Related Articles

Hardware & Test Notes

• AMD Athlon 64 3000+ (Socket 939) @ 250 MHz HT/RAM (2250 MHz)
• Soltek SL-K890Pro-939 motherboard (VIA K8T890)
• Soltek K8TPro-939 motherboard (VIA K8T800 Pro)
• 2x512 MB PC4200 EL Platinum Edition (Dual Channel)
• Gigabyte GV-RX80256D X800 256MB
• Gigabyte Passive 6800 AGP
• NVIDIA 6600GT PCIe
• Sapphire X800 XL
• Western Digital WD1200JD 120 GB SATA
• OCZ ModStream 520W power supply

These particular cards were shown as they are representative of cards in the same general price range - the 6600GT is around the 199$ mark while the 6800 is around the 299$. ATI's SRP for the 128MB version of the X800 is 199$ also but a lot of manufacturers seem to be doing the 256MB route which ups the price. Finally, the X800 XL was also included as it has a SRP of 299$. In reality it is a lot higher but assuming the price does drop to the 299$ mark, it will make things a lot more interesting.

We have our Athlon 64 3000+ Winchester running at 2.25 GHz on a 250 MHz HT bus to remove it as much as possible from the performance equation. Gigabyte's passively cooled GeForce 6800 is the only AGP card shown in the benchmark graphs but we did run a few tests with the 6600GT AGP. The memory clock on the 6600GT AGP was raised to 1000 mhz to make it equal with the PCIe version. In most cases, benchmark scores were almost dead even. In some instances (Far Cry in particular), there was up to an 8% difference favoring the AGP card. If the trends remain the same, we can operate under the assumption that a PCIe version of the 6800 may be slower than its AGP counterpart due to differences either in the motherboard chipset or with the cards themselves.

We benchmarked all of the ATI cards with 2005's first Catalyst release, version 5.1 and the NVIDIA cards with the 66.93 drivers. VIA's 4.55 Hyperion 4-in-1 drivers were used for both boards.

Benchmarks

• 3DMark 2005
• Aquamark 3
• Call of Duty
• Jedi Knight 2: Jedi Academy
• Halo
• X2 Rolling Demo
• Splinter Cell
• Doom 3
• Half-Life 2
• Far Cry