Well, today we have every right to state that Slot A mainboards built on VIA Apollo KX133 chipset have finally become an ordinary thing for the mainboard market. Almost all the mainboard manufacturers, who wished to design their products on this core logic, have introduced their boards. So, if you are looking forward to getting an AMD Athlon system your choice isn't limited to AMD 750 chipset any longer: you have the whole bunch of mainboards to choose from.

As is known, KX133 is destined to provide the market with a lot of low-cost mainboards. In fact, the first reference design on AMD 750 implied a six-layer PCB, which is more expensive and harder to produce than the ordinary four-layer boards. However, the mass supplies of KX133 were slightly delayed and AMD managed to offer another reference design of AMD 750 based mainboards based on a four-layer PCB. Nevertheless, due to VIA's numerous contacts with the mainboard manufacturers most of them preferred to deal with VIA Apollo KX133 chipset. In this article we will try to compare the available mainboards based on this particular core logic. If you are interested in comparing the features and the performance of the two different Slot A chipsets, AMD 750 and VIA Apollo KX133, take a look at our ASUS K7V Review, where we discussed this matter in detail.

So, let's not veer away from our primary goal and talk briefly about VIA Apollo KX133 chipset. Compared to AMD 750, KX133 can boast two main peculiarities: PC133 SDRAM and AGP 4x. No wonder these two features turned up in VIA's new chipset, because KX133 is none other but our good old buddy VIA Apollo Pro133A modified for EV6 system bus.

In other words, KX133, as well as Apollo Pro133A, follows asynchronous principles and hence the memory in systems on this chipset can work at a frequency different from that of FSB. It means that the memory frequency can be 100 or 133MHz while the EV6 bus in Athlon systems is clocked for only 100MHz. This allows getting greater bandwidth for the memory subsystem than in systems based on AMD 750: 1.06GB/sec compared to 800MB/sec. In practice however, the performance of AMD 750 based system with PC100 SDRAM hardly differs that much from the performance of VIA KX133 based system with PC133 SDRAM: the difference makes only 5%. Note that this number is valid only if Super Bypass option of AMD 750, which reduces the memory subsystem latency, is disabled.

The theoretical bandwidth of AGP bus can be also enlarged when working in 4x mode while AMD 750 supports only AGP 2x. Here the difference will be twofold: 1.06GB/sec against 528MB/sec. However, this impressive superiority is noticeable only when you operate theoretical peak bandwidths. In practice none of the modern applications is capable of making full use of AGP 2x bandwidth at least, not to mention AGP 4x. This happens mostly because the rendering data are stored in the local graphics memory on the graphics card. These were the major peculiarities of KX133 North Bridge, VT8371.

There are a few things worth mentioning about the South Bridge as well, VT82C686A. It seems to us there is no need to dwell on a set of ordinary features, such as 4 USB ports, AC'97 link, UltraDMA/66. We'd better focus on the following two details. First, VIA 686A South Bridge includes a hardware monitoring controller, which means that almost all KX133 based mainboards will allow controlling voltages, temperatures and fan rotation speeds, because they get this function "for free". Moreover, there is a PCI-ISA Bridge integrated into this South Bridge, which saves additional effort for actualizing ISA slots on the board unlike all chipsets from Intel's 800 family, which require an external PCI-ISA Bridge.

All in all, the features of VIA Apollo KX133 made a good impression. However, the problem hid in an absolutely different place. This core logic turned out not to support Socket A, AMD is planning to implement and promote actively in the near future. Unfortunately, while VIA was designing its KX133, nobody had even mentioned any 462-pin Socket A. That is why it appeared that the existing North Bridge package, 516-pin BGA, doesn't allow making a Socket on a four-layer PCB. To solve this problem VIA had to design another chipset Apollo KT133 (former KZ133), which features a 522-pin North Bridge and allows making Socket A on a four-layer PCB. So, KX133 and KT133 chipsets don't have any differences other than that.

