The MSI K8N Neo2's BIOS is based off of the Pheonix AwardBIOS. All the usual suspects are present, such as Standard CMOS and Power Management features, but MSI did make several changes in favour of the enthusiast, most of which are accessible in the Cell menu page.
In the Cell menu, all your CPU and memory adjustments can be made here. For those of you who don't wish to dig through the BIOS, setting the System Performance to High Performance and enabling the Aggressive Timing option will optimize the system chipset's timing. Depending on how conservative (or aggressive) the rest of your tweaks are, you can gain a performance boost although stability may be suspect.
In the DRAM Configuration page, you can configure the maximum memory speed depending on the ram you have. Setting it to Auto will default to to the CPU's FSB. If you have anything slower than PC3200, such as PC2700, you can hard lock it to 166MHz and so forth. Other than the memory speed, you can adjust the CAS latency (Tcl), RAS to CAS delay (Trcd), Min RAS active time (Tras) and the RAS precharge Time (Trp). Lower numbers will result in better memory performance at the expense of stability, depending on your ram's willingness to be tweaked.
For the overclocking newbies, the D.O.T. Ranger option is the CoreCell's dynamic overclocking function. There are six options available, with the Private giving a 1% boost and the Commander providing a boost of up 11% over the set FSB.
The CPU Overclock option allows you to make adjustments to the FSB. The current shipping BIOS caps at 250FSB, but there will be a future BIOS that caps at 300FSB. We received this BIOS, which we'll outline in more detail in our overclocking discussion, but whether or not you can attain a 300FSB overclock will depend on your cooling and ability to unlock the CPU (all retail CPUs are locked) via the CPU Ratio option.
Speaking of overclocking, the great thing about the K8N Neo is the ability to lock down the speeds of several peripherals. We already talked about the memory, but I should add that once you push the FSB past 200, you can manually set the memory speed to 200 (a number not present unless you overclock). The AGP Overclock allows AGP adjustments to be made (up to 100MHz), and keeping it in spec (66MHz) will also keep the PCI in spec at 33MHz.
Your voltage options are decent, though nothing over the top. For DRAM, it caps at 2.85v, which should be enough for all brands of PC3200 ram. Unless you're dropping PC4400 in, this should be all you'll need. For your AGP, you can move up to 1.85v, which is quite a bit and it's unlikely you'll need anymore than this for current and upcoming AGP cards. CPU voltage caps at 1.85v, which for Socket-939 A64s, should be plenty.
Overclocking with the K8N Neo2 and our Koolance EXOS were met with mixed results. I was able to get a system POST at 240FSB with our Athlon 64 3500+, but Windows would never start up. At 230FSB, we got as far as the Windows log in screen before blue screening. No amount of voltage tweaking seemed to work, and by the fourth reboot, a reinstallation of Windows was required as the system files became corrupted.
The end result was a 216FSB overclock, with translates into a 176MHz bump in speed. An Athlon FX may result in better OC success as the multiplier can be adjusted, but we were not able to do that with our CPU. The system would POST and get into Windows, but the system was very unstable at anything between 217 - 225FSB, regardless of voltage. Temperatures were in the high 40s, so heat was not an issue, so depending on your CPU, your mileage may vary.
MSI K8N Neo2 Platinum: Athlon 64 S939 3500+ (11x200: 2.2GHz), 2 x 512MB Corsair TWINX PC4000 Pro, AIW Radeon 9600 XT, 120GB Maxtor SATA 7200rpm, Windows XP SP1, NVIDIA ForceWare 5.10, ATI Catalyst 4.9.
MSI K8N Neo: Athlon 64 S754 3200+ (10x200: 2GHz), 2 x 512MB Corsair TWINX PC4000 Pro, AIW Radeon 9600 XT, 120GB Maxtor SATA 7200rpm, Windows XP SP1, NVIDIA ForceWare 5.10, ATI Catalyst 4.9.
Test software will be:
SiSoft Sandra 2004
Business Winstone 2004
Unreal Tournament 2003
Quake 3: Arena
The comparison motherboard will be the MSI K8N Neo, running an Athlon 64 Socket-754 at 10x200. Cooling for both A64 motherboards was provided by the Koolance EXOS. The Corsair modules will be run at PC3200 with 2.5-3-3-5 memory timings. We will also be displaying the overclocked performance of 216FSB and 432MHz memory.
All our benchmarks were run on a 32-bit version of Windows XP. The 64-bit Windows still isn't ready for prime time, and we chose not to use the beta version for our tests as it may not be a true indication of the motherboard's performance. According to AMD, we may get a significant performance boost in a true 64-bit environment. In anycase, the Athlon 64 (A64) runs 32-bit code natively with no emulation.
SiSoft Sandra 2004
Although a synthetic benchmark, it's a popular one, freely available if you wish to make comparison benchmarks. We will be testing the CPU, MMX, and memory speeds, using the 32-bit 2004 version. We do have a 64-bit copy, but unfortunently it won't work on our current version of Windows.
CPU Arithmetic Benchmark
CPU Multimedia Benchmark
As expected, we see a nice performance boost with the 3500+ over the 3200+. The difference between the three test beds is proportional to the increase in clock speed. The Dual Channel memory controller present n the newer Socket-939 CPU gives the 3500+ a decisive edge in the memory benchmark.
ZD Business Winstone 2004
The ZD Winstone suite is a script that runs a series of actions and calculates a final score that measures a PC's overall performance.
Like we've seen with SiSoft, the performance gain with the newer CPU is quite obvious.
Sysmark 2004 Office Productivity
Sysmark 2004 is BAPCo's latest revision of the mainstream office productivity and Internet content creation benchmark used to characterize the performance of the business client. It uses a number of real-world applications and runs them through a series of tests. We tested with the office, and content creation benchmarks.
Things are not as clear cut with Sysmark, but nonetheless, the increase in clock speed reflects in the difference in performance. Let's move on to real-world testing.