Motherboard Comparison Epox EP-7KXA vs. Asus K7V – THG.RU

Introduction

Before we will publish a large KX133-motherboard review, I decided to pick two particular candidates with this chipset, which are both available in the shops right now. VIA’s Apollo KX133 chipset for AMD’s Athlon processor was launched in early February this year and it is currently the best platform for this CPU on the market. I published a comprehensive review and a few follow ups about the KX133 platform, which you might want to have a look at:

• New Hope For Athlon – The VIA Apollo KX133 Chipset
• NVIDIA’s New NT-Drivers for Athlon on KX133
• Irongate with Super Bypass vs. VIA Apollo KX133
• The Impact of the AGP-Speed

VIA’s Apollo KX133 offers the following important features:

• AGP4X support
• PC133 SDRAM support
• 2 USB Hubs = 4 USB ports
• ATA66 IDE interface
• AC97 low-cost audio and MC-97 low-cost modem support
• 2 GB max memory support

Later on this year VIA will add the ‘Apollo KZ133’ and the ‘Apollo KM133’, both coming with integrated 3D-graphics (SMA = ‘shared memory architecture’):

After that, supposedly around September, VIA will finally offer Athlon-chipsets with DDR-SDRAM support, probably called KX266, KZ266 and KM266.

The Boards – Epox 7KXA

Epox was the first motherboard maker that was able to supply a platform with VIA’s latest Athlon chipset. This board is now available for more than a month. Being the first does not always mean being the best, which turned out to be true with the 7KXA. Many owners of this motherboard are complaining about its instability.

The 7KXA comes with all the goodies you would expect from a modern Athlon motherboard:

• 5 PCI Slots, all are PCI Masters
• 1 AGP Slot
• 1 AMR Slot
• 1 ISA Slot
• 3 DIMM Slots
• 2 ATA66 IDE Ports
• 4 USB Ports, but cable for additional two USB-ports is NOT supplied!
• Switching Power Supply
• Up to 768 MB RAM
• ECC Support for Main Memory
• Award BIOS
• All the Latest ACPI Functions
• PC Health Monitoring

The manual is very detailed and covers really everything you can ask for. The BIOS allows you to switch the memory clock between 100 and 133 MHz and Athlon’s FSB can be chosen from 83/88/90/95/100/110/115 MHz, which is not quite good enough for overclocking, because the increments are clearly too large. Epox included a special jumper on the motherboard that can be used to choose between 100 and 133 MHz FSB for the Athlon. This jumper is obviously for future processors, because it is rather unlikely that any current Athlon processor would put up with that high bus clock. I tried an Athlon 750 that I deliberately modified to use a multiplier of only x5. The system would not boot, although the Athlon 750 I used should have easily reached the 133 MHz x 5 = 667 MHz. I recon that the board design does not allow bus clocks as high as 133 MHz, so the jumper is merely showing the designer’s wishful thinking.

Another positive thing about the 7KXA is that it comes with Ghost 5.1. This software lets you save backups of hard drive partitions to another drive and you can use it to clone a hard drive as well.

My tests showed the same problems that people are going on about in several news groups as well. In terms of stability, the 7KXA is not exactly impressive. I encountered several crashes when running the benchmarks and I can only recommend that the owners of this board set the memory settings to the slowest values if stability is in any way important to them. I personally could not live with the unreliability of this Epox board, since I prefer to write my reviews only once. As if this wasn’t bad enough news, the 7KXA is also a pretty bad performer.

Another interesting thing happened when I tested the 7KXA under Windows 2000. Win2k was unable to recognize the CPU-to-AGP-bridge due to strange device numbering of the PCI-devices of VIA’s Apollo KX133 chipset. Thus the AGP is not running under Windows 2000, resulting in horrible 3D-performance.

Luckily there is a way around it. Let me first explain what the problem is though.

The Chipset Device Numbers and the Story Behind Them

Each device connected to the AGP or PCI bus needs to be identified by the system, so that it knows how to operate it. This includes graphics cards, chipsets, network cards, sound cards,… For this purpose each of those devices has got two 32-bit long numbers that identify it. The first number symbolizes the vendor. Intel has e.g. hex ‘8086’, VIA has got hex ‘1106’ as its identifier. The second number symbolizes the device. While devices like graphics cards or network cards represent only one device, chipsets can represent a few at the same time, simply due to their complex integration. A modern chipset usually consists of two chips, formerly known as north and south bridge and traditionally connected to each other via the PCI bus. Those two chips include several devices though. You will find the host-bridge, the CPU-to-AGP-bridge, sometimes a PCI-ISA-bridge, the IDE-controller and more. If you are running Windows9x or Windows2000 you can have a look in the system manager and you will find a whole lot of system devices, which are represented by the chipset and all other chips found on the motherboard..

