System setup and notes
Here's a quick rundown of the test system should you wish to compare benchmark results with your own.- AMD XP2400+ CPU @ default 2GHz (133FSB [15x] and 166FSB[12x])
- ABIT NF7-S nForce2 Motherboard (in dual DDR266, single DDR333, and dual DDR333 modes)
- EPoX 8K9A2+ KT400 Motherboard (K9A22A25.BIN BIOS, in DDR266 and DDR333 modes)
- Taisol 760 cooler
- Intel 2.53GHz Northwood CPU
- Iwill P4HT-S i845PE motherboard (run in DDR333 mode)
- Thermaltake S478 cooler
Common components
- ATi Radeon 9700 Pro (324/320)
- 256MB Corsair XMS3200 C2 run at 2-5-2-2 at DDR266/DDR333 for all motherboards (another 256MB of Mushkin PC3200 at 2-5-2-2 for nForce2 dual-channel modes)
- 61.5GB IBM 120GXP Hard Drive.
- Liteon 16x DVD
- Samcheer 420w PSU
- Samsung 181T TFT monitor
Software
- Windows XP Professional Build 2600.xpclient.010817-1148
- NVIDIA nForce2 1.13 drivers
- VIA 4-in-1s, 4.43
- Intel 4.00.109 chipset drivers
- Intel application accelerator drivers
- Plutonium XP 8.1 Radeon Drivers (based on ATI CATALYST build 6166)
- Pifast v41
- Lame v3.91 MP3 encoding with Razor-Lame 1.15 front-end using U2's Pop album
- Virtual Dub 1.4.10 DVD encoding, DivX 4.12 CODEC
- OcUK SETI benchmark
- 3DMark 2001SE
- UT2003 Demo
- Comanche 4 benchmark
- Serious Sam 2 Demo
- Quake 3 v1.30
Notes
Because of the daunting number of combinations that motherboards featuring single and dual-DDR capabilities offer, I've opted to test the ABIT NF7-S at the test XP2400's native FSB of 133 with both single-channel DDR333 and dual-channel DDR266. Why have I chosen dual-channel DDR266 over dual-channel DDR333 at 133FSB ?. The answer lies in the fact that we actually lose performance by running dual-channel DDR333 at 133FSB. The asynchronous memory speed provides more bandwidth than dual-channel DDR266, but latency is increased by having to account for the asynchronous memory and CPU FSB clocks.
This increased latency manifests itself in slower overall performance. After all, with dual-channel DDR, we have more than enough bandwidth to go around. Running in single-channel mode, however, DDR33's increased bandwidth over single-channel DDR266 is enough to make it the better choice, even when we discount the latency effect. I've also chosen to run the XP2400 at 12x166FSB to show the benefits of running a faster FSB and faster memory. With the CPU's 166FSB it makes implicit sense to run dual-channel DDR333; it perfectly complements it.
I've sought to compare the nForce2's performance against a VIA KT400-based motherboard. The comparison EPoX 8K9A2+ will also be run in both 133FSB and 166FSB CPU modes. Lastly, a 2.53GHz P4 on top of an Iwill i845PE motherboard should show where we stand in a CPU comparative sense. Phew!.
Stability
Considering the harshness of the RAM timings used, albeit with excellent PC3200 RAM, and the innate complexities of ensuring that a dual-channel motherboard is stable, the ABIT nForce2 was superbly stable irrespective of what I threw at it. The fact that it's a new chipset appears to be irrelevant here. Remember that the two modules of test RAM, although rated at PC3200, have different internal timings, so them running together in dual-channel mode, with performance-enhancing timings, at DDR333, tells you all you need to know about stability under pressure.
Overclocking
Overclocking has been extremely successful on the P4 front, why ? - basically due to the design of the chipsets themselves, most notably from Intel. With rampaging Northwood processors and motherboards that are proven to lock the sensitive PCI/AGP busses, overclocking is just a matter of raising the voltage and FSB a step at a time.
The nForce2 chipset inherently features an AGP bus locking feature. That would infer a lockable PCI bus at half the AGP's speed. But this is not a VIA platform. It's quite conceivable that the AGP and PCI busses work from independent clock generators. What we can say, though, is that with DDR333 AMD CPU's, the nForce2 has to be in perfect equilibrium at 166FSB (66/33 AGP/PCI).
With that in mind, and the ABIT NF7-S completely unlocking our XP2400+, I used a single module of Corsair XMS3200 that has seen 225MHz+ at weak timings and 2.7v. Even though I could boot into BIOS perfectly at 212FSB (10x212), the motherboard steadfastly refused to boot correctly into Windows XP above 188FSB. It wouldn't even get to the Windows XP splash screen. Considering that entering BIOS can be done without having a hard drive installed, I have a sneaking suspicion that the PCI bus is unlocked above 166FSB. I also tried 205FSB in the hope that a 6:1 FSB:PCI ratio was present. No go there either. I'm a little disappointed due to the expectation of benchmarking at 10x200+ with dual-channel DDR400. This will need further investigation. It looks like a PCI-related limitation, as BIOS is fine at 200FSB+. Here is the BIOS at 212FSB. Note that it has actually booted in at those speeds shown below.
