EPoX 8K9A2+ Gamers Review


A while back, I was lucky enough to get ahold of the 8K9A2+ motherboard from Epox. I initially wasn't planning to review this board, but after reading a bunch of other reviews on the web, I felt the need to address some issues, both strong points, and week points that I haven't seen covered. I am not claiming to have a better or more complete review on this board, I am simply covering it from a slightly different angle.
Package & Contents:

When Epox came to market with this product they wanted to appeal to the consumer in every way imaginable. Not only is this board feature-packed, it's also comes in a pretty package. Epox apparently decided to save trees with the 8K9A2+, and this board comes in a funky plastic box.

Who really cares what the box looks like? It's what's inside that counts!

Inside you'll find quite the bundle. The Deluxe version of the board comes with a mobo manual, HighPoint manual, Serial ATA manual, Driver CD, 3.5" HighPoint drivers, 3.5" Serial ATA Drivers, Rear USB Bracket, Rear Game Port, Serial ATA cable, IO shield, a Rounded ATA133 Cable, and a Rounded Floppy Cable.

The version I received has all of this, minus the Rounded Cables. . . it came with regular ribbon cables. Either way, it's quite a bundle for a motherboard that is priced $60-$80 less that a comparable ASUS, ABIT, or MSI motherboard.

Board Info and Features:

First, here's a shot of the board for you to size up.

This board is based on the KT400 chipset from VIA. The KT400 is hailed by some as a flop by VIA, and others see it as a positive step forward. Here the info from VIA on their KT400 chipset.

  • North Bridge - VT8377
  • Front Side Bus - 200/266/333MHz
  • Memory Bus - 200/266/333MHz
  • AGP Support - AGP 8X
  • Memory Type - DDR200/266/333 SDRAM
  • South Bridge - VT8235
  • North/South Bridge Link - VIA V-Link 533MB/s
  • Integrated Audio - AC '97 6 Channel
  • Integrated Modem MC '97
  • Integrated Network Controller - VIA 10/100 Ethernet MAC
  • IDE - ATA 33/66/100/133
  • USB - USB 2.0 6 Ports, 3 Controllers
  • Super I/O - LPC Super I/O
  • IO APIC - Yes
  • Power Management - ACPI/APM/PCI/PM

Other features found on this board include:

  • 3 DIMM Slots that can handle 3 double sided DIMMS
  • AGP 8X
  • 6 PCI
  • 6 USB 2.0
  • 2 ATA 133 Connectors
  • 2 HPT372 ATA 133 Connectors
  • 5.1 Channel Sound
  • HPT372 controller for DMA-133 supports RAID 0, 1, 0+1, and JBOD.
  • 2 Serial-ATA ports for serial ATA-150 drives.
  • Onboard Via based 10/100 network.
  • Magic-Light Blue LED Northbridge FAN
  • Unofficial Support for DDR400
  • And then some. . .

All in all, this is pretty much a loaded board that should be able to hold the most aggressive enthusiast for quite a while. It supports, Duron, Athlon XP, Thoroughbred cores, as well as Barton cored CPU's.

As you can see, the board has a pretty clean layout. It has the much sought after 4 mounting holes for some of the larger HSF's or waterblocks. I must admit, the holes are quite a bit smaller than on my KT7A-Raid, but the DangerDen Maze2 still mounted okay. My biggest beef with the layout is the location of the power connector. . . In order to keep your HSF free, you'll probably need a zip tie.

Now that we know a bit about the board, lets head on over and check out the Test Setup and start working this board!

Test Setup:
The test setup I used for this review goes something link this:
  • Athlon XP 1800+ (Thoroughbred)
  • EPOX 8K9A2+
  • 512MB PC2700 Infineon RAM CAS 2.5
  • Radeon 9700Pro
  • Sound Blaster Live 5.1
  • 40GB Maxtor Diamond Max Plus (IDE 1, Master) OS Installed
  • Pioneer 16x DVD ROM (IDE 1, Slave)
  • MSI 8348 DragonWriter (IDE 2, Master)
  • 40GB WD 8MB Cache 7200 RPM (IDE 3 - "RAID 0" Master)
  • 40GB WD 8MB Cache 7200 RPM (IDE 4 - "RAID 0" Master)
  • Windows 2000 SP2
  • Catalyst 3.1
  • Via Hyperion 4.45 (4.46 fixes AGP 8X issue)

This is a Gamers Review, so I haven't included a lot of Business Benchmarks such as Adobe, Winstone, etc. I've included some of the more popular Synthetic Benchmarks as well as some game-based benchmarks. All Benchmarks were run at the highest possible detail level with the drivers at their default settings.

