Six VIA KT400A Mainboards Roundup

Although VIA KT400A chipset is not the brand new solution any more, we tested a few mainboards based on it. As our tests showed, these solutions can become an excellent basis for inexpensive Socket A systems.

by Grigoriy Gubankov
08/22/2003 | 11:08 PM

Notwithstanding VIA’s recent announcement of the new KT600 chipset, the previous product from this manufacturer, KT400A, should not be considered out-dated. The functionality of KT400A can satisfy most users as Athlon XP processors for the 200MHz system bus are still not very widely spread in the market (because they belong to the upper price range), while PC2700 memory, with which KT400A operates quite well, and the 166MHz-bus processors are widespread. Another advantage of KT600 (besides its higher system bus frequency) is Serial ATA support, but, in my opinion, it is not a crucial factor: Serial ATA drives are quite rare. By the time they have ousted Parallel ATA altogether, you will probably need to upgrade the mainboard. Moreover, since KT400A is no fresh novelty in the chipset market anymore, it has another advantage: mainboards based on it are going to cost less than KT600-based ones.


So, these are the reasons I wrote this roundup now. We will discuss six mainboards based on the VIA KT400A chipset. After studying their main features, we will talk about our overclocking experience with them and their performance level.

We won’t go into detail about the chipset itself, as we already did it in our MSI KT4 Ultra Mainboard Review. So, let’s jump straight to the mainboards.

Mainboards Specifications

First, let’s have a look at the characteristics of the reviewed mainboards.









Socket A AMD Athlon XP/Duron



DDR DIMM slots




AGP slot

AGP 8x

Expansion slots (PCI/ACR/CNR)







USB 2.0 ports







IEEE1394 ports







Additional IDE-controllers



HPT372N Marvell 88i8030




Serial ATA 150

2 ports

2 ports

1 port


2 ports


Integrated sound

5.1, ALC650

5.1, CMI9739A

5.1, ALC650

5.1, VIA VT1616

Integrated LAN

10/100 Mbit

10/100 Mbit, 2 controllers


Additional features

Voice diagnostic system


RAID 1.5


External Serial ATA

RedStorm Overclocking



As you see, the mainboards differ a lot in their features. We have a rather unpretentious Soltek and EPoX boards and a somewhat more functional ABIT next to the masterpieces of the modern engineering thought from AOpen, Gigabyte and DFI. That’s why we cannot directly compare the mainboards among themselves as the latter three products obviously provide richer opportunities. So, the only criterion for comparison is overclocking potential.

Now, we will talk about each mainboard in more detail.

AOpen AK77-400 Max

Like many other mainboards, this product from AOpen may come in various flavors. We happened to snatch the most advanced variant. There are also AK77-400N without IEEE1394, Dr. Voice diagnostics system, Serial ATA and the backup BIOS chip and the simplest model, AK77, that doesn’t also have the network besides the features mentioned above.

The package is smaller than usual: the numerous accessories are packed like sardines in a tin box. However, we didn’t find any treasures inside. The user’s manual is made in an original format, while a double-sided A2 poster with colorful photos can serve as a guide to various settings and connections.

AOpen AK77-400 Max: Features

As I have already mentioned, this mainboard offers a wide range of functions, although most of them don’t need many comments. Yeah, we have Serial ATA, FireWire, 6-channel sound with S/PDIF support (and an appropriate bracket for the case rear panel), but we see the same things in other mainboards, too. So, we’d better talk about the unique features of this particular product.

First of all, it is the voice diagnostics system called Dr. Voice. The technology works when there is some problem during POST. The mainboard stops starting up and informs you about it in a woman’s voice through the PC speaker or the external speaker set connected to the sound card. In the first case, the quality of sound is very poor, and sometimes it’s not quite clear what the mess is all about (I deliberately tried this function by alternately uninstalling memory and the graphics card and over-overclocking the processor). The mainboard offers you one more doctor – Dr. LED. This chubby guy does the same job as the diagnostics LEDs in MSI’s mainboard. He is smarter, though. The problem is indicated by means of a LED with an appropriate caption so you don’t have to refer to the manual to decipher it. Unfortunately, I couldn’t test Dr. LED in action: although the necessary connector was onboard, the Dr. LED bracket was missing in the package. Anyway, I had enough fun with Dr. Voice shouting at me through my own speaker set.

The second technology to draw my attention was Die-Hard BIOS: two BIOS chips onboard to restore one another in case of failure. Such technologies (for example, DualBIOS from Gigabyte) were quite popular during the epidemic of the Win32.CIH virus that damaged BIOS chips. In a while, the epidemic subsided and the technology was abandoned: it did make the mainboard more expensive. Anyway, the problems with BIOS remained acute. For example, in case of accidental power disconnect during BIOS re-flashing. So the Die-Hard BIOS from AOpen is a definite advantage. By the way, now that we are talking about BIOS, I should mention the VividBIOS technology: the mainboard allows selecting the startup logo. The logo can be in the GIF format and even animated!

