Review of MSI KT4A Ultra Mainboard on VIA KT400A Chipset

by Grigoriy Gubankov
07/13/2003 | 11:41 PM

It is no secret that NVIDIA nForce2 is the fastest solution for Socket A platform in almost all cases. This success can be explained by a very successful memory controller implemented in this chipset. It is not just a dual-channel controller. It also supports very interesting DASP algorithm, which allows the chipset to save some data, which the CPU might need later, from the system RAM into the buffer in advance. That is why when the CPU really does request this particular data, it will be able to get them not from the slow RAM but from a much faster internal buffer, which will reduce the latencies this way.

However, one of the major advantages of this chipset, its dual-channel design, is at the same time one of its major drawbacks. This is quite true, as this feature reduces the memory configuration flexibility, because the memory modules used in the system should better be the same, i.e. you have to replace them in pairs, which can hardly be considered convenient. Moreover, the use of two memory channels sets much stricter requirements to the PCB layout and its components, which definitely affects the production cost a lot. Since the mainboard manufacturers are no altruists, the higher production expenses influence the final product price for the end-users, of course. And keeping in mind that computers have already become one of the common home appliances, every user will surely pay close attention to the end-system cost. So, nForce2 will no longer look that attractive, from this point of view. Besides, some business users working on AMD platforms are hardly very eager to pay extra money for an office PC, especially if they never actually use their maximum performance.

For some of the reasons mentioned above, single-channel chipsets are still the most attractive solution for many Socket A system users. NVIDIA also realized this and launched a single-channel version of its nForce2 solution aka nForce2 400, which supports 200MHz bus of Athlon XP processors and the newest PC3200 memory. However, NVIDIA is not the only chipset maker to support this memory type. VIA and SiS also introduced their solutions of the kind: VIA KT400A and SiS746FX. So today we are going to talk about these two babies, mostly about the VIA KT400A, of course :)

Closer Look: VIA KT400A Chipset

As you may have already guessed from the chipset name, it is a successor of KT400 solution, which was at first supposed to support PC3200. However, shortly after the chipset launch, there appeared some problems with the implementation of this support: the chipset performance when working with PC3200 turned out much lower than in case of PC2700. Moreover, KT400 chipset didn’t know to clock the memory at 200MHz when the system bus frequency equaled 166MHz. These pretty strange phenomena could be explained by the absence of a general JEDEC standard for PC3200 (they couldn’t even call it PC3200, but only DDR400). A little later, mostly due to Intel’s initiatives, PC3200 DDR SDRAM turned into a standard and the manufacturers got the opportunity to announce support of this memory type without being afraid of any compatibility issues.

It is interesting that VIA with its KT400A followed NVIDIA in a way, which had implemented the data prefetch DASP algorithm even in the first nForce chipset. VIA called its technology FastStream64, which was probably supposed to imply the 64bit memory bus of the KT400A chipset unlike 128bit memory bus of the nForce2, which could also ensure high level of performance. However, this doesn’t change the idea of the technology behind the name. FastStream64 is a much bigger number of internal chipset buffers compared with what the other VIA’s products can boast.

Judging by the description on the company site, VIA was at first going to supply KT400A with the new VT8237 South Bridge supporting SerialATA. However, their plans changed and we can see the new VT8237 only with the KT600 chipset, the next solution from VIA for Socket A platform (read our article called Contemporary Socket A Chipsets: NVIDIA nForce2 Ultra 400, NVIDIA nForce2 400, SiS748 and VIA KT600 for details). Besides that, KT600 will also support AMD Athlon XP processors with 200MHz bus and will boast “enhanced” PC3200 DDR SDRAM support. All in all, this will be just another chipset remake, which VIA has always been famous for. We should say that VIA is still a little behind SiS in terms of support for Athlon XP processors with 200MHz bus. SiS has already officially launched SiS748 chipset supporting these processors, while VIA is only getting close to KT600 release. Although we wouldn’t call this a dramatic lag. The thing is that SiS has always been slow in bringing the freshly announced solutions to the mass market, and even though the company is now using UMC’s production capacities, the actual selling schedule for SiS748 based mainboards is still unclear. Besides, the CPUs with 200MHz bus are not very widely spread yet.

All in all, KT400A (even the way it is now) is quite functional, and its pin-to-pin compatibility with the previous chipsets from VIA should make it much easier for the mainboard manufacturers to shift to production of new boards on this core logic set. Anyway, the puzzling names of the new chipset do not make the mainboard manufacturers and the users any happier that is why KT400A is not that popular yet. Many mainboard makers decided simply to ignore the VIA KT400A, believing that it doesn’t differ that much form the KT400. Nevertheless, we got a few mainboards on VIA KT400A in our lab and we are going to tell you a little more about them. The first one we would like to introduce to you today is MSI KT4A Ultra.

