The situation with the DDR chipsets for Socket A platforms looks funnier and funnier with the time. As it appeared later on, neither of the chipset manufacturers managed to develop the memory controller, which could make use of the entire DDR SDRAM potential. As a result, some manufacturers had to undertake the second try before they could start mass production of DDR Socket A chipsets showing acceptable performance. The first chipset like that was AMD-760 launched by AMD in order to demonstrate the advantages of the new DDR technology. However, instead of leaving the stage and letting other products with higher performance and richer features replace it, AMD-760 remained an unattainable example of fast DDR solutions for a considerable while. Neither ALi with their ALi MAGiK 1 chipset, nor VIA with their KT266 could beat AMD-760 for a long time. Only SiS managed to offer a worthy competitor to AMD-760 about half a year after its official launch. The newcomer was called SiS735 and proved quite successful. It was exactly due to SiS that we finally found out how fast Athlon based systems with PC2100 DDR SDRAM should actually be.

It looks as if not only we should thank SiS for opening our eyes. Other chipset makers also had to draw not very positive conclusions about their products and as a result, new versions of older chipsets were announced. The newcomers featured enhanced memory controllers and hence boasted higher performance. The first to wake up and to introduce a modified Socket A DDR chipset appeared VIA with its KT266A (see our VIA KT266A Chipset Review). This core logic managed to outperform SiS735. However, this chipset revision has never been made in mass quantities because of numerous stability issues. But, the enhanced solution from ALi, ALi MAGiK 1 Rev.C, which started shipping in mass to the mainboard manufacturers about a month ago already, turned out more successful. Those mainboard manufacturers, which had already dealt with ALi MAGiK 1 chipset before, cheered up and sat down to the production of new versions of their nearly forgotten products on ALi MAGiK 1 chipset. ASUS, Iwill, SOYO and Transcend resumed the shipments of their ALi based solutions (which used to be the slowest among all the existing DDR products for this platform), but with the new Rev.C chipset this time.

Of course, we couldn't disregard the launching of the new ALi MAGiK 1 chipset revision. We managed to get hold of a new Socket A mainboard based on the new ALi MAGiK 1 Rev.C chipset and supporting DDR SDRAM: Iwill XP333-R. Just like a year ago, when Iwill was the first in the world to introduce a mainboard supporting DDR SDRAM, this company managed to leave behind its competitors once again and turn out the first one to launch a mainboard based on the new revision of this chipset. Moreover, Iwill XP333-R is of interest to us not only due to the chipset. The manufacturer claims that the board is the first in the world to support the new DDR333 memory. So, it should be now more or less clear why we decided to review Iwill XP333-R mainboard. Let's pass over to it!

Closer Look

As we have already told you, our today's hero is called Iwill XP333-R. Its specs look as follows:

Iwill XP333-R
Supported CPUs	AMD Athlon XP/Athlon/Duron (200/266MHz FSB)
Chipset	ALi MAGiK 1 Rev.C
FSB Frequencies	100-233MHz
Overclocking Friendly Features	CPU clock multiplier adjustment Vcore and Vio changing
Memory	3 184-pin DIMM slots for PC1600/PC2100 DDR SDRAM
Expansion Slots (AGP/PCI/ISA/ACR)	1/5/0/0
USB Ports	4
Integrated Graphics	No
Integrated Sound	6-channel C-media CMI8738 sound controller
Additional Features	Integrated HPT372 ATA/133 IDE RAID controller
BIOS	Award BIOS v6.00PG
Form-Factor	ATX, 305mm x 244mm

You should have guessed from the mainboard name already that Iwill will be also shipping a cheaper modification of its XP333 solution, which will have no onboard IDE RAID controller.

We would like to start with the name of the mainboard. Having selected "XP333-R" name for its product, Iwill's marketing department tried to kill two birds with a stone. Firstly, it is evident that the manufacturers wanted to be somehow associated with the super popular Windows XP operation system, even though the abbreviation in the mainboard name stands for "eXtreme Performance". And secondly, by adding 333 to the name Iwill stresses that their solution supports the new DDR333 memory. Moreover, Iwill also wanted you to consider their solution an ideal choice for use with the new AMD Athlon XP processors. However, the most interesting thing is different. If you take a closer look at the mainboard, you will notice that the name "XP333-R" is printed on a small green paper stuck to the PCB. Following a pretty natural desire to remove the sticker and to see what they've got under it, we will find out a lot of interesting things:

  

Yeah, under the sticker we can see a totally different name of the same mainboard: KA266Plus-R. And it means that the XP333-R mainboard in our hands is none other but a new version of our good old buddy KA266-R, the world's first mainboard on ALi MAGiK 1 chipset (see our article called World's First DDR Platform: Iwill KA266-R Mainboard). However, you shouldn't be very much upset, as there is nothing bad about it. It simply means that Iwill decided to use the good design, which stood the test of time, for its new mainboard. It is not Iwill's fault that KA266-R couldn't boast very high performance. The older mainboard was built on the first ALi MAGiK 1 chipset revision. The new XP333-R is based on the third revision of this chipset, which can boast much better performance.

Besides, if we compare the design of the older KA266-R and the new XP333-R, we will see that the chipset revision is not the only difference between them. Let's briefly dwell on all the differences we discovered.

First of all, it caught our eye that Iwill turned the Socket A by 90 degrees. As a result, there are no more problems with cooler fastening and you are free to use any massive CPU cooling solution you like: there is enough free room on all sides of the processor socket. Here we would also like to point out that Iwill placed a special plastic band right under the hooks on one side of the Socket A, which should protect the PCB surface from damages and scratches when the cooling system is mounted.

The second important peculiarity is a different power supply circuit used. XP333-R features new voltage regulators for the processor core, DIMM slots and AGP slot. As for the processor voltage regulator, it has become three-phase and uses much more powerful components, which ensures stable power supply even with 46A. It means that the power supply regulation has become much better. Therefore, XP333-R will be able not only to support the fastest Athlon and Athlon XP processors, but also to grant higher stability for all other processors working in it. At the same time, Iwill paid special attention to "refining" the power going to DDR DIMM slots, so that to make sure that all DDR333 memory modules will work well with this board.

And the third peculiarity worth mentioning is the fact that Iwill gave up using an AMI IDE RAID controller on its mainboard. Now the company integrates controllers from HighPoint, which let them benefit significantly. HighPoint has already got a new HPT372 chip supporting new ATA/133 in their product range. Due to this controller, the mainboard supports not only ATA/133 HDDs (as the controller integrated into the chipset supports only ATA/100 interface), but also 0, 1 and 0+1 RAID arrays with "hot swap" in case the chassises are there. If you are wondering whether the bandwidth provided by the new ATA/133 interface can find its applications nowadays, please, see our article called Maxtor D740X-6L HDD Review: First Look at ATA/133. However, you should anyway bear in mind that this is a very good feature for a mainboard, which is intended to serve you a longer while.

These are all the differences between the mainboards. All the rest is the same. However, there remains one more thing we should point out, though it mostly has to do with the package rather than with the mainboard features. Anyway, as you know, almost all Iwill's mainboards, are equipped with an on-board C-media CMI8738 sound controller. XP333-R is also not an exception:

This is a pretty nice sound chip, which supports 3D sound and allows connecting 6-channel acoustic systems to the board. However, Iwill used to provide the layout only for the primary speakers set, which undoubtedly limited the potential of CMI8738 chip making it as good as an ordinary stereo system. Now XP333-R is shipped together with a special bracket, which fits into the case rear panel instead of one of the expansion cards. This bracket is provided with an extra pair of output ports for the 6-channel acoustic systems.

Moreover, Iwill also prepared a set for advanced users: an enhanced SuperAudio bracket with RCA support and an optical SPDIF In and Out besides the already described features. So, we have every right to state that Iwill XP333-R boasts very rich audio options.

Unfortunately, when putting into the package an extra bracket for 6-channel sound, Iwill totally forgot about one more very important bracket, which could allow connecting another pair of USB ports to their mainboard. Therefore, although Iwill XP333-R officially supports 4 USB ports, you will be able to use only two of them, if you don't have this bracket of your own.

Yeas, we nearly missed the most important thing: Iwill XP333-R is based on the new ALi MAGiK 1 chipset revision C:

The chip is provided with an active cooler of a pretty original shape, which should ensure stable operation of the system built on Iwill XP333-R.