The last technology I would like to say a few words about is SilentTek. As you may have guessed, it is intended to reduce the noise from the system case. You need a special Windows utility for this technology to operate properly (the utility, of course, comes on the CD with the mainboard). Well, it may be better to adjust all the settings in BIOS, but I doubt all the options could be fully taken into BIOS. I won’t describe the settings and controls you can apply to the fans, just believe they are numerous and very flexible.

Summing up the mainboard’s features, I should say that AOpen did make an extraordinary mainboard and provided it with a number of original technologies, which definitely deserve paying attention to.

Let’s go on now and discuss the PCB design of AOpen AK77-400 Max.

AOpen AK77-400 Max:  PCB Design

This mainboard is well designed. Cables attached to connectors won’t hinder the airflow inside the case or PCI cards installation. Anyway, I found a thing to get annoyed with: the first two DIMM slots stand too close to each other, so if you use modules with heat-spreaders (for example, those from Corsair), there will be little space left between them. In this case, the heat-spreaders may make cooling worse rather than better.

This mainboard is also not free from the widespread problem of modern products: the installed AGP graphics card blocks DIMM slot latches. A slight drawback is the placement of one USB connector and the game port between PCI slots, so you may encounter difficulties in plugging-in an expansion card. However, these are all minor inconveniences.

The power supply of the mainboard is up to the mark, too. The CPU is powered via a three-phase circuit. Higher power voltage purity is achieved by using three high-value capacitors. Besides, the power supply circuit is impulse rather than linear as in most modern mainboards. Although, this is not critical, that’s a pleasant fact. By the way, notwithstanding the great attention they paid to power circuitry, they didn’t solder up an additional 12V connector.

The only thing that disappointed me in this board was the necessity to set up the FSB frequency with the jumpers. Not the exact frequency, of course, but the range we will work with in the BIOS. Well, actually three mainboards of the reviewed six can be blamed for that. You can set any FSB frequency directly from the BIOS only in the mainboards from ABIT, EPoX and Soltek. However, AOpen set me the hardest task with this jumper: its location is described in the enclosed poster I told you about, but not in the user’s manual. So, I spent about half an hour trying to launch the processor at its proper frequency. Of course, that was my mistake as I could be smart enough to check the poster, but why didn’t they duplicate the info in the manual? I think they should have done it and hopefully they will make up for this drawback soon.

Summing up, I’d like to say that the mainboard leaves a good impression, slightly spoiled by the accident with the FSB-setting jumper.

AOpen AK77-400 Max: BIOS

The BIOS of the AOpen mainboard is remarkable for two sections. First of them is the exclusive Silent BIOS/HW Monitor section (only the first part is “exclusive”, to be more exact). Here we have the basic SilentTek settings, including fan rotation speeds for mainboard startup, for the POST and for OS boot-up and operation. AOpen warns that setting the fan rotation speeds without installing the SilentPC utility may lead to CPU and/or system overheating. All changes you do to the cooler speeds are enabled immediately, so you can check the noise level right in the BIOS and correct it to your taste.

The second remarkable section is Frequency/Voltage Control. All frequencies are shown right after you make any changes, with an appropriate multiplier. For example, the memory frequency is shown as 166x2.00=333MHz. The AOpen mainboard seems to be the only one in this review to tell the multiplier instead of hiding it behind the memory type name. I guess that’s a significant advantage.

FSB, memory, AGP and PCI frequencies as well as four voltages (memory, AGP, PCI and processor) are displayed during the POST procedure with their nominal values so that you could see how far this or that parameter goes beyond the nominal (if you forgot this nominal). This monitoring is not as detailed as by EPoX mainboards, but still is quite enough for checking the system operation at a glance.

That’s about all I’ve got to say about the AOpen mainboard. Let’s get to the next one – ABIT KD7-S.


This creation from ABIT is a more functional variation of KD7: two SerialATA and one ParallelATA channels were added. However, it is slower than the KD7-G model in the Ethernet speed (100Mb/s against 1Gb/s). It’s a curious fact ABIT didn’t change the name of the mainboard on transition to the KT400A chipset. Once, it was rumored that ABIT wouldn’t produce KT400A-based mainboards at all (see this news story), so it’s quite possible that we have got nothing but a sample. It means ABIT’s KT400A-based mainboards may not make it to the shops, or at least there won’t be too many pieces selling. However, since we have this product, let’s review it.

The package we got the thing in included nothing besides the mainboard itself, so we are not sure what accessories are supposed to come with it. Anyway, I doubt they will differ much from the accessories of the KT400-based ABIT KD7 mainboard and the only thing interesting in it is the Serillel ATA adapter intended to connect Parallel ATA devices to Serial ATA connectors. We have already met this gadget in our ABIT NF7-S Mainboard Review.

ABIT KD7-S: Features

This section won’t be long very long as there is nothing very outstanding in ABIT KD7-S to talk much about. It’s not because the mainboard lacks certain features, but just the features can hardly be called unique or exclusive. This is a well-done fully-functional mainboard, although devoid of FireWire. That’s a questionable solution, but on the other hand, FireWire peripherals are less widespread and also quite expensive.

