ECS NFORCE3-A motherboard mini review

[imported_jz]

Hi Zap & furballi,

Thank you for your help! It turns out the cd rom drive plugged in to the IDE 1 was bad, but I didn't discover this until I unplugged everything. This for some reason this caused the bios to hang at this memory line . . . weird.

Everything is working now and I'm enjoying my new computer.

Thanks again.

 

[Zap]

Originally posted by: jz
It turns out the cd rom drive plugged in to the IDE 1 was bad, but I didn't discover this until I unplugged everything. This for some reason this caused the bios to hang at this memory line . . . weird.

Cool deal, thanks for updating us and hopefully you stick around and participate some more in the forums.

Why it would "hang" right after the memory count is because that is when it is getting ready to poll the IDE drives, and as one was faulty, made it get stuck.

 

[furballi]

This is one of my favorite board. It's cheap and very "hardy". It doesn't like to run two sticks of Kingston ValueRAM at 1T timing. But then that's a minor complain considering the dirt cheap price.

All the capacitors on the board are rated at 105C. Most cheaper boards (like e-machine) combine 85C and 105C capacitors. ECS also makes a NForce3-A939 version of this board. I suspect it will perform equally well with the dual core/dual channel A64 processors.

 

[mikeford]

Topic Title: Frys: ECS nForce3-A board & Sempron 2800+ $69.99
Topic Summary: SoCal ad., 4/12&13 only
Created On: 04/12/2006 10:59 PM

http://forums.anandtech.com/messageview...atid=40&threadid=1841874&enterthread=y

 

[wazzledoozle]

Ive gotten two of the ECS NF3-250 boards/Sempron 3100 combos now, one for a friend to upgrade and one to upgrade my mom's computer. Both are 130 nm Paris cores with 64 bit extensions enabled (I have tested with Ubuntu 64 bit) The sempron in my mom's computer does 2.2 GHz with +1.7% vcore, and it could probably do more but no need

 

[furballi]

Can't push my 2800 on this NF3 board past 2.1GHz. The same board and RAM ran the 3100 at 2.51GHz.

 

[Zap]

Like lapdances, overclocking is YMMV.

 

[saabman]

Having a problem getting my Fry's combo deal (ECS NF3 +3100+) to power up with my Seasonic S12-330 PSU.

I only have the CPU (w/zalman 7000 hsf), memory (one crucial 512 DDR400 stick), and 20 pin + 4 pin PSU cables connected to the MB.

With only the + and - power switch jumper connected to the MB, I plug the PSU power cable in, switch PSU rocker switch to on, and the case power button does nothing (switching + and - power switch jumpers doesn't matter).

It almost seems like the PSU is dead, except it does power on with an older 20 pin Tyan MB as soon as I turn on the PSU rocker switch, but the MB jumpered case power switch doesn't seem to work.

I've also tested the ECS MB with an older PSU and it powers up the CPU fan correctly with the jumpered MB case power switch.

Any ideas what is going on?

 

[cvrefugee]

I had the same problem. What I did was disconnect the extra 4 pin molex and it booted up just fine.

 

[Zap]

I forgot which of you... but one had mentioned in another thread that they were plugging the +4 part of the 20+4 ATX connector into the 4 pin +12v and that apparently doesn't work. There is a dedicated 4 pin connector for that which isn't a part of the 20+4. Hopefully this info helps the "other" guy.

 

[GML3G0]

I'm considering this for a budget build. No overclocking. Is it reliable, or is a capacitor going to blow up on me?

Or am I better off going with the Biostar Tforce 6100 939 with just integrated graphics and an A64 3000+?

RAM or should I just get some Corsair value RAM?

And the Fortron AX450 will suffice?

 

[Zap]

The ECS board is fine for a budget build. No known capacitor issues.

Value RAM is always a "better" value no matter what else you get.

That PSU should be fine.

What will this budget build be used for? The Biostar board with IGP should be fine even for gaming (albeit at lower quality/resolution). If you want really cheap, then the ECS board with CPU is available for $80 at Outpost (as of today available again) and just add $10 for a cheap HSF and $20 for a cheap AGP card and you're set for a budget rig. The socket 939 setup would be closer to around $75 more, but would be "better." Thus, depends on how "budget" you want.

 

[saabman]

Zap,

Yea the 4 pin molex goof was mine. As I said in the other thread, I hadn't built a new system for the last 5 years and thus wasn't familiar with the breakaway 20/24 pin power connector on the new PSUs. It just seemed automatic to use the breakaway 4pin molex from the 24pin to plug into the 4pin receptacle on the MB right next to the 20pin MB connector (WRONG, because it's a PCI express power connector as you indicated).