Well, now that we have given credit to all the advantages of the new Iwill XP333-R mainboard, it's high time we added a fly to this ointment and said a few words about the bottlenecks of this solution. Especially, since we don't need to search for them too much… As you may have noticed already, we didn't include the new PC2700 SDRAM (DDR333) memory into the list of supported memory types in the beginning of the review. To tell the truth, we did it on purpose, despite the fact that the manufacturer claims the support of this memory has been implemented.

The matter is that DDR333 support on Iwill XP333-R is a purely marketing trick. In reality, there is simply nothing of the kind. When we booted the system built on XP333-R in our testlab, we first of all tried to check how well it worked with DDR333 memory. However, though we did our best we failed to find any jumpers of settings in the BIOS, which could allow us clocking the system bus and the memory bus asynchronously. So, the memory on Iwill XP333-R works only synchronously with FSB, just like on older AMD-760 based mainboards. As a result, if you use XP333-R with the Duron CPUs and the youngest Athlon processors, the memory will appear working at 100MHz. When the system is built with the latest Athlon or Athlon XP CPUs, the memory works at 133MHz (we are talking about the nominal mode here). And it means exactly the support of PC1600/PC2100 DDR SDRAM.

What then made Iwill believe that their new XP333-R mainboard supports DDR333, you would probably ask? We managed to find an answer in the official XP333/XP333-R FAQ, which you can download here. Take a look at the quote below:

Question-10: How can I get my DDR333 memories to work at 333MHz?
Answer: You need to set your FSB to 166MHz, and your DDR memories will work at 333MHz (166x2=333).

To tell the truth, we were impressed with this pretty cynical explanation. Of course, the memory on XP333-R will work at 333MHz if you overclock the FSB frequency up to 166MHz. But the same thing will be valid for any other mainboard supporting synchronous clocking of the memory and processor buses. So, Iwill's approach to the matter allows us to state that all the existing DDR mainboards for Athlon CPUs theoretically support DDR333. However, using the power of the faster DDR333 to the full extent is a totally different question. As you know, AMD doesn't produce any CPUs supporting FSB frequency over 133MHz. The 166MHz Iwill is talking about are none other but pure overclocking. We would even call it extreme overclocking, as it will undoubtedly require unlocking and reducing the CPU clock frequency multiplier, because far not all AMD processors will be able to work at the frequency 25% higher than the nominal value. Maybe some day AMD will launch processors supporting 166MHz FSB (333MHz DDR) on Thoroughbred core, for example, and Iwill simply wants us to know in advance that their product will support them. But, the next line of the official XP333-R specification saying "Supports 266/200MHz FSB" destroys this supposition completely...

We will discuss a little bit later how well Iwill XP333-R works when overclocked (for instance, up to the notorious 166MHz), and in the meanwhile we would like to dwell on a couple of other items.

Unfortunately, Iwill XP333-R inherited some design drawbacks from the predecessor, KA266-R. First of all, it is extremely inconvenient location of the power supply connector, which is situated near the AGP slot in such a way that the power supply cable appears hanging above the CPU hindering proper cooling. Moreover, it may also prevent you from installing full-size expansion cards in some cases.

However, BIOS of XP333-R based on Award program code is very rich in different settings, including a great number of those dealing with the memory controller. Nevertheless, Iwill didn't make its solution completely jumperless. There are sets of jumpers responsible for disabling the onboard sound controller or IDE RAID controller, for instance.

The mainboard supports hardware monitoring functions implemented via ALi M5879 controller. The mainboard is equipped with two thermal diodes. The CPU temperature is taken not from the diode integrated into Athlon XP processors but from the one located in the middle of Socket A, which doesn't ensure very accurate measurements. The second diode is placed in the upper left corner of the mainboard. Besides, the board can control 4 voltages and rotations of 2 fans (there are three fan connectors on the board altogether).

Overclocking

Checking the overclocking potential of Iwill XP333-R was of particular interest to us for the following reason. We believed that if the manufacturer claims that the board allegedly supports DDR333 and suggests setting the FSB frequency to 166MHz to get this support, then the mainboard should definitely run stably at least at 166MHz. And if it is true, then Iwill deserves all the laurels, as the mainboards we tested in our lab very rarely proved capable of running stably at the FSB frequency exceeding 160MHz. And as we have told you earlier, Iwill took good care of the stability issue having enhanced the power supply circuits, optimized the mainboard layout and provided the chipset with an active cooler.