However, this mainboard still does have one peculiar feature: fully-fledged six-channel sound. Many mainboards with a six-channel audio codec have just three mini-jack connectors, two of which are line-in and microphone-in by default. For six-channel sound output you must have three connectors, of course. So, you have to change the function of the two other jacks in the drivers so that they outputted the remaining sound channels, or use additional brackets (sometimes not coming with the mainboard) or connect the mainboard to appropriate connectors in the system case, which is not always possible. None of these methods is handy. In ABIT KD7-S, we have all five connectors required for true six-channel sound output and also an S/PDIF connector. That’s a very pleasing fact. However, it has its drawback: the back panel of the mainboard simply couldn’t accommodate four USB ports, so it has only two ports instead of four.

ABIT KD7-S: PCB Design

First things to draw my attention in the PCB were the active cooling system mounted on the North Bridge and four memory slots. I can’t definitely say how useful the active cooler on the North Bridge is. KT400A-based mainboards don’t generally suit for FSB overclocking because of a limited set of multipliers: you cannot reach high frequencies. However, the improved cooling of the chipset may help during the “approved” overclocking in system cases with poor ventilation. The chipset fan of ABIT KD7-S is not too noisy. When the mainboard is installed inside the system case, you won’t probably hear it at all.

As for the four memory slots, I had some concerns about them. All other mainboards we review today have only three DIMM slots, and I thought the fourth one was added by ABIT just to make an impression and was afraid to discover some instability with four memory modules installed. I was wrong. Plugging in four memory modules (three Corsair XMS3200C2 and one Corsair XMS3500C2) and thus having 1GB of memory, I had no stability- or performance-related issues. By the way, the memory power supply circuit is impulse as well as in AOpen AK77-400 Max, which may be needed for normal operation of four memory modules. Besides, ABIT KD7-S carries an additional 12V power connector.

I would also like to say a few words about the opportunity ABIT mainboard offers to set any bus frequency you like (within the allowed frequencies range, of course) without any manipulations with the jumper set. At first sight this seems to be a really easy task to fulfill, but five mainboards of the six tested today didn’t have this option. By the way, it is very convenient to work with the Clear CMOS jumper on ABIT KD7 mainboard, because it has a special tail, which makes all manipulations with the cap much easier.

The layout of onboard connectors is good enough. I don’t only like the primary ATX connector set at the front. As a result, the mainboard power supply cable will go next to (in some system cases, above) the CPU socket hindering proper airflow and threatening to get inside the CPU cooler blades. Well, this is a bit of exaggeration, of course, still I think it would be better if they pushed the CPU power connector to the front and the mainboard one – to the back and below. Overall, I grade the PCB design of ABIT KD7-S as good.


This mainboard follows the tradition of most (if not all) ABIT mainboards in using BIOS from Phoenix/Award. There is nothing extraordinary about the BIOS, but let’s dwell on a couple of details anyway. Firstly, let’s go to the SoftMenu III section, which is responsible for overclocking functions in modern ABIT mainboards. The voltage sent to the CPU may vary from 1.1V to 2.325V – quite enough for any manipulations with the CPU, considering the maximum nominal voltage of modern AMD processors being only 1.65V. The memory voltage doesn’t spoil the picture: the range is 2.55V-3.25V. This is more than enough also, as the memory nominal equals 2.5V (or 2.8V for super-overclocking memory). Note that the memory in ABIT KD7-S is always powered a little bit above the nominal. Moreover, the default value is 2.65V. I don’t quite understand the reason for the increased voltage – they must have wanted to ensure stable memory operation this way. As for AGP voltage, the ABIT mainboard doesn’t allow adjusting it for some reason.

The DRAM Timing Control section contains numerous memory-timing options. I’d rather say they are too numerous: I met half of them for the first time. Of course, I didn’t know what their change was going to lead to. Still, if you are not sure about the meaning of this or that parameter, just let it be. Memory-setting gurus will have a grand time here, I guess.

The PC Health Status section informs you of two CPU temperatures: one is taken from the external thermal diode (CPU Surface Temperature) and the second one - from the thermal diode integrated into the core (CPU Core Temperature). There has been a lot of argument about the measurement error of the external sensor, now you can see it yourself. As far as I could notice, the biggest difference between the two temperatures was about 15oC.

One thing I have to complain about here, is just the same as what we saw by MSI mainboard, reviewed earlier): it is the memory frequency indication. The DRAM Timing Control also keeps the memory clock-rate settings, and it shows the “base” frequency for a given interval rather than the real bus frequency. That is, if you set 180MHz memory, you will see just 166MHz in BIOS and at startup. That’s even more strange as all ABIT mainboards I’ve ever encountered showed the FSB and memory frequency in the normal way. Maybe the reason for that is the presale nature of the BIOS; the ABIT website even had no BIOS for the KT400A-based mainboard available for download.

We leave ABIT KD7-S for now to return to it (and others) later, in the “Overclocking” section. The next mainboard we are going to take a closer look at is quite a curious one.


This is the only standalone mainboard in this review meaning it doesn’t belong to any model family – it comes just as it is, no variations. DFI LAN PARTY is a unique product and that’s exactly how it is positioned in the market.

The mainboard draws our attention from the very beginning: its package is much larger than usual. Does the huge size imply numerous accessories? Yes, of course, we have got a lot of them. First, it is a bag called PC Transpo for transportation of the system case (to those LAN parties, I may suppose). A brief test I literally performed myself showed that this bag is fine for carrying the PC for short distances.

