First lets go through the list of components used in the test system:

• EVGA H55 (123-CD-E635-KR Motherboard)
• Intel I5 661 Processor
• G.SKill 2X2gb DDR3 2000 (PC3-16000) Ripjaw F3-16000CL9D-4GBRH
• PC Power & Cooling 750 Watt Power Supply
• Western Digital Blue 500gb SataII Hard Drive
• Thermalright HR-01 Plus CPU Cooler
• Visiontek HD3870 Video Card

We had the intention of using Everest’s (version 5.3) full battery of CPU and memory tests, but that plan was quickly thwarted as we discovered the lack of support for both the H55 chipset and the I5 661 processor. It was determined the best way to move forward was to use Sisoft Sandra’s suite of tests instead.

The first test is a quick run of HDD Tach to test the function of the Intel controller. The only BIOS option for the Intel controller is IDE mode, and below are the results of the 8mb and 32mb runs. The test results are as expected and the Intel controller appears to operate quite well. As mentioned earlier in this review, if a raid array is something a user must have, then that option is available using the  JMicron controller.

A ten pass run of LinX was also run to check for stability at stock speeds, no problems were encountered.

Once confident of stability, the Sisoft Sandra suite of tests were run. The CPU tests include the Inter-Core Bandwidth and Latency test, the Drystone ALU (GPS) and Whitestone FPU (GFLOPS) test, and the Multi-Media Int (MPixel/s) and Float (MPixel/s) test. Comparison systems built in to Sisoft Sandra vary depending on the test run. Each test was left at the default comparisons provided. The results show this motherboard/CPU combination can hold it’s own against compared systems until it runs up against X58 based systems, hardly unexpected though. Keep in mind on a couple of these tests, the lower number the better.

The Sisoft Sandra memory tests include memory bandwidth and memory latency. Again this combination did quite well, providing good numbers. Because the system is set at default BIOS settings the memory is running well below it’s rated speed. These scores should get quite a bit better as we raise the memory speed during the overclocking attempts.

To round out the stock testing a quick SuperPI 1M benchmark was performed. Not a bad showing at all for a stock system!

Immediately jumping in to a pretty hefty overclock of 4.0 Ghz, this combination had no problem at all getting there. A very small bump in CPU voltage (+.30mV) and a small bump in CPU VTT (+.100V) had the system purring along nicely. Again, we ran a quick ten pass LinX stability test.

The same three Sisoft Sandra CPU benchmarks were run again using the same comparison systems. As you can see by the results, the scores were much better at the higher clock speed.

The following memory tests again showed increased performance as the speed was raised.

And finally a SuperPI run at 4.0 GHz, looking better all the time!

If you have read my previous motherboard reviews you know the target speed for a review is a solid 1GHZ overclock. Achieving this goal took quite a bit of research and trial and error. Once you get to 4.oGHz using the 2:10 memory divider and the CPU turbo mode function, you’re going to run in to problems getting any higher of a clock unless you lower the divider to 2:8.  The reason for this is because the Integrated Memory Controller (IMC) is pretty weak on most i5 and i3 processors and begins to struggle at any memory speed over 1600 MHz. The other factor is the QPI link Speed, you need to keep that as close to your CPU speed as possible. I set this value to the max available which is 6.400 GT/s. The CPU voltage was set to +90mV and the CPU VTT was set to 1.20v and we were able to run a full twenty pass LinX Stability test with no problems at all. The bad news is that your memory bandwidth and latency scores obviously take a hit because of the lower memory divider used. The good news is that you now have additional headroom with your memory speed to work your way up beyond 4.33 GHz CPU speed.

First is a LinX stability screen shot at 4.33 GHz.

Next is the same Sisift Sandra CPU tests as run above and again using the same comparison systems.

As you can tell by the tests below, the memory bandwidth and latency take a small hit because of the 2:8 divider used here.

And finally a run of SuperPI 1M finally broke the ten second barrier.

So, there we have it, a solid 1 GHz overclock, albeit having to suffer a bit of memory performance to get there. As you can see by the LinX run screen shot above, there is still some temperature and voltage room to work with. A much higher clock speed is a definite possibility, and this test system is on air!