System Setup
Hardware
- nForce3 250 'Qualification Sample', Socket 754, nForce3 250, Athlon 64
- ASUS K8V Deluxe, Socket 754, VIA K8T800 Athlon 64, 1005.011 BIOS
- AMD Athlon 64 Model 3200+, 1MB L2, 10 x 200MHz
- Corsair XMS3200LLPT, DDR400, 2-2-2-6 (K8V), 2-3-3-8 (nForce3 250, according to CPU-Z)
- Intel Pentium 4 3.2GHz, 512KB L2, 16 x 200MHz
- Western Digital WD360 Raptor, SATA, 36.2GB
- ATI Radeon 9800XT (412/730)
Software
- Windows XP Professional w/SP1
- ATI CATALYST 4.1 and Control Panel
- NVIDIA nForce3 250 Platform Driver
- Audio driver version 4.09 (WHQL)
- Audio utility version 4.09
- WinXP ethernet NRM driver version 4.16 (WHQL)
- Network management tools version 4.16
- GART driver version 3.77 (WHQL)
- SMBus driver version 4.04 (WHQL)
- WinXP IDE 2.5 driver version 4.12 (WHQL)
- VIA Hyperion 4.51v (ASUS K8V)
- HEXUS Pifast v41
- Simplisoft HDTach 2.61
- Kribi Bench 1.19
- Sciencemark 2.0
- 3DMark 2001SE v330
- Quake3 v1.30 HQ (four demo)
- LAME 3.92MMX MP3 Encoding(192CBR, U2's Pop album)
- Realstorm Ray Tracing
- X2: The Threat - Rolling Demo
The usual HEXUS test suite as far as chipset or system level testing goes, with the nForce3 250 up against K8T800 in the form of my usual ASUS K8V reference base.
As usual, all tests were performed three times with the upper and lower results discarded, leaving the median result for reporting. If any of the test results deviated from any of the others by more than 2%, the entire result block is discarded and a new triplet of scores obtained from scratch.
If you have any query about the tests or methods we use in our benchmarking, hit our forums to discuss them with us.
Before we move onto the results themselves, a quick note about what to expect from nForce3 250, performance wise. The simple answer is not very much at all. The memory controller and CPU to memory controller interface are all on the CPU, the chipset has nothing to do with things in that respect, bar setting up the connected memory modules correctly, allowing decent timing and voltage adjustment.
The AGP implementation is unchanged from that which appeared in nForce3 150 and with that part of nForce3 150's performance being just fine, err, expect it to be just fine here. nForce3 250 isn't about performance as such, it's about features and forward thinking CPU support.
While the new bus width on the downstream HyperTransport link will help system performance under heavy peripheral I/O loads (read high bandwidth USB devices such as hard disks and WiFi networking), it's not something we specifically test with the HEXUS suite of benchmarks. I'll attempt to cover it as and when I can though.
Onwards to the graphs.