After looking at the other PSU cables, I found the correct 4pin molex (with lock tab) the goes to the 4pin receptacle on the MB.

These forums are just great for issues like these. Thanks to everyone for their help, and this might just help the next person who makes this same mistake.

By the way, I'm using the ECS nforce3/AMD 3100+ combo for an HTPC (in a Silverstone LC17 case) running Linux FC4/mythtv and all seems to be going well so far (still working on ivtv installation).

 

[CorrND]

I just bought the Fry's ECS nForce3-A plus Athlon 64 3000+ kit for $119.99 + tax yesterday. Code on the CPU indicates it's a Venice 3000+. I'm a little disappointed that I'm buying into the soon-to-be-obsolete socket 754 but this deal was not to be passed up. By the time I'm ready for my next upgrade, I figure it will be a system overhaul anyway.

Bought the combo at 1pm, spent 4+ hrs troubleshooting what I eventually decided was a defective board. No big surprise, given ECS's track record! I can provide details if people are interested. I went back to Fry's for a replacement board (I love Fry's return policy/customer service) and the original mobo was definitely defective. I had the new system up and running by 7.

I still had a little excitement after WinXP was in and doing the first start-up. The system wouldn't get past the Windows splash screen and would simply restart. I was able to start in safe mode and it turned out that it was just in need of up-to-date nForce3 drivers. I burned the drivers and a copy of DirectX9 needed for the drivers onto a CD with another computer and installed them. No other hiccups to report. Still running stable today so my next step is to see what kind of OCing I can do.

 

[furballi]

AMD will phase out 939 BEFORE 754! Let use know how far you can push that Venice. I believe that chip has 10x multiplier for a 2.5GHz overclock with the BIOS. You can still push beyond 250MHz FSB with Clockgen in Windows.

 

[CorrND]

Originally posted by: furballi
AMD will phase out 939 BEFORE 754!

Is that true?! The only thing I can think is that this will allow them to produce ultra-low cost computers on the 754 socket and just move the high end/939 to AM2. I can honestly say that I don't need anything more than this 754 setup is providing me.

Originally posted by: furballi
Let use know how far you can push that Venice.

I certainly will.

 

[jk147]

A weird issue came up. While I can do 250 fsb on one stick of 512, I can't do it when both memory sticks are in there.. topping out at 230. Anyone exprienced this issue?

My setup:
Sempron64 2800
ECS Nforce-3a
512x2 Gskill 2.5 cas
9800 Pro

 

[furballi]

Use 166 memory divider. Two modules may be too much for the memory controller on the CPU to handle. Another possibility is that the 2nd stick of RAM is not very good. My CORSAIR VS can go up to about 240MHz FSB at 1:1 memory divider.

 

[CorrND]

Originally posted by: furballi
AMD will phase out 939 BEFORE 754! Let use know how far you can push that Venice. I believe that chip has 10x multiplier for a 2.5GHz overclock with the BIOS. You can still push beyond 250MHz FSB with Clockgen in Windows.

(Before I write, I should say that I haven't done any overclocking since I my Celeron 533A back in 1999, so I'm a little rusty and not up with all the latest settings.)

I let the CPU burn in for a couple days to make sure everything was stable before doing any overclocking. Over the weekend, my Venice 3000+ was able to hit 245MHz FSB / 2450MHz CPU. I was only using the +1.7% voltage setting but upping it to +3.4% and +5.1% to hit 250MHz FSB wouldn't even get it to POST. I'm guessing it's the RAM (1 x Corsair Value DDR400 512MB), but upping the voltage on the RAM didn't help either.

2 Questions:

1. In terms of performance, is it worth reducing the RAM frequency to 166MHz to gain a couple hundred MHz on the CPU? Or will I be crippling the ability to get data into the CPU and negating the increased CPU performance?

2. What is a safe max operating temp for an Athlon 64? Running Stress Prime at 2450MHz, the CPU temp stabilizes at 49C, which I'm guessing is well below the max.

 

[furballi]

You're at the limit of the Corsair VS RAM. Suspect you may gain another 100 to 150MHz core speed with 166 divider.

1. 166 divider will cost about 50MHz core speed. Therefore, if you can increase the FSB by at least 5MHz with 166 divider, then you will gain MORE SPEED.

2. The upper limit of the CPU is around 50C, +/-5C, at 75F room temperature. Bumping Vcore up by +3.4% is SAFE!

 

[CorrND]

Thanks for the advice! I've got some more questions about a RAM upgrade. My brother and I both have 512MB of memory right now and we're thinking about buying 1GB and splitting it, to save a couple bucks. As I said, my memory is 1 x 512MB Corsair VS DDR400, which I think is 2.5-3-3-8. His is 2 x 256MB OCZ Enhanced Latency DDR400, which I think is 2-3-2-6. I want to get more Corsair VS, but I've got a couple questions:

1. If we get a 2x512MB kit, my brother's timings will have to go down. I'm sure that affects performance to a certain degree, but will it be significant? I mean, will he really notice the difference?