Closer Look: MSI KT4A Ultra

Of course, KT4A Ultra is not a single product name but the name of the entire product family including three models: KT4A Ultra, KT4A Ultra-SR and KT4A Ultra-FISR. The difference between these models is the following. The first product is the basic model with minimal features set. The second one supports two SerialATA channels due to the onboard Promise PDC20276 controller. And the third one also supports Gigabit LAN due to onboard Broadcom 5702 controller and FireWire interface due to VIA VT6306. We managed to get the “SR” board. Unfortunately, the box didn’t say anything about it and I guessed that it is the SR model only after I had taken a really close look at it and consulted the MSI web-site. Another strange thing about the mainboard model variations had to do with the Bluetooth connector, which was called an optional feature in the user’s manual. However, the company web-site didn’t say which mainboard models have no Bluetooth connector. Anyway, the package didn’t include the Bluetooth module, so I really doubt that this feature is of much practical value, because I hardly ever saw these modules selling separately.

By the way, do you remember that I told you about the pin-to-pin compatibility of the VIA KT400 and KT400A chipsets? We got a very clear evidence proving this point much better than any VIA’s statements. Look at the mainboard: the model name is written on a piece of paper stuck to the PCB. I couldn’t help removing this paper sticker and... This is what we saw under it:

Of course! As you have already guessed, MSI decided to save time and money and didn’t even bother to rewrite the model name on the mainboard PCB. What a smart and economic move!

However, this is just a sample and I suppose that mass units will have the model name rewritten, of course. But even if they don’t do it, this is not much of a deal. Let’s better check the mainboard specs:

The package bundle is pretty nice, nothing necessary is missing, besides, there is a couple of pleasing trifles in it. I think you will agree that it is very nice to have a bracket for the rear case panel, two SerialATA cables a FDD cable and a round UltraATA cable and two user’s manuals: one for the mainboard and one for the SerialATA controller integrated on it, and a CD-disk with the drivers and additional software. Also, there is a panel with the USB ports on it and a D-LED diagnostic system, with the explanations for all LED combinations listed in the user’s manual. One of the USB ports is covered with a film warning that you shouldn’t use it when the Bluetooth-module is used: “do not remove when using Bluetooth”. This whole thing is called D-Bracket 2. Besides, the package also includes a sound panel for the case rear called S-Bracket, which features S/PDIF and additional audio outputs. I am going to talk more about it later in this review.

At first let’s have a look at the mainboard ports panel:

As we see, there is nothing strange there, only except 4 USB ports instead of traditional two on other mainboards. I think this is a very smart solution, because devices with USB interface get more and more popular today. That is why 2 USB ports are very often insufficient for the users’ needs. And taking into account the additional panel with USB-ports included into the mainboard package we get the total of 6 USB ports, which is a definite advantage of this product.

As for other mainboard features, we could hardly call them remarkable or rich especially from the today’s point of view. First of all, we were very upset to see not integrated network onboard, especially since only a lazy manufacturer doesn’t integrate network onto his products today. Even the reference SiS746FX mainboard, which we will also use in our comparative testing today, features onboard LAN. Maybe MSI wanted to make this solution as inexpensive as possible, as the additional Ethernet controller costs some money. However, if this is the case, then I can’t understand why they integrated the Bluetooth controller, which is also not a free piece of hardware. And to tell the truth, I doubt that Bluetooth is more demanded today than Ethernet.

As I have already said, the mainboard we are reviewing today has no FireWire controller. This MSI’s move certainly makes a lot of sense, because peripheral devices with FireWire interface are pretty expensive and those users who have enough money to afford them, will definitely be able to get a more sophisticated mainboard.

Now let’s check what we get from MSI KT4A Ultra-SR in terms of work with the disk subsystem. Of course, the mainboard is equipped with two Parallel ATA connectors implemented via the VT8235 South Bridge. Also there are two SerialATA connectors and one Parallel ATA aka IDE3 provided by the Promise controller. You can connect only one Parallel ATA hard disk drive to the IDE3 connector. The onboard Promise controller supports RAID 0 (stripe) and RAID 1 (mirror). You can build an array from any two HDDs (which means that you can use one Parallel ATA and one Serial ATA HDD, for instance), however, it certainly would be best to use two similar HDDs.