Next a set of round IDE cables comes. This set includes two UDMA and one FDD cable. They look tasty due to their metal armoring. I don’t know what is the purpose of this, but it looks nice, anyway.

But these are ordinary supplements coming with any decent mainboard, yeah? So what’s so exceptional about this one? The mainboard package included a device called FRONTX to be installed into a 5”-bay (like the Media XP module from ABIT). However, the FRONTX is narrower in functionality compared to Media XP; for example, it lacks card-reader capabilities. On the other hand, FRONX allows you to change the allocation of connectors to your taste. The connectors have lugs, while the FRONTX panel – slides. Well, it’s hard to explain, just look at the snapshot.

But this is not the end. The package also includes two LAN PARTY stickers for the system case. One is a usual shield, while the other must be intended for the system case. Besides the stickers, there is a HDD power adapter from the ordinary 4-pin connector to the Serial ATA one, and a data Serial ATA cable. And all this serves to connect a single Serial ATA drive (the mainboard just doesn’t support any more)!

So, DFI LAN PARTY comes complemented with various necessities. More is better than less.

DFI LAN PARTY KT400A: Features

So, this mainboard has everything you might wish as far as accessories are concerned. The situation with its functionality is about the same. Firstly, there are ultra-violet highlighted PCI, DIMM and IDE slots and connectors. The special IDE and FDD cables add to the impression. I doubt that it has any favorable effect on the mainboard functions, but what a treat for people who have windows in their system cases! Secondly, the mainboard carries numerous red light-emitting diodes. Four of them serve diagnostics purposes, so they are similar to Dr.LED from AOpen or POST-codes from EPoX we will meet later. Two more LEDs indicate the power status of PCI and memory slots thus helping to prevent them from damage. The LEDs are quite bright and can even light up a dark room. I guess they look cool in a system case with a side window.

One more nice thing about this mainboard is the power and reset buttons. This feature will be of little use during regular work, but may come in handy when you work with the mainboard not installed into the system case, because it will save you time and trouble searching for the proper pins to be closed with the tip of the screwdriver.

What is next? Well, LAN is Local Area Network, so DFI LAN PARTY should offer something exceptional in this respect. And it does with its two Ethernet controllers. The first of them is a VIA VT6103 PHY-controller; the second is RTL801L from Realtek. Thus, if you need, you may use this mainboard as a router for your home LAN. The Ethernet ports are placed at the back panel, so there was no place left for the game port. It comes on an additional back panel bracket. Furthermore, the mainboard couldn’t accommodate all mini-jack connectors, so you will have to mess up with the drivers assigning the FRONTX connector for microphone in, or use an S/PDIF-input speaker set. The bracket with this input port as well as output port is enclosed with the mainboard.

Now, to the HDD subsystem. The mainboard uses an advanced IDE RAID controller from Highpoint – HPT372N. This chip is better than PDC20376 used in Gigabyte’s products (we will talk about it a bit later). So, the HPT372N controller supports RAID arrays of levels 0 (stripe), 1 (mirror), 0+1 (stripe + mirror), JBOD (Just a Bunch Of Disks) and RAID 1.5. The first four array types are quite well-known, while the last one is used less often.

The capacity of a RAID 1.5 array equals the capacity of each of the drives (two identical drives are required). The data are stored on one half of each drive, while the second half mirrors the first half of the other drive. Thus, the data are fully mirrored and the data security is analogous to RAID 1 (mirror). Also in some conditions the data can be read from two HDDs simultaneously, like in Stripe, which should theoretically ensure aster work of the entire disk subsystem. It is true that RAID 1.5 does perform better than RAID 1, though not in all cases. Nevertheless, this technology can ensure a performance boost at the same price (RAID 1.5 requires two HDDs, i.e. the same as RAID 1).


Let’s see how smartly all the connectors are located on the PCB. Running a little ahead, I must confess that most mainboards we reviewed today provoked no criticisms in terms of PCB design. There is already a common standard most companies follow when designing their products, and any deviations are quite tiny all in all. DFI had to bend aside because couldn’t fit the game port anywhere. That’s why the PCB carries one more connector, while the USB port, which would have been in its place, moved in between the third and fourth (counting from below) PCI slots. It may cause some difficulties when plugging the cable into it. By the way, ABIT also had no place for the game port in its KD7-S, but made it easy: they just abandoned it altogether thinking no one would ever need this port. Well, this sounds logical enough.

But back to DFI LAN PARTY. The mainboard package doesn’t include the USB bracket (well, they had to forget about something!). This means they suggest you connect the USB ports of the system case to the onboard connector. I also think that the location of the 12V power connector next to a high capacitor is not very convenient: if you are unlucky enough, you may accidentally tear it off trying to unplug the 12V power cable.

As for good things, you can install an AGP graphics card of any length and it won’t block the DIMM slots clips: the memory slots are placed at a distance from the AGP slot.


