NVIDIA GeForce GT 330M vs NVIDIA GeForce GT 320M

NVIDIA GeForce GT 330M

The Nvidia GeForce GT 330M is the successor of the GT 230M and technically a faster clocked GT 240M (but it should still maintain the power envelope of the GT 230M as the naming suggests). Therefore, the GT330M is a mid-range laptop graphics card that offers DirectX 10.1 effects.

The performance of the GT 330M is similar to the GeForce GT 240M and therefore located in the range of the Mobility Radeon HD 4650. The card supports DirectX 10.1 and all the features of the GT 230M / 240M (as it is based on the same GT216 core). The modern ATI Radeon HD 5650 offers DirectX 11 effects and performs better. 

Modern and demanding DirectX 10 and 11 games (like Crysis or Risen) can only be played fluently with medium detail settings and resolution settings. Less demanding games like Sims 3 run in high details and resolutions. See the gaming list below for detailed benchmarks.

The 48 shader cores of the GT216 are based on the desktop GTX 200 architecture and are therefore a bit improved compared to the 48 cores of the GeForce 9700M GTX. According to Nvidia, the micro architecture was improved regarding battery runtime and performance. The stream processors / shaders are 1-dimensional (1D) and can therefore not directly compared to the 5D shaders of current AMD / ATI cards like the Mobility Radeon HD 5650.

To use the calculation performance of the GeForce GT 330M for other applications, the card supports CUDA, OpenCL, and Direct Compute. For example encoding videos using the stream processors is considerably faster than using a fast CPU.

The GT330M supports the hardware decoding of HD videos using the integrated PureVideo HD engine. The integrated VP4 video processor is able to fully decode H.264, VC-1, MPEG-2, and MPEG-4 ASP (e.g., DivX, xVID). Hardware decoding of MPEG-1 encoded videos is not supported, but this can be handled by every CPU without high load. In conjunction with an Core i5-520M, the GT 330M decoded H.264 with 1-6%, VC1 in 5-10%, and WMV with 3-9% CPU load (1080p over HDMI on a Sony Vaio Z11). 

As with the GT 230M, the performance depends on the used graphics memory. With slow DDR2, the card can be up to 20% slower than with faster memory. Usually the GT 330M is paired with cheap and relatively fast DDR3 memory. GDDR3 would be a bit faster. GDDR5 is not supported according to Nvidia.

A low power version of the GT 330M is also available for laptop manufacturers featuring a minimal slower clock rate and less power consumption (Codename N11P-LP).

NVIDIA GeForce GT 320M

The Nvidia GeForce GT 320M or GT320M was presented in the HP dv6t and dv7t series and should be based on a stripped down GT216 core with 24 instead of 48 shader cores.

Beware: The GT 320M should not be confused with the newer GeForce 320M in the Apple MacBook 13" 04/2010 laptops, which is a chipset graphics card.

In conjunction with the Nvidia 9100M G integrated graphics, the GT 320M supports Hybrid-SLI (only HybridPower). HybridPower is a technique to choose between the integrated and dedicated graphics core, if performance or battery runtime is needed. This works only in Windows Vista. Up to now the user has to use a tool to switch between the GPUs. Later Nvidia wants to switch automatically in the drivers. GeForceBoost is not supported with this card, as there would be no performance gain.

The GT 320M features 24 stream processors that do the graphic work of the vertex- and pixel-shaders. Nvidia Shaders are 1-dimensional and can therefore not be directly compared to the 5-dimensional shaders of AMD (HD 4650 e.g.).

The performance of the GT 320M is nearly on a level of the old GeForce 9600M GT (altough the 9600M GT features 32 shaders). Demanding Direct X 10 games like Crysis or Dirt 2 can only be played in low details fluently. Older games like Left 4 Dead can be played with high details fluently. More game benchmarks can be found below.

The GeForce GT 320M supports PureVideo HD (VP4) to decode HD videos with the GPU. H.264, VC-1, MPEG-2, and MPEG-4 ASP (DivX or xVID) can be fully decoded by the GPU. Futhermore, CUDA and DirectCompute are supported to use the processing power of the graphics card for other applications (with modern drivers). PhysX is currently not supported because of the low shader count.