A very easy way to see all the PCI/AGP-devices in your computer with their two identification numbers is a program called ‘WPCEDIT’, found at www.h-oda.com. This website contains tons of excellent tweaking software and is second to none in the world. WPCEDIT is actually designed to let you change all the settings of your PCI/AGP devices and it is a very powerful tool. You better know what you are doing with it though, because there’s nothing easier than causing a system crash with this lovely software. We only want to use it for supplying us with a list of all the PCI/AGP-devices though, which is completely harmless. Clicking the ‘Devices’ submenu of the ‘Edit’-menu on a normal KX133-system gives you the following list:

Those numbers might not look very helpful, but you can certainly spot all the devices that start with the ‘1106’, which represents VIA. All those ‘1106’-devices are part of KX133’s VT82C8371 north bridge and the VT82C686A south bridge. What we are interested into is the host bridge ‘0691’ and particularly the CPU-to-AGP-bridge ‘8598’, falsely called ‘PCI-PCI-bridge’, because WPCEDIT doesn’t recognize it yet. If the motherboard’s BIOS reports those device numbers, you get the correct designation of the devices in Windows98, and Windows2000 is able to recognize the AGP.

Here is what it looks like in Windows98:

You can clearly see the ‘VIA Tech CPU to AGP controller’ and the ‘VIA Tech VT83C69x CPU to PCI bridge’.

This is the system manager of Windows 2000:

You can see the same thing as above. If Win2k reports all the four VIA-devices of your KX133-system you can be sure that AGP is actually working.

How To Run the 7KXA under Windows 2000 with AGP Turned On

For reasons unknown to me the 7KXA reports different device numbers than other KX133-motherboards. See for yourself:

For some reason the host bridge has the number ‘0391’ instead of ‘0691’ and the CPU-AGP-bridge uses number ‘8391’ instead of ‘8598’. Under Windows98 this doesn’t have any negative effect, because you need to install the ‘viagart.vxd’ AGP-miniport driver for VIA chipsets anyway. Thus AGP works even if Windows98 doesn’t know what kind of devices ‘0391’ and ‘8391’ are.

This is what the system manager under Windows98 looks like on the 7KXA:

You can see that Windows98 doesn’t recognize the two, thus it installs them as ‘PCI standard host CPU bridge’ and ‘PCI standard PCI-to-PCI bridge’. As already said, under Windows98 this is simply a cosmetic problem, it doesn’t have any other impact. Things are different however with Windows 2000. This is what the Win2k system manager reports:

Again you cannot find the CPU-to-AGP bridge, because Windows2000 doesn’t recognize it. The result is that Windows2000 does not install and run the ‘viaagp.sys’ driver for the AGP and so you won’t have any AGP-functionality under Windows 2000.

How To Run the 7KXA under Windows 2000 with AGP Turned On, Continued

The way out of this dilemma is very simple though.

All the system devices Windows 2000 knows are listed in a file called ‘machine.inf’, which you can find in the ‘WINNTinf’ folder (it’s actually a hidden folder, so you might have to alter the view properties of your Explorer). In this text file, which can easily be edited with Notepad, you will find a ‘chapter’ called ‘VIA_SYS’, looking like this:

Here you find all device numbers of VIA-devices that are known to Windows 2000 listed and described. Now all you’ve got to do is find the line for ‘0691’ and ‘8598’. Select those and copy them right above or underneath, so that you have the lines twice. Edit one of the ‘0691’ lines and replace ‘0691’ with ‘0391’. The same you do with one of the ‘8598’-lines, you replace it with ‘8391’. Alternatively you can just copy the two lines here and add them:

Now look for the ‘chapter’ called ‘VIA_MFG’. This is what you’ll find:

You do the same as above, double the 0691 and 8598 lines and replace them with 0391 and 8391. Or you simply add those two:

Save your new and updated ‘machine.inf’ in the WINNTinf-folder.

