facebook rss twitter

Review: NVIDIA's GeForce 6600 GT AGP

by Ryszard Sommefeldt on 16 November 2004, 00:00

Tags: NVIDIA (NASDAQ:NVDA)

Quick Link: HEXUS.net/qa4r

Add to My Vault: x

System Setup and Notes

AGP vs PCI Express, NV43 vs R360

In the PCI Express board's article, I pitted it against the other main NVIDIA PCI Express products to see where it stacked up in their grand scheme of things. This time I pit the AGP version against the PCI Express version, using very similar Intel-based systems. ATI's mid-range AGP Radeons also get thrown into the mix. I use a Radeon 9800 XT 256MB to provide results from that Radeon SKU, and it's also downclocked to 380/350 to simulate the rare and slightly expensive Radeon 9800 PRO 256MB, in the absence of a 128MB regular PRO.

The results from the PCI Express board's article are thrown into the mix. 6800 GT, 6800 Ultra and 5900 PCX from that article are using 61.75, 6600 GT on PCI Express uses its launch driver - 65.76 - with the AGP 6600 GT using 66.93, NVIDIA's recent WHQL release. The two ATI boards use ATI's CATALYST 4.11 release.

All driver optimisations were left at their defaults. There's a new setting in 66.93 that allows you to force the hardware to do automatic mipmap generation even if the application doesn't request them, along with the filtering mode used to sample from them.

Driver opt
Click for a lossless version

Driver Info and Clocks

The stock clocks for the board, despite the same GC20 DRAMs as the PCI Express version, are 500/450 (900MHz DDR, some 100MHz shy of the default clocks that the PEG16X board uses). Testing was therefone done at 500/450 (500/900) and 500/500 (500/1000), to see what the performance delta is.

Reference clocks

Driver

Hardware

  • NVIDIA GeForce 6600 GT Reference Board, NV43 + BR2, 128MB, AGP8X, 500/900 and 500/1000
  • NVIDIA GeForce 6600 GT Reference Board, NV43, 128MB, PEG16X, 500/1000
  • NVIDIA GeForce 6800 GT Reference Board, NV45, 256MB, PEG16X, 350/1000
  • NVIDIA GeForce 6800 Ultra Reference Board, NV45, 256MB, PEG16X, 400/1100
  • NVIDIA GeForce FX 5900 PCX Reference Board, NV35, 128MB, PEG16X, 390/700
  • ASUS ATI Radeon 9800 XT, R360, AGP8X, 412/760 and 380/700
  • Intel Pentium 4 Extreme Edition, 512KB L2, 2MB L3, 3400MHz
  • Albatron 915P LGA775 Mainboard, LGA775, DDR400, PEG16X
  • Shuttle SS77G5, LGA775, DDR400, AGP8X
  • 512MB Corsair XMS3200LLPT, 2-2-2-6, DDR400
  • Western Digital WD360G Raptor, SATA

Software

  • Windows XP Professional w/SP2
  • DirectX 9.0c Runtime
  • NVIDIA Detonator Release 60, 65 and 66; 61.76, 65.76, 66.93
  • 3DMark 2001SE Build 340
  • 3DMark03 v3.4.0
  • Unreal Tournament 2003 v2225
  • ShaderMark 2.0
  • D3D RightMark 1.0.4.9 PB3
  • Realtime High Dynamic Range Image-Based Lighting v1.2

Notes

A subset of tests are run to establish the performance of the AGP 6600 GT variant, compared to the PEG version and its immediate peers. Testing is done at 500/900 and a PEG-matching 500/1000. Driver differences can also be compared, between 65.76 and 66.93, using the 6600 GTs.

The image quality of 66.93 appears to be almost 100% static, compared to 65.76, at a cursory glance.