Test setup

For our current DFI adventure we put together the following test system:

Test Setup
CPU	Intel E8600 Core 2 Duo (rev. E0) Intel QX9650 Core 2 Quad (rev. C0) Cooled by TTIC "Big"
Mainboard	DFI Lanparty UT X48-T3RS
Memory	2x 1Gb OCZ PC3-16000 Flex II 2x/4x 1Gb OCZ PC2-12800 EB Platinum
Other	Silverstone TJ09 Temjin Tower Case Leadtek GTX280 NEC DVD-RW drive Wester Digital Raptor WD740 SATA Gigabyte I-ram 4Gb OCZ ProXstream 1kW Watt PSU

As you can see, we have used only the current motherboard for our testing, but varied it with different cpu's and memory. We did not run a complete benchmark suite as I didn't find the stock results very interesting. The overclocked results on the other hand could be interesting, but the results were not as consistent as I would have hoped. Therefore, we will stick to the overclocking of the board itself, showing you some results of what it is capable of when equipped with quad core or dual core cpu's, and with different kinds and loads of memory.

For those that do wonder if the board is fast when compared to other boards, I can assure you : yes it's as fast as the other DDR3 offerings, as long as you keep the performance level low enough (stock setting is PL6) and the memory timings low enough. In my opinion, differences with other boards at stock performance are too insignificant to show, so I didn’t' bother.


Dual core performance

Let's start off with the dual core performance of the board. For this little adventure we've put our hands on Intel's latest Core 2 Duo CPU: the E8600. As many of you know, the E8600 is part of a new stepping of Core 2 Duo cpu's, the E0 stepping : E0 promises better overclocking results and cooler running CPU's. With this in mind it was with great anticipation that we got started. Small difficulty was the fact that the early bios of the board did not support the E0 just yet, a second difficulty the fact that no info on overclocking E0's on this board was available yet. In other words, we started with a blank sheet of paper.

When starting our overclock, we wanted to test several things. How high will the CPU go in absolute speed? What's the front side bus limit of the CPU or the board? How high will the memory scale? Every one of these questions has a different methodology, so we'll have to take them one at a time.

One of the main choices you have to make when overclocking is which memory divider to choose. With DDR3, frequencies are higher, so we got a few more dividers to choose from. Here's where the problems really started though, as all the biosses we had at our disposal had breached memory dividers, one bios worse than the other. In the end, only two dividers could be used: either the 1:2 dividers or the 3:5 divider. Of these two, only one divider really was trouble free, and that is the 1:2 dividers. 3:5 was usable only when certain options were set in bios, another thanks to Praz over at ocxtreme to find this one out. I was told that Oscar chose to make the dividers accessible one at a time, so he tried to get them right one at a time. The choices he made to give us access to the 1:2 divider first, the 3:5 second, and the others maybe on a later date, are actually the correct choices and should give you ample ways to overclock your CPU. The problem is: there shouldn't be any choices to be made at all, all of the dividers should "just work".

When starting the testing on the board, I quickly came to the conclusion that the bios (or the board) fails one really important setting, and that's an "auto-recovery" function. Whenever an overclock fails, and the board fails to boot, one of the features that comes in handy is a function that makes the board go back to the "optimized default" setting automatically. The DFI board does not feature this function at all, so whenever an OC went bad, I had to hard reset the bios. Luckily, you don't have to remove the bios jumper any more but you can do this by pressing the "power" and "reset" button on the board (or your case) at the same time for a couple of seconds. You do have to power down the board first in many occasions though. A button on the back plate, as Asus and some others started offering not long ago, is also a great alternative, but also not present. All of this is a real shame, especially because overclocking proved to be a very time-consuming process.

As in the early stages of my testing only the 1:2 divider was working properly, this was what I've been testing. My hopes were up very high as I used the OCZ PC16000 Flex II memory (Samsung based) in my setup, to get the board to 500FSB and beyond, while clocking the memory over 2000Mhz, something they should easily be capable of.



Can you imagine the disappointment that, even after several weeks of tweaking, I couldn't get my board up to 500FSB (or 2Ghz on the ram). I can assure you, it hasn't been for a lack of trying, as I've tried every trick up my sleeve, tuned the memory in every way I thought possible, and watched the forums of the "big guys" trying to make this board behave every day. In the end, Praz did manage to get his board to 500FSB/2Ghz on the ram, completely stable (but not any higher than that). Only Anand has shown us a small preview of the Corsair DDR3-2100 memory running at a FSB of 533 (with 1:2 dividers). I've never seen this repeated though, and a definitive Anandtech X48-T3RS review never made it to the site (so far). Myself, I had to settle for 485FSB Orthos stable. Might have squeezed a little more out of it, but 490 was not stable in the long run. The screen shows a multi of 8.5 in this case, but in case you wondered: the multies did not have a huge impact and the E8600 chip is capable of far more than 4.2 Ghz.

With bios 725 (and especially 829), and the help of ocxtreme, finally came support of the 3:5 divider. Although the 3:5 setting has been a mixed bag on several motherboards, due to occasional misreporting of DDR3 speeds, it seemed to work fine for me. The divider finally opened up the FSB tuning, so have a look where the new bios got us :



530FSB fully Orthos stable is not too bad, but also not enormous when looking at offerings from the competition. Where the DFI board did shine though is in the performance level it could sustain when clocked this high. 530FSB was possible using an tRD setting of 9 on reasonable voltages, and even PL8 on rather high voltages (too much for 24/7 use anyway, unless you put it on water-cooling, which I didn't do). Only a couple of days ago Praz was able to tune the board up to 540FSB (PL9), but even Oscar himself says that we shouldn't expect anything beyond this level. The thumbnail below, to the right, is the Orthos screen at 530FSB. It's a shot at only 25 minutes; you'll have to take my word for it that it ran like this for about two hours before I shut it down.

The question remains: is 530-540FSB good enough ? With the E8600 which has a maximum multiplier of 10, this board can reach 5.3 to 5.4 Ghz. While this is already quite high, I can imagine reaching this speed with an E0 stepping on my Single Stage Phase cooler already. For people with access to dry ice or LN2, this is certainly not high enough compared to the performance of many P45 motherboards, or the Rampage Extreme for example.


max cpu speed @ 1.375V, max FSB screen (click to open)

Finally, I did a quick test at the maximum frequency the CPU could run at. As I was using water-cooling, I was expecting it to hit around 4.5Ghz. I was pleasantly surprised to find out it scaled nicely to 4.7Ghz at only 1.375 Volts. No doubt it has more in it; this setting was nice and stable at only 49°C for the CPU cores.

Let's see how the quad cores do >>>