ASUS P5NSLI: Core 2 Duo and SLI on a Budget

FSB Overclocking Results

Front Side Bus Overclocking Testbed
Processor:	Core 2 Duo - E6300 1.86MHz Core 2 Duo - E6700 2.66MHz
CPU Voltage:	1.325v
Memory Settings:	3-4-3-10 at 667MHz
Memory Voltage:	2.1v
SPP Voltage:	1.5V
FSB Voltage:	Default
Cooling:	Scythe Infinity
Power Supply:	PC Power and Cooling 850 SSI
E6300 Overclock:	321fsb x 7 (2247MHz) +21%
E6700 Overclock:	320fsb x 10 (3200MHz) +20%

This board underperforms as an overclocker when comparing it to the Intel chipset boards with our best current in-house P965 boards hitting 7x535FSB and 10x395FSB with the same setup. At these settings the system was able to complete our benchmark test suites three consecutive times along with Dual Prime95 and SuperPI 32M without issue.

We had the LDT set at 3.5X and tried 3X but our board would not even post at 325FSB with either processor. Based upon our previous overclocking results with the C19A boards we are not surprised by the results. The current generation NVIDIA Intel Edition chipsets have never been known for their high FSB overclocking ability and we will have to wait for the next generation chipset to see better results. We will continue working with the board and will provide an update on our results once we have completed additional overclocking tests.

Memory Stress Testing
Memory Tests

This memory stress test looks at the ability of the ASUS P5NSLI to operate at the officially supported memory frequencies of 533MHz DDR2 at the best performing memory timings the TwinMOS Twister DDR2-667 will support. Note this memory is rated at 4-3-3-10 timings for 667MHz operation and was required to match our previous test results. Our Transcend memory we utilized in our previous Core 2 Duo memory performance article would not operate properly on this board at the same settings it was capable of in the VIA or Intel chipset boards we tested. This is not an issue with the Transcend memory but the memory tuning on this board.

ASUS P5NSLI Stable DDR2-533 Timings - 2 DIMMs (2/4 slots populated - 1 Dual-Channel Bank)
Clock Speed:	266MHz (1066FSB)
Timing Mode:	533MHz - Default
CAS Latency:	3
RAS to CAS Delay:	3
RAS Precharge:	3
RAS Cycle Time:	9
Command Rate:	2T
Voltage:	2.0V

The ASUS was very stable with two DDR2 modules in Dual-Channel at the settings of 3-3-3-9 at 2.0V. We were able to hold 3-4-3-10 2T at DDR2-667 and 4-4-4-12 2T at DDR2-800 with this memory set at 2.1V. The board would run 1T timings at DDR2-533 with our G.Skill, Corsair, and OCZ PC26400 modules but the differences in performance were minimal. We are still running tests at DDR2-667 with 1T timings where there is a measurable difference. However, there were several inconsistencies with other memory modules on this board. We noticed if we strayed too far away from the SPD settings that the board would lock up, generate memory errors, and corrupted a drive image that we are still investigating.

Even though several of our memory modules would easily do 3-3-3-9 at DDR2-800 on other boards we would notice memory errors at relaxed timings of 4-3-3-10 and could not run the board consistently at DDR2-800 without 4-4-4-12 timings. We feel like the issue is attributed to our 2.1V limit and the initial BIOS release that appears to be geared for conservative timings at this point. We are working closely with ASUS and NVIDIA on this issue currently. Our advice at this time is not to push the memory too far on this board until additional tuning and testing has been completed.

We will now install DIMMs in all four available memory slots, as that results in more strenuous requirements on the memory subsystem than testing two DDR2 modules on a motherboard.

ASUS P5NSLI Stable DDR2-533 Timings - 4 DIMMs (4/4 slots populated - 2 Dual-Channel Banks)
Clock Speed:	266MHz (1066FSB)
Timing Mode:	533MHz - Default
CAS Latency:	3
RAS to CAS Delay:	4
RAS Precharge:	4
RAS Cycle Time:	12
Command Rate:	2T
Voltage:	2.1V

The ASUS was completely stable with four DDR2 modules in Dual-Channel operation at the settings of 3-4-4-12. We tried several combinations of memory settings and memory modules at lower timings but the board was not stable enough to complete our test suite. Overall, the memory tuning on this board needs some additional work. Once you dial the memory in then the board is extremely stable, but we know the chipset is capable of a little more performance. We will continue our memory testing with other modules that have arrived recently and hopefully we can get a BIOS update shortly.