2. An extra 512MB stick will fill all 3 slots on my brother's Abit nForce2 motherboard. Is that going to be a problem? I remember reading a while back that the nForce2 can be pretty picky about memory, especially with all 3 slots filled. Would he be better off ponying up some extra cash and getting a stick of 512MB OCZ with matching timings?

3. If mixing 512MB Corsair VS with my brother's OCZ is going to be fine, I could alternatively buy 1 stick of 1GB Corsair VS for myself and give him my 512MB stick. It looks like Corsair's Value Select 1GB is CAS 3, though. Will I see a real-world performance difference going from CAS 2.5 to 3? On the other hand, would I be better off in overclocking having 1x1GB instead of 2x512MB?

Thanks for the help so far!

 

[furballi]

1. The Corsair VS should run at 2.5-3-3-6-1T. The drop in speed can only be measured with benchmark applications.

2. Some boards may complain with three modules. Plus, the Corsair may not get along with the OCZs. It may be necessary to drop the timing from 1T to 2T. Obtaining the same part number from the mfr (different lot #) not guarantee compatibility.

3. Again, the speed drop is small (about 20 to 30MHz core speed). I doubt that there will be any noticeable drop in performance. The Corsair VS 1GB module may come with Samsung UCCC chips. This stick should be able to run at 2.5-3-3-8-1T with a good board and CPU.

There will be less stress on most systems with a single 1GB module in DIMM 1. This may translate to higher overclock speed.

 

[CorrND]

Thanks again for the answers! This really helps a lot. One more clarification question:

1. My brother's computer is running a Barton XP 2600+ with no overclocking, so the FSB is only 166MHz. Will that increase the odds that mismatched memory will work together? Or does mismatched memory not work together regardless of bus speed?

 

[Zap]

With running memory at SYNC speeds on a lower FSB, I'm thinking that will increase chances of the stuff working together. Here's my suggestion for you and your brother. The stick of RAM you have should go to your brother. IIRC Nforce2 can still run dual channel with 256x256x512. For your system either a 1GB dual channel kit or a single 1GB module. Since you are on single channel and thus negating any benefits of having two sticks of RAM, perhaps go for a single module. That way you still have an upgrade path in the chance that 1GB ends up not enough some time in the future before you upgrade the board.

As for your overclock and memory settings, the Corsair RAM is supposed to run at 400MHz. How that works is this:

DDR400/PC3200 = 2 x 1/1 x HTT (BIOS setting of 200MHz RAM)
At default clock speed of 200MHz HTT, 200 times 1 divided by 1 times 2 (for DDR) = DDR400/400MHz.

DDR333/PC2700 = 2 x 5/6 x HTT (BIOS setting of 166MHz RAM)
At default clock speeds of 200MHz HTT, 200 times 5 divided by 6 times 2 (for DDR) = DDR333/333MHz.

DDR266/PC2100 = 2 x 2/3 x HTT (BIOS setting of 133MHz RAM)
At default clock speed of 200MHz HTT, 200 times 2 divided by 3 times 2 (for DDR) = DDR266/266MHz.

What happens when you overclock? You just use the same formula...

245MHz times 1 divided by 1 times 2 = 490MHz. Wow, that's a heck of an overclock for that RAM.

If you change your RAM setting to 166MHz you'll have 245 times 5 divided by 6 times 2 = 408MHz. That's 8MHz more than the RAM is supposed to run so not only is it running closer to spec (meaning it's no longer holding you back) but it's still overclocked a bit so you do benefit.

 

[CorrND]

Thanks for the advice! Now for an overclocking report. Turns out that the RAM divider wasn't the limiting factor. By turning HT down to x3 (from default x4) I was able to hit the BIOS maximum of 250MHz FSB, amazingly while keeping the RAM 1:1 with the FSB. After I hit that, I went to clockgen. Here's a chart of what I was able to hit with various BIOS settings:

FSB MEM HT Vcpu
---- ----- --- ------
245 200 4x 0 (default, 1.4V?)
253 200 3x 0
261 166 3x 0
267 166 3x +1.7%
273 166 3x +3.4%

The 253 setting is truly amazing -- the RAM is operating at 506MHz. Corsair VS is incredible! I didn't do any stability testing at any of those settings. Is it safe to go to +5.1%? I'm all for getting some free performance, but I'd rather not take a risk of frying the CPU. Any other recommended setting tweaks I should try?