However, this is not enough. Every extreme overclocker needs not only granted stability, but also a rich set of overclocking tools to put its "dirty deed" into life. Iwill engineers did their best to anticipate the overclockers' wishes and provided their XP333-R with a full set of opportunities an overclocker can ever dream of. Thus, XP333-R allows:

• Changing the FSB frequency in the interval from 100MHz to 233MHz with 1MHz increments (MicroStepping technology).
• Adjusting the CPU clock multiplier (for processors with the unlocked multiplier).
• Changing the processor Vcore from 1.125V to 1.85V with 0.025V increments. Increasing the Vcore above the nominal value improves the system stability when working at higher frequencies.
• Increasing chipset Vio and Vdimm from the standard 2.5V to 2.6V or 2.7V. It improves the stability of the overclocked system.
• Changing the PCI bus divider. The possible variants include: FSB/3, FSB/4, FSB/5 and FSB/6. This is very important as it allows to retain the working frequency of PCI and AGP devices close to the nominal.
• Setting the memory CAS Latency and controller registers configuration (FailSafe, Slow, Normal, Fast, Ultra è Ultra2) manually. These settings allow choosing the most optimal working mode for the memory from the performance and stability points of view.

Please, take note that all the mentioned settings except Vio are available in BIOS Setup. Vio can be set with the jumpers.

And now let's find out what we can achieve with all these numerous tools. The most important feature for an overclocker board is good stability at high and super-high FSB frequencies. That is why we carried out a number of experiments aimed at figuring the maximum clock frequency, at which XP333-R works stably and reliably. For these experiments we took AMD Athlon XP processor with the unlocked clock multiplier (with closed L1 bridges on the processor package) and the memory of best quality we had at our disposal - Kingmax Tiny BGA DDR333.

And here is the result: Iwill XP333-R is much more stable than the predecessors at higher FSB frequencies. We managed to increase the FSB frequency up to 189MHz (378MHz DDR), which is 42% higher than the nominal value:

This is an excellent result for a Socket A mainboard. We haven't ever seen a mainboard boasting a brilliant overclocking potential like that before. Moreover, we can't state that it is the maximum we managed to achieve. It may be the memory that failed to work at higher frequency in our case and not the mainboard, because the nominal frequency for DDR333 that is 166MHz has also been exceeded gravely.

This way, Iwill XP333-R deserves being called one of the best Socket A mainboards for processor overclocking. Besides, the possibility to set the PCI bus divider manually alongside with perfect stability at higher FSB frequencies allow us to suppose that XP333-R will be able to work well with AMD processors using 333MHz bus, if they ever appear. And in this case the name of XP333-R is 100% justified.

Testbed and Methods

There is only one more aspect left, which we need to consider before making the final verdict about the new Iwill XP333-R mainboard. It is the performance. Especially since we will be able to get an idea of the efficiency of the enhancements and optimizations made to the ALi MAGiK 1 chipset (the current Rev.C), as Iwill XP333-R mainboard is the world's first solution based on this chipset revision.

We decided to compare the performance of ALi MAGiK 1 Rev.C with that of SiS735 and VIA KT266A. These three chipsets are the only ones, which can be called "contemporary Socket A chipsets supporting DDR memory". As you see, we excluded AMD-760 and VIA KT266 chipsets from the "contemporary" ones, because the manufacturers have already stopped supplying them.

As for the mainboards, we used MSI K735 Pro (MS-6517) on SiS735 chipset, which is the today's best brand name product on SiS735, and MSI KT266 Pro2 on VIA KT266A, which turned out the fastest VIA KT266A based solution after we updated the BIOS.

As a result, we assembled the following tesbeds:

The test systems worked under Miscrosoft Windows XP.

Performance

Since we were interested in checking the performance of ALi MAGiK 1 Rev.C chipset, we resorted to synthetic benchmarks to test the memory subsystem performance:

The results shown by ALi MAGiK 1 Rev.C in this benchmark, especially the practical bandwidth of its memory controller, show very well how hard ALi's engineers worked on it. There is not a single trace of the former slowness left. In memory reading and writing this chipset almost catches up with the fastest DDR Socket A chipset - VIA KT266A. We should also point out that the results of our today's hero look much better compared with those shown by ALi MAGiK 1 Rev.B (see our Contemporary Socket A Chipsets Comparison). As for the latency, it is a very old problem of ALi's. So, ALi MAGiK 1 Rev.C is also not free from it, despite the fact that it managed to get very close to SiS735 here.

