RSS Feed

HCW Tech Blog

For the latest info on computer hardware, tech, news, video games, software tips, and Linux, check out our new improved front page: HCW Tech Blog

Reviewed by: Norman Tan [10.04.05]
Manufactured by: Leadtek



Analog connectors are quickly becoming a thing of the past and if one can afford this video card, then they can surely afford monitors with DVI connectors. These DVI outputs are powered by dual 400mhz RAMDACS which can support resolutions up to 2024x1536 @ 85hz. Unfortunately, dual-link DVI support was not included with this card, so those with 30" ACD will be forced to look elsewhere. For mere mortals, this will be more than enough.

In an effort to differentiate themselves from the competition, Leadtek also included this breakout dongle that supports S-video out, Component video out, composite video in and s-video in. To compare the quality of the component out, a Dell 24" LCD was used and movies were compared using the DVI and Component (1080i) inputs. While one could see the differences between the DVI and Component inputs, there was minimal issues with the screen flickering and movies did not exhibit any issues.

In addition to the cables, Leadtek also included a very decent software package that included the games Splinter Cell: Chaos Theory and Prince of Persia: Warrior Within. For multimedia purposes, PowerDVD 6, DVD MovieFactory 3 SE and Videostudio 8 Se DVD were also included.

The Specs

As mentioned earlier, the Leadtek Extreme Edition is a highly overclocked version of the reference 7800GTX design which is clocked at 430mhz/1200mhz. The Leadtek on the other hand clocks in at a whooping 490mhz/1250mhz which is a very healthy overclock. By comparison, the current top of the line ATI card, the x850 XT PE checks in with a core clock of 540mhz and a memory clock of 1180mhz. All cards in this comparison feature 256 mb of onboard video memory though it is expected that a 512 mb variant of the 7800GTX will be released in the future.

Of course, comparing those numbers is a little meaningless without knowledge of the Pixel pipelines and memory bus width. In the case of the 7800GTX, we have 24 pixel pipelines along with a memory bus width of 256-bits. The ATI card, brings to the table 16 pipelines and a 256-bit memory bus.

Putting it all into a convenient chart, we see that on paper at least, a stock 7800GTX has better numbers than the x850 XT PE. The Leadtek with its higher clocks is even faster.

  Core Clock Pixel Pipelines Multitexture Fillrate Memory Clock Mem Bus Width Memory Bandwidth
Stock 7800GTX 430 MHz 24 10320 MP/s 1200 MHz 256 bit 38.5 GB/s
Leadtek 7800GTX Extreme 490 MHz 24 11760 MP/s 1250 MHz 256 bit 40 GB/s
X850 XT PE 540 MHz 16 8640 MP/s 1180 MHz 256 bit 37.8 GB/s

The Features

As many of you will know, the 7800GTX core codenamed the G70 isn't necessarily a new core. If anything, it is closer to the NV40 than one would think and rumor has it that at one point, it was named the NV47. Of course, taking an old design and tweaking it isn't necessarily a bad thing as the NV40 was already very good at what it did. In short, the G70 gains an additional 8 pixel pipelines over the Geforce 6 series 16 and an additional 2 vertex shaders for a total of 8. These additions resulted in a transistor count of 302 million which is 80 more than the 6800 Ultra.

With an larger transistor count, it may come as a surprise that the G70 actually consumes less power than a 6800 Ultra. This is due in part to the switch from IBMs 130nm process for the 6800 Ultra to TSMC's 110nm process. This allowed for a modest bump in speeds without affecting power consumption too much. Of course, much of the performance increases in the 7800GTX are due to the increases of parallel pipelines which enable more to be done in less time.

While throwing more pixel pipelines at everything will indeed add performance, Nvidia also set out to make the architecture of the G70 core more efficient. Nvidia sampled shader code from 1,300 of the most common types of shader computations in all sorts of applications and games. They found that the general trend was that applications were taking advantage of shaders which meant that performance was bound by the speed at which the GPU could calculate these.

Why the GPU? The use of shaders increases the number of calculations that need to be done per pixel thereby putting stress on the GPU instead of the memory. As we see textures get more complex and push more pixels, memory performance will become more important and we may start to see video cards with more video card memory. For this release though, it is clear that Nvidia has been concentrating on the GPU side of things.

In addition to Shader performance, a word commonly being heard nowadays is High Dynamic Range or HDR. In short, an integer 32-bit color can only be described with RGB values that range from 0-255. With HDR, 16 floating bits are allocated to four channels (red, green, blue, alpha) which allow the 7800GTX to get that much closer to photo realistic rendering. For interests sake, this standard is what the movie houses use to render their 3d movies.

FInally, there are improvements in the way that anti-aliasing is done. Traditionally, with anti-aliasing, the algorithms would only take into account the edges of polygons. That meant that things such as plants and chain link fences would not benefit from AA as they were often just textures with transparencies in them. Nvidia developed a way to anti-alias the pixels inside of these transparencies which results in better image quality as thin things such as chain-link fences will blend better into backgrounds.

Next Page: (Benchmark Setup; 3DMark 2003)