No matter, if we want it or not, but the next industrial standard in the memory market is most likely to be DDR333. And in the first place it is true for Socket A platforms.

Pentium 4 platforms will not start using DDR333 memory in mass, because Intel, the chipset leader for this platform is not going to support DDR333 this year. However, the chipset makers, such as SiS, VIA and ALi, have already announced their Pentium 4 chipsets with DDR333 support.

As for Socket A, the powerful ideological stimulus here is the upcoming launching of the Hammer processor family with the DDR333 memory controller embedded into the CPU core. Although Hammer platforms do not have anything in common with Athlon XP platforms, AMD and the developers of core logic solutions for AMD CPUs started pushing forward DDR333, so that by the time ClawHammer is due in the end of this year, DDR333 could be quite widely spread already. As a result, there are two DDR333 chipsets in the market already: VIA KT333 and SiS745.

At first sight it seems that the use of DDR333 memory in Socket A platforms doesn't make much sense. The bandwidth of the processor bus in Athlon XP systems working at 266MHz is equal to 2.1GB/sec. AMD has no plans to increase either the frequency or the bandwidth of the Athlon XP bus in the near future. For example, all AMD processors based on the new 0.13micron Thoroughbred core will keep using 266MHz bus with 2.1GB/sec bandwidth. PC2100 DDR SDRAM, which has been used in all Socket A systems until nowadays, features the same bandwidth. The bandwidth of the memory subsystem built on DDR333 makes 2.7GB/sec. That is why the use of faster memory than PC2100 DDR SDRAM may seem absolutely inefficient with Athlon XP processors, because they will be unable to make use of the entire memory bus bandwidth.

However, the tests we carried out for VIA KT333 showed that in reality things are completely different. The use of DDR333 memory instead of DDR266 increases the performance of Athlon based systems by 3%-5%. It happens because the protocols transferring the data along the processor bus and the memory bus are different and the memory subsystem features much higher latency than the CPU.

Besides, the cost of DDR333 and DDR266 memory modules differs just a very little bit, so we can expect DDR333 to get really popular for Socket A systems. This is one of the reasons why we decided to test a couple of new Socket A mainboards supporting this memory type. These mainboards are based on chipsets from VIA and SiS. Since we haven't yet had the chance to get into details with the SiS' DDR333 solution for Socket A, we will also try to make up for it here and to compare the performance of VIA DDR333 chipset and SiS' one.

Closer Look: Chipsets

At present there are two Socket A chipsets supporting DDR333 memory that are shipping in mass quantities to the mainboard manufacturers. They are VIA KT333 and SiS745.

VIA KT333 is none other but a very familiar VIA KT266A with an enhanced memory controller supporting not only PC2100/PC1600 DDR SDRAM, but also DDR333 memory.

As for SiS745, this chipset is a successor to SiS735. However, it is not only the memory controller that has been changed in the new solution. Besides the DDR333 support, SiS integrated into its chipset the IEEE 1394 controller supporting up to 3 ports. Unfortunately, SiS' engineers didn't manage to implement an integrated IDE controller as well that is why SiS745 doesn't have any ATA/133 protocol support. Moreover, SiS745 was deprived of network support.

The table below sums up the features of both DDR333 chipsets:

When we received the press presentation of VIA KT333 chipset, we noticed that there was an interesting message under one of the slides with a comparative table of the Socket A chipset parameters. This message seemed to be intended only for the presentation maker, however, for some reason they forgot to remove it. Here it is: "You might want to remove this slide Richard… on paper, the KT266A doesn't have much of an advantage over the SiS735". In other words, VIA guys automatically admitted that the features of SiS735 looked more attractive than those of their VIA KT266A. Besides, from the performance point of view SiS735 looked not bad at all. We still remember very well that the launching of SiS735 pushed VIA to quickly redesign its KT266 chipset, because it appeared hopelessly slow against the background of SiS' competitor.

And what do we see now? The features of VIA KT333 are better than those of SiS745 in most cases. While VIA enhanced its South Bridge with ATA/133 support, SiS decided to leave its South Bridge as it is and simply replaced the network controller with IEEE 1394 controller.

