Performance and Overclocking
We tested Foxconn ELA mainboard in an open testbed with the following hardware and software components:
- Mainboard: Foxconn ELA, BIOS P03 from 09.23.2008;
- CPU: Intel Core 2 Duo E8400 (3.0GHz, 333MHz FSB, 6MB L2 cache, Wolfdale, rev. E0, 1.25V Vcore);
- Memory: 2 x 1024MB DDR2 Corsair Dominator TWIN2X2048-9136C5D (1142MHz, 5-5-5-15, 2.1V voltage);
- Graphics card: Nvidia GeForce 8800 GTS 320 MB;
- HDD: Samsung SP2504C (250GB, SATA II, 7200 RPM, 8MB, rev. A);
- CPU cooler: Zalman CNPS9700 LED;
- Thermal compound: Noctua;
- PSU: Antec NeoPower HE 550 (550W);
- OS: Microsoft Windows Vista Ultimate SP1 (32-bit).
Foxconn ELA mainboard first boot at nominal speeds went on well, without any problems. The mainboard reports sufficient information on boot-up including the BIOS version and release date, processor model name, its current frequency, memory timings…
The mainboard loaded the OS successfully having set for 800MHz memory frequency nominal timings; the Performance Level was set at 7 for one channel and at 8 for another. However, it turned out that we can’t find out the current processor core voltage not in the PC Health Status section of the mainboard BIOS, not in the Foxconn’s brand name FoxOne program, not in any of the third-party utilities. The Vcore was formally changing depending on what was set in the BIOS, but then remained the same independent of the power-saving technologies or the absence of the processor workload. Maybe the absence of processor Vcore voltage monitoring is the problem of our particular mainboard sample. Well, let’s try measuring it “by touch”.
We have already mentioned in the beginning of our article that the board is equipped with FSBSEL1 Connector set of jumpers that allows changing the startup bus frequency…
Besides Intel default frequencies of 266, 333 or 400MHz, you can also set 450MHz right away. By lucky coincidence, it is exactly the bus frequency at which our Intel Core 2 Duo E8400 processor can work even without increasing its Vcore. We set the jumper accordingly, start the system normally. However, the FSB frequency remained at the nominal 333MHz. As you remember, Foxconn ELA mainboard with BIOS version P03 sets very low processor core voltage, therefore we need to increase it to the nominal value first. We did, and then we raised it even more, but the board kept booting in its default mode. We set the FSB frequency to 400MHz using the jumpers, but nothing changed: FSBSEL1 Connector turned out simply non-operational. It was frustrating, but not fatal; let’s try overclocking our system using good old BIOS technique. This is when we immediately find out that the board is not cut for overclocking.
Many manufacturers make certain routine BISO procedures automatic. Take, for instance, “Spread Spectrum” parameter. It allows “spreading” the electromagnetic distortions spectrum emitted by the mainboard working at high frequencies thus lowering the harmful EMI for surrounding components, so it is enabled by default. However, this parameter has very serious negative effect on the mainboard’s overclocking potential, so it would normally be disabled during overclocking. Knowing this, many mainboards from other manufacturers would set “Spread Spectrum” to “Disabled” on their own once you start increasing the FSB frequency. If you only increase the FSB frequency a little bit, the negative effect from “Spread Spectrum” parameter will not affect overclocking, so it can be turned back on.
Foxconn ELA mainboard will have you disable “Spread Spectrum” parameter manually. However, it will automatically disable all processor power-saving technologies by setting “Disabled” for “C1E Function” and “EIST Function” as soon as you increase the system bus frequency by at least 1MHz. You can turn them back on, but the board will disable them again after your restart the system.
The most interesting thing here is that Foxconn ELA mainboard ignored any FSB settings changes in the BISO, just like it did with jumper settings. The FSB frequency always was at the nominal 333MHz. There are mainboards that are great for overclocking, there are some that overclock worse than the others, but there are no boards that cannot overclock CPUs at all while all corresponding settings are there. So, after a while we managed to locate the source of the problem: it turned out “Auto Detect PCI Clock” parameter, which is pretty harmless and even useful on other mainboards. It allows disabling the idle PCI slots: if during system boot-up it detects an expansion card installed into the slot, this slot receives power, and if the slot is empty, it needs no power and none goes its way. We didn’t think it could cause any problems and set it to “Enabled” right from the start, and then it turned out that we shouldn’t have done it. Only when we disabled “Auto Detect PCI Clock” FSB frequency started changing – looks like it was yet another BISO bug. Just in case, we decided to check the FSBSEL1 Connector jumpers once again, but in this case it was not the fault of “Auto Detect PCI Clock” parameter: setting the FSB frequency with these jumpers didn’t work again.