This mainboard is controlled by a BIOS from Phoenix/Award. There is nothing wrong about this particular BIOS, except a strange location of settings that enable/disable the mainboard’s integrated controllers. You usually find them in Integrated Peripherals, but DFI somehow chose to include them into Genie BIOS Setup – the page that contains overclocking and voltage options and so on. The name itself – Genie – is not too appropriate, I think. I only guessed its purpose by its place in the main menu – it was the same as in the BIOS’s of other mainboards.

Otherwise, the BIOS of DFI LAN PARTY is quite standard. So, let me finish our story and go over to the next mainboard.


EPoX offers two KT400A-based mainboards: EP-8K9A9I and EP-8K9A9+. The latter features a Serial RAID controller (VIA VT6420) and two Serial ATA ones. Both mainboard models use the same PCB, of course.

As for the accessories, there is nothing remarkable about them. Well, the mainboard is a mainstream product, and EPoX has never had the fame of an extravagance manufacturer. The advantages of EPoX mainboards are usually hidden inside – in the overclocking settings and good manufacturing quality.

EPoX EP-8K9A9I: Features

Again, this mainboard is a pretty ordinary product and the remarkable feature worth mentioning separately is the POST-code indicator that signals which POST procedure is currently executed and what’s wrong with the mainboard if anything fails. Compared to the voice or LED diagnostics by AOpen, this approach has both advantages and drawbacks.

The advantages are obvious: the POST-codes are more numerous than system components and information representation with the codes is more informative to an experienced user. On the other hand, the idea behind POST-codes is not as evident as a voice message or a lighted LED next to the “Memory” label. You will have to consult the user’s manual to decipher the code until you learn all of them by heart.

EPoX EP-8K9A9I: PCB Design

The premise that all mainboard makers now follow a widely-accepted PCB design standard holds true in this case, too. As the EPoX mainboard is a pretty ordinary product, its connectors layout is free from any evident drawbacks. Even the graphics card doesn’t block the DIMM slots clips as the mainboard carries only five PCI slots. Even the FDD connector is shifted down, not to hinder the FDD and HDD connection zone. The only thing I didn’t like that much was the undersocket thermal diode that looks too fragile. However, as long as does its job well, its looks doesn’t matter. By the way, none of the manufacturers except ABIT and Soltek, specifies explicitly how the CPU temperature is measured. EPoX, unfortunately, is no exception here.

Well, there is one weird thing about the EPoX mainboard: there is a hole under the CPU socket. It is probably intended to eliminate excessive warm air pressure under the CPU, however, I really doubt that this hole is so badly needed, especially since the processor socket is not absolutely hermetic and warm air has where to flow. One more possible way to use this hole, which I managed to invent, is the installation of your own thermal diode under the CPU. Quite exotic, you know.


The BIOS tells all the frequencies and tells them right. That’s good. Moreover, if you change the settings, it shows what frequencies the AGP and PCI buses will work at after the next startup. In half of the mainboards we reviewed today, the frequencies are reported only after they have been activated, or in the worst case, in Windows with the help of various monitoring utilities. However, you should keep in mind that in some cases, you may fail to get the Windows start, as the data may disappear from the HDD much earlier.

There are two things I found strange in the BIOS. First, the Power BIOS Features page has a setting called ROMSIP Table. It can be Normal, Fast or Ultra. The user’s manual doesn’t tell what it is for, but as I found out this setting is responsible for the S2K bus timings (S2K is the bus between the memory controller and the CPU). Second, the memory/FSB frequency ratio (EPoX calls it Memory Type) can take values named by the memory types, that is, DDR266, DDR333 and DDR400, although the frequencies can be quite different during overclocking. Yes, this representation is very illustrative from the visual point of view, but if you set up high clock for the system bus, the DDR333 memory may not work, although the Memory Type parameter is set exactly to this value.


Besides this model, Gigabyte offers the one without the “Ultra” suffix and without Serial ATA (like ABIT mainboards with and without the “S” suffix). It seems like the Serial ATA controller and connectors become the principle factor for positioning a mainboard in the market.

The mainboard of course comes with everything necessary for efficient work, and a few more items. First of them is the bracket with two FireWire ports. Have a look:

FireWire uses two types of connectors: 6-pin and 4-pin ones. Regular 6-pin connectors are used for all computer devices, which you usually come across. While 4-pin connectors are used for digital cameras. However, this shouldn’t be a problem even if one FireWire connector is not enough for you, because you can always get a 4-to-6 pin cable adapter. So, the availability of two FireWire connector types is a definite advantage, as it makes this product really excellent for work with any sort of FireWire devices.

The second item is really unique: a bracket for external Serial ATA devices. As you get a Serial ATA power adapter with the mainboard, you can easily use a Serial ATA HDD as an external storage device. The length of the Serial ATA cable (about 1 meter) allows doing this.


The specification of the mainboard lists two additional IDE controllers: PDC 20276 from Promise that supports two Parallel ATA channels and can unite them into RAID of levels 0 and 1. It doesn’t support RAID 0+1, but you can do nothing about it: Gigabyte is stubborn in using Promise controllers in its mainboards. The second controller, Sil3112 from Silicon Image, supports two Serial ATA channels with the option of uniting them into RAID 0 or 1. Thus, this Gigabyte mainboard can take as much as 10 IDE devices (and eight of them can be united into RAID arrays). More than enough, in my opinion.