As a result, it is evident that VIA and SiS have shared the DDR333 Socket A chipset market. VIA KT333 is a more expensive chipset with more functions, while SiS745 is a low-cost DDR333 solution with cut down features. We will see later in this review how the things with the chipsets performance stand. And before we start, let's have a closer look at the mainboards, which took part in our tests.

Closer Look: Mainboards

ASUS A7V333

Usually ASUS mainboards cost more than the competing products from other manufacturers. ASUS A7V333 is no exception here. Though its high cost can be partially explained by too much extras added by ASUS engineers. The mainboard features an integrated 6-channel C-media CMI8738 PCI sound controller, VIA USB 2.0 controller providing 2 USB 2.0 ports, IEEE 1394 controller from Texas Instruments providing 1 IEEE 1394 port, Promise ATA/133 IDE controller supporting RAID 0 and 1. By the way, some A7V333 mainboard revisions may go without the integrated IDE RAID controller and IEEE 1394 controller.

No wonder that ASUS A7V333 is one of the largest mainboards of all our testing participants. However, even despite large size, the developers couldn't fit all the integrated chips conveniently, so the PCB design appeared not that impeccable in places. For instance. To our great disappointment, the FDD connector on A7V333 is located in front of the fourth PCI slot. But this is not the worst thing. The most significant drawback is the too close location of Socket A to the DDR DIMM slots. This may cause some problems not only during cooler mounting. Besides that, the DIMM modules will be heated up by the warm airflow going from the processor cooler, which is certainly no good.

Since the mainboard is equipped with a lot of integrated controllers, it should also have many output connectors. To locate all the connectors ASUS had to change the rear panel design completely and to provide the board with a bracket of their own.

Also ASUS A7V333 package contains some additional brackets fitting into the case rear panel instead of the expansion cards. These brackets feature 2 USB ports and 1 IEEE 1394 port.

ASUS A7V333 is the only mainboard of all our testing participants equipped with an AGP Pro50 slot. It is also equipped with voiced diagnostics system based on Winbond W83791SD chip, which we have already met by other ASUS products. We would also like to mention that the mainboard supports Smart Card Reader and the devices for reading Secure Digital and Memory Stick cards.

In its VIA KT333 based mainboard ASUS implemented some very interesting and unique technologies dealing with hardware monitoring. First of all I would like to mention that A7V333 appeared the first Socket A mainboard on VIA KT333, which doesn't use the thermal sensor integrated into the processor Socket A to measure the CPU temperature. It takes the temperature from the thermal diodes built into Athlon XP and Duron (Morgan) processors. This makes the temperature monitoring on ASUS A7V333 much more exact than on other mainboards. By the way, it is not for the first time that ASUS uses the mentioned monitoring scheme with the built-in processor thermal diode. The same feature has already been implemented in ASUS mainboard on NVIDIA nForce chipset: A7N266-E and A7N266-C.

The second interesting feature is Q-Fan technology. It implies that the rotation speed of the processor cooler fan varies depending on the CPU actual temperature. In reality it means that once the CPU temperature goes below 50 degrees Centigrade, the fan connector receives lower voltage, which in its turn reduces the noise level noticeably (mostly in case of slower Athlon, Athlon XP and Duron).

Overclockers should be pretty happy with ASUS A7V333 mainboard, as they will get some really popular overclocking friendly functions at their disposal in the mainboard BIOS Setup. The mainboard allows setting the FSB frequency within the interval from 100MHz to 227MHz with 1MHz increment. Note that when the FSB frequency lies between 100-120MHz, the PCI divider is equal to 1:3, in case of the 120-160MHz interval - to 1:4, and in case of frequencies exceeding 160MHz - to 1:5. Also the mainboard allows adjusting the CPU clock frequency multiplier (for the CPUs with the unlocked multiplier). Note that the mainboard supports the maximum multiplier of 14x. To improve the mainboard's stability during overclocking, you can increase the processor Vcore by up to 0.1V with 0.025V increment.

ASUS A7V333 uses Award medallion BIOS featuring a lot of settings for memory timings configuring.