AMD, however, is interested in manufacturing Socket A processors in the first place. The thing is that both new reps of Athlon family, Spitfire (Duron) and Thunderbird, will be made with 0.18 micron technology and possess an integrated on-die L2 cache. So, there appears no more need in a processor board for the external L2 cache microchips, as in ordinary Athlon CPUs. Since they can do without this processor board and the processor cartridge, all Athlons are currently packed in, the manufacturing expenses can get lower. Moreover, since AMD doesn't manufacture any of them itself: the company simply buys them from some other manufacturers. Besides, socketed processors allow introducing much more efficient CPU cooling scheme. So, a shift to Socket A will prove highly positive for AMD that's why they try to do it as soon as possible.

Having seen Intel's sufferings with pushing forward its Socket370, AMD probably decided to avoid making the same mistakes and to complete the transition as quickly as possible. That's why both new AMD processors, low-cost Spitfire (Duron) and more expensive Thunderbird, will be available only as Socket A versions. The only exception will be made for some AMD's OEM-partners. AMD will provide them with a certain limited amount of Slot A Thunderbird CPUs.

So, AMD is very anxious to see mainboard manufacturers designing and producing Socket A mainboards. Especially since they have already had the same unpleasant experience in the beginning of AMD Athlon sales, when it was extremely hard to find a corresponding mainboard. Therefore, now AMD is propagandizing actively its solution among mainboard manufacturers as well as among their potential customers. They are propagating two myths about the drawbacks of Slot A and disadvantages of VIA Apollo KX133 chipset, which has immediately become very popular in the market. To prove its point AMD says that Slot A CPUs based on a new Thunderbird core don't work in KX133 based mainboards. And secondly, they claim that it is impossible to design a Slot A-to-Socket A converter similar to a very widely spread Slotket developed for Intel platforms.

That is why we have every reason to announce that the future of VIA Apollo KX133 is predetermined and it will hardly live a long life. Although, AMD is very likely to be misleading everybody and to be guided mostly by its own marketing ideas. Let us explain.

First, AMD claims that Thunderbird doesn't work on mainboards based on KX133. However, before AMD could say knife, the leading manufacturers of KX133 based mainboards for Slot A, ASUS and ABIT, immediately released new BIOS updates supporting these CPUs. So, if there were any problems with Thunderbird not working on KX133 based mainboards, we shouldn't blame the chipset for them. If worst comes to worst, they will be caused by the faulty mainboard layout, otherwise, the problem should be with the absence of the corresponding support in the BIOS. In either case VIA is not the one to blame. Besides, AMD has recently announced that only the Thunderbirds, which were made in Austin with aluminum compound technology, will have problems when working on KX133 based boards. Copper Thunderbirds will work fine on these boards. In fact, even if AMD's attacks against KX133 based boards are quite grounded, the owners of these mainboards will hardly be very upset about the problems with Slot A Thunderbird processors, because these CPUs will be available in very limited amounts only for OEM and will never appear in retail. Nevertheless, with this "horror-tale" AMD managed to exert a certain influence over some mainboard manufacturers. Some of them (MSI, for instance) decided not to deal with KX133 at all and all the rest started moving towards Socket A mainboards.

Another argument AMD applied to speed up the shift to Socket A mainboards dealt with the Slot A-to-Socket A converter. As to AMD, it was just impossible, or at least very hard, to design one like that. However, we really doubt if this statement is credible. Just imagine: AMD is going to manufacture two versions of its Thunderbird - for Slot A and for Socket A. However, it will turn not only pretty expensive, but also quite silly to develop two different dies, especially taking into account a relatively short life of Slot A Thunderbird. That is why it is evident that Slot A Thunderbird will use the same die as its Socket A version. So, Slot A Thunderbird processor board will be none other but the notorious converter, which strikes AMD as something so impossible. Moreover, as to a certain source, some mainboard manufacturers have already succeeded in designing these converters. Well, this AMD's argument also seems to stand no criticism.

