Basic Features: Epox 9NDA3+

 Epox 9NDA3+ Motherboard Specifications
CPU Interface Socket 939 Athlon 64
Chipset nVidia nForce3 Ultra
Bus Speeds 200MHz to 400MHz (in 1MHz increments)
PCI/AGP Speeds 66MHz to 100MHz (in 1MHz increments)
Core Voltage +.05V, +.10V, +.15V, +.20V
DRAM Voltage 2.5V to 2.8V in 0.1V increments
AGP Voltage 1.5V-1.8V in 0.1V increments
Chipset Voltage 1.6V-1.75V in 0.05V increments
Hyper Transport Ratios 1X to 5X in 1X increments
CPU Ratios Auto, 8X to 25X in 1X increments
DRAM Speeds Auto, 100, 133, 166, 200
Memory Slots Four 184-pin DDR Dual-Channel Slots
Unbuffered Non-ECC Memory to 4GB Total
Expansion Slots 1 AGP 8X Slot
5 PCI Slots
Onboard SATA 4-Drive SATA by nF3-250Gb
Onboard IDE Two Standard nVidia ATA133/100/66
(4 drives)
SATA/IDE RAID 4-Drive SATA Plus 4-Drive IDE can be combined in nVidia RAID 0, 1, 0+1, JBOD
Onboard USB 2.0/IEEE-1394 8 USB 2.0 ports supported by nF3-250Gb
2 1394A FireWire ports by VIA VT6307
Onboard LAN 1 Gigabit Ethernet on-chip
by nF3-250GB and Vitesse VSA8201 PHY
Onboard Audio Realtek ALC850 8-Channel with UAJ
SPDIF coaxial in and out
BIOS Award 10/07/2004

The single chip nVidia nForce3 Ultra is used for the 9NDA3+. For more information on this chipset, you can refer to our launch articles for the chipset:

nForce3-250 - Part 1: Taking Athlon 64 to the Next Level
nForce3-250 - Part 2: Taking Athlon 64 to the Next Level

The nF3 Ultra is targeted at Socket 939, with the nF3-250Gb and basic nF3-250 aimed more at Socket 754. This distinction will likely blur due to the recent introduction of the nForce4 chipset, but there is no real performance difference in the nF3-250 and nF4 chipset families. The major difference is that nForce3-250 chipsets are for AGP video and nForce4 is for PCI Express video.

The current Epox catalog shows the 9NDA3+ that we are testing, and an upcoming 9NPA+ that is based on the nForce 4 chipset.

Our 9NDA3+ test board was a full retail package and included:
  • A "Power Pack" that includes a user's manual, quick installation guide, installation screwdriver, 8 mini heatsinks for Mosfets and VGA DIMMs, Epox logo badge, and a driver CD
  • Two round red IDE cables
  • One flat floppy cable
  • Two Serial ATA data cables
  • Two Serial ATA power cable
  • One USB bracket with 2 ports
  • One Firewire bracket with 2 standard ports
  • One Accessory Bracket with Serial Port 2 and a Game port
  • One "nVRAID Driver" diskette
  • One I/O shield

Epox uses a large heatsink on the single chip nF3 Ultra with active fan cooling. The active cooling solution did a decent job of keeping the nF3 Ultra chip cool during both normal operation and overclocking.

Many of the manufacturers of nF3-250 family chipsets have chosen the Realtek ALC850 7.1 audio chip. The ALC850 is also the choice of Epox, which includes both Optical and Coaxial SPDIF connectors on the rear IO panel. This 8-channel audio codec is fully AC '97 2.3 compliant and features 16-bit 8-channel audio and auto-jack sensing with support for a full range of analog and digital IO. A wide range of sound standards are supported including:
  • EAXTM 1.0 & 2.0 compatible
  • Direct Sound 3DTM compatible
  • A3DTM compatible
  • I3DL2 compatible
  • HRTF 3D positional audio
  • SensauraTM 3D Enhancement
The ALC850 Codec provides four pairs of stereo outputs, with 5-Bit volume controls and multiple stereo and mono inputs, along with flexible mixing, and gain and mute functions. Two 50mW/20ohm headset audio amplifiers are integrated at Front-Out and Surround-Out, and both amplifiers are selectable for Front-Out, Line-In and Mic-In as a Universal Audio Jack.

You can find more information on the recently released ALC850 at Realtek.

