Thursday, June 22, 2006

First Attempts to Overclock Quad Core Processor. More on Intel Kentsfield

5:15 pm | Aleksey Razin

Of course, those readers who are interested in the new generation Intel processors would hardly be satisfied with Kentsfield screenshots only. Everyone wanted to see some real overclocking results and the reports about some corresponding experiments have already hit the pages of XtremeSystems.org forums.

Before we start investigating the overclocking potential of this CPU let’s take a closer look at one very interesting peculiarity of this processor. Just like its dual-core predecessor it allows disabling “excessive” cores. Although in this case three cores out of four will be disabled. As you know, Conroe has shared L2 cache that is still available as a whole even if one of the cores has been disabled. Kentsfield consists of two Conroe dies, each featuring two cores and 4MB of L2 cache. By the way, the CPU-Z utility reflects L2 cache as 4MB, while it simply doesn’t see the second 4MB at all :)

So, if you disable three cores of the Kentsfield processor, then the remaining core will only work with 4MB L2 cache that remained in the active half of the CPU.

Maybe there will be a way to disable only two cores of the Kentsfield CPU. However, there will hardly be the need for that, mostly just to increase the overclocking potential.

With the nominal Vcore, the Kentfield processor overclocked from 2.4GHz to 3.2GHz with the air cooler. It is quite nice for a quad-core CPU, I should say. C1 core stepping sample should be able to go farther than that with higher Vcore settings and more powerful cooling solution.

By the way, they didn’t increase the multiplier of the Kentsfield processor during their overclocking experiments. Actually, all Extreme series CPUs should have unlocked multipliers. Kentsfield should also have one. Maybe the CPU owner simply didn’t check out this feature…

First Screenshots from the Kentsfield Processor. Intel's Quad-Core on the Way!

10:55 am | Aleksey Razin

We never doubted that there were engineering samples of the Kentsfield processor out there. They have already been briefly shown at Intel’s booth during Computex 2006 and have also been floating around the company’s partners. Of course, the information leaks couldn’t be prevented completely and one of the Kentsfield based processors appeared in XtremeSystems.org forums.

Take a closer look at the marking of this engineering sample: Intel does specify the L2 cache size (8MB), while in fact the cache works as 2 x 4MB. This Kentsfield sample is none other but two Core 2 Duo E6600 (2.4GHz) cores combined in a single package, which explains the strange description on the CPU-Z screenshot:

The processor is based on B0 core stepping, but there is still at least half a year left before the launch, so the stepping may change more than once until then. As for the TDP value, there is no exact data about it yet, but Kentsfield processors work in the currently existing mainboards supporting Core 2 Extreme. The latter however require the mainboards to comply with 05A (TDP=95W) specification, although the actual TDP value for Core 2 Extreme X6800 (2.93GHz) doesn’t exceed 75W. In this particular case Kentsfield worked in MSI 975X Platinum mainboard revision 2.08.

As you see, there are four actual cores in the system while there is only one CPU. Looks like we will very soon see not only the first Kentsfield benchmarks but also will be able to estimate its overclocking potential.

Celeron D 356 (3.33GHz) Reaches 5.4GHz. Intel offers Overclockers 50%-60% Frequency Gain

10:52 am | Aleksey Razin

Now that there is so much stir around the upcoming Conroe processors launch, some overclockers may have forgotten that there have recently appeared 0.065micron Celeron D models that are still based on CedarMill C1 core stepping. In the meanwhile this core stepping is quite rare among retail Pentium D 9xx CPUs, however Celeron D 356 (3.33GHz) and Celeron D 352 (3.2GHz) have it by default. Moreover, these CPUs should become cheaper in the end of July and by fall they may even acquire new D0 core stepping.

By the way, a lot of mainboard makers mention in their compatibility list that there might be 0.065micron Celeron D processors with B1 core stepping. Of course, this is no catastrophe, but these processors may overclock slightly worse. From the economical standpoint it is more beneficial for Intel to use old core steppings for budget CPUs.