As for the support of different memory types, the situation here looks as follows. According to the table in the user’s manual, the mainboard supports PC2100 (DDR266) and PC2700 (DDR333) with the 200MHz FSB. When the bus frequency equals 266MHz, the mainboard can work with PC2100, PC2700 and PC3200 (DDR400). Moreover, they guarantee proper system functioning only if you use memory modules specifically approved by MSI. I can’t say anything about those memory modules, which are not recommended by MSI, but the approved Corsair XMS3200C2 worked perfectly well in our system.

The most interesting things, however, take place when the bus frequency grows up to 333MHz. In this case, you can feel safe and sound only with PC2700, because PC2100 is not supported at all, and as for PC3200, you can allegedly try it but the company doesn’t promise anything. And frankly speaking, it is very good that they do not promise anything, because we failed to make the mainboard work in this mode. I assume that the promised improvement of the memory support, which is to come in the KT600 chipset, should imply proper work of the system with PC3200 DDR SDRAM. We will continue talking about the implementation of different memory types support later in this article when we come to describing the BIOS Setup features.

Now a few words about the implementation of sound on MSI KT4A Ultra-SR board. This mainboard supports six-channel sound due to C-media CMI8738 codec. The sound quality it provides is a subjective parameter, so we won’t discuss it here. We’d better check the way this six-channel sound is output. Do you still remember about the S-Bracket? Now we are going to dwell on it a bit more.

So, as you remember there are three connectors on the back panel of the mainboard. By default these are Line-Out, Line-In and Mic. Of course, these connectors are too few to get impressive six-channel sound. So, if you want to have four- or six-channel sound output you have to sacrifice Line-Out and Mic by making them work as a sound output for the second pair of speakers and the central channel or subwoofer. This is where you will appreciate the availability of S-Bracket, which is supplied together with the mainboard: it will prevent you from resorting to such inconvenient tricks. This bracket is equipped with two mini-jacks for the central channel and the second pair of speakers, and with an optical and coaxial S/PDIF Outs, which will be helpful if your have speakers support this interface. Of course, S-Bracket should be connected to a special port on the mainboard.

PCB Design

From my point of view, the PCB design of MSI KT4A Ultra-SR is not bad, though there is nothing really remarkable about it. It has a few definite advantages, but at the same time is not free from some upsetting drawbacks. Among the advantages we would like to mention very convenient positioning of the connectors for the rear panel brackets of the PC case, namely for the USB, D-Bracket and S-Bracket (the KT4A Ultra-FISR mainboard model will also get FireWire connectors).  All connectors for these brackets are located closer to the back edge of the PCB, so that the cables will lie quietly at the bottom of the case and will rise up to the connectors only at the brackets, which will also cause absolutely no problems, as there will be no PCI cards in the corresponding slots. The only thing I don’t quite understand is why the cables are so long, especially those of the D-Bracket. But this doesn’t matter, as they will never leave the bottom of the case, for sure. The S-Bracket cable is also a bit too long, I would say. By the way, the user’s manual says that the only USB connector available on the mainboard serves for connection to the USB ports of the case. Nevertheless, as I have already said the bracket with the additional USB ports is supplied with the mainboard.

Now I have to mention a few things, which disappointed me a little bit, as far as the PCB design is concerned. To begin with, the ATX power supply connector is located in the upper part of the PCB, very close to the edge (besides, there is no additional 12V power supply connector). As a result, you will have to tie up the power cable into a very tight loop, which will definitely hinder the air circulation inside the case, and may even get into the processor cooler fan. Besides, there are three pretty tall capacitors next to the power supply connector, which can be very easily damaged when plugging in the ATX power cable and especially when unplugging it.

The FDD connector is also located not in the best way. In fact, it is pretty nice that they placed it in front of the DIMM slots, however, its close location to the IDE connectors may cause certain problems when plugging/unplugging the FDD cable and IDE cables too, in case both IDE connectors are occupied. And one more thing about the DIMMs: the installed AGP graphics card will make it a really hard task to install or remove the DIMM modules from the slots, even if the card is not a long one, but of a regular size like RADEON 9700 Pro, for instance.

The CPU power supply circuit is a dual-channel one, which means that MSI KT4A Ultra is no overclocker’s product, because all overclocking-friendly solutions usually use three-channel or even four-channel power supply. The memory power supply circuit also doesn’t push us to think about overclocking, because it is linear and not pulse. However, we are going to go into details about overclocking in the next section of our review. So, read on!

BIOS and Overclocking

MSI KT4A Ultra uses BIOS from AMI. However, there are not too many differences in settings of this BIOS and the competing BIOS from AWARD. The same colors, font and even items names. The main BIOS window looks like this:

Please, pay attention to the item dealing with the voltage and frequency settings. It is a definite misprint. It is interesting that the same misprint is also on the screenshot in the user’s manual (though it is absent in the text). I wonder how come that no one noticed it!. Anyway, this is not the only typo in the BIOS. A very similar mistake can be found in the Advanced BIOS Features, where we see “Boot Sequency” instead of “Boot Sequence”.