Nevertheless, it is VIA Apollo KT133, which will become the most widely spread chipset of the near future, although it is practically identical to KX133. However, there will hardy be any alternative to it till autumn. And in September we will witness a new wave of the competition between AMD and VIA. They will introduce their new AMD 760 and VIA Apollo KX266 chipsets supporting 266MHz FSB and DDR SDRAM. And in the meanwhile, when there are no products on chipsets other than KX133, we would like to offer you a comparison of the following mainboards on VIA Apollo KX133 chipset:

• ABIT KA7
• AOpen AK72
• ASUS K7V
• Chaintech 7ATA
• EPoX EP-7KXA
• Gigabyte GA-7VX
• LuckyStar K7VA133
• Tyan S2380 Trinity K7

To begin with, take a look at the comparative features chart below:

And now let's discuss the peculiarities of the tested mainboards in a bit greater detail.

ABIT KA7

KA7 is the first experience in manufacturing mainboards for AMD CPUs of ABIT Company, which has already won overclockers' love and respect. Unfortunately, ABIT didn't make an AMD 750 based board, it seems to be the six-layer design that prevented them from manufacturing it. And now ABIT is devoting a lot of time to mainboards on VIA chipsets, so KX133 is not an exception.

KA7 is quite an extraordinary board, and undoubtedly its main trump is SoftMenu III technology, which allows jumperless CPU configuring via BIOS Setup. In fact, many of the today's mainboards allow operating the FSB frequency through BIOS Setup, but KA7 can boast a bit more than that. First, it is the only mainboard where you can change the FSB frequency with 1MHz step within the interval between 100 and 183MHz. Besides, you can also control Vcore and Vio, which is also quite valuable for CPU overclocking. Due to these unique functions ABIT KA7 deserves a worthy place in overclockers' PCs. By the way, a new BIOS released for this mainboard will let you change L2 cache frequency divider via Setup.

Another unique feature of KA7 is four DIMM slots. Theoretically, VIA Apollo KX133 chipset can address up to eight memory banks, i.e. nothing prevents this board from supporting four memory slots, however, the reference design and PC133 specification recommend mainboard manufacturers to equip their products with only three slots for higher stability. Nevertheless, ABIT chose a different way and provided its mainboards with a memory buffer, like those for Intel processors. These six chips by Texas Instruments located right in front of the DIMM slots ensure KA7's stable functioning with four SDRAM modules.

The board has 6 PCI slots, which is the greatest amount among all the boards tested. These slots as well as 1 ISA slot offer very good expansion options. However, we have to point out that KA7 doesn't have any integrated sound and AMR slot.

Besides, we should also mention that ABIT KA7 supports officially Slot A Thunderbird processors.

Also ABIT has recently announced a new revision of its KA7, KA7-100. You can distinguish this board from its predecessor by the presence of UltraDMA/100 controller (probably with RAID support) and two extra IDE channels supporting this protocol. The ordinary KA7 has just a couple of empty spots instead.

However, to our great disappointment, we have to point out that this mainboard proved the least stable of all the participants.

AOpen AK72

AOpen mainboards have always been known for their high stability and excellent construction due to very high quality components. And what about AK72?

At first sight your eyes rest on a very thoroughly designed voltage regulator module composed of 14 capacitors 2200uF each and 6 transistors with massive heatsinks. AMD Athlon CPUs, especially their elder models, consume a lot of power that is why the quality of a voltage regulator is of great importance. We think it is thanks to this regulator module that AOpen AK72 proved very stable and reliable throughout the tests. Besides, we should also note that this mainboard has shown very good performance. In particular, AK72 very often appeared among the leaders in gaming applications.

BIOS Setup of this mainboard is slightly different from what we saw by all the others and features a couple of extra functions responsible for saving the current settings in EEPROM and restoring them from there.

The package includes a special bracket with a pair of additional USB ports. It can be installed into the case instead of an expansion card.

Unfortunately, AOpen KA72 can't boast an impeccable design. Even if we close our eyes to the extremely large size of this mainboard (it is so big that some cases may turn out unfit for it), there will be still a few more drawbacks to mention. The major problem is a power supply connector, which is located between DIMM slots and Slot A. In fact, it is not the best location because the cable may hinder air circulation and prevent you from installing larger cooling devices.