We know of a few cases when Celeron D processors on 0.065micron CedarMill core overclocked to 5.0GHz with powerful air coolers. It was not only the significant frequency potential of the 0.065micron core, but also the high nominal clock frequency multiplier of the CPUs. XtremeSystems.org forum reports the results of Celeron D 356 (3.33GHz) C1 core stepping overclocking. The CPU reached 5.4GHz frequency with only an air cooler.

To be more exact, they only managed to take a screenshot at 5.4GHz frequency, however the Vcore was relatively low: 1.4V. And at 5.1GHz frequency and nominal Vcore the CPU passed all the tests flawlessly. The 50%-60% frequency increase without much effort – far not every CPU can boast something like that.

By the way, it was a regular boxed processor that they got from a regular computer store. Here is what the high nominal clock multiplier of 25x can do: to reach 5.4GHz frequency they hade to raise the system bus frequency only up to 218MHz. Celeron D processors formally support 533MHz bus, which corresponding to 133MHz system bus frequency. Celeron D 3xx models based on 0.065micron core have twice as large L2 cache than the predecessors: 512KB against 256KB. Larger cache also affects the performance level, and of course so does the excellent overclocking potential.

Conroe Overclocking Experience: Up to 500MHz Bus Frequency. Voltmodding and Liquid Nitrogen Work Wonders.

10:47 am | Aleksey Razin

As we have already seen, the overclocking potential of Conroe processors improves every time they switch to a new core stepping. However, the range of supported nominal frequency multipliers remains unchanged. In this case it will be the ability of the mainboard to support high bus frequencies that will play the crucial role in successful processor overclocking. As we have already said several times in our previous news stories, the mainboards specifically designed for Core 2 Duo and Core 2 Extreme have some real advantages over the slightly modified i975X based boards that have been so desperately wanted by the first lucky Conroe owners.

In the meanwhile, voltmodding and extreme cooling solutions may always help make up for the insufficient overclocking potential of the mainboard you have. For example, XtremeSystems.org forums describe a very unusual result of Conroe B0 core stepping overclocking reported by a Japanese user. His Intel D975XBX mainboard revision 302 modified to support Conroe processor allowed reaching 500MHz bus frequency:

So far we believed that 400MHz is the necessary “overclocking minimum” that should allow younger Conroe models to reach more or less significant frequencies. If the mainboard allows reaching 500MHz bus frequency, Core 2 Duo E6300 (1.86GHz) processors will work at 3.5GHz frequency with 7x clock multiplier. This is the maximum you should be able to achieve with a default cooling solution and nominal voltage settings. In other words, if the mainboard permits, you will be able to achieve very good overclocking results even with the youngest Conroe models.

By the way, next year Intel will launch Core 2 Duo E4200 (1.6GHz) processor that will support 800MHz bus and 8x clock frequency multiplier. It will overclock pretty well in less extreme mainboards that offer modest ranges for bus frequency increase. The mainboards better suited for overclocking that will support high clock frequency multiplier settings will make this CPU a real overclocking hit (thanks to its low price).

In the case we are talking about right now they used Core 2 Duo E6700 (2.67GHz) overclocking with a liquid nitrogen cooling system and performed significant voltage regulator modifications on the mainboard. As a result, the bus frequency hit 500MHz, which is a great result for a pretty old mainboard, I should say. The frequency multiplier reduced to 9x made the CPU run at 4.5GHz. It is quite normal result for Conroe B0 core stepping and with liquid nitrogen cooling system the platform can run stably under workload.

Of course, not every Conroe owner will be brave enough to modify the voltage regulator on his mainboard or can get hold of liquid nitrogen. However, this example shows very well that if you wish you can squeeze quite a bit of speed even from the mainboard that is initially not intended to support Conroe processors.

Archive
All Latest News