by Grigoriy Gubankov
08/27/2003 | 02:06 PM
Mainboards from different manufacturers are becoming more and more like each other. There are several reasons for it. The first and most obvious is that the chipset makers are packing more and more functions into their products. So, mainboard designers receive a typical collection of functions with every new chipset. This collection of features is constantly expanding and now covering nearly all user requirements. The second reason for the mainboards to look similar to one another lies in the wild competition in the market today. Since designing a mainboard is no easy and cheap task, many companies choose to use the reference design from the chipset maker in their PCBs, tailoring it to their own needs. The third reason is the fact that many mainboard makers (or maybe we should call them sellers?) don’t produce their products anymore, but pass the production to special companies, like ECS. And those companies have more money issuing one big batch of the same mainboards rather than a few batches of different ones.
<%BANNER[article]%>So, we shouldn’t really be surprised at the news that two mainboards selling under different trademarks may in fact be the same. Moreover, the companies’ officials don’t refute the news, but point at the same three reasons, which I have just mentioned.
This all-out unification can be a trap for mainboard makers, though. The user wouldn’t actually care which mainboard to buy as long as they are all the same. Minor differences are hard to notice and, well, do you really bother to read up the specs if you know that the same hardware lies in every package? It is especially hard for the minor players. The retail market is vitally important for them as it promises higher profits, but they cannot rake in more money by increasing production volumes as giants like ASUS or MSI do. This has the following consequence: minor mainboard makers are more interested in making their products attractive, memorable in the eyes of the customer. So, today we will have a look at a product like that from the Taiwanese DFI Company. The mainboard we got for our tests is called DFI LAN PARTY NFII Ultra. As you may guess, it is based on the NVIDIA nForce2 400 Ultra chipset.
DFI LAN PARTY NFII Ultra | |
CPU | Socket A AMD Athlon XP/Duron |
Chipset | NVIDIA nForce2 |
FSB frequency | 100-300MHz |
DDR DIMM slots | 3 |
AGP slot | AGP 8x |
Expansion slots (PCI/ACR/CNR) | 5/0/1 |
USB 2.0 ports | 6 |
IEEE1394 ports | 3 |
Additional IDE-controllers | HighPoint HPT372N, Marvell 88i8030 |
Serial ATA 150 | 1 port |
Integrated sound | 5.1, ALC650 |
Integrated network | 2 Ethernet 10/100 ports |
Additional features | RAID 1.5 |
BIOS | Phoenix – AWARD |
Form-factor | ATX |
This specification table already mentions some features, which will undoubtedly make this mainboard stand out against others. However, we will discuss them later. For now, let’s open the package to see what dainties DFI put there for us.

The package is big, as all the accessories are neatly assorted into smaller boxes. The biggest of these “internal” boxes conceals a bag to carry the system case and keyboard in. It is called PC Transpo. We already posted a snapshot of this bag “in operation” in our Six VIA KT400A Mainboards Roundup, but here it is once again:
By the way, we don’t think anyone can wear the PC Traspo in the manner depicted on the DFI mainboard package.

As you see, they suggest that you should wear the PC Traspo on your back, like a satchel, but it’s really impossible to do it with a bag that has only one strap. But back to the topic.
The next biggest “box in a box” is a repository for round IDE cables. They look quite pretty, at least better than those included with MSI KT4A Ultra. Take a look yourselves:
Moreover, the cables from DFI shimmer in ultraviolet light. Our normal sunlight includes ultraviolet rays, so these cables look slightly highlighted. A stylish thing.
One more item DFI thinks necessary to add to the mainboard is a device named FrontX. You let this thing creep into the 5”-bay of your PC and it will provide you with one FireWire and two USB ports as well as mic-in and headphones jacks. The connectors are not glued to their places, but inserted on slides. It means you can change their layout to your own taste.


The last box goes under the name of Accessory Kit. There is a bunch of small items accompanying any mainboard. To be more specific, we have two brackets for the back panel of the PC case with a game port (it just didn’t fit into the mainboard’s back panel) and two FireWire ports, one Serial ATA cable along with a Serial ATA power adapter, a bracket with S/PDIF in- and output (both – non-optical) and a bracket with holes for the mainboards’ own connectors. The last thing is necessary, since the connector panel is non-standard.