All in all, it seemed to me that they were really in a rush with this BIOS, so that no one even bothered to read the text. Unfortunately, this rush told not only on the text, which we will show you later. Now let’s take a closer look at the notorious “Boot Sequency” item, because it is the first interesting thing we come across in the BIOS of our board:

We can select three boot-up devices and define whether the system should be searched for on “other devices”. By default the first device is left empty. In fact, among the “devices” can also be USB flash drives, for instance. Here I would like to say that I managed to boot from a flash-drive successfully, although I had to enable in advance the USB Legacy Support, which is located in a totally different menu page, namely in Integrated Peripherals. Of course, the manual didn’t say a word about it. Also, note that the 1st Boot Device may be set directly during the POST procedure by pressing F11 key.

Now I would like to dwell on the item, which will be of interest to fine tuning and overclocking fans. It is Advanced Chipset Features Setup, to be more exact – DRAM Timing Control page. AGP Timing Control is hardly of any interest to many users, especially, since the parameters available there are not very familiar to the majority of users. This is what the DRAM Timing Control page looks like:

Besides a great number of timing settings on this page, you can also adjust the memory frequency. And here we have every reason to take a few big pokes at MSI. The first poke: at any bus frequency, the list of available memory frequencies includes not only those ones, which are actually supported but all the other ones, too. This way, we strongly recommend to refrain from changing the memory frequency without the table of correspondences from the user’s manual. If you set the wrong frequency, the mainboard will not boot and the diagnostic LEDs will indicate failed memory detection. The second poke: the memory frequencies are always shown as 266MHz, 333MHz or 400MHz, even if the bus frequency is 140MHz, for example, which you can see on the screenshot. Moreover, they are displayed like that not only in the BIOS, but also during POST. Well, this is the asynchrony from MSI.

As I have already promised, I have to say a few words about AGP Timing Control. Here we can enable/disable Fast Writes mode, change AGP Aperture Size, and AGP time settings, such as inserting an idle clock cycle before writing in and reading from the AGP for higher stability. The first thing mentioned should be responsible for manual changing of the AGP working mode, but as you can see from the screenshot, it is frozen at “Auto”. Well, this is one more strange peculiarity, which is definitely not mentioned in the user’s guide as well. In fact, we are not longer surprised with things like that...

The next undocumented surprise is waiting for us in the Power Management Setup section. Take a look at the CPU HALT Command Detection parameter. The name is far from evident, don’t you think so? The manual doesn’t have any description for this item. So, we can only guess what this item actually stands for. The most logical supposition in this case is that this is none other but the well-known S2K Bus Disconnect technology, which allows reducing significantly the AMD CPUs temperature in idle mode. However, our experiments didn’t show any temperature reduction when we enabled this option in the BIOS. So, this is either not the Bus Disconnect technology, or it doesn’t work, which is equally unpleasant for the end-user. By the way, the way this mainboard measures the CPU temperature is still a mystery for me: on the one hand, I didn’t find any thermal diodes on the PCB under the processor socket, but on the other hand 47^oC AMD Athlon XP 2700+ temperature seems to me a bit too low, especially when the CPU utilization is 100% and the regular boxed cooler is used. However, the anti-overheating circuit used on this board is very likely to be using the internal CPU diode for temperature measurements.

The next page is titled PC Health Status. I thought I should still mention it, although there is nothing unusual there. In fact, it is pretty hard to think of anything new there, the most important thing is that all voltages and temperatures are available. There is also Chassis Intrusion item, although I don’t really know how the board detects this intrusion.

Now let’s take a look at such an important page of the BIOS Setup as Frequency/Voltage Control.

What can we say here? There are quite many controls, especially for one mainboard, which is not positioned as an overclocker’s choice. There is also a pretty exotic item called DDR Termination Voltage, which actual efficiency is quite doubtful to me. However, we will not just enumerate all settings. Let’s better check their quality and efficiency, i.e. the increments for the adjustable parameters.

So, FSB Clock can be adjusted from 100MHz to 280MHz; the CPU ratio – from 6 to 15; the CPU Vcore – from 1.575V to 1.7V with 0.025V increment or Auto (too low); DDR Voltage – from 2.5V (Auto) to 2.8V with 0.1V increment; Termination Voltage can be set to 1.27V, 1.29V or Auto; the AGP Voltage can be set to Auto and to any value from 1.6V to 1.8V with 0.1V increment. These are pretty worthy intervals, really, except the processor Vcore, which has been reduced to unbelievably narrow interval for some unknown reason.

