Overclocking Experiments

We ran our tests in an open testbed configured as follows:

• Mainboard: Biostar TP45 HP, rev. 5.0, BIOS from 07.10.2008;
• CPUs:
  • Intel Core 2 Duo E8400 (3.0GHz, 333MHz FSB, 6MB, Wolfdale, rev. C0);
  • Intel Core 2 Quad Q9300 (2.5GHz, 333MHz FSB, 6MB, Yorkfield, rev. M1);
• Memory: 2x1024MB Corsair Dominator TWIN2X2048-9136C5D;
• Graphics card: NVIDIA GeForce 8800 GTS 320MB;
• HDD:  Seagate Barracuda 7200.10 ST3320620AS (7200RPM, 16MB, SATA 320GB);
• CPU cooler: Zalman CNPS9700 LED;
• Thermal compound: Noctua;
• Antec NeoPower HE 550 PSU (550W);
• Microsoft Windows Vista Ultimate 32-bit.

During the first run of tests Biostar TP45 HP mainboard didn’t really impress us with its performance. Our Intel Core 2 Duo E8400 CPU can overclock to 4.05-4.1GHz, which we have already proven in numerous previous test sessions. With the nominal clock frequency multiplier of x9 it means that the FSB speed will be increased to 450-455MHz. we also increased the processor Vcore, Vchipset and Vmem and set the minimal memory frequency using 1:1 divider. The mainboard would start fine and load the OS, but would fail all stability tests. We tried increasing the voltages a little more to stabilize the system, but nothing helped. However, the problem turned out to be in the DDR2 Enhance Mode parameter. Once we changed it from Auto to Disable, all stability tests at 450MHz FSB were passed with flying colors.

Now that we know the settings for the board to successfully overclock our processor to 4.05GHz frequency, we decided to find the maximum FSB frequency when the system would remain stable with the CPU overclocked to its maximum. At first we have to find out the maximum bus frequency at which the board can boot and work normally. We lowered the processor clock frequency multiplier to the minimal x6 setting and set the FSB at 510MHz…

From that moment on our experience with Biostar TP45 HP turned into some kind of overclocking frenzy. Every time I was setting the FSB frequency at a higher value I expected that the board wouldn’t boot any more or if it booted, to fail loading Windows. When my expectations didn’t come true, I launched stability tests expecting to see an error message. But, things were going smoothly. And even is there were a few glitches they could be cured by raising the voltages a little bit. It was a really outstanding experience.

I believe I don’t have to describe all the details here. I just have to say that we stopped at 600MHz FSB –my personal record so far. This is when the system failed to pass our stability tests. The most important thing is that it was not the limitation imposed by Biostar TP45 HP mainboard. The board performed very confidently and everything indicated it could keep going. What prevented us from the triumphal continuation were our DDR2 memory modules that couldn’t work stably at 1200MHz. This is one of those situations when we wished the boards could support DDR3 SDRAM, which can easily go past this frequency.

As you remember, our Intel Core 2 Duo E8400 processor worked with x6 clock frequency multiplier, i.e. was only partially overclocked.  A few quick calculations showed that we could overclock our CPU to the maximum frequency of 4.05GHz by setting the FSB to 579MHz and the clock multiplier to x7. However, the board would slightly lower the bus frequency, so we set the FSB to 580MHz.

We were very pleased to see that the system passed our stability test.

It is interesting that processor power-saving technologies usually get completely or partially disabled when you increase the processor Vcore or change its clock frequency multiplier. However, despite both these factors Biostar TP45 HP lowered the processor core voltage and reduced the multiplier to minimal x6 in idle mode! Looks like Biostar can provide training to all other mainboard developers out there who can’t implement this feature in their products yet.