We have no USB bracket, which is quite natural: the mainboard’s panel carries four USB ports, and FrontX – two more (the MCP-T South Bridge supports 6 USB ports in total). The onboard USB connectors can also be output to the system case ports – such cases are quite popular today. So, DFI is quite right in not giving us that bracket.
Some items didn’t find a place in any of the boxes. These are a sticker with “LAN PARTY” caption, a shield with a similar caption and “It’s party time” logo, a several-page “Quick Installation Guide” and five jumper caps with tails to catch them by. Curiously enough, the mainboard itself has ordinary, tailless jumper caps. So if you use Clear CMOS too often, just replace the jumper cap with a more convenient one. Besides, inside the box you will also find a pack of thermal paste, but we didn’t check its efficiency. Anyway, if you happen to have no paste at hand, this one will do.
As I have already mentioned above, this mainboard boasts certain unique features. First of all, it has two Ethernet 10/100 controllers. Physically, one of the ports is implemented via a PHY-chip from ICS, and the second one – via a Realtek controller. So, the manufacturer didn’t use the ability of the MCP-T South Bridge to support two Ethernet controllers; the Realtek controller is connected via the PCI bus. This must be less expensive to implement.
Another peculiar feature of the mainboard is the IDE RAID controller, HPT372N from HighPoint. It provides the user with two extra Parallel ATA channels. Besides ordinary RAID 0, 1 and 0+1, this controller supports RAID arrays of level 1.5. RAID 1.5 is in fact a slightly modified RAID 1. With RAID 1, we have data “mirrored”, that is, it is all stored on one drive, while the other is a copy. With RAID 1.5, half of the data is stored on one drive of the array, and another half – on another drive. The remaining space on each drive is used for mirroring the data stored on the other drive accordingly. Here is an illustration to what has just been said:

Of course, the use of two disks allows reaching higher read speed by alternating the read requests between the two drives (reads alternation is also used in regular RAID 1 arrays). According to Tom’s Hardware, RAID 1.5 works faster than RAID 1, but this performance gain is very small and you don’t always see it. Anyway, since these two array types are formed by the same number of HDDs, it makes sense to decide on RAID 1.5, because it is sometimes faster. However, RAID 1.5 is now supported by the HighPoint PHT372N controller only. You should take this into consideration, since further mainboard upgrade may be a problem if you want to keep your RAID 1.5.
Of course, we have to draw your attention to the fact that all connectors of the mainboard shimmer in ultraviolet light, although this doesn’t have any influence on its performance. Anyway, exotic looks make DFI LAN PARTY NFII ULTRA a unique product. As for us, we like everything beautiful unless it conflicts with the technical characteristics.
As for performance, we will discuss it a bit later. And now, we still have a few more exciting features to mention. They are definitely targeted at enthusiastic users, if not overclockers. Thus, the PCB carries two buttons: power-on and reset. To tell the truth, you really start appreciating these buttons when you use the mainboard without the PC case, that is when you undertake numerous overclocking experiments or play with the memory timings and have to close up the respective contacts with the screwdriver tip. Here, you just push the button: very handy!
DFI LAN PARTY NFII Ultra also features six LEDs onboard. Four of them help you to keep track of the POST procedure. Unfortunately, the light code definitions are explained somewhere in the middle of the user’s manual. It could be better to have this table in a separate leaflet, or in the Quick Installation Guide, which is only four pages long.
The remaining two LEDs indicate if the PCI and DIMM slots are powered, which should prevent the corresponding devices from damage during installation. The first thing is more necessary as the PCI slots are usually powered even when the system is shut down, but not disconnected from the power supply. By the way, all the LEDs emit bright red light – they can illuminate your room no worse than a Christmas tree.
This mainboard leaves an impression of a reliable product. The CPU power circuit is a three-phase one with huge induction coils: the wire is really thick even according to the mainboard standards. A small heatsink is mounted on a transistor, which is a base for one of the supply circuits (evidently, for the chipset). It’s no secret that the supply circuit transistors of modern mainboards heat up a lot, so it definitely made sense to use a heatsink here. Especially, since the transistor we are talking about was really warm during work.


