AN35N Ultra -- how to tell if DC enabled? Also, Geil RAM timings?

[Matthias99]

Hi. I apologize if this has been asked 20 million times before, but I tried the search function and nothing turned up. I've emailed Shuttle support, but it's been a few days and I haven't heard anything.

I've got a new AN35N Ultra-400 motherboard, and two 256MB sticks of Geil PC3500 (DDR433) RAM. The RAM is in slots 2 and 3 (I also tried 1 and 3). There's also a Barton2500+ (locked) with a Vantec Aeroflow stuck on there, if anyone cares.

Plugged everything in and it fired right up. Went into the BIOS and looked at the settings -- the SPD picked up the RAM as 400Mhz, 2.0-3-3-6. That's correct, and I didn't see anything in the BIOS about dual channel, so I let it boot. Now, I know some motherboards give a specific message during the POST that dual-channel is enabled. However, I don't know if the AN35N-Ultra is supposed to or not! Mine just says "Memory at XXXMhz" (whatever speed I have it set at). SiSoft Sandra and MemTest86 report memory bandwidth that seems correct (over 1000MB/sec, in line with the DC NForce2s in Sandra), so I guess it's working. I was just wondering if anyone could verify that it's set up right.

Also, while the system was capable of running the full Memtest86 test suite overnight without a hitch, and 3DMark01/03 seemed fine (once I sorted out my motherboard drivers, that is; took me a while to figure out why I was getting *lower* performance than on my old KT133A), it bombed out of Prime95's "blend" torture test in just a few minutes at DDR333/2.0-3-3-6. The in-place tests (which don't stress the memory) work fine, so I'm pretty sure something's a little off in the RAM. I know sometimes NForce2 boards sometimes need a higher, uh, CAS-to-RAS delay (or whatever the last number is; I can never keep them straight), so I left it running this morning at 2-3-3-11 to see if it would work. I'll have to try playing around a bit more and/or trying each stick separately if that doesn't hold up, I guess. Hope I didn't get a dud stick, because this stuff was a great deal, assuming it runs as advertised.

It's this RAM: http://www.newegg.com/app/ViewProdu...144-514&catalog=147&manufactory=BROWSE&depa=1

Anyway, has anyone used this RAM in an AN35N-Ultra? If so, what settings did you find best? Should I try the voltage at 2.7? Thanks in advance,

 

[Balrog]

Welcome to Shuttle Forum Matthias! Yes one of the things that Shuttle skimped on was the little notification in BIOS of Dual-Channel mode, but you are correct and running dual with memory in slots 2 & 3. I use Buffalo/Winbond PC3200 personally, but I know that many people have problems running CAS2 on this board even if the RAM supports it. Often it will run CAS2 later after the RAM and MoBo have been burned in but to start with try 6-3-3-2.5 or 7-3-3-2.5 to get things running smooth, then tighten them down later. Good Luck!

Regards, Balrog

 

[Noodles]

I had the same problem as you. I wanted a message that said it was working like it should. Download cpu-z. This program will tell you if you are running in dual or single channel mode under the memory tab.

See if you can get the ram running at tight timing on another board if thats an option. Otherwise, the board is notoriously picky about ram. Burning it in might help like balrog said, I have no idea. I run my buffalo at default timing currently. I might try tightening it up a little later when I do some more upgrading.

 

[Matthias99]

Thanks for the quick reply and the info. DC seems to be working just fine; it's just baffling that they didn't have any indication of its state. It seems like it would take all of five minutes to add that to the BIOS display on startup.

Anyway, I found that 2-3-3-6 was slightly unstable, but 2-3-3-11 seemed to work just fine at 333. Now I'm working on OCing the FSB -- I've got it running memtest today at 2.0-5-5-11 at 200FSB with 2.7VMem, and it seems pretty stable (passed an overnight run of Prime95 at max priority at those settings). 3DMark01 was giving me some slightly wacky behavior (once it ran the test but then wouldn't produce a score, and another time it produced a score but ORB refused to take it), but I've never tried running it on a heavily OCed CPU, so I'm not sure what to expect here. I'm not complaining about picking up 3000 points, though. I'll have to try looping it overnight to check for stability, I guess. 3DMark03 seems fine, and I went from about 5800 to 6100 without changing my video settings!

