The mainboard industry has faced a confusing situation lately. For the most part, it's for the blazing rivalry between AMD and Intel: each of them requires not only special platforms for its processors, but also some particular memory types. Intel favors RDRAM and offers just one Pentium 4 chipset, which doesn't work with any other memory type but RDRAM. AMD pursues another policy trying to boost DDR SDRAM, which is supposed to provide some performance gain to Athlon systems. At the same time AMD keeps in mind compatibility with the older memory types, such as PC133 SDRAM. None of these two different policies, however, seems to be a success so far. Neither RDRAM, nor DDR SRDAM have won the users' hearts. There are several reasons for that. The market outlook for RDRAM is rather somber because of its disagreeably high price, whereas DDR SDRAM is troubled by a too low performance gain it provides the mainboards.

That's what we would like to highlight now - the problem of the insufficient performance gain brought by DDR SDRAM. It sounds logical that a combination of DDR SDRAM and Pentium III CPUs can't generate an appreciable effect. The bandwidth of Pentium III CPU bus is much lower than that of DDR SDRAM. That's why the architecture of Pentium III doesn't let us get a performance increase, which can arise if the CPU-to-memory data stream flows faster. Unfortunately, no Pentium 4 chipsets with DDR SDRAM support are available yet, so it's impossible to assess the contribution of DDR SDRAM to the performance of this promising platform. Our only hope stays with Athlon CPUs from AMD. Their bus is originally built on DDR architecture, and its bandwidth is equal to that of DDR SDRAM. Therefore, DDR SDRAM has a good chance to perform well on this platform.

However, this case is not cloudless either. The core logic manufacturers have failed to develop an effective DDR SDRAM interface in their Socket A chipsets. Today there exists only one mass DDR chipset, which can bring about a tangible performance gain. We mean AMD 760 initially created for marketing purposes, i.e., to demonstrate the capabilities of the DDR technology. At first AMD wasn't going to keep producing it in mass for long, but the reality made this company radically alter its position. Since neither VIA KT266, nor ALi MAGiK 1 are able to catch up with AMD 760 in performance, AMD goes on producing the chipset in mass quantities. We can't say whether AMD will stick to this idea in future, but still... In theory, SiS has prepared a chipset, which can oppose AMD 760 as the fastest DDR Socket A chipset. Then, there is VIA working on a new DDR SDRAM controller with higher performance to build it into the upcoming revisions of its KT266. So, those of you who are willing to get a high-performance Socket A DDR system should still focus on AMD 760 now.

The mainboard manufacturers can't remain aside, as far as the DDR chipset fuss goes. Taking into account that AMD 760 seems to be lasting much longer than it has been expected, some of the manufacturers set up a second generation of AMD 760 based mainboards. Unlike those boards, which appeared shortly after AMD 760 had been announced, the new ones promise to be cured from their predecessors' main problem - a too high price. One of the reasons why the first AMD 760 mainboards were so expensive was their six-layer PCB design. From now all the AMD 760 based mainboard have a cheaper 4-layer design. In one of our recent reviews we discussed one of these AMD 760 4-layer mainboards, Chaintech CT-7KJD, and today we would like to turn to one more product of the kind.

Closer Look

In this review we'll tell you about another new mainboard based on the already widely-spread AMD 760 chipset, EPoX 8K7A. As well as Chaintech CT-7KJD, which we dealt with last time, EPoX 8K7A was announced in 2000, but it appeared only this year. Unlike its predecessors, this board is designed on a 4-layer PCB and boasts a bunch of smart functions. But let's discuss everything step-by-step.

As a commentary to the specs, we'd like to add that EPoX offers one more mainboard based on the same PCB as EP-8K7A. It is known as EP-8K7A+, a modified EP-8K7A with an integrated ATA-100 RAID controller from HighPoint - HPT370. That is why please take note that these two products feature absolutely identical characteristics and hence everything we'll say about 8K7A will be true for 8K7A+ as well.

A short glance at the specifications and the photo of EPoX EP-8K7A is enough to notice that this mainboard is somewhat bigger in size than its counterparts from other manufacturers. You know, it's typical of many other mainboards from EPoX. The reason is pretty simple - the company's engineers stuff the products with a lot of interesting features and do their best to make them as user-friendly as possible. For example, those of you who make up their mind in favor of EPoX EP-8K7A will never have any problems with installing any of the existing CPU cooling systems. There is enough space around Socket A for any popular cooler. All the capacitors are moved away from the socket and round it there are four holes to fasten some cooling systems to the PCB.

To stabilize the CPU power supply, EPoX introduced a two-phase circuit of eight transistors. Four of them are provided with huge heatsinks:

It makes EP-8K7A a perfect solution for all the today's processors and even for the upcoming models. Although Athlon CPUs have now acquired a new Palomino core with 20% lower heat dissipation, elder Palomino models will still need rather strong power supply. No doubt, EP-8K7A will manage to meet their requirements.