This touching concern about a single transistor could imply that we would see a huge cooler installed onto the chipset North Bridge. However, the North Bridge of the chipset has no active cooling. It is no secret that nForce2 Ultra 400 generates less heat than the “older” nForce2. And even nForce2 often came just with a passive heatsink.
As for design layout, it’s good overall. The cables that go to the USB and FireWire connectors won’t interfere with the expansion cards or hinder airflow. Parallel IDE connectors are grouped in pairs and placed at a distance, so that you had no heap of IDE cables to mess up with. IDE connectors are taken to the front of the PCB, so that the IDE cables don’t hinder proper airflow even if you use traditional cables (not the round ones that come with the mainboard). The installed graphics card doesn’t block the DIMM slot clips. The DIMM slots themselves are set at a distance, so you will have no problems with memory modules equipped with heat-spreaders.
It seemed like everything was excellent, but we still found a drawback. The main (20-pin) ATX power supply connector is shifted to the front part of the PCB and the power cable has to go all the way through the system case. Moreover, this connector landed before the FDD connector. Thus, the FDD connection/disconnection has become no trivial task. Maybe the designers just had place for this connector in its usual location, which was now taken up by the CPU power circuitry components. But at least they might have put it down behind the FDD connector.
The mainboard’s BIOS is made by Phoenix/Award. Overall, the BIOS is quite ordinary, but let’s anyway page through some of its sections like Advanced Chipset Features, PC Health Status and Genie BIOS Setting.

So, the first of the sections, Advanced Chipset Features, includes numerous settings: you even have to scroll down to access some of them. They are far-reaching, too: from memory timings to the AGP port settings. As you can see in the snapshot, these settings are more than enough for tuning the system up. Most users won’t ever need them really.
By the way, the original BIOS was not free from some bugs, particularly connected with memory timing settings, so we had to re-flash a new version (the fresh BIOS is always available at DFI website). However, the new version was also far not quite perfect. For example, I discovered that some settings affect the range of available DRAM and FSB frequencies in the Genie BIOS Setting section. I won’t go deep into detail about this influence, but it shows up when you change the System Performance and Memory Timings presets.
As for AGP settings, they are rather scarce for this mainboard. On the other hand, there is an opportunity to set AGP latencies, which is quite useless for most users.

Now, we are in the PC Health Status section. Regrettably, DFI didn’t manage to avoid the common mistake: they don’t tell what thermal diode the CPU temperature is taken from: the one integrated into the CPU core or the one placed under the CPU socket. Since there is some element under the CPU on this mainboard, I assume it could be a thermal diode. So, we are not sure which way the CPU temperature is measured here. Another questionable option is CPU Shutdown Temperature. You cannot change the default value – 110oC. Meanwhile, the official maximum temperature for AMD processors is 95oC. So, we have a too high shutdown temperature: the processor may be damaged even if the overheat protection system works correctly. It would be better if DFI allowed changing this value.