Now you only need to ‘uninstall’ the ‘PCI standard host CPU bridge’ and run the ‘hardware wizard’. It will from now on recognize the CPU-to-AGP-controller of the 7KXA and your software can take advantage of AGP4x.

The Boards – Asus K7V

I don’t want to go on about it much; Asus is well renowned for its stable and reliable motherboards that stand out of the crowd quite significantly. Recently Asus added the ‘jumperfree’-design to the BIOS of all of its new boards, so that even overclockers should be very pleased with Asus now. Stability is of the highest importance to all the people who do serious work on their systems, but even overclockers have an easier life if the board supplies superior stability. Another thing worth mentioning about Asus is the fact that they have the best BIOS-support of all motherboard makers in the world. As soon as there is a new CPU, microcode or other update out from AMD or Intel, you will find a new BIOS-update for your motherboard to keep it up to date. Abit may be trying to compete with Asus in the overclocker scene, but they’ve got a lot to catch up with in terms of BIOS updates. It’s one of the weakest spots of Abit.

The K7V is following the usual Asus-tradition. It’s well equipped and well designed. See for yourself:

• 5 PCI Slots, all are PCI Masters
• 1 AGP Pro Slot
• 1 AMR Slot
• 3 DIMM Slots
• 2 ATA66 IDE Ports
• 2 or 4 USB-Ports, cable for additional two is optionally supplied
• Switching Power Supply
• Up to 1.5 GB SDRAM or VCRAM
• ECC Support for Main Memory
• Award Anti-Virus BIOS with Phoenix look
• All the Latest ACPI Functions
• PC Health Monitoring
• Onboard Aureal Audio (optional)

The board manual is top-notch Asus standard. It describes all that you need to know and then some, e.g. BIOS beep codes. The K7V follows Asus’ ‘jumperfree’ design, which gives you the option to either adjust your board manually with dipswitches or use the comfort of the BIOS setup instead. You can adjust the I/O-voltage on the board from 3.3 to 3.56 V with a jumper and the rest really should be done in the BIOS setup, unless you want to avoid that your customers overclock their system. The BIOS setup offers setting for the CPU-voltage in 0.05 V increments and the following FSB-speeds: 90/92/95/97/ 100 /101/103/105/107/110/112/115/117/120/122/124/127/130/ 133 /136/140/145/150/155 MHz. This should be just about good enough for most of the overclockers out there. You can of course also choose between 100 and 133 MHz memory clock. One other thing may be important to overclockers as well. Once a ‘jumperfree’ Asus board does not reach POST, mostly because the overclocking was taken over the top, you only need to turn off your system and the next time the board will start at a safe setting and bring you right into the BIOS-setup where you can adjust the CPU-speed. I can remember other motherboards with BIOS-setup adjustable CPU-clock that need to get their CMOS cleared to make it reboot after an unsuccessful overclocking attempt. The solution that Asus offers is a lot smarter and more comfortable.

Once I found out that I had to use the BIOS 1004.02A instead of the latest 1004.04A, the K7V ran absolutely stable with each of the three operating systems I used. The performance was excellent too. It outperformed the 7KXA by a respectable amount and it was faster than VIA’s excellent reference board as well.

Test Setup

I thought it would be best to use the Giga-Athlon for the test, since its power hunger would stress the boards particularly hard. I tried two different PC133 DIMMs, one 128 MB CL2 DIMM from Crucial Technologies (Micron) and one 128 MB CL2 HSDRAM DIMM from Enhanced Memory Systems. Both DIMMs are excellent performers in terms of reliability.

For the benchmarks I still used Sysmark 2000, although I’ve got my doubts that this benchmark is particularly useful for a motherboard evaluation. The office applications used in this benchmark are hardly able to stress the memory, and they leave the AGP completely alone. Therefore you can hardly spot performance differences between boards with Sysmark and you still don’t know if your memory and AGP performs stable. I added Quake 3 Arena, Expendable and SPECviewperf 6.1.1 to the benchmark suite, because those applications make heavy usage of the memory and the AGP. Especially SPECviewperf will easily crash a system with instable memory. It can also point out performance deltas of the memory system and the AGP very well.

Windows98 Testing

I included the Asus K7M with AMD’s 750 chipset in those benchmarks, so that you have the chance to compare the KX133 motherboards with this older platform. The AMD750 chipset on this K7M board was rev. C, stepping 6 and SuperBypass was enabled.