The mainboard offers a rather strange way of changing the CPU multiplier. I didn’t think a modern Athlon XP mainboard might have a multiplier set by a DIP-switch (by the way, the maximum multiplier the mainboard allows to use is 18x). I’m quite at a loss guessing the reason for it. Maybe Gigabyte didn’t want the multiplier to be reset after BIOS re-flashing, but it would be no hard task to set it up again in the BIOS. Overall, a questionable solution.

The Gigabyte board also features DualBIOS technology. Its key point is the same as with the Die-Hard BIOS from AOpen: two BIOS chips onboard to stand up for each other in case of emergencies. By the way, there is one more BIOS-related technology – Q-Flash – that helps to update BIOS from a floppy, without booting the OS.

Now, about some unpleasant things. The mainboard carries no jumpers for choosing the FSB frequency range to be later fine-tuned in the BIOS. So, you set the FSB frequency as follows: you install the CPU and power the system up, then the mainboard automatically determines the FSB frequency of the CPU and adjusts the FSB range accordingly. This way, if you happen to have a processor intended for 133MHz bus, the mainboard won’t allow you to use it with a 165MHz-clocked bus. Considering that most AMD processors today have an unlocked multiplier, I would like to “thank” Gigabyte on behalf of all overclockers. The only thing you can actually do with the frequency auto-adjustment is to force the 100MHz bus frequency (and increase it further to 132MHz). It is evidently not the best thing to be done.


The design of this mainboard is not free from some drawbacks. The ATX power connector is at the very front of the PCB, so the ATX cable goes next to the CPU and over the memory slots, which is no good. The memory slots, in their turn, sit too close to each other (although there are only three of them). So, if you install modules with heat-spreaders (for example, those from Corsair), they will be nearly touching each other. Again, it creates no good thermal environment.

That’s not all. Gigabyte 7VAXP-A Ultra has all four HDD connectors placed in a row. This means you will have a neat heap of cables to mess up with. If the mainboard is installed into a pretty small system case, the process of connecting/disconnecting IDE-devices may become quite an exciting occupation and bring you a lot of fun and joy. Curiously, the Gigabyte 7VRXP mainboard I have in my workstation exposes the same design flaws. Maybe the company considers them not flaws, but features?


The known peculiarity of BIOS’s from Gigabyte is you have to press Ctrl+F1 to access advanced options. So, I punch the keys and get to the Advanced Chipset Features page that allows adjusting memory timings and AGP settings. The options are numerous, although the AGP frequency setup is missing. Somehow this setting moved to the Advanced BIOS Features section, under the name of Flexible AGP 8x.

The next thing was waiting for me in the Frequency/Voltage Control section. Overall, it has the ordinary settings for FSB frequency, FSB/memory frequency ratio (memory goes under the name of its type) and voltages. Strangely enough, all voltages are given in Volts, while the CPU voltage is shown in percent. Moreover, the range is rather narrow – 5, 7.5 and 10%, which is not enough for serious overclocking. On the other hand, 10% voltage increase is considered safe, and Gigabyte must have followed this rule.

That’s all about the Gigabyte product. The last mainboard to be reviewed is Soltek SL-KT400A-L.


Soltek manufactures two KT400-based products: with and without the integrated network. The presence of the networking capabilities is marked by “L” in the name of the mainboard – this is our case. Curiously, there is a place on the PCB for a Serial ATA controller and ports, although Soltek has never produced mainboards on VIA chipsets with Serial ATA support. Maybe the company intended to roll out one more, Serial ATA-enabled, version of the mainboard, but gave up the idea.

The package included the same stuff you get with any other mainboard. The only “feature” is a thick book on accompanying software utilities: it is a longer read than the user’s manual. I wonder what Soltek is up to: are they going to start selling software?

SOLTEK SL-KT400A-L: Features

As we don’t have any non-standard accessories with the mainboard, its functionality is not exceptional, too. The only thing worth mentioning is the house technology from Soltek – Red Storm Overclocking Tech. Its key point is that the mainboard itself is trying to overclock the CPU to the maximum frequency, without losing any stability. However, this stability is determined by the mainboard, so this technology will be of no help if the system cannot boot up Windows or fails in the middle of the tests. All my attempts to feel the difference with auto-overclocking ended in hang-ups or system reload during the work of Red Storm. Seems like Soltek still has quite a bit to do with this technology.

Overall, Soltek SL-KT400A-L mainboard is for a home user who’s not a dedicated overclocker and quite reasonably wants to save some money. Although, the BIOS still includes a few settings, which are definitely intended not for a casual user (it’s hard to tell who they are intended for, actually). We will discuss them in a moment.


The design of this mainboard resembles that of the EPoX product discussed above, although there were a few evident differences. For example, the FDD and HDD connectors are placed in a row, which is not very good. If all three cables are plugged in, you will have trouble plugging and unplugging any of them. Still, this is better than by Gigabyte’s board.

I guess the SCR1 connector is not in its right place, too. According to the user’s manual, it serves for smart-card reader connection. The most logical place to install a smart-card reader is the front part of the system case, so the cable will wind through the entire case. Maybe that’s not a serious drawback, but it is still a bit upsetting.