In conclusion I would like to mention that the chipset North bridge is equipped with a massive aluminum heatsink. Besides, you can see plastic film stuck to the PCB under the fastening clips of the processor Socket A, which prevents the PCB from damage during cooler installation.

The stability and performance of ASUS A7V333 mainboard didn't arouse any complaints during our tests.

EPoX 8K3A+

EPoX's mainboard differs from other VIA KT333 based solutions participating in our test session by 6 PCI slots in the first place. Though on the other hand, this mainboard doesn't have an onboard USB 2.0 controller, so EPoX 8K3A+ doesn't support USB 2.0. As for other integrated controllers, the mainboard features an integrated ATA/133 IDE RAID controller from HighPoint (HPT372), which supports not only RAID 0 and 1 like Promise PDC20276 used on most other mainboards, but also supporting RAID 0+1. by the way, another version of this mainboard called 8K3A doesn't have this onboard ATA/133 RAID controller.

Although EPoX 8K3A+ seems to be not so packed with all sorts of chips as ASUS A7V333, its design can hardly be called very convenient. The graphics card installed into the AGP slot, blocks the clips of the DIMM slots, all IDE connectors are placed in front of the PCI slots and ATX power supply connector is located in such a place that the power supply cable will anyway go above the CPU cooler, no matter where the PSU is actually situated. Of course, this cable will hinder proper processor cooling. Moreover, the PCB size of EPoX 8K3A+ is a bit larger than usual, so it may not fit easily into any case.

To control the processor temperature the mainboard uses a special diode in the middle of Socket A. And under the processor socket clips there is a plastic film preventing the PCB from damage during cooler installation.

Although we've been criticizing EPoX 8K3A+ from the very beginning, we should definitely point out that this solution is very clearly positioned as an overclocking friendly product and it is not for nothing. Firstly, the CPU voltage regular on this mainboard is implemented very thoroughly of 6 transistors. Although 3 transistors of the six available are located on the reverse side of the PCB, which may be not very convenient during the mainboard installation into the PC case.

Secondly, there is a lot of room around Socket A, which makes this mainboard fit for use with processor coolers of any size.

Thirdly, the mainboard BIOS Setup based on Award Medallion BIOS V6.00PG features a lot of settings and allows changing any timings manually.

As for some particular overclocking options, EPoX 8K3A+ is an indisputable leader here. The mainboard allows changing the FSB frequency between 100MHz and 200MHz with 1MHz increment and adjusting the CPU clock frequency multiplier (the maximum is 15x). However, the most interesting thing is the incredibly wide ranges for different voltages adjustment: processor Vcore can be changed from 1.4V to 2.2V with 0.025V increment, and Vmem can be increased by up to 0.7V over the nominal value with 0.1V increment.

I believe it is not surprising that when we tested the overclocking potential of our mainboards, EPoX 8K3A+ appeared the best of all. Athlon XP 2000+ CPU (working at 1666MHz) overclocked up to 1825MHz, which is a very good results for CPUs of the kind. The FSB frequency in this case was set to 146MHz, while the other mainboards allowed the maximum of 143MHz only. Of course, in our case the overclocking was also limited by the potential of the processor, as its clock multiplier was locked, but the victory of EPoX 8K3A+ is undeniable.

Also 8K3A+ is equipped with so beloved by many overclocking fans POST-controller. The two-digit indicator located in the lower left corner of the PCB you can see POST codes displayed during system boot-up. These codes help to identify the nature of hardware problems, which may arise on boot-up.

However, when it came to performance tests of EPoX 8K3A+ some blunders still occurred. It turned out that the mainboard BIOS up to version 2304 inclusively doesn't allow setting CAS Latency = 2. in other words, even though you set it equal to 2 in the BIOS Setup, the real state of things remains unchanged and the board keeps working with CAS Latency 2.5. Starting with BIOS version 2312 the problem has been eliminated, however, in this case the board runs very unstable with CAS Latency = 2. Hopefully, the new BIOS releases (the last one we tested was 2326) will be free from problems like that.