The next thing we'll do, just before we jump into some benchmarks, is take a look at the BIOS.

Bios Features:
The BIOS on the 8K9A2+ board comes pretty much loaded. It has just about everything a man could want in a BIOS, including multiplier selection, adjustable fsb from 100 to 255, vcore adjustments, memory voltage adjustments, and myriads of ram timing tweaks. Instead of me just telling you about it, lets take a look. All thumbs are clickable.
Bios Features - - - - Peripheral Settings
AGP Settings - - - - PC Health Screen
RAM Timings - - - - Voltage Settings
CPU Frequency and Multiplier

The biggest thing that doesn't impress me about theBIOS is that when an AGP 8X card is detected it doesn't let you select your AGP rate. It must run at AGP 8X. This was causing a serious problem with my Radeon 9700 Pro. It wouldn't run stable on this board at AGP 8X. I had to use PowerStrip to lock my AGP rate to 4X.

That being said, the newest VIA Hyperion drivers, the 4.46 actually have fixed the AGP 8X stability issue.


This board is one heck of an overclocker. I am using a T-bred A chip (133FSB) on this board and it doesn't really seem to have limits. The big downer on the KT400 based boards is that when you overclock above 166FSB, you start to overclock your PCI and your AGP. There is no way to "lock" it at a certain frequency, and the boards lack of a 1:6 divider, definately make overclocking a little more limited. The FSB is keyed in and is selectable from 100 to 255 FSB, and any multiplier from 5 - 24 can be chosen in increments of .5.

That being said, this board still rocks in the OC department. With my cheap PC2700, I've had it to 205FSB, and my 1800+ overclocked to 2.26GHz. . . not stable mind you, but it's posted and booted into Windows. I was getting memory errors.

For this review, my MAX STABLE overclock was at 190x11=2.09GHz (Programs detect it at 2.1GHz), and this was definately due to my RAM. I had set to Ultra Timings and volted up to 2.8v, and the CPU was overvolted to 1.7v.

Now lets get on with some benchmarks!!!

SiSoft Sandra:
* * * All overclocked results on all benchmarks are at 190x11 on the CPU. The GPU remains at stock speeds at all times.
First we take a look at the common Memory Bandwidth Benchmark. The CPU here is set at 133 FSB and the ram is running at it's stock 166FSB. . . All settings normal and by SPD. It seems a little on the low end of the scale, but I am running cheap RAM at CAS 2.5.

This next shot, shows how much of improvement their is running the RAM at 190FSB and Ultra settings, with the CPU running 190 FSB as well. That's a little better. . . That overclock bumped up the bandwidth by 40%. Nice!

Now we'll take a look at SiSoft's CPU Arithmatic Benchmark. According to the chart, the stock CPU is sitting slightly below the database CPU.
When we overclock the CPU to 190x11 we see a dramatic increase in performance, and the chip outscores a XP2400+.

The last SiSoft Benchmark I ran was the CPU Multimedia Benchmark. Once again, at stock speeds the 1800+ CPU shows a slightly lower performance than the database.
By overclocking though we see an easy 36% perfomance increase in this synthetic benchmark. Gotta love that "free" performance.
PCMark 2002:
PCMark is a benchmark that shows overall computer performance, and this is how it looks on the 8K9A2+. From overclocking the CPU gets an extra 34%, Memory - a 20% boost, and even the Hard Drives feel the benefit to the tune of 9%. Not bad.

OpenGL Benchmarks:

* * * All overclocked results on all benchmarks are at 190x11 on the CPU. The GPU remains at stock speeds at all times.

The first of the OpenGL Benchmarks is the Synthetic GLExcess. Benchmark settings were at maximum quality, and driver settings were at default. Overclocking does offer a nice performance boost, although not nearly as much as in the SiSoft Benchmarks.
The second OpenGL Benchmarks is the ever-popular Vulpine GL Mark. Once again, this benchmark was ran at maximum quality setting with the drivers set at default. The numbers pretty much speak for themselves.
The last OpenGL Benchmark is the game Jedi Knight II: Jedi Outcast. The game is patched up to version 1.03, and all detail is turned on to maximum. The map ran is the Massassi Temple. This game is very CPU oriented. The scores at 1024x768 and 1280x1024 are almost identical. It's the CPU speed that determines the score.
We've looked at some OpenGL results, now lets head on over and check out the always famous, and sometime controversial 3DMark Benchmarks.