ASUS K7V

ASUS was one of the first to launch its mainboard on VIA Apollo KX133. This time there was no conspiracy (remember the situation with AMD 750 based boards): ASUS was openly promoting its K7V.

Among the main advantages of this board we should first of all mention good performance of this product. It proved the best of the best in office applications.

Besides that, K7V can boast a number of unique features. First, ASUS equipped its mainboard with an AGP Pro slot, which is not only AGP 4x fully compatible but also supports power-hungry graphics cards.

Secondly, ASUS decided not to use the hardware monitoring controller integrated into KX133 South Bridge. It uses their own controller. That's why K7V can boast a bit richer monitoring options compared to all the other mainboards on the same chipset.

Like AOpen AK72, ASUS K7V is accompanied by a bracket with two additional USB ports, which can be installed instead of some expansion cards. Like ABIT KA7, K7V officially supports Slot A Thunderbird in its BIOS.

And overclockers should be very happy with a great number of the supported FSB frequencies and the possibility to increase Vcore and Vio.

The main drawback of this board we would like to mention is extremely large PCB size, which can cause certain problems during the board installation into some cases. Besides, some of you may be also disappointed with the absence of ISA slots on the board.

Chaintech 7ATA

The main peculiarity of this Chaintech board is 3BIOS technology including two main components. First, there are two Flash ROM microchips on the board. The BIOSes stored in these chips are absolutely independent and stand by each other against damaging. Second, the BIOS contains a special flashing utility, which prevents the user from using some incomplete or faulty flashing utilities.

Also we can't leave out the fact that Chaintech 7ATA is of quite an ordinary size that is why it looks just tiny against the background of all other monstrous mainboards.

Besides, 7ATA may feature a fully-fledged PCI hardware sound controller, Creative CT5880, which is optional though. Compared to the software AC97 sound used on other mainboards considered, it for sure provides much better sound quality and doesn't overload the CPU.

Unfortunately, Chaintech 7ATA has no specific options for extreme CPU overclocking and doesn't allow changing Vcore and Vio, that is why we don't think it can be called overclocking friendly.

During our tests we managed to find out that in some cases the BIOS of this board didn't read the memory module parameters from SPD correctly. However, we hope that this bug will be eliminated in further BIOS versions.

EPoX EP-7KXA

EPoX EP-7KXA makes a very good impression due to the accurate and tidy design and high quality voltage regulator module including 18 capacitors, 2200uF each. All in all, EP-7KXA corresponds fully to the reference-design that's why we can't point out any significant defects of the PCB. Moreover, the PCB size is quite standard for an ATX board, which is a valuable thing for KX133 based mainboards. In general, EPoX EP-7KXA is an ordinary working product, aimed at nothing exalted, which fulfils its functions 100% correctly.

Gigabyte GA-7VX

This mainboard stands out among all the rest due to its brand DualBIOS technology. Like Chaintech 7ATA, this board also has two BIOS microchips, so that the contents of the first chip can be always reflashed with the contents from the reserve one. This protects the board against breakdown in case Flash ROM gets spoilt.

The second distinguishing feature, not that significant though, of Gigabyte GA-7VX is the retention mechanism fastened to AGP slot, which serves to maintain the graphics card in a proper position.

Like Chaintech, Gigabyte equips its mainboards with a hardware PCI sound controller from Creative, that is why they can boast integrated sound of much higher quality than that provided by the competing products.

Gigabyte GA-7VX is the only mainboard using AMI BIOS. Although this BIOS and the old Award 4.51PG are as like as two peas, AMI BIOS features a bit fewer options for memory timings configuring. On the one hand, it's good, because it prevents inexperienced users from being misled, but on the other hand, it makes the settings less flexible.

In our tests GA-7VX proved a highly stable and very reliable mainboard and showed very good performance.

