AN35N Ultra FSB speed affects XMS3200?

Hi,

I my mobo is running the latest BIOS and the latest hardware drivers (nVidia nForce2 400 chipset drivers, etc.). WinXP is fully updated, and the system worked adequately with the Corsair VS 512MB modules and the Athlon XP 2200+ processor.

I needed to upgrade my system (bite tongue) so I purchased a 3200+ and 2GB of XMS memory (trying not to repeat what is in my profile... ). I am posting from the upgraded system right now - BUT the upgrade process is not done. Here is a quick synopsis of what I submitted to Corsair:

If I set the FSB to 200 in the BIOS the mobo identifies both the processor and the memory speed correctly, but the OS does not successfully load and the system reboots when the black XP screen tries to switch to the (light) blue XP user login screen.

If I set the FSB to 166 in the BIOS the mobo identifies the (same) processor as an Athlon XP 2500+ and the (same) memory's speed as 333MHz, but the OS does successfully load - and here I am posting.

I should add that I have let the mobo fully test the memory (disabled quick power on self test) - 2GB does take awhile - and the mobo thinks its "OK". And when XP was failing to load I tried the XP install disc and - THIS IS KEY - XP could not load the drivers into memory and gave me the dreaded blue screen of death. When XP loads drivers from the CD they are loaded into memory so I am guessing there is a problem with the AN35N Ultra using the XMS3200 modules at 200MHz FSB.

I've been reading a bit more, and supposedly Corsair's XMS memory is supposed to work with Shuttle's AN32N Ultra. Even though nVidia doesn't list it as compatible with the nForce2 400 Ultra chipset, a fella posted that Shuttle said they tested XMS and it worked (I know, heresay... ).

There are some current threads on XMS3200 memory having problems with FSB400 (200mhz) and being fine at 166mhz, and others have said that dual channel is only a modest performance gain, so perhaps I will have to try the 1GB DIMMs in slots 1&2 to drop dual channels.

My other thoughts were to manually set the BIOS, with the stock settings I have read here. Maybe I can overcome Shuttle's inability to properly autosense components (like the processor biz).

My last thought it I am not sure if this is a "matched pair" of DIMMs, suitable for dual channel. Could that make the combo unstable at higher FSB speeds?

Guaranteed-Compatible Memory for your Shuttle AN35N

FYI some of their older DDR 1 products may not be listed on that because they don't make them anymore.

Thanks for that, and I am guessing that my "CMX1024-3200" DIMMs are not the same thing as "TWINX2048-3200C2". Maybe my DIMMs are not a matched pair and I should only run them in slots 1&2? Strange thing is Shuttle directs AN35N users to nVidia's website for nForce2, which does not list any Corsair compatibility.

Maybe some details might help: CPU-Z says this about the Memory Timings:

DRAM Freq 167

FSB:DRAM 1:1

CL 3.0

tRCD 3

tRP 3

tRAS 7

and I thought this memory's SPD was set to 3.3.3.8 (CPU-Z SPD tab says so too)? I am wondering if the AN35N is not reading the SPD correctly; other things are odd to me: should a processor be identified as a different "model" depending on the FSB setting in the BIOS?

FSB set at 166MHz and processor is called an AMD Athlon XP 2500+

FSB set at 200MHz and processor is called an AMD Athlon XP 3200+

Seems like it should be an XP 3200+ no matter what the mobo FSB is set at, but I am not sure...

Ah, so the CMX1024-3200 DIMMs are not matched eh? Shame, but maybe I should run them in slots 1&2 so they are on the same memory controller and not running "dual channel". I've read its a 5% performance loss, but if I can run the FSB at 200MHz, the processor runs as a 3200 instead of a 2500, and the memory would be 400 too - could be better than Dual Channel eh?

Corsair Tech Support recommended the memory voltage be set manually to 2.7V. I am trying that and FSB 200MHz before I bother them again. Just not sure why the mobo sets the memory timings differently than the SPD.