I feel like I'll probably need more VMem to go much higher -- these chips are specced to 2.9V. I'm not sure I'm up to volt-modding my brand new motherboard yet, though -- I think I'll wait for it to burn in and see if anything improves.

 

[Noodles]

Try running auto voltage. Balrog seemed to get more stability using that then 2.7v. I'm not sure, but the auto voltage may allow for higher voltages than the bios has options for.

 

[drshivas]

Welcome to the Shuttle forum!

You'll also want to mod your NB and SB. This will give you more FSB stability.

I, like many here, took off the NB heatsink and replaced the thermal pad with some AS3 or some paste- after lapping. Barlog has detailed post around here somewhere about the lapping. He also put a fan on the HS as well.

And I put a passive heatsink on the SB with thermal tape-that little guy gets REALLY hot.

I too have the Buffalo WinBond 3200 CH-5 chips (2x256) in DC mode. I am at 2.5.3.3.8, running a 2100+ @11x200 8.25vcore- (got one of the last guaranteed TbredBs from newegg too).

And ditto on setting the vdimm to auto. It *seems* to have worked for me, albeit I am not OCing that far.

 

[Matthias99]

Well, I may get around to adding some improved NB/SB cooling, but thus far I've just been trying to find my current limits.

I'm still at 200FSB, 2.0-5-5-11 -- any tighter on the timings at CAS Latency 2.0 and Prime95 chokes. I'll have to see if I can get, say, 2.5-3-3-8 running and compare them in the benchmarks. I may also try pushing it to 210FSB or so (I have a locked CPU, so I can't push the FSB much past 200, I'm guessing).

 

[Matthias99]

Update:

Didn't see any differences in the Auto/2.7V memory settings. At least it didn't improve my overclock any; I'm not pulling out the voltmeter and poking my motherboard's memory sockets yet. I can't get under 2.0-5-5-11 @ 200FSB (even 2.0-4-4-11 dies in prime95), although 2.5-3-3-X works fine. The CL2.5 settings give me a higher memory bandwidth (according to SiSoft Sandra), but my performance is lower in all my real-world benchmarks and games. CL2.0 it is!

I'm currently testing some 210FSB settings. Hopefully I can squeeze a few more Mhz out of everything in my box.

Also, I can't get anything to work right with the "Aggressive" CPU Interface settings on (at least at or above 200FSB) -- anyone know anything about those?

 

[Balrog]

CAS2.0 @200FSB is good but I've found CAS2.5 @220FSB to be better in all benches.

Regards, Balrog

 

[drshivas]

Try setting everything to "Expert." Even though those settings (optimal, aggressive, turbo) are templates for suggested settings, the expert mode is the only one that lets you control all settings. I suspect that any mode other than expert does something behind the scenes.

And agreed with Barlog. Higher FSBs are more valuable than lower latencies. A lower latency at the same FSB is of course better, but once you relax those latencies, you can free your FSB to go higher.

Some of the super fast PC3400+ memory has timings of 3-4-4-8 and higher. Catch the drift?

 

[Matthias99]

I'm not sure if my proc will run at 220FSB -- that's 220x11 = 2400Mhz? For a Barton 2500+ On air cooling? Eh...

The only settings for "CPU Interface" are "Optimal" and "Aggressive". It doesn't say much about it in the manual or BIOS screens other than that "Optimal" uses "default" values and "Aggressive" uses "over-clocked" values for the CPU interface. Very helpful. My memory timings are already on Expert; that was the first thing I did after installing the board and firing it up.

And yes, I realize that a higher FSB with looser timings will probably be better overall than a lower FSB with tighter timings. However, I was being advised to run 2.5-3-3-8 above, and I was merely reporting that I could successfully run 2.0-5-5-11 at DDR400, which gave me better performance overall.

 

[drshivas]

IC, you're right. The CPU interface only has aggressive and optimal. And it would be nice if they listed what the actual differences are. I remember while OCing my Epox, the FAQs said to set it one way or another. Not sure if it applies to all NForce2 boards though.