SiSoft Sandra is based on a more complex Stream algorithm (compared with the memory read/write). In this benchmark ALi MAGiK 1 appears almost as fast as SiS735 because of the relatively high latency. These both chipsets fall behind the leader, VIA KT266A, by about 10%.

And now let's see how the performance in synthetic benchmarks corresponds to that in real applications.

In the popular SYSmark2001, which measures the performance in business and content creation applications, ALi MAGiK 1 proves almost as fast as its competitors. The performance differences make only 2-3 points, so we can't call any of the testing participants a leader.

In content creation applications, for instance, ALi MAGiK 1 manages to outperform all the rivals.

And in the office part of this test, ALi MAGiK 1 Rev.C falls behind SiS735 as well as behind VIA KT266A.

Aiming to get an idea about the chipsets' performance in office applications, we measured the time they needed to compress a huge amount of data with the popular WinZIP 8.0. The idea of this test is that apart from constantly addressing the memory, the compression utility keeps on working with the disc subsystem. These test results will enable you to assess also the quality of the IDE Bus Master drivers. The diagram above shows how long it took the systems to compress a directory with Unreal Tournament. The shortest time stands for the best result.

As we see, ALi chipset is again the winner. Maybe it is due to ALi IDE Miniport boasting some special caching features like Intel Application Accelerator.

The tasks of video processing require fast operation with streaming data that is why the results of video encoding into MPEG-4 format shown by testing participants got positioned according to the practical bandwidth of their memory subsystems, which we have already measured.

In Quake3 Arena ALi MAGiK 1 Rev.C did really great! It fell only a trifling 1% behind VIA KT266A.

The increase in screen resolution didn't uncover any problems with the AGP driver by ALi MAGiK 1 Rev.C, which made the competitors run more or less equally.

In Unreal Tournament, like in the majority of tests, ALi MAGiK 1 Rev.C appears somewhere in between KT266A and SiS735.

One more game and the results are again the same. ALi MAGiK 1 Rev.C outperforms SiS735 by 1.5% and falls behind KT266A by 2.5%.

Well, did you expect to see a different picture here? We didn't :-)

And now a quick look at scientific and professional apps.

Science Mark hardly revealed anything different from what we had already seen.

SPECviewperf, which results depend on the memory subsystem performance (on its bandwidth in the first place) quite tangibly, showed that ALi MAGiK 1 Rev.C managed to beat even VIA KT266A in some tests. It means that in those applications where memory bandwidth matters most of all, ALi MAGiK 1 Rev.C can theoretically grab all the laurels from the recognized leader, VIA KT266A. In other words, if only ALi engineers could do something about the high latency of their memory controller, then MAGiK 1 could be an absolutely impeccable solution: the world's fastest Socket A DDR chipset.

Conclusion

First of all we would like to say a few words about ALi MAGiK 1 Rev.C chipset. As our tests showed, the third revision of this chipset, which used to be the slowest of all, has finally shown good performance. By "good" we imply that ALi MAGiK 1 Rev.C chipset outperforms SiS735 with the same cost in all the tests. Unfortunately, our hero didn't manage to catch up with VIA KT266A in all the tests, but nevertheless, it will undoubtedly find its niche and its customers. Firstly, the lag is not that great: ALi MAGiK 1 Rev.C is only 2% at the most behind the today's No. 1, and secondly, ALi MAGiK 1 Rev.C based mainboards will definitely cost much less than the corresponding solutions on VIA KT266A.

As for the mainboard on ALi MAGiK 1 Rev.C considered, namely Iwill XP33-R, it made a twofold impression, to tell the truth. Surely, it is a remarkable product with tremendous overclocking opportunities, which is the world's first mainboard with the integrated ATA/133 IDE RAID controller onboard. However, on the other hand, we had a very uncomfortable feeling about the marketing trick with the DDR333 memory support, which is not there, as we have found out. As a result, we can call Iwill XP333-R a great product, which unfortunately, doesn't meet its own claimed specification.

So, let's sum up.

Highs:

• Great stability during overclocking;
• Good stability in nominal mode;
• ATA/133 IDE RAID support;
• 6-channel sound support.

Lows:

• Incompliance with the specification: no DDR333 memory support in nominal mode;
• Some design drawbacks.