Gigabyte GA-7VRXP

According to the list of supported features, VIA KT333 based mainboard from Gigabyte has every chance to compete even with ASUS' solution. Gigabyte GA-7VRXP is equipped with a four-channel Creative CT5880 PCI sound controller, VIA VT6202 USB 2.0 controller, which ensures the support of 4 USB 2.0 ports and Promise PDC20276 ATA/133 IDE RAID controller supporting RAID 0 and 1. Just like ASUS A7V333, it supports Smart card Reader, Secure Digital and Memory Stick Card reading devices. If we go on comparing the features of KT333 based mainboards from ASUS and Gigabyte, we will have to point out that GA-7VRXP doesn't support IEEE 1394 ports, but on the other hand boasts an onboard Realtek RTL8100BL network controller.

Despite a great lot of integrated chips and onboard connectors, which Gigabyte engineers had to locate somehow on the PCB, the design deserves our praise. We would even call it a reference design within our roundup. ALL IDE connectors, alongside with FDD and ATX power supply connector are located in front of the DDR DIMM slots. There is a lot of free room around processor Socket for massive cooling solutions and the DIMM slots clips never get locked by the graphics card installed into the AGP slot.

Note that Gigabyte provided us with two versions of its GA-7VRXP of different revisions: 1.0 and 1.1. It turned out that the new mainboard revision differs a significant lot from the older one. For the better, of course :) Although both mainboard revisions look similar GA-7VRXP rev. 1.1 boasts much better performance, is more stable and features richer overclocking options. In our tests we used the new mainboard revision 1.1.

As for the BIOS, we were a little bit disappointed. It is based on AMI BIOS v2.0 code. The BIOS Setup setting s are very poor and doesn't allow even changing some memory timings. The Setup offers some setting sonly for CAS Latency and SDRAM Commend Rate, as well as some mysterious Fast Command with the Ultra/Fast/Normal options. Besides, you will not be able to change the Memory Bank Interleaving scheme and CAS to RAS Delay and RAS Precharge Time. Memory Bank Interleaving is always set to 4-way, and CAS to RAS Delay and RAS Precharge Time - to 2, which is the most optimal setting, I should admit.

As for the CPU overclocking potential of GA-7VRXP, we can make a few comments here, too. Far not all the functions are implemented via BIOS Setup. For instance, you can change the CPU clock frequency multiplier, but only with the help of dip-switches and only up to 12.5x. The FSB frequency can be adjusted between 100MHz and 200MHz with 1MHz increment, but the shift between 100MHz-133MHz and 133MHz-200MHz intervals should also be performed by dip-switches first. The processor Vcore can be increased by 5%, 7.5% and 10%; Vdimm and Vagp can be increased by 0.13V with 0.1V step. So, to tell the truth, if it were not for the dip-switches, we could call Gigabyte GA-7VRXP a great overclocker's choice.

Like all other Gigabyte's products targeted for the high-performance sector, GA-7VRXP supports DualBIOS technology and features two Flash-memory chips onboard storing two copies of the BIOS (primary and reserve).

To control the processor temperature the board uses a special thermal diode located in the middle of Socket A. And under the processor socket clips there is a plastic film preventing the PCB from damage during cooler installation. The chipset North Bridge of Gigabyte mainboard is equipped with some weird-looking gold heatsink fastened onto a sticky pad with pretty doubtful thermal conductivity.

During the tests we discovered one more strange thing about Gigabyte GA-7VRXP. The last BIOS version F6a was slowing down the mainboard dramatically. The memory performance dropped down nearly twice. That is why we had to use not the latest but the last but one BIOS version M5, which is supplied together with the mainboard.

MSI KT3 Ultra-ARU

Like other similar products, the mainboard based on VIA KT333 from MSI is equipped with a number of additional controllers. MSI engineers provided their mainboard with an additional USB 2.0 controller from NEC, which implements 4 USB 2,0 ports, and ATA/133 IDE RAID controller from promise supporting RAID 0 and 1.

Also KT3 Ultra-ARU boasts an outstanding sound solution. The board is equipped with a 6-channel Realtek ALC650 codec, which also supports digital SP/DIF interface. That is why the package includes a special bracket installed into the case rear panel instead of one of the expansion cards, which is equipped with an optical and digital SP/DIF interfaces.