The last page we are interested in is Genie BIOS Setting. This is the most important section of the entire BIOS. Guided by some unknown logics, DFI included here the settings for frequencies and voltages as well as the settings dealing with enabling/disabling various integrated controllers (except the integrated audio codec that somehow crept into the traditional “Integrated Peripherals” section). Maybe the company wanted to stock up more settings in one place for you to navigate less along the BIOS pages. But why didn’t they include the settings for integrated sound then? I have no idea...
Anyway, it’s more important how flexible the settings are. So, the Vcore can vary from 1.1V to 1.85V with 0.025V step and from 1.85V to 2.0V with 0.05V step. The DIMM slots are powered with 2.5V-2.8V voltage with 0.1V increment. Vagp changes from 1.5V to 1.8V with 0.1V increment. The chipset voltage varies from 1.5V to 1.9V. These ranges are a treat for an overclocker, although the memory voltage range might have been wider. The fastest modules may require 2.9V. The chipset voltage range is impressive, really.
Now let’s pass over to the frequencies settings. The FSB frequency can be set from 100MHz to 300MHz with 1MHz increment. Possible AGP frequencies are 50MHz or 60-100MHz with 1MHz or 2MHz increment. As for the number of possible FSB/memory frequency ratios, DFI mainboard offers wide opportunities here just like any other nForce2-based product. I was very pleased to see that the memory frequency was shown right there. You don’t have to take guesses or make calculations to know which frequency the memory will work at on the next startup. By the way, you cannot type in the frequency value directly, but choose it from a scroll-down list.
Now, we will do a brief overclocking test. First of all, let’s see how close the mainboard is to the claimed upper limit of the FSB frequency (300MHz). It’s no secret that mainboard makers are prone to specify fantastic bus speeds, which are of no use, however. So, we reduce the CPU multiplier to 7x and reduce the memory frequency as greatly as possible, at the same time raising the memory timings. The chipset voltage will be set to 1.9V. After that we started steadily increasing the FSB frequency, to see when the mainboard would give in. we tested the mainboard stability with 3DMark2001 SE (run three times in a row), all demos from Unreal Tournament 2003 Demo and a full set of tests from the OpenGL SPECviewperf 7.1 benchmark.
Of course, the mainboard didn’t reach the desired 300MHz, although the achieved result was very nice: 241MHz. We also learned that 1.8V on the chipset didn’t affect the system stability. So, we guess that 1.9V voltage is redundant for the nForce2 chipset.
The second overclocking test aimed at checking the memory overclocking potential. We checked how high the mainboard allows overclocking OCZ PC3700 EL DDR memory with 3-4-4-8 timings in the dual-channel sync mode. We ran the same set of benchmarks.
Again, we achieved 241MHz as during asynchronous work. It looks like the overclocking potential of this mainboard is limited by the chipset itself, or by the FSB layout of the PCB rather than by the memory subsystem.
Performance tests traditionally end up our reviews. We configured the following testbed:
Memory worked in sync mode with 2-2-2-5 timings. The testbed was controlled by Windows XP Professional with Service Pack 1. We used 9.0a DirectX version.
As for the version of the nForce2 driver, we used v.2.03, because v.2.45 didn’t include Accelerated IDE driver. Without it, the system performed slowly in Winstone tests that actively involve the disk subsystem. In other tests, the difference between versions 2.03 and 2.45 was pretty insignificant.
So, below are the performance results. We include the numbers for ABIT NF7-S 2.0 for better comparison. You should have already read about them in our “Modern Socket A Chipsets Review”. The testbed we used then was nearly the same as the today’s one.
DFI LAN PARTY | ABIT NF7-S 2.0 | |
Business Winstone 2002, Score | 36.7 | 39.2* |
Content Creation Winstone 2003, Score | 40.9 | 41.2* |
3DMark2001 SE, Score | 16504 | 16570 |
3DMark03, Score | 5033 | 5010 |
3DMark03, CPU Score | 641 | 669 |
PCMark2002, CPU Score | 6826 | - |
PCMark2002, Memory Score | 6242 | 6468 |
SiSoft Sandra Int RAM Buffered Bandwidth | 2897 | 3064 |
UT2003, dm-antalus, 1024x768x32 | 72.26 | 72.09 |
Serious Sam SE, The Grand Cathedral, 1024x768x32 | 111.9 | 113.2 |
* - The results of these tests shouldn’t be compared as the ABIT NF7-S testbed had a different hard disk drive.
As you see, there is an evident difference in the results of memory bandwidth tests, in Business Winstone and the CPU tests from 3DMark03. As for Winstone, it would be hard to draw any definite conclusions, as the testbeds were different. The hard disk doesn’t affect the performance in memory tests, so we can state that DFI board was not the fastest here. ABIT must have taken the trouble to fine-tune the chipset parameters, while DFI didn’t. It must be lower memory bandwdith that puts the DFI mainboard down in the 3DMark03CPU tests, too. However, you shouldn’t worry too much about that, since DFI does quite well in gaming benchmarks (3DMark tests should also be considered here), running as fast as the ABIT mainboard. Overall, the performance of DFI LAN PARTY is up to the mark.
Summing all up, we would say that DFI LAN PARTY NFII Ultra is a very good product. It does have certain shortcomings, like only one Serial ATA port and inappropriate ATX connector placement, as well as a few minor drawbacks in the BIOS. However, when we look at all the advantages the mainboard from DFI can boast, such as a rich set of supported features, very nice PCB layout, pretty good overclocking-friendly options and functional BIOS Setup, the above mentioned drawbacks seem really trifling.
There is one thing however, that some of you may consider a great advantage, and some may regard as a real drawback of this mainboard. The accessories and extra features of the product, like the system-case bag and highlighting, cost extra money. As a result, this mainboard will be an impressive piece and a great buy on the one hand, but will cost you more than any competing product with similar features on the other. So, this is actually the main point to consider when shopping: do you want to pay extra for extra features and rich accessories pack? If yes, then don’t hesitate and go for DFI LANPARTY NFII Ultra!