at 350 MHz official name will be « NVIDIA GeForce 6200 with TurboCache supporting 128MB @ 10.8 GBps ».
First point: the 128 MB. Of course it is memory addressable by the GPU, but 112 Mb of this memory comes from the system. This 128 MB figure will however be on every box, and might be highly misleading even if the word “supporting” will also be included.
Second point: the 10.8 GBps. To reach this figure NVIDIA cumulates the graphic card memory bandwidth (2.8 GBps) with the PCI Express bandwidth (8 GBps). And except for using the system memory as video memory, this figure will never be reached, because an nForce 4 means using an Athlon 64 and PC-3200 memory. So the highest figure possible is 6.4 GBps, and we don’t take in consideration the fact that the GPU needs memory bandwidth and that the graphic cards also needs PCI-E bandwidth for other tasks. In addition with a chipset Intel configuration, NVIDIA indicates that this bandwidth is reduced from 8 to 4 GBps.
However except for these considerations, the only real important matter in the end is performances. And for performances the promising GeForce 6200 TurboCache should be able to provide a better performance/price ratio than current entry level graphic cards.