My mobo (Shuttle AN35N Ultra - nVidia nForce2 Ultra 400 chipset) has the following choices:

FSB Frequency: when I set it to 200Mhz XP wouldn't load, but at 166MHz it works fine. I was going to get the memory sorted and then try to switch this back to 200...

CPU Interface (optimal/agressive): "use most stable or use overclocked CPU/FSB parameters" I am not sure what this means.

Memory Frequency: "select DDR SDRAM frequency" By SPD, or 50-200% & it tells you the "Resulting Frequency"

Memory Timings: the board was setting these to 2.5, 3, 3, 7 (DIMMs are marked 3, 3, 3, 8) so I wonder if this mobo can't read the SPD... ?

FSB Spread Spectrum - grayed out

AGP Spread Specturm - grayed out

AGP Aperture Size: mobo chooses 64MB (maybe because of the Raedon 9800?)

AGP 8x support - enabled

AGP Fast write - enabled

System BIOS cacheable - disabled

Video RAM cacheable - disabled

CPU Thermal Throttling - 50% (set by mobo) I read its for processor activity in suspend mode.

I can also set the CPU Ratio and Voltage manually if need be; currently my 3200+ is sensed as a 2500+ when FSB is set to 166MHz.

I thought the model/part #s would differ between them.

OK, so I manually set the DIMM voltage to 2.7V as recommended by Corsair TS, and I also manually set the memory timings to 3,3,3,8 as marked on the DIMMs, but this did not make the configuration stable at 200MHz FSB. My next move is to move the DIMMs to slots 1&2, disabling Dual Channel to obviate the possibility that these DIMMs are not a matched set.

RAM GUY, shouldn't 3,3,3,8 be more stable than your recommendations or do I have that reversed?

One side note, I tried to up the CPU ratio to 14 so my 3200+ ran at 2200MHz when the FSB was set at 166MHz, but CPU-Z says the ratio is still at 11... and the mobo did not "sense" the processor as a 3200+ either. Odd, or is this some "feature"? :|

The processor's tag has AXDA3200DKV4E, which indicates it is a 3200+ Barton Core (2200MHz model). Could the tag be a fake? I guess someone could glue the tag on a 2500+, but then why would the BIOS show it as a 3200+ when the FSB is set at 200?

http://en.wikipedia.org/wiki/List_of_AMD_Athlon_XP_microprocessors

I thought that if the FSB is running at 166 I could bump the multiplier to 14 so the processor could run at 2200MHz - but CPU-Z showed the multiplier was still at 11X. I am tinkering with this bit while I try to get the memory working at 200FSB. Again, when I set the FSB to 200 the processor is sensed as a 3200+... strange that the FSB would change the processor model.

When I get a chance to swap the DIMMs to slots 1&2 I'll get back with results.

EDIT: one thing RAM GUY - my modules are marked "CMX1024" but you have never called a module that (I searched with your user name for CMX1024); the only 1024s are TWINX; all the CMX have lower numbers, like 512... can you help me out with that?

Here is what's on the DIMM tags:

CMX1024-3200.................................XMS3200v1.3

...........................................................0614343-1

XMS3200............1024MB........400MHz........3.3.3.8

I suppose this is them: http://www.corsair.com/_datasheets/cmx1024-3200.pdf So the DIMM tag doesn't state the Part Number per se... okay.

The other module has the same "lot code" 0614343-1; Corsair TS said the first 4 numbers will be the same on matched DIMMs. Well I tried slots to make the memory run single channel just to check and it didn't help: still will not load WinXP (auto reboot). I will be testing individual DIMMs as requested by Corsair TS.

So I tested the DIMMs individually as requested, in both memory channels (1/2 and 3) and got mixed results. I began to think that they were both not going to work, which usually indicates a problem with other components/setup. But I did get one module to boot XP at 200MHz FSB and run for awhile.