The boards performed all pretty stable, only the 7KXA had to be rebooted in one of the Sysmark2000 runs. Unreal Tournament was another obstacle for the 7KXA, it could only run UTBENCH when the CAS latency of the memory was switched from 2 to 3.

The other boards did not crash in any benchmark, although they were of course tuned for highest performance.

Windows NT 4 SP6 Testing

I included Windows NT for those who still prefer this operating system to the new Windows 2000. My only interest was to see how the boards scored in memory and AGP-intensive tests, so that I only ran SPECviewperf and Quake 3 Arena. It turned out that both benchmarks scored their highest results under NT. Quake 3 Arena gained some 5 fps on the Asus board.

None of the boards crashed and all could be kept at the highest performance settings.

Windows 2000 Professional

This new operating system is a tough customer. First I had to make sure that the Epox 7KXA would run with AGP turned on, as you could read above. Then I had several stability issues. In most of the cases the motherboards failed to even boot to the Windows 2000 desktop. Finally I found the source of evil. None of the boards would run stable unless the ‘SDRAM Interleaving’ in the BIOS setup was turned off. Still all boards finally failed SPECviewperf by freezing the screen unless the CAS-latency was switched to ‘3’. After this final adjustment, Windows 2000 ran fine.

Why Was Irongate Not Included in the Win2k and WinNT Benchmarks?

I am afraid I have to admit that I was unable to turn on AGP2x for Irongate in the GeForce driver. Running Irongate with GeForce against the other three KX133 boards would have been rather unfair and would have made Irongate look worse than it is. Although ‘EnableIrongate2x’ as well as ‘EnableIrongateSBA’ is included in both rev. 5.13 drivers for GeForce, I failed miserably in turning them on. I tried everything in the registry that I could, but still Win2k as well as WinNT would boot GeForce with only AGP1x.

Benchmark Results – Windows 98 Benchmarks

I already said that I don’t consider Sysmark2000 a particularly helpful motherboard benchmark. You can see that the motherboards perform close to identical, only the 7KXA falls back a tiny bit.

In all games the K7V is clearly the winner. However, the distance to its competitors isn’t particularly far. It is worth mentioning however, that the Epox 7KXA performs even worse than VIA’s own reference board.

In SPECviewperf the picture is the same as above. The K7V is clearly the fastest board in this test, while the 7KXA is just about as fast as VIA’s reference board. The K7M with AMD750 ‘Irongate’ and enabled SuperBypass falls back in this test, because it doesn’t come with AGP4x support and its memory is only clocked at 100 MHz.

Benchmark Results – Windows NT 4 Benchmarks

Well, you can see that the frame rates have gone up a bit compared to Windows98. The K7V is the performance leader once more and the 7KXA falls back behind VIA’s reference board.

Somehow the K7V is able to smoke the competition in Design Review and Data Explorer. Those two viewperf-benchmarks are particularly memory bandwidth dependent and care less about the AGP-performance. It seems as if the K7V has a superior memory performance.

Benchmark Results – Windows 2000 Professional Benchmarks

Again the Asus K7V leads the pack and the Epox 7KXA comes last in those two games. Please note that Q3A scores about 5 fps less than under NT.

Once again the K7V performs better than the rest, but the differences are marginal. The scores are again lower than under NT4.

Conclusion

The benchmark results show it once more, motherboards with VIA’s Apollo KX133 chipset should definitely be preferred over motherboards with AMD’s old 750 chipset. It is certainly true however, that there’s no urgent need for an upgrade from an AMD750 to a KX133 motherboard right now, as long as you are happy with the performance of your current system. The improvements of KX133 lie mainly in the area of software that can take advantage of AGP4x. The higher memory bandwidth provided by KX133’s PC133 support can only show a little in 3D-applications. People who are buying a new Athlon system however, should only go for KX133. There is simply not even one single reason to buy an AMD750 based board anymore.

As for the few two contenders in this test, it’s pretty obvious that my recommendation goes towards the Asus K7V. It provided excellent stability combined with very good performance. So far it’s the fastest Athlon-platform that I have tested. Our upcoming large KX133 motherboard test will show if Asus can pull off another overall win. So far it proved to be a very promising candidate for winning the ‘Best Athlon Motherboard’ contest.

Follow-up by reading the article ‘VIA’s Apollo KX133 Chipset and Windows2000‘.