Epox 9NDA3+: A New Socket 939
by Wesley Fink on October 25, 2004 12:01 AM EST- Posted in
- Motherboards
Memory Stress Testing: Epox 9NDA3+
The memory stress test is very basic. It simply tests the ability of the Epox 9NDA3+ to operate at its officially supported memory frequency (400MHz DDR) at the lowest memory timings OCZ PC3200 Platinum Rev. 2 modules will support. All DIMMs used for stress testing were 512MB double-sided (or double-bank) memory. Memory was only tested in Dual-Channel mode using either one dual-channel (2 DIMMs) or 2 dual-channels (4 DIMMs).| Stable DDR400 Timings - One Dual-Channel (2/4 DIMMs populated) |
|
| Clock Speed: | 200MHz |
| CAS Latency: | 2.0 |
| RAS to CAS Delay: | 2T |
| RAS Precharge: | 10T* |
| Precharge Delay: | 2T |
| Command Rate: | 1T |
Using two DIMMs in Dual-Channel 128-bit mode, the memory performed in all benchmarks at the fastest 2-2-2-10 timings at 2.6V.
| Stable DDR400 Timings - 2 DIMMs (2/4 DIMMs populated) |
|
| Clock Speed: | 166MHz |
| CAS Latency: | 2.0 |
| RAS to CAS Delay: | 2T |
| RAS Precharge: | 10T* |
| Precharge Delay: | 2T |
| Command Rate: | 2T |
Tests with all four DIMM slots populated on the Epox board were something of a disappointment, however. Most 939 boards that we have tested require 2T Command Rate with 4 DIMMs in two dual channels. Not surprisingly, 2T was also required with 4 dimms on the 9NDA3+. However, this Epox is the only 939 board we have tested that absolutely forces DDR333 when four DIMMs are installed. Some other boards that we have tested also try to force DDR333 with 4 DIMMs, but this can normally be corrected to DDR400 by selecting that speed in BIOS. However, the Epox sets DDR333 with 4 DIMMs even if you specify DDR400. With no way to even try DDR400 with 4 DIMMs, we are left with the forced DDR333 setting with 4 dimms in our memory tests.
Epox can possibly correct this strange 4-DIMM memory behavior with a BIOS update. For now, if you absolutely need a 939 board that supports 4 DIMMs at DDR400, you will need to look at other 939 motherboards, since the Epox is the only 939 board that we have tested which cannot run at DDR400 with 4 DIMMs. Of course, we did not need to slow memory timings at all with 4 DIMMs, though we would hardly expect that with memory running at DDR333 instead of DDR400.
Facebook
Twitter
RSS
36 Comments
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.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.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.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!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):Wesley,
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)
http://www.hardocp.com/article.html?art=Njc1LDM=
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?