Zotac GeForce GTX 460 1GB Graphics Card Review

👤by Sahil Mannick Comments 📅09-09-10
Specifications

The GTX460 is based on the GF104 chip, one that Nvidia designed from the ground up to target the mainstream. While the GTX465 was still essentially a GF100 core with disabled stream multiprocessors (SM), the GF104 is a completely new core, yet still based on the Fermi architecture. So we can expect a lot of the benefits of Nvidia's new architecture including full support for Microsoft's DX11 API, excellent tessellation performance and CUDA implementation for PhysX and other applications.

The core boasts an impressive 1.95 billion transistors on a 40nm manufacturing process. Compared to the 3 billion transistors on the GF100 core, the new one should produce much less heat and naturally the die size is much smaller, more akin the ATI's Cypress core which packs a similar 2.15 billion transistors. In terms of the internal core layout, GF104 is similar to the bigger GF100 core with the same Polymorph engine for each SM and a Raster engine for every 4 SM (4 SM make up a Graphics Processing Cluster (GPC)). The difference comes in the number of stream multiprocessors and their arrangements. Instead of a total of 16 on the GF100 core, GF104 only sports half the amount at 8. As such, the number of GPCs has reduced from 4 to 2. This time however, each SM has been configured to have 48 CUDA cores instead of 32. On top of that, each SM now has 8 texture units instead of 4. The implication of this is that each SM of the new core has more compute and texture power than the original. The result is a core organised in an 8x48 block to give a total of 384 CUDA cores. Additionally, there's the L1 Cache for each SM and a common L2 Cache shared the by the core's 8 SMs.

GF100: 16 SMs x 32 = 516 Cores, 16 SMs x 4 = 64 texture units
GF104: 8 SMs x 48 = 384 Cores, 8 SMs x 8 = 64 texture units

When I reviewed the GTX465, I noted that Nvidia had simply disabled the number of SM units to produce three different SKUs from the GF100 core. The flagship GTX480 had one SM disabled for a total of 480 CUDA cores and 60 texture unit, and on the GTX465, 5 were disabled to give 352 cores and 44 texture units. This time, Nvidia have followed the same trend and disabled one SM unit on the GF104 chip to create the GTX460. The end product therefore sports 336 CUDA cores and 56 texture units.

Another notable difference of the GF104 core is the number of memory controllers. The core has 4 in total, instead of the 6 found on the GF100 core, for a 256-bit memory interface. The GTX460 makes full use of the 4 memory controllers for 1024MB of memory whereas the 768MB variant has one disabled. To compensate for the narrower memory bus, Nvidia have teamed the core with GDDR5 memory for much higher bandwidth that could previously be achieved using the same bus width.

GTX480: 15 SMs x 32 = 480 Core, 15 SMs x 4 = 60 texture units
32-bit channel x 12 = 384-bit interface, 256MB x 6 memory controllers = 1536MB memory

GTX465: 11 SMs x 32 = 352 Core, 11 SMs x 4 = 44 texture units
32-bit channel x 8 = 256-bit interface, 256MB x 4 memory controllers = 1024MB memory

GTX460: 7 SMs x 48 = 336 Core, 7 SMs x 8 = 56 texture units
32-bit channel x 8/7 = 256/192-bit interface, 256MB x 4/3 memory controllers = 1024/768MB memory

The card we are looking at today is the 1024MB variant without the "crippled" memory interface. On initial comparisons, we can see that the GTX465 and GTX460 are trying to achieve the same goal using different routes. Despite the lower moniker of the GTX460, which is also reflected by its number of CUDA cores, the card boasts more texture units than the GTX465 and its smaller design will no doubt contribute to lower power consumption and thermal output. Let's see what other changes Nvidia have made to the clock speeds and how the core/texture unit count differences will impact raw performance.


What the GTX460 lacks in stream processors, it more than makes up in clock speed compared to the GTX465. We can immediately see that the GTX460 ends up the superior offering. The 16 CUDA cores deficit is compensated with an increase of 11% and 12.5% on the core and memory clock speed respectively. This translates to 11-12% increase in memory bandwidth and pixel fill rate over the GTX465, and 6% increase in shader operations. The texture fill rate sees a significant 40% boost thanks to the higher number of texture units on the GF104 chip. This performance gain comes with the added benefit of lower power consumption.

Based on those numbers, Nvidia look to be onto a winner and they have priced the card accordingly to make it very competitive in the mainstream market. The only direct competition that ATI currently offer is the HD5830 and going up the price band, there is of course the popular HD5850. Current pricing puts the Zotac GTX460 at around £180 and the cheapest ATI HD5850 at around £200 so in this review, we will see if Nvidia can claim back this market.

Product Specifications

Graphics Engine: GeForce GTX460
Model: ZT-40402-10P
Bus Standard: PCI Express x16 2.0 (Compatible with 1.1)
Memory Size(MB): 1024 GDDR5
Memory Interface: 256-bits
Core Clock Speed (MHz): 675
Stream Processors: 336
Shader Clock (MHz): 1360
Memory Clock Speed (MHz): 3206

3D API: Direct X 11 & OpenGL 4.0
DVI Output: 2
HDMI: 1 (HDMI 1.3a w/audio (8 channel LPCM & bitstream) + DisplayPort)
HDMI-Output: 2 (optional, via DVI to HDMI adaptor)
VGA: Supported using DVI-to-VGA adapter)
HDCP Support: Yes
Dual Link DV1: Yes
HDMI Support: Yes
Display Output (Max Resolution): 2560 x 1600
SLI/3-Way SLI support: No
Card Dimension (mm): 209.55 x 111.15




25 pages 1 2 3 4 > »

Comments