They use AMIBIOS. The Advanced Chipset Features section with the memory timings and AGP settings cannot blow your brains out. Among memory timings, you can only change CAS Latency and Command Rate. You can also enable/disable Bank Interleave. That’s all. Compared to the ABIT mainboard, that’s simply nothing. Well, if Soltek doesn’t want to give the user access to fine memory settings, it is OK. But why do they offer full AGP settings then?

Next goes the Frequency/Voltage section with those “strange” settings I mentioned above. In fact, the general idea behind them is quite clear, but it is really hard to think of a situation when you could use them efficiently. For example, you can set DDR Stability to Auto, 00, 04, 08 and 0C. What’s the highest stability – 00 or 0C? Yeah, let’s read the manual: “this item allows you to configure the register for DDR stability enhancement”. Hm, no answer to my question, actually. One more: the “Rank Interleaver and Timer Control” can be set to 10, 11, 00 or 01. The user manual says this setting is somehow connected with the performance, but doesn’t say explicitly what we shall do to increase this performance. They are quite mysterious at Soltek, you know.

Maybe the company wanted to fend off a casual user with the fearful names and parameters, so that he didn’t have the mood to go and change them. But why did they add these parameters and their ridiculous descriptions to the user’s manual then? People who are into experimenting or who are experts in memory fine-tuning won’t buy this mainboard anyway: it is evidently not for them.

Among good things about the BIOS, I would single out the CPU temperature monitoring by the external and integrated thermal diodes, like in the ABIT board. However, Soltek also invented a few really outstanding names for these functions. The temperature taken from the built-in thermal diode is called “ABS II Current Temperature”. ABS II is the second version of the CPU overheating protection system aka ABS (Anti-Burn Shield). The system follows the AMD recommendations and uses the temperature measurements taken from the thermal diode integrated into the processor core, that is, ABS II Temperature is the true CPU temperature. So why did they give the setting this name? Maybe to advertise their ABS II once again.

Summing it up, I would say that if it were not for the mess with the names, the BIOS would be quite all right, although devoid of fine-tuning options.

So, this was the last mainboard we wanted to review today. Now we are going to present the results our testing participants showed during overclocking and in performance tests in regular mode.


Before going over to the benchmarks, I would like to give you a brief description of the test methods. First of all, we will try to see whether the mainboards could make an Athlon XP 1700+ start up and work at 2250MHz (166x13.5, “2800+” rating) with 1.85V Vcore. Thus, we will check the ability of the mainboard to support high-end AMD processors, even with some reserve left. The system stability was tested with 3DMark 2001SE tests and dm-antalus from UT2003 Demo. If the mainboard fails to work properly, we have milder testing conditions: 2167MHz with 1.8V. To clear any doubts about sufficient cooling and other things, I’d like to say that these testing conditions were checked on one of the participating mainboards beforehand.

The second check is more traditional: we will reduce the CPU multiplier and will be increasing the FSB frequency until the system becomes unstable. Note: KT400A doesn’t suit for FSB overclocking as it has AGP and PCI frequencies dependent on the FSB frequency. It means we are likely to reach the top of supported AGP frequency rather than the mainboard stability peak.

Unfortunately, I cannot present any overclocking results for the DFI mainboard. This product was unstable in its regular mode, not to mention the results during overclocking. I suppose there is a faulty component on the mainboard, preventing it from proper operation, though, I can not provide any solid proofs for this assumption.

Anyway, I was unable to make the mainboard work stably in any mode and, again, you won’t see it in the tables. It’s rather sad, as it looked a promising product to me.

Now, let’s get to the tests.

The first test proved impossible to pass for some mainboards. The first one to fail is Soltek that couldn’t start the CPU as 2800+ with 1.85V Vcore. The system was only stable when I reduced the frequency to 2167MHz (“2700+” rating), but still with certain issues: the hot restart (from Windows) sometimes led to system’s freezing during the POST. In this case, even the Reset button did not help, and I had to power the system off and then on again. This problem vanished after the frequency dropped to 2083MHz (2600+).

The second unlucky participant is, surprisingly, the AOpen mainboard. Notwithstanding its cool power supply circuit, this mainboard could only work properly with the CPU overclocked to 2600+ rating with 1.8V core voltage. However, this one was somewhat better than the Soltek board: it could work stably with 2800+ CPU, but had the same hot restart problem.

The Gigabyte mainboard didn’t pass in this test. First, the mainboard determined the CPU frequency and didn’t allow clocking the system bus over 165MHz (in our case). Second, 7VAXP-A Ultra simply cannot set 1.85V or 1.8V core voltage. The maximum you can get from the mainboard is 10% above the nominal voltage, which is 1.5V for our Athlon XP 1700+. So, the overclocking capabilities of this mainboard remained a mystery. However, judging from the voltage range and FSB frequency setup approach, they wouldn’t present us anything exceptional.

The remaining mainboards, those from ABIT and EPoX, passed the test quite well. EPoX mainboard, that uses two-channel CPU voltage regulator like the board from Soltek, shows that the quality of the channels is as important as their quantity.

