#981 ·
Read this if you actually want a decent motherboard,
Overclocking -
Overclocking a Core 2 Duo processor is a relatively straightforward task, though I suppose it largely depends on your motherboard—specifically, its ability to push the CPU to its absolute limits. Our first attempt at overclocking was, frankly, a disaster. The Epox board we were using, which featured a P965 chipset, just wasn't up to the job. It lacked any real overclocking potential, and even though that chipset is usually reliable in this department, things were made worse by some questionable design choices regarding the memory. For some strange reason, the manufacturer decided to omit the current memory frequency relative to the FSB in the BIOS. This meant the user was forced to manually calculate the clock speeds based on the FSB and the memory multiplier. It’s a clumsy setup, especially for anyone just starting out. In the end, we gave up on Epox and switched to the aforementioned Abit board, which we knew could actually handle some pressure.
Hitting 2.8 GHz without raising voltage
When you're overclocking, there are a few things you really ought to watch out for. Ideally, you should find the maximum FSB frequency the motherboard can handle right away. To do this, you should lower the CPU multiplier so the actual operating speed doesn't compromise system stability. You also have to be mindful of memory ratios (1:1, 4:3, 5:4, etc.) and the NB strap values in the BIOS; you essentially try to keep the memory frequency as close to its rated speed (533, 667, or 800 MHz) as possible by balancing those two values. Experienced users might also identify the maximum possible memory frequency, which becomes a key target. Once that's settled, we return the CPU multiplier to its highest setting and gradually ramp up the FSB, keeping a close eye on the memory frequency and testing the overall system stability, particularly the CPU. Just to warm things up a bit, without bumping up the voltage, we pushed the CPU from its default 1.8 GHz to 2.8 GHz, roughly 1000MHz. The processor handled the strain without much fuss, which led us to get a little too optimistic about what was possible.
Ceiling at 3.3GHz
As you can see from the results, 3.3 GHz was the final stable frequency we managed to squeeze out of this processor. Beyond that, it simply wouldn't go, regardless of how much voltage we threw at it or what combinations we tried. From what I've read online, these processors can reach 3.6 GHz or even higher, but our very early Intel test sample wasn't exactly in the mood for extra effort. The last frequency where we could actually boot into Windows was slightly under 3.6 GHz, but there was no way it would stay stable enough to run a single test. The ideal scenario for overclocking is when you hit the maximum stable CPU clock combined with the highest memory frequency at the lowest possible latencies. In practice, if you have memory that hits a max speed of 900MHz at 4-4-4-15 latencies, you try to maximize the CPU speed while maintaining those specific memory settings. That way, you end up with a high CPU clock paired with fast, low-latency memory for the best results. But that's just the tip of the iceberg; performance depends on countless other details we won't dive into here.
Super PI at 3.3GHz
Even though the voltage shown in the image is 1.55 volts, this test sample is actually capable of running normally at 3.3 GHz with just 1.45 volts; the higher voltage in the photo was merely an attempt to chase even higher speeds. The maximum temperature the CPU reached at this clock speed was 42 °C, though that was achieved in an exceptionally well-ventilated case (with 3x120 mm fans) and under a Zalman cooler, so the conditions were more than adequate. One of the main complaints in other tests is usually the operating temperature at max clock speeds, but that didn't seem to be an issue for us. Interestingly, the CPU hits 3.0 GHz easily at factory voltages, but once you try to go faster, you immediately need to bump the voltage above 1.4 volts. Judging by the various results floating around the web, I guess that's just a general rule of thumb.
Overclocking -
Overclocking a Core 2 Duo processor is a relatively straightforward task, though I suppose it largely depends on your motherboard—specifically, its ability to push the CPU to its absolute limits. Our first attempt at overclocking was, frankly, a disaster. The Epox board we were using, which featured a P965 chipset, just wasn't up to the job. It lacked any real overclocking potential, and even though that chipset is usually reliable in this department, things were made worse by some questionable design choices regarding the memory. For some strange reason, the manufacturer decided to omit the current memory frequency relative to the FSB in the BIOS. This meant the user was forced to manually calculate the clock speeds based on the FSB and the memory multiplier. It’s a clumsy setup, especially for anyone just starting out. In the end, we gave up on Epox and switched to the aforementioned Abit board, which we knew could actually handle some pressure.
Hitting 2.8 GHz without raising voltage
When you're overclocking, there are a few things you really ought to watch out for. Ideally, you should find the maximum FSB frequency the motherboard can handle right away. To do this, you should lower the CPU multiplier so the actual operating speed doesn't compromise system stability. You also have to be mindful of memory ratios (1:1, 4:3, 5:4, etc.) and the NB strap values in the BIOS; you essentially try to keep the memory frequency as close to its rated speed (533, 667, or 800 MHz) as possible by balancing those two values. Experienced users might also identify the maximum possible memory frequency, which becomes a key target. Once that's settled, we return the CPU multiplier to its highest setting and gradually ramp up the FSB, keeping a close eye on the memory frequency and testing the overall system stability, particularly the CPU. Just to warm things up a bit, without bumping up the voltage, we pushed the CPU from its default 1.8 GHz to 2.8 GHz, roughly 1000MHz. The processor handled the strain without much fuss, which led us to get a little too optimistic about what was possible.
Ceiling at 3.3GHz
As you can see from the results, 3.3 GHz was the final stable frequency we managed to squeeze out of this processor. Beyond that, it simply wouldn't go, regardless of how much voltage we threw at it or what combinations we tried. From what I've read online, these processors can reach 3.6 GHz or even higher, but our very early Intel test sample wasn't exactly in the mood for extra effort. The last frequency where we could actually boot into Windows was slightly under 3.6 GHz, but there was no way it would stay stable enough to run a single test. The ideal scenario for overclocking is when you hit the maximum stable CPU clock combined with the highest memory frequency at the lowest possible latencies. In practice, if you have memory that hits a max speed of 900MHz at 4-4-4-15 latencies, you try to maximize the CPU speed while maintaining those specific memory settings. That way, you end up with a high CPU clock paired with fast, low-latency memory for the best results. But that's just the tip of the iceberg; performance depends on countless other details we won't dive into here.
Super PI at 3.3GHz
Even though the voltage shown in the image is 1.55 volts, this test sample is actually capable of running normally at 3.3 GHz with just 1.45 volts; the higher voltage in the photo was merely an attempt to chase even higher speeds. The maximum temperature the CPU reached at this clock speed was 42 °C, though that was achieved in an exceptionally well-ventilated case (with 3x120 mm fans) and under a Zalman cooler, so the conditions were more than adequate. One of the main complaints in other tests is usually the operating temperature at max clock speeds, but that didn't seem to be an issue for us. Interestingly, the CPU hits 3.0 GHz easily at factory voltages, but once you try to go faster, you immediately need to bump the voltage above 1.4 volts. Judging by the various results floating around the web, I guess that's just a general rule of thumb.