There is another peculiar thing about the power supply circuit. Four of the eight transistors used in the circuit to minimize the conductors length are located on the board's rear side. To our great disappointment, it can be hardly called the best way-out because these transistors may appear in the way in some PC cases.

As for the supported memory, 8K7A has two 184-pin DIMM slots for DDR SDRAM. Unfortunately, AMD 760 supports more modules only if they are Registered ones. So, like most other manufacturers of AMD 760 based mainboards, EPoX rejected the idea to lay out a couple of extra DIMM slots. Perhaps, it was a mistake. Registered DDR SDRAM modules are available in retail and are only 1.5 times more expensive than the common DIMM modules. Consequently, it would have been wise to implement two more slots counting on Registered SDRAM users. This way of thinking was apparently taken up by ABIT as its recently announced AMD 760 based product, KG7-RAID, is equipped with four slots for DDR SDRAM against only two slots by EP-8K7A. It means that EP-8K7A supports only 2GB DDR SDRAM at the maximum.

We are sorry to state that AMD 760 chipset, EP-8K7A is built on, cannot clock the memory and CPU buses asynchronously. As a result, the board's memory subsystem can reach its top performance only in case of Athlon CPUs with 266MHz bus and PC2100 DDR SDRAM. On the other hand, the synchronously working buses let AMD 760 support the Super Bypass mode. On 8K7A this mode can be enabled via BIOS Setup. It brings about a slight performance upturn bypassing some delays needed to synchronize the buses.

Speaking of expansion slots, we should say that EP-8K7A is provided with only 1 AGP 4x and 6 PCI slots. EPoX 8K7A is intended first of all for the DIY market and not for the OEMs, so the board doesn't have any AMR, CNR or ACR slots as well as ISA slots to support older devices. This way, if you venture to assemble a system based on EP-8K7A, you'll have to confine yourself to PCI expansion cards alone.

The AGP 4x slot of EP-8K7A is equipped with a traditional retention clip to fasten the cards in proper position. But this clip won't allow you to install AGP Pro graphics cards which can work well even in common AGP 4x slots. Nowadays these cards are rare to meet, though.

A nice thing about EPoX 8K7A is that the AGP slot is located far enough from the DIMM slots. It means that only "long" graphics cards installed into the AGP slot will block the DIMM modules, while cards of a conventional size won't, saving EP-8K7A from this widely spready annoyance.

But things are not that impeccable and there are problems of another sort to upset you. The layout of the DDR DIMM slots requires special termination and an individual voltage regulator, making all the neighborhood crammed with electronic components. As a result, there is no space left for the FDD and IDE connectors to be located. That is why on 8K7A these ports are moved to the left edge of the board disagreeing with ATX specs. So, now the cables inside the PC case are located in far not the best way and full-size cards can never fit into some of the PCI slots.

The ATX power supply connector is also recommended to be placed in front of the memory slots, but EPoX put it between the DIMM slots and Socket A. This time the design change deserves only favorable comments. The cable won't hang over the CPU and nothing will hinder proper processor cooling.

Like on its all other mainboards for progressive users, EPoX implemented on 8K7A its traditional P80P Debug (POST) controller. The main idea of this controller is that in the beginning of each initialization procedure the BIOS displays a dual-digit code on the special indicator situated in the lower left corner of the board. This code defines the purpose of initialization and the device, which is to be initialized. If everything is OK, the BIOS starts initializing another device. If the attempt fails, the following devices are not initialized and the BIOS either stops or goes on trying to initialize the troublesome device. In case any problems arise, the indicator shows the code of the latest device initialized. Using a code table provided in the user's manual, you can find out, which of the devices was a failure. This way the POST controller comes in handy when you assemble and set up the system.

Despite all the might of the P80P Debug (POST) controller as a good diagnostics means, it has some drawbacks as well. Firstly, the indicator can be seen only when the case is opened; secondly, such a powerful diagnostics system can easily confuse an inexperienced user. For this reason EPoX 8K7A features voice diagnostics as well, which delivers voice messages through the speakers connected to the Outs of an integrated AC'97 audio controller. Of course, the number of these messages is not big at all (amounting to 10 only). Besides, the voice produced also leaves much to be desired, but it still helps to get an idea of the problem. It is for the first time that EPoX introduced a combination of POST controller and the voice diagnostics called POSTMAN to its mainboards. Due to POSTMAN technology we can call EPoX 8K7A a mainboard with the best diagnostics abilities today.

As a South Bridge, EP-8K7A has 686B chip from VIA. The trend goes that this chip is the most popular companion for AMD 760, though AMD also offers its own South Bridge. Naturally, the manufacturers' love to 686B chip is explained by their vital desire to reduce the cost of AMD 760 based mainboards, as long as the South Bridges from AMD is pricing higher.

