MSI K7N2 & Amd Athlon XP 2600+ (Barton) = unstable?

Hi there all ,

I have this problem with my current setup.

First some specs :

Motherboard : MSI K7N2 (MS-6570) 333FSB Max
Memory : 2 x 512 PC27000 333MHz
VGA card : 1 x Aopen Aeolus GeForce4 Ti4200 AGPx8
Sound card : on board disabled. Philips Acoustic Edge PCI
Controllers : NForce2 on board , Adaptec 2940 + BIOS PCI
Harddisks : 1 x IBM 40Gb , 1 x Maxtor 40Gb , 1 x Maxtor 80Gb All are ATA100
DVD Player : Toshiba M1502
Floppy disk drive : Samsun 1.44mb
Rounded cables , 4 fans in total (including cpu fan)
Processor : 1 x Amd Athlon XP 2600+ Barton core (166FSB)

The thing is … when i install this pretty cpu in my sweeet configuration shown above… it works.

As long as i put the FSB on 133MHz (making it a Athlon 2000+ in comparison). Windows XP SP1 works flawlessly. Dos boots , everything is happy in paradise.

When i put the FSB on 166MHz in the Setup (some people would say the bios/cmos) , Windows XP SP1 produces blue screens , Dos fails after 60 seconds , hangs in the setup , and what completely baffled me :

The Windows XP bootable setup CD … BLUE SCREENS !

The cpu has been tested by the supplier and he claims it works perfectly. MSI tested a reference motherboard for me with one of their own Amd Athlon XP 2600+ Barton cores and claims it works perfectly.

I’ve changed VGA cards , swapped memory , removed disks , put fans on the PSU instead of the motherboard (high power consumption). Updated the bios to version 3.4 (latest since couple of days) and nothing aids in getting rid of this problem.

Can anyone help me with this very weird , yet highly annoying problem ?

Have you tried running Memtest86 at FSB166? This sounds like a memory problem. I’ve got my Corsiar XMS PC3200 running at 185 FSB, I’ve heard a lot of complaints about memory instability with some of the 333mHz boards.

It’s not likely a memory problem, cos it works when the CPU is 133MHz.

The Ram speed is asynchronous (166) to the CPU’s (133/166)FSB.

I don’t suppose the heatsink/cooler is perhaps a little under spec? or worse, maybe u used a heat sink compound which isn’t upto the job?

The newer athlons really need arctic silver, or similar ripoff.

Arctic silver does 9W of heat / square cm, I think.
Dodgy ripoffs approximately 7.5W/ sq cm.

Heatsink compound from tandy/dick smith/jaycar/radioshack approx 1w/sq cm.

Just another thought. Your board doesn’t have some sort of jumper to set the default divider for the CPU’s front side bus?

I remember my little Gigabyte 7VRXP had a jumper to set the default CPU FSB as 100/133.

I could run the memory speed independently at 100/133/166

The CPU though had a default of 100 when the board was shipped and refused to work at it’s proper 133 (athy 1800) even after setting it to 133 in the bios.

Then I found the jumper, switched it to a default of 133 and everything is a soruce of joy again :slight_smile: And then bumped it up to 145 :bigsmile:

Just a quirk in the design I suppose.

But it’s a via chipset.

not likely a memory problem, cos it works when the CPU is 133MHz

Actually, if you read the AMD forums, there are an abundance of reports of just this problem. Not all DDR is created equal, and some boards demand that you use only the best. Using 400 mHz rated DDR on a 333 board has solved quite a few stability problems. Reading the ASUS forums, I see a high percentage of the stability problems resulting from use of PC2700 DDR.
Belvadere’s symptoms are pretty typical for memory problems.
Mr B, see if you can borrow some high quality PC3200 to test with, but first try the PC2700 in Memtest86 at 166 FSB (not asynch). If you want to improve CPU stability, I suppose you could lower the multiplier at 166 to 12. The Barton should be stable to at least 185x12 with the right DDR.
I’m not familiar with that board, but are there any PCI or AGP settings that would need to be adjusted with changes in FSB?

My mainboard is pretty much the same (MSI K7N2G-ILSR) and doesn’t have any problems my “cheap” memory, but according to quite some forums, memory is a problem on the Nforce2 boards. Some even suggest that the NForce2 chipset overclocks memory by default, to gain a higher performance. That shure would explain the compatibility issues with cheaper memory, and the fact that 400Mhz memory runs better…

But ok, we can test this. If you set the BIOS to manual memory settings and asycn CPU-MEMORY mode (get rid of the 1:1 setting), you can put your memory on any speed you want (only use certified speeds, because values in between can be the cause of many errors); I’d say 133Mhz. You could try to raise your latencies as well (CL4 or so). Test it…

Another thing that comes to my mind: do you have a PSU with the additional power cable (originally intended for the P4)? The mainboards need this to provide enough power…

I do assume that your cooling, as well as your powersupply doesn’t have a lack of capacity… right?

I wouldn’t actually run the DDR in asynch mode due to the performance hit, but it’s a good way to test it out. Set the CPU at 166 and set the DDR at 80% just to see what happens. Lowering the timings would be good too, for a start.