I don’t know if you noticed it or not, but I haven’t yet said a single word about the opportunity to change or check the frequencies and dividers for the AGP and PCI buses. It is not because I forgot about it, but because there is no way to check or even change these frequencies and dividers. Absolutely no way! You can’t see or adjust them in the BIOS, you will never notice them during POST, nowhere. The mainboard must be setting these dividers on its own, which might be correct, but the available divider set is very limited. Also, starting from a certain FSB frequency, the AGP and PCI speeds will grow up continuously until they turn pretty dangerous. However, the user will learn about it only when he sees some weird artifacts on the monitor or receives a message from the HDD check utility about the FAT defect.

Having seen everything I have just mentioned above, I got very well prepared for the overclocking tests: I didn’t expect any high results. For our tests we took a specially prepared AMD Athlon XP 1700+ processors, reduced the multiplier to 9, the Vcore grew to 1.7, the memory timings were set to 2.5-3-3-7-1 (we used two Corsair XMS3200C2 memory modules) and we kept increasing the FSB frequency until the system started stumbling at list in one of the tests. We used the following tests: 3DMark2001 SE and SPECviewperf 7.1. Actually, I was right when I hoped for relatively low results: the system worked stably only at 177MHz FSB, which is not a great achievement.

Performance

In this section of our review we will discuss the performance of VIA KT400A chipset in MSI KT4A Ultra and SiS746FX in the reference mainboard from SiS. Our testbeds were configured as follows:

Now a few words about the testing methodology. At first I was going to compare the chipsets performance with PC2700 and PC3200. However, I had to give up the latter idea, for the reasons mentioned above, so with the PC3200 we will only have SiS746FX.

Anyway, the tests of SiS746FX reference mainboard also didn’t go that smoothly. The thing is that its BIOS contained a vast amount of memory timings settings. Moreover, the system didn’t always start when I set some timings from the available list, even if they looked pretty safe. That is why I gave up the idea to fine-tune the memory subsystem and set the fastest Ultra mode for PC2700 and one step slower Turbo mode for PC3200. As for the memory subsystem settings of the VIA KT400A based system, the timings were set to the minimum there. We also enabled 4-way interleaving and SDRAM 1T Command. Fast Command was set to Fast and Fast R-2 Turnaround – disabled.

Before we start discussing the actual results, have a look at two really cool screenshots below, which were taken when we set the bus frequency of the tested systems to 166MHz:

VIA KT400A

SiS746FX

As you see, both testing participants have slightly exaggerated the bus frequency and hence the processor frequency. To be fair, we have to say that MSI increased the frequency by only 1/3 MHz, while SiS raised it by almost 1.5MHz. These are the extraordinary ways of proving the superiority of the product.

So, now that we keep this fact in mind, let’s check out the numeric results of our tests:

First of all, we have to point out that SiS746FX is noticeably far ahead in Business Winstone, which is probably provided by the high quality IDE driver, because the work with the disk subsystem takes a considerable part of the test (saving and loading files). In other benchmarks the chipsets perform more or less equally: the maximum performance difference throughout the whole test session was not more than 4.2% in the synthetic PCMark2002 test. Also it is worth pointing out that SiS746FX didn’t gain a lot when the memory was replaced with the faster PC3200. It is probably the bus of Athlon XP processor, which features lower bandwidth than the memory with all the resulting consequences. It would be also very interesting to take a look at the KT400A performance in this case, because the FastStream64 algorithm should definitely contribute to the performance improvement. Unfortunately, we can’t check it out this time.

Conclusion

After reading this article you may have though that the reviewed MSI mainboard is not a very successful product. It is not quite correct. As you can see, the major problems we discovered have to do with extreme overclocking settings. When you work in the nominal mode, there are no problems at all. Maybe the only thing to be mentioned is the use of PC3200 and 333MHz bus. So, if you are looking for a reliable basis for your home system, and if you are not going to torture your mainboard in extreme overclocking attempts, then this MSI product is a good choice. However, if you are an overclocking fan, we wouldn’t recommend buying this solution, which costs around $80-$90.

As for the competition between VIA KT400A and SiS746FX, there is not much to be said here. Both solutions perform almost equally fast and offer similar features set with the only exception of PC3200 support. Both these products are not brand new already, because VIA KT600 and SiS748 will very soon conquer the mass market and offer the users support of AMD Athlon XP processors with 200MHz bus. However, if you don’t need this support, you can feel quite at home with the solutions discussed today in this article: they are better value and offer practically the same set of features.