Thanks to VIA 686B chip EP-8K7A can support ATA-100 device, 4 USB ports, AC'97 sound and hardware monitoring. It is noteworthy that EPoX is one of the few manufacturers that ship their mainboards with a special bracket with an additional pair of USB ports, which is installed instead of one of the expansion cards. So, the statement about EP-8K7A supporting 4 USB ports is quite correct. To implement software AC'97 sound EPoX has taken the most primitive dual-channel codec from VIA, VT1611A. The monitoring system of EP-8K7A controls two temperatures (the sensors are situated in the center of Socket A and near the South Bridge), the rotation speed of two fans (there are only three connectors for them on the board) and five voltages.

The North Bridge of EPoX 8K7A is equipped with an active cooler from AAVID settled on a layer of thermal paste. Together with the high-quality power supply circuit and well-done layout, it makes 8K7A outstandingly stable. Apart from that, to improve the mainboard stability, EPoX resorted to a tricky thing. The default voltage coming to DDR SDRAM modules on 8K7A is 2.6V and not 2.5V as it is given in the specs. As a whole, the trick is not shifty at all, as the 4% voltage increase is definitely within the acceptable range for the memory.

Overclocking

After a confusion with SoftMenu on EPoX 8KTA3, the company didn't want to push its luck another time and provided 8K7A with a set of jumpers serving to adjust the CPU settings. As you may remember, on EPoX' previous mainboard, 8KTA3, where the CPU multiplier was set via BIOS, the CPU overclocking was sometimes hindered by the board trying to boot the processor with a default multiplier at the startup, and only a while later it changed the multiplier according to the SoftMenu settings. Sometimes, when the multiplier was to be decreased for the sake of the FSB frequency growth, the system simply wouldn't start up because the CPU couldn't work even less than a second with a standard multiplier and an overclocked FSB.

EP-8K7A doesn't have any of these problems any longer. The CPU multiplier is set by the hardware, with the help of a dip-switches set on the left side of the board. Surely, the multiplier can be adjusted only when it's unlocked like it is on all the recent Athlon CPUs, fortunately. Anyway, it's not that difficult to unlock the CPU multiplier if needed (click here to read how it can be done).

On 8K7A the FSB frequency is changed in two stages. The basic frequency (100 or 133MHz) is selected with a jumper, finer adjustment is done via BIOS Setup, which allows changing the frequency up to 166MHz with 1MHz increment. Please, don't forget that if the basic frequency is set to 100MHz, the mainboard takes a 2:3 multiplier for the AGP frequency and the BIOS Setup allows increasing the FSB frequency only to 132MHz. If the basic frequency is 133MHz, a 1:2 multiplier is used for AGP and the frequencies from 133MHz to 166MHz become available in BIOS Setup.

For extreme overclockers EP-8K7A has a couple of jumpers for the Vcore and Vmem adjustment. In both cases the default voltage can be increased by 0.1V, 0.2V, 0.3V or 0.4V.

Due to the possibility to increase the Vcore and Vmem, which both were important for us, we managed to squeeze the maximum of 158MHz out of the FSB (the voltages were set to the maximum). Note that the system worked perfectly stably. Indeed, it's a splendid result confirming this mainboard's brilliant stability and reliability.

Testbed and Methods

EPoX EP-8K7A is not the first AMD 760 based mainboard investigated in our test lab, so we won't waste time comparing it to the performance of other platforms. Those readers who are interested to see what AMD 760 based mainboards look like in comparison with other chipsets are highly welcome to consult one of our latest reviews on this topic - VIA KT266: Final Verdict. MSI K7T266 Pro Mainboard. During this test session we'll merely look at the performance of EPoX EP-8K7A compared to other AMD 760 based boards. There were 4 systems taking part in the runs:

We tested in Microsoft Windows 98 SE.

Performance

Now let us have a look at the performance. Here is what our memory test showed:

As you see, in SiSoft Sandra 2001 all the racers demonstrate similar memory capabilities. Well, let's find out how the things stand in real applications.

First come office applications:

In office applications ASUS A7M266 is the fastest. Still, all the others don't fall behind by more than 3%.

Now it's high time for gaming applications:

In gaming applications EP-8K7A shows a moderate result.

In conclusion we'd like to say that performance is not the main argument for this or that mainboard when you choose between several boards based on the same chipset. The performances of different products do not differ that greatly for you to be guided by benchmarks only. It's much more important to pay attention to the board's other features as well.

Conclusion

For the time being, EP-8K7A is one of the leading mainboards for Socket A processors with excellent performance and a broad variety of functions. This board works with DDR SRDAM and is based on the fastest core logic for Athlon CPUs, AMD 760. Boasting all the necessary CPU overclocking tools, advanced diagnostics system, prominent expansion abilities, high stability and an optional RAID controller, this board has every chance to become a favorite for many progressive users.

Highs:

• Four-layer PCB and high-quality CPU power supply stabilization circuit;
• High stability;
• Built-in POSTMAN diagnostics system;
• The whole bunch of CPU overclocking functions.

Lows:

• Some design inconveniences.