I would like to point out that MSI also ships a mainboard revision without the additional onboard controllers. It is called KT3 Ultra.

As for the mainboard PCB design. there are a few comments to make. Although two IDE connectors and the FDD connector are located in front of the DIMM slots, the IDE connectors of the RAID controller are just in front of the last PCI slots. Also the ATX power supply connector is placed in far not the best position: just behind the Socket A, that is why the power supply cable will be hanging above the processor cooler inside the PC case thus hindering proper CPU cooling. although there is a lot of free room around Socket A, it is moved so close to the mainboard edge that once the mainboard is installed into the case you will be unable to reach the cooler retention mechanism clips. Moreover, the AGP graphics card will block the DIMM slots clips as well.

Although VIA KT333 North Bridge doesn't heat too much. MSI equipped it with a silver colored active cooler. This is the only mainboard of all our testing participants that is provided with an active cooling solution for the chipset North Bridge. To control the processor temperature the mainboard uses a special diode in the middle of Socket A. And under the processor socket clips there is a plastic film preventing the PCB from damage during cooler installation.

Also MSI KT3 Ultra-ARU supports MSI's brand name D-Bracket diagnostic system. It allows detecting errors during system boot-up with the help of a special bracket installed instead of one of the expansion cards. It has 4 dual-color LEDs showing the POST status. The device works just the same way as D-LED, which MSI used to implement in its older mainboards, however, in this case the leads are located on a separate bracket also equipped with a pair of USB ports.

The BIOS of KT3 Ultra-ARU mainboard is based on AMI BIOS v2.0 however, unlike the BIOS of Gigabyte GA-7VRXP, it Setup is fully packed with all the necessary functions. The only thin we can complain about is the inability of disabling the USB 2.0 controller from the BIOS Setup.

All overclocking friendly options of MSI KT3 Ultra-ARU are implemented via BIOS Setup. For those processors, which boast unlocked clock frequency multiplier, the mainboard allows changing it up to 15x at the maximum and changing the FSB frequency from 100MHz to 220MHz with 1MHz increment. Also there is the whole bunch of means for voltage adjustment: Vcore may be increased by 0.1V with 0.25V increment, Vagp and Vmem - by 0.3V with 0.1V increment. BIOS Setup also allows changing any memory timings.

The test showed that when we use the memory with CAS Latency = 2, Vdimm should be better increased by 0.1V. Only in this case MSI KT3 Ultra-ARU is stable enough.

ECS K7S6A

ECS K7S6A is the only mainboard of all out today's testing participants based on SiS745 chipset. As we have already stressed above, SiS745 chipset is positioned as a low-cost DDR333 Socket A solution. This idea was implemented brilliantly by ECS in their K7S6A mainboard equipped with no additional controllers. Even the IEEE 1394 controller integrated into the chipset wasn't laid out on the mainboard.

However, this ascetic design of K7S6A had its indisputable advantages. The average cost of VIA KT333 based mainboards we tested makes around $140, however, you will be able to buy ECS K7S6A for half the price: $70. So, the major advantage of ECS SiS745 based solution over the competitors is, of course, it extremely low cost.

Since ECS K7S6A costs a really miserable sum of money, we don't feel like being too critical to the mainboard PCB design. Nevertheless, we can't help mentioning that all the connectors on the board, such as IDE, FDD and ATX, are located very inconveniently. On the other hand, this way they managed to free quite a lot of space around Socket A, which will allow the user to install a cooler of any size. Besides, there is a special plastic film under the Socket clips to protect the PCB surface against damage. Since SiS745 chipset is made of only one chip and there are no additional controllers onboard, ECS K7S6A is of relatively small size: only 220mm wide.

The BIOS of ECS K7S6A is based on Award Modular BIOS v6.00PG and boasts a wide range of setting for key SDRAM timings.

As for CPU overclocking, ECS mainboards are not that famous for this kind of things. K7S6A is also not an exception. The mainboard has a few jumpers, which allow changing the CPU clock frequency multiplier (12x at the maximum), and via BIOS Setup you can change the FSB frequency and set it to one of the following rates: 100, 103, 107, 110, 112, 124, 133, 137, 143, 147, 150 and 166MHz. Also there is a special option for processor Vcore adjustment between 1.1V and 1.85V with 0.025V increment. To control the CPU temperature the mainboard uses an external thermal diode built into the middle of Socket A.

