How the cards match up...
Since these two graphics cards are based on a similar architecture to that of their flagship parts, we can anticipate they are quite different in design.
The Radeon HD 2600XT is built around ATI's RV630 architecture, while the GeForce 8600 GTS is based on the Nvidia G84 design. Both are built predominantly for use with the PCI Express x16 bus, though the Radeon HD 2600XT also supports the older AGP interface - Nvidia just recently expressed interest in releasing the GeForce 8 series for those still using AGP through the use of a bridge chip.
Officially the GeForce 8600 GTS is designed to only support 256MB of on-board memory, but that has not stopped a few select manufacturers from doubling that number. At this point we do not believe the extra 256MB really helps the performance of the GeForce 8600 GTS and for the price premium the GeForce 8800 GTS (320MB) would become a much smarter option. The Radeon HD 2600XT can come with either 256MB or 512MB of memory and there is no real serious price premium if you go for the latter.
Lets talk core and memory speeds as this really is where the two begin to show their differences. The GeForce 8600 GTS comes clocked at 675MHz with a memory frequency of 2000MHz. This allows for a theoretical memory bandwidth of 32.0GB/s which is still impressive for a mid-range contender (compare that to the GeForce 8800 Ultra's 103.7GB/s).
The Radeon HD 2600XT on the other hand comes clocked at a staggering 800MHz, while the memory frequency does vary from 1400MHz to an amazing 2200MHz, depending on the memory used. The GDDR4 versions of the Radeon, which cost significantly more, come with the 2.2GHz memory, while the GDDR3 versions like the one we are testing today generally come clocked at 1.4GHz.
The Radeon HD 2600XT features 120 SPs (Stream Processors) which is considerably less than the 320 found on the higher-end and more expensive Radeon HD 2900XT. There are just 8 TMUs (Texture Mapping Units), half that of the 2900XT and 4 ROPs (Raster Operation Pipelines).
This still makes the Radeon HD 2600XT look somewhat more powerful than the GeForce 8600 GTS on paper. The GeForce 8600 GTS only features 32 SPs - though they are clocked at 1180MHz whereas the 24 shader processors (120 stream processors) of the Radeon HD 2600XT are clocked at 800MHz. The GeForce 8600 GTS has an advantage when it comes to its texturing capabilities with twice as many TMUs built in this GPU.
The GeForce 8600 GTS also has a serious advantage when it comes to pixel outputting as it features twice as many ROPs. The 8600 GTS boasts two ROP partitions that can output four pixels per clock, while the Radeon HD 2600XT has just the one supporting four pixel pipes. The Radeon HD 2600XT runs its ROPs at 800MHz for a 3200 Mpixels/sec fillrate, while the GeForce 8600 GTS comes out on top working its ROPs at just 540MHz for a 4320 Mpixels/sec fillrate.
In terms of raw computing power the Radeon HD 2600XT can achieve 192 GigaFLOPS, while the GeForce 8600 GTS can produce just 113.3. When talking about Z-only processing however, the GeForce 8600 GTS is far superior as it can perform 32 pixels per clock while the Radeon can only do 8.
With so many differences to talk about, this battle cannot be played out on paper, therefore we must now turn to real-life tests to determine which of these two products comes on top and deserves the recognition as the best mainstream graphics solution currently available.