To find out the maximum FSB frequency at which the mainboard can work stably, we usually increase the chipset voltage, but lower the memory frequency and processor clock multiplier to their minimums. In this case it turned out that Foxconn ELA mainboard has problems with arithmetic. The “Target CPU Core Speed” informational parameter still promised us the same processor frequency as if the multiplier had never been lowered at all. Moreover, the same higher processor frequency is displayed on first system start, and only after a successful reboot, the mainboard seems to be suddenly recollecting and starts showing the correct value.
I specifically stressed that only after a successful boot-up. Shortly after we started our overclocking experiments, we discovered that Foxconn ELA mainboard doesn’t really know what “Watchdog Timer” is. This technology has many names, individual for each mainboard maker, but the idea is the same for all: watching out for startup POST process. If the booting stalls at some point, usually if an optimistic overclocker set too high frequencies or didn’t increase the voltages enough, Watchdog Timer detects the problem and reboots the system in safe mode. The user gets a chance to access BIOS Setup and correct the wrong parameter settings.
Unlike many other contemporary mainboards, Foxconn ELA has only two possible ways to act. First – the mainboard doesn’t start at all. In this case there is not much you can do and you have to resort to Clear CMOS button and then reset the BIOS parameters all over again, because the board doesn’t allow saving any settings profiles. Second – the board boots and passes POST successfully. However, don’t start celebrating just yet: it doesn’t matter if the board can load the OS and pass stability tests. The most important thing is that it survives the next system reboot. Unfortunately, you may have already guessed that the board would hang during most reboots and wouldn’t start again. And in this case, see above: everything ended with desperate beeping of the PC speaker, Clear CMOS button and another BIOS settings configuration session.
The experiments aimed at finding the maximum FSB frequency by lowering the processor clock multiplier also ended with no success – the board was very unstable. Overclocking with the nominal multiplier was not any better – the board behaved absolutely unpredictably. Even at not very high for our processor FSB frequency of 400MHz the board could only boot once, refused to reboot second time and unable to start third time. As a result, when the annoying beeping of the PC speaker drove me nuts and I couldn’t handle resetting the BIOS settings and adjusting them over and over again, I decided to give up all overclocking attempts. Yes, overclocking is an important and remarkable part of today’s computer reality, many computer enthusiasts purchase expensive mainboard specifically for overclocking , but not everyone out there is an overclocking junky and some people do not even want to read about it.
We know a lot of mainboards that overclock processors nice and easy. Well, Foxconn ELA is not one of them, but it is not a big deal. Let’s give up trying to make it work at higher speeds and focus on optimizing its performance in the nominal mode. Take, for instance, DDR2 Corsair Dominator TWIN2X2048-9136C5D memory modules that we used in our tests. Many mainboards can read the info from the extended SPD registers of these memory modules – these mainboards would set higher frequency and voltages for these modules on their own. Foxconn ELA mainboard can’t do it on its own, but it has all the necessary functionality and we are going to take advantage of it right now.
So, we increased the voltage on the memory modules and the chipset, just in case. We set 1:3.2 divider for the memory, which corresponds to 1067MHz frequency at the nominal FSB of 333MHz. The system started fin e and loaded the OS. Performance Level setting turned out even lower that at 800MHz memory frequency: 6 for one channel and 7 for another. It was way better that with 1:1 memory divider, because in this case the Performance level was usually set at 9. However, all utilities for some reason reported 888MHz memory frequency instead of 1067MHz as was set in the BIOS. 888MHz memory frequency corresponds to 1:2.665 divider, so what will happen if we set it like this? What memory frequency will the board boot with? However, we couldn’t satisfy our curiosity, because the board refused to boot, which meant that we had to go back to clearing the CMOS and resetting all BIOS parameters again. And we have been trying so hard to avoid this notorious procedure! It’s a pity that Foxconn ELA mainboard wouldn’t let us do even that little…