* * * All overclocked results on all benchmarks are at 190x11 on the CPU. The GPU remains at stock speeds at all times.
First, we look at the aging 3DMark 2001. Many of todays games are still based in DirectX 7 tech, so I thought this definately had a fit in a gamers review. At 1024x768 the default score is 12,662, and overclocking adds 18% performance at reaches an amazing 15,028. At 1280x1024 default score is 11,447 and overclocked it scores 12,729 an extra 11%.
One of the more controversial benchmarks around has got to be 3DMark03. Regardless of the controversy, it shows that faster systems score higher - and really isn't that the purpose of a benchmark? So here it goes. . . Running the default benchmark yields some decent scores: 4509 @ 1024x768, and 3565 @ 1280x1024. Overclocking doesn't result in that big of a boost. It only gets an extra 10%. Clearly this is more taxing on the GPU. Regardless, running 560MHz above stock speeds does help.
Now that we've got these out of the way, let check out the rest of the DirectX benchmarks!

DirectX Benchmarks:
* * * All overclocked results on all benchmarks are at 190x11 on the CPU. The GPU remains at stock speeds at all times.
The first of these benchmarks is the DX8 Benchmark from Code Creatures. It is a really intense "Nature" type benchmark. This is how it stacks up. Not a whole lot of performance boost from overclocking the CPU as it is really GPU intensive. Less than 4FPS average.
The Comanche 4 Demo/Benchmark is another benchmark that shows itself to be CPU intensive. The scores at 1024x768 and 1280x1024 are almost identical. Overclocking in this benchmark give us an extra 32% perfomance boost. Not bad at all!
Unreal Tournament 2003 is one of the latest games that really work a computer. It is based on DX7 and DX8, and can be a real workout. It comes with a benchmarking utility that is less than adequate. I've opted to use Nicao's UTBench2K3 for this review as it possibly the easiest, and most accurate UT2003 Benchmarking program available. It offers many features that other programs don't have, such as "average results by second". That gives accurate results like in 3DMark 2001. Normally UT2003 displays your results, such as highs and lows, based on getting results 30 times a second. This utility can average (it can be turned off) those 30 times/sec into one result. Much more accurate and realistic. Here's a shot of the built in graphing feature. Well get into UT2003 more on the next page.

UT2003 Benchmarks:
* * * All overclocked results on all benchmarks are at 190x11 on the CPU. The GPU remains at stock speeds at all times.
I've opted to use Nicao's UTBench2K3. It's an awesome, accurate, easy to use utility that even has HardOCP compatible benchmarks. Go get it here. For ease of understanding and comparison, I've used the HardOCP compatible settings on all of these Benchmarks. They are all done at 1024x768 at High Detail. All driver settings are default.
The first test is the CPU test. The two sets of results on the left, are taken without averaging the results by second. The second set is averaged by second, and reflects actual game performance. Overclocking gives quite a boost as you can see.
Next are some UT2003 Map Benchmarks. The results speak for themselves. Again, the two sets of bars on the left of the graph are displayed as HardOCP compatible, and the sets on the right have all the same settings, but are averaged by second to reflect actual game performance.
These test all show some advantage of overclocking, and what this board can do. On a side note, one of the most dramatic tests to demonstrate the "real in-game" performance benchmarking of Nicao's UTBench2K3 is Inferno. Apparently for 1/30th of a second the framerates drop quite low, but you'd never see it in game, as reflected by his utility.
Now, on the conclusion and final thoughts.


Although this board performed extremely well in all of the tests, it does have some issues. Until the 4.46 Hyperion drivers, AGP8X wouldn't work very well with the Radeon 9700 series, and you can't select AGP4X in the BIOS. Now that the drivers have apparently fixed this issue, it really isn't relevant. However, there seems to be a conflict between the HighPoint controller and the USB 2.0 controller. If you have any USB 2.0 removable storage (CD-RW or card readers) plugged into your USB 2.0 ports, and you are using RAID, you won't be able to boot. It will lock at the RAID initialization screen. Not a big deal to me, but it's a deal nevertheless.

So here is how it sums up:

This board is a great board. It is rock stable, overclocks well, has tons of features, and has some great performance. It does have a few minor flaws, but nothing that I would cry home to mama about. If you're looking for a top-end board with RAID, you can't get it in the nForce line of mobo's so you might want to check out this board. After getting the latest VIA drivers, I could definately recommend this board to my friends.


  • Great bundle of features.
  • Excellent performance and stability.
  • Great Overclocker.


  • USB 2.0 Removable Storage conflicts with HPT372 controller.

Thanks to Dallmann Computers for setting me up with this killer mobo.

If you have any questions or comments about this review, please Email me, or better yet, join the forum to discuss it with others.

Ben's Custom Cases gives the EPoX 8K9A2+ a 9/10.