Still unstable but I need to try and be systematic with the BIOS changes and see if I can pinpoint the issue. One thing that bugs me, based on some reports that the AN35N boards undervolt components, is CPU-Z reports the processor is getting less than 1.65V and I set it to give 1.65V. It fluctuates between 1.62 and 1.65, but since settings are available by 0.25V increments I am guessing that is significant variability. I tried upping the CPU voltage but got instability again - so I'll be doing additional testing.

It's annoying to not be overclocking but having to do the OC song and dance...

Not another system, but I do have another AN35N Ultra 400 mobo that I could swap in - I also have the old Corsair VS memory (2x512MB DIMMs) that I could swap in and try running at 200Mhz FSB (but I have to check it's ratings).

Thanks for the reminder; I'll do some swapping, but family vacations are looming in the way.

Not another system, but I do have another AN35N Ultra 400 mobo that I could swap in - I also have the old Corsair VS memory (2x512MB DIMMs) that I could swap in and try running at 200Mhz FSB (but I have to check it's ratings).

 

Thanks for the reminder; I'll do some swapping, but family vacations are looming in the way.

Enjoy the vacations and let us know how it goes when you are able to do the troubleshooting.

Well, I would like to test more configs, but Corsair has asked me to return them for replacement. They have been gracious enough to answer my ceaseless questioning... so I am not going to ask them what I said made them think it was the memory.

It could be that an RMA was easier/cheaper than working with me :cool:

P.S. Yellowbeard's one of my favorites, and finally available on DVD via Netflix! Mr. Prostitute... don't laugh!!!

Well, I would like to test more configs, but Corsair has asked me to return them for replacement. They have been gracious enough to answer my ceaseless questioning... so I am not going to ask them what I said made them think it was the memory.

 

It could be that an RMA was easier/cheaper than working with me :cool:

 

P.S. Yellowbeard's one of my favorites, and finally available on DVD via Netflix! Mr. Prostitute... don't laugh!!!

Trivia question; what was Mr. Prostitute in charge of :laughing:

As far as replacing your modules, we are interested in keeping you as a long term customer. At this stage, it makes sense to replace the modules and go from there.

hello, unless your barton is unlocked or a mobile CPU you cannot go up with the multi, only down.

you can either paint the bridges with rear window defogger paint (read a guide on it, it ain't that easy or just buy a 2600M barton. its unlocked and clocks great on the cheap.

Trivia question; what was Mr. Prostitute in charge of :laughing:

 

As far as replacing your modules, we are interested in keeping you as a long term customer. At this stage, it makes sense to replace the modules and go from there.

I have always heard good things about Corsair and thought folks held them in high regard, so it is strange that nVidia didn't get Corsair on their compatability list for the nForce2 400 chipset. I hope that is not the problem but the Corsair tech working with me thought not, as have others who have posted here or there on the web.

Mr. Prostitute's duties were, ahem, not appropriate for discussion here... but I have failed the forum Trivia :sigh!: I can't remember any stated duties.

Synthohol, I have an actual AMD Athlon XP 3200+ Barton (PN AXDA3200DKV4E), unless someone was cheeky enough to put a different sticker on it. I am not trying to overclock anything; just trying to run them at their rated speeds (200MHz FSB). Since you seem to know AMD chips, riddle me this: why does Phoenix (CMOS) and CPU-Z see the CPU as a different model depending on the FSB settings of the MOBO? I guess it makes sense, but not really since other things (like Corsair's memory) are identified properly no matter what the FSB is set to in CMOS.

 

Mr. Prostitute's duties were, ahem, not appropriate for discussion here... but I have failed the forum Trivia :sigh!: I can't remember any stated duties.