LuckyStar K7VA133

Very often LuckyStar products arouse numerous complaints about the quality. But at the same time the mainboards by LuckyStar are slightly cheaper than those manufactured by their competitors. Well, let's see what this product is worth. At first sight we had an impression that LuckyStar engineers decided to save a bit of money at the expense of the board's voltage regulator module, which has very few capacitors. Theoretically, this fact should greatly harm the mainboard's stability, especially when used with faster AMD Athlon processors. However, our tests didn't reveal any serious stability problems and the mainboard proved as good as most of its more expensive competitors. But this seems to be the only positive thing about it. The performance in office applications leaves much to be desired. Besides, there is no integrated sound and the dip-switches, which you can see on the board, are mentioned neither in the manual nor on the PCB.

Tyan S2380 Trinity K7

Tyan Company has become a VIA addict. And as to them, they have never regretted it. That is why no wonder they decided to manufacture a product on VIA Apollo KX133 for Slot A.

Although Tyan S2380 Trinity K7 can't boast any peculiar features, it stands out due to its high quality voltage regulator module composed of 6 transistors with huge heatsinks and 17 1500uF and 2200uF capacitors. Even though the North Bridge of S2380 Trinity K7 has no heatsink at the top, it seems to have no negative effect on the entire mainboard stability.

The board has 6 PCI and 1 ISA slot and hence features a bit more expansion options than a lot of its competitors.

Unfortunately, the tests showed that Tyan S2380 Trinity K7 has some problems with correct reading of memory parameters from SPD, though it is not that dramatic.

Performance

The testing system was configured as follows:

• AMD Athlon 800 CPU
• Creative 3DBlaster Annihilator graphics card
• Creative Sound Blaster Live! sound card
• IBM DJNA 372200 HDD
• 128MB PC133 SDRAM by Micron
• MS Windows98 SE

We began with the performance in office applications.

This test show that ASUS K7V is an indisputable leader. Since office applications are the best way to check how the board operates memory and processor L2 cache, we can say without doubt that ASUS managed to achieve the most optimal functioning of the memory bus and FSB.

Here we have three leaders: ASUS K7 (for the second time already), ABIT KA7 and Gigabyte GA-7VX.

Now let's take a look at the games and the performance of our mainboards there. We will be able to feel not only the work of the memory subsystem but also that of the AGP bus.

Do you see how greatly the situation has changed compared to the previous tests? And it is the influence of AGP port and its timings. According to this test, EPoX EP-7KXA is No 1.

The smaller gets the amount of data transferred via AGP bus, the more convincing becomes the leadership of AOpen AK72.

Unreal Tournament belongs to the games, which load the CPU and the processor bus in the first place. Here two mainboards from the bunch prove the fastest: ASUS K7V and AOpen AK72.

If we set a lower resolution in Unreal Tournament, the group of leaders will include one more member - Chaintech 7ATA.

Expendable loads not only AGP bus but also the memory bus that's why the results obtained in this test are very similar to what we saw in office applications. Again ASUS K7V is ahead of all.

At lower resolutions Chaintech 7ATA joins the leader in Expendable.

However, as you can see from the tests all VIA Apollo KX133 based mainboards perform very close to each other. So, we suggest bearing in mind a combination of features the board boasts and not its performance.

Conclusions

First, let's sum up:

And now our conclusions. We would recommend you the following two mainboards on VIA Apollo KX133:

• Gigabyte GA-7VX. This high quality mainboard proved the most stable of all and showed nearly the best performance. And the unique DualBIOS technology makes the board even more attractive, because it protects the mainboard against such viruses as CIH and against errors during new BIOS versions reflashing. The only drawback of GA-7VX is processor overclocking, which is practically impossible. So, Gigabyte GA-7VX will be the best choice for those users who appreciate system stability and reliability above all.
• ASUS K7V. ASUS stuffed its mainboard with the whole bunch of features, which are pretty valuable for dedicated overclockers. Vcore settings, wide range of supported FSB frequencies and progressive hardware monitoring will hardly remain unnoticed by those of you who are used to squeezing the maximum out of your PCs. Besides, this board proved one of the fastest, which is also a definite plus. However, you will have to put up with a couple of design inconveniences: extremely large size and the absence of ISA slots.