facebook rss twitter

Review: NVIDIA GeForce 7950 GX2

by Ryszard Sommefeldt on 5 June 2006, 14:18

Tags: XFX (HKG:1079)

Quick Link: HEXUS.net/qafvm

Add to My Vault: x

System Setup and Notes

All cards were tested on modern AMD Athlon 64 FX-62 platforms, following AMD's recent launch of their new CPU socket. ATI hardware was benchmarked on the Radeon Xpress 3200 reference board, the NVIDIA cards were tested on an ASUS M2N32-SLI nForce5 590 SLI mainboard. Full specs are as followed.

Hardware

  • NVIDIA GeForce 7950 GX2 Test Systems
Processor(s) AMD Athlon 64 FX-62, 2.8GHz, 90nm, Windsor, 1MiB-core, AM2
Mainboard(s) ASUS M2N32-SLI Deluxe (nForce5 590 SLI)
ATI Radeon Xpress 3200 Reference Board
Memory 1GiB (2 x 512) Corsair XMS2-8500
2GiB (4x 512) Crucial Ballistix PC8000
Memory Timings 4-4-4-12-1T @ 400MHz
4-3-3-8-2T @ 400MHz
BIOS Versions 0017
08.00.13
Disk Drive 160GB Seagate Barracuda ST3160812AS 7200.9 SATA2
Graphics Card(s) NVIDIA GeForce 7950 GX2 (500/600)
XFX GeForce 7950 GX2 (570/775)
GeForce 7900 GTX x 2 (650/800)
ATI Radeon X1900 XTX x 2 (650/775)
Graphics Driver NVIDIA ForceWare 91.27 (GTX), 91.29 (GX2s)
ATI CATALYST 6.5
Operating System Windows XP Professional, SP2, 32-bit

[advert]

Software

  • Game Software
  • Far Cry
  • Splinter Cell
  • Quake 4

Notes/Excuses

Yes, the ATI system has 2GiB of system memory and yes, we didn't retest the GTX boards with 91.29. For our sins, we're convinced that neither make a difference big enough to affect the overall picture we'll present at the end. Feel free to chastise in the HEXUS.community, and as your article author I entirely sympathise.

It's our usual slimmed-down games test suite for this one (sadly), given the hilariously long amount of time we had to spend with it, before Computex. Competition is fierce; single board solutions fight single board solutions, and dual GPUs fight dual GPUs, so I've split out the graphs to make it easy to follow.

Otherwise, the usual rules apply. All benchmarks were run three times and the median value reported. If any of the collected figures deviated by a substantial amount we ran them all again to generate reliable results.