However, despite the inspiring success, we still had some doubts about the reliability of the system in this mode. And again our concerns were connected with the Corsair Dominator TWIN2X2048-9136C5D memory modules we were using. We have already tested these modules before and with 5-5-5-15-2T timings they could only work at 1155MHz maximum. Now the frequency was 1161MHz. To see if our concerns were justified, we launched Prime95 utility and the results turned out quite predictable:

We tried several times with different memory settings. The screenshot above shows the “best” results, when the error message appeared 9 minutes into the test. Usually we could last 5 minutes tops and sometimes the system couldn’t run even for a minute. In other words, Intel Burn Test utility may be a good stability testing tool for an overclocked processor. Due to extremely high CPU temperatures during the test it may suit perfectly for processor coolers testing. However, when it comes to memory stability, it is not efficient enough, so it can’t be used to evaluate the overall system reliability.

This was a great opportunity for us to check out OCCT utility as well, as we were not quite happy with it before. Yes, our concerns proved absolutely justified: this program reported that the system was completely stable and a 30-minute test run was successfully completed.

The screenshot above shows that the processor frequency has been lowered by the processor power-saving technologies after the test has been completed. Unfortunately, OCCT is also not an ideal application for complex stability tests.

We have slightly increased the processor frequency by setting the FSB to 580MHz. However, we couldn’t achieve stability even when we lowered this setting.

When we set the FSB to 579MHz in the BIOS, the board lowers this setting a little bit and in fact the FSB speed is 577MHz. the CPU frequency also appears a little lower than the 4.05GHz we promised you, and the memory frequency is at 1155MHz, which is the maximum it successfully passed the tests at before. At that time, however, we had an Nvidia nForce 680i SLI based mainboard. Despite various issues with Nvidia chipsets, they are known to be more flexible when it comes to memory than Intel chipsets. So we failed to repeat the former success.

Well, looks like we will have to reduce the FSB frequency and increase the CPU multiplier because of the memory this time. With the multiplier set at x7.5 we had to increase the bus speed to 540MHz in order to overclock our processor to the promised 4.05GHz.

This time the memory should do just fine and all the test applications we use including Prime95 confirmed that system was absolutely stable.

So, Biostar TP45 HP mainboard demonstrated truly phenomenal results during dual-core processors overclocking. However, quad-core CPUs turned out a serious problem for it. And not just in terms of overclocking, but in terms of regular operation, too.

Before we installed Intel Core 2 Quad Q9300 CPU we used the jumper to clear CMOS, but the board didn’t start. The power was on, fans were spinning, but the board didn’t star. We checked the contacts at the bottom of the CPU and the socket pins, we reinstalled the processor, but nothing changed. Sometimes the mainboards get stuck like that and nothing helps except replacing the CPU. The same happened this time: we returned back Intel Core 2 Duo E8400 CPU for a few seconds, merely enough for the board to boot, and the next moment it started working just fine with Intel Core 2 Quad Q9300. So, at first we didn’t pay any attention to this matter.

Our quad-core Intel Core 2 Quad Q9300 processor can overclock to 490MHz FSB on very few mainboards, such as Gigabyte GA-EP45-DQ6 or ZOTAC nForce 790i-Supreme. However, on most mainboards quad-core processor overclocking stalls at 450MHz FSB or close to that. Keeping in mind how remarkably Biostar TP45 HP mainboard performed during dual-core processor overclocking, we decided to start a little higher than the average: at 460MHz FSB. That appeared too much, the board rebooted automatically, but then refused to boot again with exact same symptoms.

Another round of joggling the CPUs and we were ready to give it a try at 430MHz FSB: the board booted Windows with a few graphics artifacts. At 440MHz FSB the board could only boot and at 450MHz it would get back into the dead freeze that could be cured only by replacing the CPU. It was interesting, however, that the board booted just fine with Intel Core 2 Extreme QX6700 processor, which meant it didn’t like 45nm CPUs in general or our Intel Core 2 Quad Q9300 in particular. It was impossible to replace CPUs after every failure that is why we decided to give up our quad-core CPU overclocking experiments on Biostar TP45 HP mainboard.