Epox provides a full selection of rear I/O ports. These include 6 audio mini jacks plus both coaxial and optical SPDIF out connectors to support the Realtek ALC850 on the back IO panel. IO also includes PS2 mouse and keyboard, parallel, 1 serial, 1 standard Firewire (IEEE1394), 4 USB, and a Gigabit Ethernet.

nVidia "Any Drive" RAID is supported on the 9NDA3+. Any of the standard 4 IDE drives or an additional 4 SATA drives can be combined in a Raid 0 (striping), Raid 1 (Mirroring), or Raid 0+1 array.

On-chip Gigabit LAN is built-in with the nForce3 Ultra and runs completely independent of the PCI bus. Epox uses the Vitesse PHY (Physical Layer) chip to provide the interface for direct communication of the LAN to the chipset.

Four DIMM slots support up to 4GB of up to DDR400 memory in a Dual-Channel memory configuration. It is worth mentioning that Dual-Channel 1 is DIMMs 1 and 2, and Dual-Channel 2 is DIMMs 3 and 4.

Index Board Layout: Epox 9NDA3+


View All Comments

  • PrinceGaz - Monday, October 25, 2004 - link

    I agree with #7 (The Plagiarmaster) that something needs to be done to improve the video encoding tests. Results with a given codec can vary considerably for a processor-family (Athlon 64, Prescott etc) depending on the front-end app used (VirtualDub, AutoGK, DVD2AVI etc).

    Its not too important for mobo articles like this where you should only really be comparing the results with the same CPU on other mobos, but its critical to the CPU articles like the recent 4000+ and FX-55 review.

    It would be better for those to take some time to find the best performing popular front-end app for a particular codec and processor family, and then use that in all tests for that processor type and codec. So for the DivX tests you might use a different front-end for the Prescott to what is used for the A64 depending on which was found to be fastest. Same for XviD depending on what was fastest for that. The codec should obviouslly be the same for all CPUs else the test would be invalid, only the front-end app could differ. At the end of the day both will produce the same results regardless of the front-end so it makes sense to use whichever is best suited to a given CPU, as the current results with only one app in isolation may be quite misleading. If you don't like the idea of using different apps depending on the CPU, then the only alternative for presenting useful video encoding results is to repeat the test several times with different apps in all CPU articles.

    And further to earlier comments, I could not recommend a mobo with the issues the Epox has on the hope that things will improve in a future BIOS revision. I might recommend it *when things have improved*, but I would never do so before. As it stands I'd steer well clear of this board.
  • rickcfer - Monday, October 25, 2004 - link

    I tried to manually set the speed, but the board would not even post until I removed the second set of dimms. Corsair xlls 3200 pro 512 dimms. Reply
  • Wesley Fink - Monday, October 25, 2004 - link

    #13 - On the K8N Neo2 you need to set the BIOS memory timings to DDR400 when you use 4 dimms. If you leave it on auto it sets 333 with 4 dimms. Reply
  • rickcfer - Monday, October 25, 2004 - link

    "However no one should need to live with a board that forces DDR333 with 4 dimms when no other 939 board does it."

    The MSI Neo2 Platinum does it too. It's documented in their manual. How did you guys get around this? I filled all dimms and it defaulted to DDR333. Bios was 1.2
  • Wesley Fink - Monday, October 25, 2004 - link

    #7 - dvd2avi shows better results with AMD than AutoGK, even though Auto GK is based on the dvd2avi engine. AMD also performs better with XVid than Divx, since Divx has many optimizations for Intel. Another huge difference is the video clip used for encoding. Certain video clips generate much higher or lower frame rates than our standard video clip. The Star Wars clip we used in the past is about 20% faster on AutoGK than the longer Sum of All Fears we now use.

    For all of these reasons we use a standard procedure with a standard file for encoding tests.

    #8 - The poor location of the ATX/12V contrasts with the best layout I've seen for the slots and SATA. I prefer the ATX upper right edge but other A64 boards also use this layout and at least there is plenty of room around the connectors. I think Epox would disagree with you that this is a positive review.

    #11 - There are some other great OC boards that have issues with reboot freezes. I had to balance the fact the Epox was one of the best 939 overclockers we've tested with the annoying reboot issue. Some can live with that, but I couldn't. However no one should need to live with a board that forces DDR333 with 4 dimms when no other 939 board does it. If you don't use 4 dimms the board may be acceptable to some.