The mainboard proved very stable throughout the whole testing session.

Testbed and Methods

One of our goals during these tests was to find out which Socket A mainboard supporting new DDR333 memory is the fastest. At the same time we compared the performances of VIA KT333 and SiS745 chipsets. In order to see if it makes real sense to use DDR333 memory we tested all the mainboards with both: PC2700 DDR SDRAM (DDR333) and PC2100 DDR SDRAM.

To assemble our testbeds we used the following hardware:

• AMD Athlon XP 2000+;
• Mainboards: ASUS A7V333, ECS K7S6A, EPoX 8K3A+, Gigabyte GA-7VRXP, MSI KT3 Ultra-ARU;
• 256MB PC2100 CL2 DDR SDRAM and 256MB PC2700 CL2 DDR SDRAM;
• Visiontek Xtasy 6964 (NVIDIA GeForce3 Ti500) graphics card;
• IBM DTLA 307015 HDD.

All the benchmarks were run in MS Windows XP.

For a better comparison we also added to the diagrams the results of the recently announced Pentium 4 2.4GHz working in EPoX 4BDA+ mainboard based on i845D chipset with DDR support. It will allow us to see whether the fastest Athlon XP processors (unfortunately, we failed to get the today's latest Athlon XP 2100+ from AMD for some reason) working in the best Socket A mainboard will be able to compete with the today's fastest Pentium 4.

Performance

First of all we took wcpuid utility and measured the frequency of the CPU working in each of the tested mainboards. Of course, the performance of this or that product depends a lot on the actual frequency of the CPU. Some manufacturers take advantage of it and raise the FSB and consequently the CPU frequencies artificially. That is why the mentioned above test will let us unmask some cunning fellows :)

It is surprising but the CPU frequency for all the tested mainboards appeared close to the nominal. That is why when we will speak about mainboards performance we will have the right to neglect the frequency measuring error.

Let's pass over to the tests. As usual, we begin with SiSoft Sandra 2002, which measures the practical bandwidth of the memory subsystems of our testing participants.

Even though the first practical results were obtained in a synthetic benchmark, they prove our suppositions mentioned above. The use of DDR333 memory doesn't have any noticeable effect on the performance of Socket A systems. The mainboards based on VIA KT333 show similar results working with DDR333 and DDR266 memory. As for SiS745, its memory subsystem performance appears absolutely the same in both cases.

As far as the relation between the memory bandwidths of SiS745 and VIA KT333 is concerned, the VIA solutions looks much more favourable. While VIA keeps improving its memory controller in every new chipset, SiS decided to stay with what they achieved in SiS735 and focused mostly on the Pentium 4 chipsets now. Therefore, the new chipset can boast only some formal DDR333 support. So, no wonder that SiS745 is positioned as a low-cost solution for Socket A platforms, as the performance is not of primary importance in this market sector.

And now a few words about the mainboards. The fastest memory subsystem appeared by MSI and Gigabyte mainboards. Moreover, in some cases MSI KT3 Ultra-ARU and Gigabyte GA-7VRXP working with DDR266 memory could outpace some solutions with DDR333 even.

Now let's have a look at the situation in real applications:

In office applications we do not see much difference between the testing participants. However, ECS K7S6A based on SiS745 falls behind the rivals, and the leadership goes to VIA KT333 based solutions working with DDR333.

The same results are obtained in Internet content creation applications. Here we would like to mention one very interesting thing: if we take a look at the performance of SiS745 based ECS K7S6A mainboard, we will see that its performance with DDR266 memory is higher than with DDR333 memory. Isn't it strange? Of course, it's not! The matter is that SiS decided to save time and trouble redesigning the memory controller and added DDR333 support to the features just formally. As a result, the DDR266 memory working synchronously with the processor bus defeats the asynchronous combination of 333MHz memory and 266MHz processor bus.

This is a pretty predictable outcome, hardly differing from what we have already seen.

