At present Intel is trying to get rid of numerous problems, which simply overwhelmed it all together. First something went wrong with the launching of 133MHz Pentium III processor, then there was some trouble with the new i820 Camino chipset, which should support an absolutely new memory type - RDRAM. So, while Intel was up to its neck in problems, its old competitor in the chipset market - the Taiwanese company VIA - turned out the first to supply a new chipset supporting 133MHz system bus. This chipset became known as VIA Apollo Pro133, and it consists of two chips: VT82C693A and VT82C686A.

The first question that crops up as soon as you read the words "supporting 133MHz system bus" is: what does this support actually mean? As is known, it is pretty hard to find now a mainboard based on the good old i440BX, which has no support for this frequency. However, everything concerning BX is not so simple as it might seem. It is not only the final CPU frequency that depends on the FSB frequency, but also the working frequencies of PCI and AGP. If the external frequency is set to 133MHz, most BX-mainboards provide the PCI bus with the required 33MHz mentioned in the specification with the divider equal to 4. While with AGP everything is more complicated. i440BX has only two dividers, which serve to get the AGP bus frequency from that of FSB. They are 1 and 2:3. In other words, there is absolutely no way to get 66MHz AGP as it is stated in the specs if the FSB frequency is set to 133MHz. The best thing you can get is 89MHz. As you see, this value exceeds the nominal frequency by 23MHz, which can lead to very unpleasant consequences: some graphics accelerators, especially those of the latest generations, may simply stop working. Though on the other hand, if the graphics card starts working in such conditions, AGP-texturing will run much faster than usual.

Well, VIA engineers kept in mind all these problems and equipped their Apollo Pro133 with additional AGP divider equal to 2, which allows getting the required 66MHz for this bus and eliminate all the trouble concerning the graphics card and its work with the 133MHz bus. Actually, this is the key issue in the whole 133MHz FSB support.

However, there appears another question: is there any real need for 133MHz FSB? To be impartial, there is. Think carefully, even if we suppose that 133MHz memory exists (PC133 SDRAM or high-quality PC100), then we will hardly find such a CPU, which would be able to work at 33% higher frequency. The only more or less acceptable way to make a 100MHz Pentium II/Pentium III work with 133MHz bus is to disable L2 cache. But we think you will agree: this idea is just crazy! Who will sacrifice almost 70% of the performance for the sake of 33% gain in processor frequency?! That is why it turns out that nowadays the only possible utilization of VIA Apollo Pro133 is to continue working with 100MHz system bus. So, we have every reason to call its coming a bit premature. And as for VIA Apollo Pro133A with the North Bridge VT82C694A, which is much more practical due to its AGP 4x support, it is still unavailable, unfortunately.

VIA Apollo Pro133 is a chipset with classical architecture. The North Bridge VT82C693A is a memory, AGP and PCI controller, which is pin-compatible with Intel's analogue i82443BX. The South Bridge VT82C686A includes PCI-ISA Bridge and IDE controller. If we compare this chipset with its predecessors, for example, VIA Apollo Pro Plus, we'll happen to know that in VIA Apollo Pro133 the thermal characteristics were regarded with much greater attention, so that to provide stable performance at higher frequencies. And as a result VT82C693A can boast a 4-level architecture compared to the 2-level one of its older fellows.

Let's try to compare VIA Apollo Pro133 With the Intel's competitor - i440BX:

Judging by the table above we can say that VIA Apollo Pro133 offers a really wide range of different features, but at the same time its memory operating speed leaves much to be desired, which is, actually, quite usual for VIA chipsets.

In fact, VIA Apollo Pro133 supports not only the official 133MHz bus, but also VCM (Virtual Channel Memory) SDRAM. This technology allows opening up to 16 independent cached memory access lines simultaneously, which helps to data access speed when the data stored in different banks is requested. According to the developer, i.e. to NEC, this new approach may increase memory operating speed by 30%!

Another advantage of VIA Apollo Pro133 in contrast with i440BX is the possibility to use the asynchronous memory bus. Here it means that the memory may receive either 100 or 66MHz if the FSB is set to 100MHz, and if the FSB frequency equals 133MHz then the memory may be 100 or even 133MHz. Such decrease of memory frequency compared to the external CPU frequency may be of great use actually. It will simply save you trouble, because you won't need any new DIMM modules while shifting to a faster system bus.

Besides, the South Bridge VT82C693A also has the support of UltraATA/66 interface, which can theoretically increase the data transfer rate from IDE-devices, from hard disk drives, for instance. However, this is not the end yet. VIA Apollo Pro133 also boasts a few nice trifles, such as 4 USB-ports support (against only 2 ones by i440BX) and integrated hardware monitoring, which doesn't require any external LM78-compatible controllers.

As its counterpart i810, VIA Apollo Pro133 also has an AC97-codec, which allows connecting a low-cost soft-modem and sound card by means of either AMR-slot, or their direct integration onto the mainboard. So, you shouldn't disregard AMR: as we see, it is supported not only by Intel, but is recognized everywhere and by all chipset manufacturers, and it is very likely to become one of the most promising technologies soon.