    I'm surprised that I point out two glaring faults on this board, qualify all my comments in light of those faults, and you still don't think I've been negative enough. I tried to keep in mind this board could become a great 939 board with a decent BIOS update and then the review would be totally misleading. It could be a great board, but it isn't there yet.

    I believe readers need the full perspective to make informed buying decisions, although we all know great put-down lines are more entertaining.
  • tagej - Monday, October 25, 2004 - link

    One of the last lines in the review: "The Epox is a great 939 choice now if you will only run 2 DIMMs and you can deal with the annoying reboot problems."

    Does this stike anyone else as crazy? How in the world is freezing and other reboot issues acceptable to anyone? Sounds to me like right now this board belongs on the junk heap, not in my PC -- at least until the issues get fixed, presumably with a bios update or two.
  • tagej - Monday, October 25, 2004 - link

    With the nforce 4 chipset boards just around the corner, it's probably wise to simply wait for those. Unless you have a significant investment in a video card you want to protect (which a lot of people do), you can just go straight to nv4 instead of getting one of these nv3 boards. If you have a great vid card you want to keep using, then nv3 939 is a relatively good path. Reply
  • jAMBAZZ - Monday, October 25, 2004 - link

    Hi Anand, please regard the fact that many people wont/dont buy the expensive RAM you are testing with. As A64 will do just as fine using HTT/Ram dividers and 200 mhz PC3200 ram at very tight timings (2-2-2-5) and is not like the previous Athlon (XP) model.

    As a note to this Epox review, it would be very nice indeed if you didnt use 1:1 HTT/RAM since you would then be able to show what kind of FSB this mobo truly allows for. You did the same in the A64 value and overclocking article which is a shame imho.

    Best Regards Mikkel Nielsen.
  • JonathanYoung - Monday, October 25, 2004 - link

    I've noticed this a lot on this and other hardware review websites... it makes me think that motherboard companies pay these sites to review their products. What I'm talking about are comments such as "The basic layout of the Epox is generally excellent, with some standout elements and a few glaring layout faults." Before reading this line I already noticed the PSU ATX connector in an ugly location, and I'm sure no one would agree that this is "generally excellent". Now, how can something tbe "generally excellent" yet still have "glaring faults"? It doesn't make any sense to me. You see this a lot on hardocp as well. There'll be a major fault or issue with a product, but they always come back with (and I'm paraphrasing here) "but other than that, it's overall excellent". If something is truly "excellent" then it doesn't have any faults, otherwise it's just average or mediocre. Okay, rant over! Reply
  • ThePlagiarmaster - Monday, October 25, 2004 - link

    Since it didn't get answered in the last article comments, I guess it's worth repeating (especially since the same results keep showing up):


    I'm wondering why the dvd2avi divx 5.1.1 show such close results here. With basically the same machines on hardocp, they show the athlon64 beating Intel's best by HUGE margins (like 20% faster than the 3.6 and almost the same over the 3.4ee)

    You guys showed the same thing no too long ago, but I can't find the article now. The paragraph under the graphs of the A64 winning said Intel lost the last thing they used to win in the benches. The article appears gone? What happened here, why so different from dvd2avi results at hardocp? They used the same divx 5.1.1 so it's not the encoder or the frontend. Heck even the 3500+ dominated the 3.6 and 3.4ee (more than 10%).

    Whatever it is, I think a bit of research needs to be done on what's best for AMD, and what's best for Intel and pit them against each other. Clearly AMD people would run the way hardocp does (though it looks no different than what's used here). While you state you can show whatever you want, perhaps you should be showing the BEST for each platform. Would people really go home and run in a way that makes their cpu look like crap?

    When you're talking about cutting 20% off of encoding, that adds up to a lot of time. I'd argue with the statement about divx 5.1.1 exploiting sse3 and making Intel a usual winner. Hardocp has been using 5.1.1 and A64's kill p4's with it (and have for a long time on their site). Is it AutoGK that throws things out of whack? Does it favor Intel so much that it causes a 20+ percent reversal? If it's truly based on dvd2avi how could it be so far out of whack compared to hardocp's scores? I think your readers would want to know which way to encode the fastest with whichever cpu they choose. Why would people want to know what the middle ground is and lose 20% cpu performance? Maybe Xmpeg for Intel, and pure dvd2avi for AMD?

Log in

Don't have an account? Sign up now