:o: LOL...I know not to bring up anything inappropriate here. Mr. Prostitute was in charge of DISCIPLINE! As was every officer on the boat. They ran a tight ship such as when they nailed that guy's foot to the deck ::pirate::

riddle me this: why does Phoenix (CMOS) and CPU-Z see the CPU as a different model depending on the FSB settings of the MOBO? I guess it makes sense, but not really since other things (like Corsair's memory) are identified properly no matter what the FSB is set to in CMOS.

ya know that is quite the funky bug.

back when i was painting bridges with conductive paint i ran int the same phenomenon without touching the BIOS. i was having difficulty getting all 5 bridges connected (even with my loope my eyes were not playing nice) and each time i installed the chip to test a different model was displayed. anywhere from a 1500+ to a 2000+. btw the chips original model was a 1700+ tbred. when the bridges were correct it should display a 2100+

it seemed that depending on the fsb reported by the chip itself regulated its model # to the BIOS. also there was a switch on the board that had t/bred-duron and you had to select duron if it was a t/bred with the bridges fooled with. that also tricked the BIOS for a reason i never looked into.

well thats my personal hypothesis anyway.

p.s. i wound up just replacing the chip with an unlocked 1800+

I got the memory back from Corsair (long time ago of course) - thank you Corsair for doing whatever you could for me. UPS did their best to destroy the shipment (the box was shredded and the buggah still left it on the porch and drove off... pathetic) but Corsair did a good enough job of packaging it so it survived the ordeal.

The new memory seems to work fine, but my mobo/cpu still cannot handle running at full speed (again, not overclocked). So perhaps the CPU is a dud, and the reason why at 200mhz FSB it crashes after fully booting into XP, or perhaps it is the Shuttle AN35N Ultra mobo problem running at full speed. It is stable at 166mhz FSB, which makes the memory run at 333 (dual channel) and the CPU register as at 2500+ instead of the 3200+ it is supposed to be...

argh, it is stable at that config! Not a hiccup with anything I do. But it just doesn't seem right to have to throttle it back. So my next move is to look for a better Socket A mobo that works with my stuff.

It could be the CPU, but I am not sure how to test that, and much of what I have read on the Shuttle mobo is that it is not necessarily stable at the specs they list for it... so be it.

Thanks wired for the edit - and I'll try to remember.

So I have abandoned Shuttle's AN35N Ultra and I am about to install the following MOBO:

DFI LanParty NFII Ultra B

eBay UK had one (finally). Some say it might have been the finest Socket A mobo produced. I am excited about hardware RAID - I hate manual HD mirroring. And it is a SATA board, so I got a new HD - 1TB (32MB buffer, 7200rpm) for 90 greenbacks. Hope wired don't have to edit this :)

When I can around the no floppy drive issue (installing SATA drivers during XP installation - I have a method to test) then I will get this baby rebuilt and update this thread with a 400mhz FSB test of the new memory Corsair sent and the processor, FINALLY.

After further testing and researching I've come to the conclusion that the issue is AMD's Athlon XP 3200+. I don't think there is anything wrong with this specific processor, meaning it has not been broken since manufacture, but rather that the entire XP 3200+ model line is probably flawed.

I have been reading some of the forum chatter at DFI's website for my mobo and the consensus is (with the forum demagogues over there) that this particular AMD processor does not play well with the DFI mobo. Since the DFI mobo is held in high regard, very high regard, and this is the second mobo to not play well with it, then I am lead to conclude that there is something about the XP 3200+ line that is not stable.

So that means I owe Corsair an apology, not for lying about my RMA - I tried to be forthright with them but they wanted to make sure it was not their memory (very impressive customer service!) - but that I even bothered them in the first place. In my defense it is not something that is easily diagnosed, since there is no way to specifically test the processor and eliminate issues that could belong to the memory.

The final bit I have here is I have to run the XP 3200+ as an XP 2500 (throttled back) for stability. That sucks since I was trying to eek out whatever life was left in my Socket A hardware before succumbing to the ever-rising tide of software needs. Fortunately for almost everyone out there this is a non-issue - likely nobody has a Socket A any longer, or would be looking at an XP 3200+... but perhaps it is a decent technical note: processor issues can appear to be memory issues, and I do not know a way to isolate the two.