The processing of streaming video depends a lot on the memory subsystem performance. That is why the results differ quire a lot here. The leadership belongs to three mainboards with DDR333 memory: MSI KT3 Ultra-ARU, Gigabyte GA-7VRXP and ASUS A7V333.

The gaming benchmarks show almost similar outcome that is why it doesn't make much sense to comment on each diagram separately. As we can see, the leaders in all the three benchmarks are again MSI KT3 Ultra-ARU and Gigabyte GA-7VRXP, which have dashed quite far ahead of all the others. We can even notice that combined with DDR266 memory these solutions beat their rivals working with DDR333. Note that these are two boards of all the testing participant that have an AMI BIOS. This observation gives us every reason to suppose that this BIOS suits much better for mainboards based on VIA KT333 than the one from Award.

As for SiS745 chipset and Elitegroup mainboard based on it, we wouldn't call it a high-performance solution. Cheap - yes, fast - no.

In order to find out how the boards will behave in professional OpenGL applications we ran a couple of tests in a popular 3ds max 4.26 application. We measured the performance in ViewPorts, as the final rendering speed depends only on the processor in your system.

The results obtained show that the situation in 3ds max is close to what we have just seen in 3D games. Again the solutions from MSI and Gigabyte are ahead of all. Then a while behind them comes ASUS, the fourth position is occupied by EPoX 8K3A+, which appears the slowest of all the KT333 products tested.

In conclusion, we ran a couple of tests with a new PCMark2002 benchmark from Madonion. On the diagrams you can see two results: CPU Score is obtained from the JPEG decompression speed, LZ77 decompression speed, text string search time, mp3 decompression and 3D calculations processing speed. Memory Score is obtained from the reads and writes speed of 6KB-3MB data packs.

We will not comment on the results, as they repeat the previous picture.

Conclusion

First of all, I would like to say a few words about the efficiency of using DDR333 memory in Athlon based systems. As we see, the use of DDR333 instead of DDR266 can provide a certain performance increase (up to 5%) in some cases. However, the fastest DDR266 mainboards can easily outperform DDR333 products. Therefore, I wouldn't consider the use of DDR333 in Socket A systems a necessary thing. You should also bear in mind the fact that everything mentioned above makes sense only for VIA KT333 based products.

As for SiS745 the situation is a bit worse here. DDR333 memory in systems built on this chipset is just a waste of money. In some tasks this combination turns out even slower than SiS745 + DDR266. All in all, I wouldn't even call SiS745 a fully-fledged DDR333 chipset. The proper description to this product will be "low-cost Socket A solution". Moreover, SiS745 falls behind VIA KT333 with DDR333 memory by 5%-15%, and behind VIA KT333 with DDR266 memory - by 5%-10%.

And now a few words about the tested mainboards. According to the above discussed results, there are two leaders in our test session: MSI KT3 Ultra-ARU and Gigabyte GA-7VRXP on VIA KT333 chipset. It was not only remarkable performance that pleased us a real lot in these products, but also good overclocking opportunities and rich integrated features.

At the same time we would like to say a few more good words about the PCB design of Gigabyte GA-7VRXP mainboard. It would be really cool if other mainboard manufacturers could follow Gigabyte's example and pay as much attention to the convenient components mounting.

ASUS A7V333 mainboard deserves the title of a "bonus winner". Its specs list is really outstanding. Unfortunately, ASUS A7V333 failed to impress us with performance against the background of solutions from MSI and Gigabyte, but still.

EPoX 8K3A+ undoubtedly deserves to be called "an overclocker's dream". We have never seen such a rich set of overclocking friendly options. Besides, EPoX paid special attention to the stability of its mainboard when working at higher FSB frequencies, which will definitely have a positive effect during extreme CPU overclocking. However, our tests showed that it is the slowest solution of the 4 boards tested.

And in conclusion a few words about the only mainboard on SiS745 chipset from Elitegroup, which took part in our tests. Of course, its performance is not so fast as that of VIA KT333 based boards. This solution can't boast any rich feature of other considered products. However, it's $70-price and excellent stability deserve our positive feedback. This way, we would recommend ECS K7S6A as a great DDR solution for Value PCs.