The results of the second test were more or less predictable: all mainboards notched somewhere around 175MHz, which proved that you shouldn’t use a KT400A-based mainboard with RADEON 9700 PRO for bus overclocking, at least in the AGP 8x mode. Maybe reducing the AGP speed to 4x could allow you to reach higher operational AGP (and, accordingly, FSB) frequencies. Once again, the Gigabyte mainboard didn’t take part in these tests as it couldn’t increase the bus speed above 165MHz.


Today, we have the following testbed:

You may consider it strange but I used a relatively slow processor. The Athlon XP 2700+ processor I usually use for benchmarking is a sample we have got after its official announcement. Maybe the sample nature of the chip caused the problem: some of the reviewed mainboards couldn’t tell the CPU temperature correctly and enabled the overheating protection system. As such systems are quite fast nowadays, I had the mainboard running for a second, no more. So, this is the reason I took Athlon XP 1700+, which I usually use for overclocking tests. However, this time I won’t overclock it as we are trying to measure the mainboard performance rather than the chipset performance, and will do our best to reveal any problems caused by BIOS, for instance.







Business Winstone 2002, Score






Content Creation Winstone 2003, Score






3DMark2001 SE, Score






3DMark03, Score






3DMark03, CPU Score






PCMark2002, CPU Score






PCMark2002, Memory Score






UT2003, dm-antalus, 1024x768






UT2003, flyby-antalus, 1024x768






UT2003, dm-antalus, 640x480






UT2003, flyby-antalus, 640x480






RTCW, Checkpoint, 1024x768, High Quality






RTCW, Checkpoint, 640x480, Fastest






Well, as you see all the mainboards are running close to each other, which is actually no great wonder. However, there are two products we can definitely call the leaders: they are the boards from Soltek and Gigabyte. By the way, I tested the Gigabyte mainboard without turning on the Top Performance option, although it boosted the speed of the product considerably. Why so? Just look at the two next screenshots (top – Top Performance enabled, bottom – disabled).

You see that the Top Performance option boosts the FSB frequency and, accordingly, the CPU clock-rate, too. That’s why the Gigabyte mainboard would be a clear winner in this mode. Moreover, I may venture a supposition that the mainboard also reduces the chipset timings and makes some other useful things, but you cannot single them out from the frequency growth without the next benchmarking round. By the way, Gigabyte doesn’t mention this overclocking anywhere, so this looks like cheating in a way. However, this doesn’t mount up to the scandal we had with NVIDIA and 3DMark03.

Last thing I’d like to say is that all the mainboards showed similar performance level. The slight difference may well be eliminated by a new BIOS version.


Now, I will try to sum up all the things I’ve already said and list all the advantages of each mainboard. So, here we go in the alphabetical order:

AOpen AK77-400 Max. The manufacturer company did everything for you to like this mainboard. It offers numerous features: full set of interfaces, voice diagnostics, SilentTek and Dual BIOS. Nice accessories, very smart PCB design and excellent frequency adjustment options – all add to the pleasant impression. Unfortunately, the overclocking tests proved hard for this product. I don’t quite understand what’s wrong with it: is it the BIOS or anything else. Overall, I recommend this mainboard as a highly functional product for use in the regular mode or slightly overclocked (for example, setting junior Thoroughbreds to 166MHz bus and increasing their multiplier a little).

ABIT KD7-S. This mainboard can’t boast numerous features, but it is simply beyond competition as it comes to CPU and memory voltages increase. So, if you don’t care about the absence of FireWire, but are fond of overclocking experiments, and don’t bother about minor PCB design drawbacks, this is your mainboard. Even though it has that BIOS thing that cannot truly say the memory frequency. One more point in ABIT’s favor is the implementation of all 5.1 sound connectors plus line-in and mic-in. The four memory slots (compared to three by the rest of mainboards) won’t spoil the picture, too.

DFI LAN PARTY KT400A. It’s true this mainboard has no rivals in functionality and rich accessories. However, it’s rather hard to say who’s going to buy it, as we couldn’t test it during overclocking. Anyway, this is your product if you have a windowed (or fully transparent) system case – ultraviolet highlighting looks definitely cool.

EPoX EP-8K9A9I. This mainboard can’t boast any reach functionality. On the other hand, such features as FireWire and Serial ATA are necessary for a limited user group. So, if you want a mainboard to serve you the next year or two, and don’t have any FireWire peripherals, it will suit you well. Besides, it has rather good overclocking capabilities.

Gigabyte 7VAXP-A Ultra. It’s all quite simple with this one. It is a very functionality mainboard with an exclusive solution for external Serial ATA devices. Regrettably, it doesn’t suit for overclocking at all: poor CPU voltage options, tricks with the FSB frequency, and the CPU multiplier adjustment with the DIP-jumpers. Overall, it’s excellent if you use it in the nominal mode, but pretty average at overclocking.

Soltek SL-KT400A-L. A reliable mainstream mainboard. However, Soltek tried to spoil an overall nice SOHO product with queer BIOS options and not working (as out tests showed) Red Storm Overclocking technology. This technology may work smoothly on other mainboards, but not on the SL-KT400A-L. So, if you buy it for an inexperienced office or home user, you’d better tell him not to mess up with the BIOS. Otherwise, the results can be unpredictable.