As for some purely technical peculiarities, we would like to say that in contrast with i440BX VIA Apollo Pro133 also supports AGP fast write, and hence it is able to transfer the data from the chipset to AGP-device totally avoiding system memory.

And now let's pass over to the benchmarks. For our tests we chose Chaintech 6ATA2 mainboard based on VIA Apollo Pro133. It is an ATX-format Slot-1 board with 3 DIMM-, 4 PCI- and 2 ISA-slots, and of course, 1 AGP- and 1 AMR-slot as well. The supported memory is restricted to 768MB. The mainboard is also equipped with the integrated sound based on AC-97 codec. Besides all the ordinary characteristics listed above, we simply can't disregard the fact that Chaitech started working out products intended for overclockers in the first place. 6ATA2 possesses a special feature, which allows increasing the processor voltage by 0.1 and 0.2V, and supports a large range of external frequencies: 66, 75, 83, 100MHz and up to 152MHz with 1MHz step, which is really very pleasant. Such great variety - 56 frequencies! - became possible due to a not so well-known clock synthesizer RTM520-39C used in 6ATA2. So, if Chaintech keeps on like that, in a while we can expect a number of supercool novelties.

We compared the performance of Chaintech 6ATA2 with a mainboard based on i440BX - ABIT BX6 Rev.2.0. Besides that, the testing system had the following configuration:

• Intel Pentium III 450 CPU;
• ASUS AGP-V3800 graphics card;
• IBM Titan DJNA 372200 HDD;
• 128MB SEC PC-100 SDRAM;
• Windows98.

The first thing we would like to check is the allegedly slower work of VIA Apollo Pro133 with the system memory. Here are the data transfer rates taken:

Well, it look as if there were no changes compared to the previous version VIA Apollo Pro Plus. The new VIA chipset works with memory considerably slower. It's hard to say why so: VIA either cannot or simply doesn't want to improve its chipset time diagram. Even ALI has done it! And VIA still persists. Let's hope for the better, anyway. Probably the next chipset version - VIA Apollo Pro133A - will change the situation.

The second cause for feeling upset is the gaming applications benchmarks. As is known, VIA is not very good at AGP-port realization. Just take a look at Quake2:

Everything follows our predictions: Apollo Pro133 is pretty slow here. And since the main contribution to the gaming results is made by the graphics subsystem, namely by AGP-port, VIA Apollo Pro133 failed here as well. All this makes us believe that despite our great hopes VIA has hardly improved anything with its AGP-port, compared to its predecessors. Again all we can do is to wait for VIA Apollo Pro133A, which is expected to have the saving AGP 4x support. Well, we'll see…

Now let's take a closer look at the average performance in office applications:

Well, lower memory operating speed tells here as well. The influence is not that significant, but still. Unfortunately, VIA Apollo Pro133, which boasts a lot of cool features, performs not that cool, in fact.

The only thing, which still gives us some hope, is that VIA Apollo Pro133 should theoretically be the leader in HDD data transfer operations according to its technical characteristics, namely to UltraATA/66 support. In order to offer you a better comparative analysis, we also tested another mainboard based on i810 with UltraATA/66 support. It appeared to be a new Intel CA810 with a Celeron 300A processor overclocked to 450MHz due to the system bus frequency increase up to 100MHz.

And again - no cause for delight. First, as you can clearly see, the enabled UltraATA/66 provided absolutely no real effect. It is not surprising at all, actually. UltraATA/33 provides a 33MB/sec bandwidth and it is more than enough for the need of today's HDDs, which can achieve maximum data read rate from the disk surface equal to about 20MB/sec. And the second thing: even with the enabled UltraATA/66 VIA Apollo Pro133 fails to surpass i440BX, though the latter doesn't support UltraATA/66 at all.

So, VIA Apollo Pro133, which looked so promising according to the specs, performed very poorly in reality.

Conclusions

So, as we have expected, the main trump of VIA Apollo Pro133 is still its price, which is twice as low as that of i440BX. And from all other points, except a great number of various integrated features, of view VIA Apollo Pro is none other than a huge disappointment. VIA Apollo Pro133 will have absolutely no practical value until the appearance of some new processors working with 133MHz FSB. And as soon as they appear, Intel will introduce its own core logic right away. VIA's chances to dominate in the chipset market are equal to naught, because their chipsets are considerably worse in terms of memory and AGP than those by Intel. As far as VIA's main argument is concerned, namely the support of a PC133 memory, which is cheaper than RDRAM, it is no longer that powerful, because Intel has also introduced PC133 SDRAM support in its i820 chipset last MCH. So, VIA Apollo Pro will most likely compete with i810e - a cheaper Pentium III chipset with the integrated graphics. Besides, we also look forward to VIA Apollo Pro133A with AGP 4x support, which is expected in VIA chipsets earlier than in its Intel competitors.

And one more "but". Intel has deprived VIA of its P6 bus license. It is very likely that we will soon witness another long-lasting trial, which will probably lead either to complete disappearance of VIA chipsets from the market or even to the total change of its products range: it will start producing K7 chipsets.