Originally posted by rdgrimes
Have you tried running Memtest86 at FSB166? This sounds like a memory problem. I’ve got my Corsiar XMS PC3200 running at 185 FSB, I’ve heard a lot of complaints about memory instability with some of the 333mHz boards.

Haven’t tried it. I will let it run this evening , if the thing will even start at 166MHz. Seems on autodetect that the DDR ram is 333

Originally posted by debro
[B]It’s not likely a memory problem, cos it works when the CPU is 133MHz.

The Ram speed is asynchronous (166) to the CPU’s (133/166)FSB.

I don’t suppose the heatsink/cooler is perhaps a little under spec? or worse, maybe u used a heat sink compound which isn’t upto the job?

The newer athlons really need arctic silver, or similar ripoff.

Arctic silver does 9W of heat / square cm, I think.
Dodgy ripoffs approximately 7.5W/ sq cm.

Heatsink compound from tandy/dick smith/jaycar/radioshack approx 1w/sq cm. [/B]

The CPU is 34 degrees (Celcius) , don’t think it’s a cooling problem , but i really appreciate the input. It’s a brand new cooler and i’ve monitored it on both the 166 and the 133 FSB. No real differences.

Originally posted by debro
[B]Just another thought. Your board doesn’t have some sort of jumper to set the default divider for the CPU’s front side bus?

I remember my little Gigabyte 7VRXP had a jumper to set the default CPU FSB as 100/133.

I could run the memory speed independently at 100/133/166

The CPU though had a default of 100 when the board was shipped and refused to work at it’s proper 133 (athy 1800) even after setting it to 133 in the bios.

Then I found the jumper, switched it to a default of 133 and everything is a soruce of joy again :slight_smile: And then bumped it up to 145 :bigsmile:

Just a quirk in the design I suppose.

But it’s a via chipset. [/B]

It has indeed such a jumper , 1-2 = 100 MHz , 2-3 = 133/166MHz. It’s jumpered correctly , otherwise the cpu wouldn’t even start at 133MHz :slight_smile:

Originally posted by rdgrimes

Actually, if you read the AMD forums, there are an abundance of reports of just this problem. Not all DDR is created equal, and some boards demand that you use only the best. Using 400 mHz rated DDR on a 333 board has solved quite a few stability problems. Reading the ASUS forums, I see a high percentage of the stability problems resulting from use of PC2700 DDR.
Belvadere’s symptoms are pretty typical for memory problems.
Mr B, see if you can borrow some high quality PC3200 to test with, but first try the PC2700 in Memtest86 at 166 FSB (not asynch). If you want to improve CPU stability, I suppose you could lower the multiplier at 166 to 12. The Barton should be stable to at least 185x12 with the right DDR.
I’m not familiar with that board, but are there any PCI or AGP settings that would need to be adjusted with changes in FSB?

Hmm , this is indeed a thing to remember ; I’ll have to look in to that.
Thanks for the info. :bow: Do you happen to have some weblinks where people describe this solution thoroughly ?

Originally posted by Dee-ehn
[B]My mainboard is pretty much the same (MSI K7N2G-ILSR) and doesn’t have any problems my “cheap” memory, but according to quite some forums, memory is a problem on the Nforce2 boards. Some even suggest that the NForce2 chipset overclocks memory by default, to gain a higher performance. That shure would explain the compatibility issues with cheaper memory, and the fact that 400Mhz memory runs better…

But ok, we can test this. If you set the BIOS to manual memory settings and asycn CPU-MEMORY mode (get rid of the 1:1 setting), you can put your memory on any speed you want (only use certified speeds, because values in between can be the cause of many errors); I’d say 133Mhz. You could try to raise your latencies as well (CL4 or so). Test it…

Another thing that comes to my mind: do you have a PSU with the additional power cable (originally intended for the P4)? The mainboards need this to provide enough power…

I do assume that your cooling, as well as your powersupply doesn’t have a lack of capacity… right? [/B]

So not the 1:1 setting , but which setting then ? I’ve tried 3:2 or something and it completely froze. I had to reset the bios using the motherboard jumper to get it to work again when i tried that.

What’s a safe factor to use ?

There’s a P4 certified dual fan 350W installed

Originally posted by rdgrimes
I wouldn’t actually run the DDR in asynch mode due to the performance hit, but it’s a good way to test it out. Set the CPU at 166 and set the DDR at 80% just to see what happens. Lowering the timings would be good too, for a start.

80% … i can only test it on some fsb/cpu factor. your recommendation would mean a 10:8 ratio ?

A big big thanks to ALL of you :bow: :bow:

The 2 x 512 MB DDR memory is now working at 133MHz (x 2 of course) and now the CPU is WORKING at a whooping speed of 2.083 GHz , making it a complete 2600+ Barton.

:bow: :bow: :bow: :bow:

I am a very happy person. Gone is the frustration and i thank you all since you saved me lots of $$$$'s. I was 100% sure the cpu was broken.

so is your DDR running in synch or asynch? Sounds like asynch, which will be about the same or even slower than running it in synch at 133 FSB. Keep playing with the timings and try to get in in synch at 166.

http://www.amdforums.com/forumdisplay.php?s=&forumid=12
http://www.nforcershq.com/forum/viewforum.php?f=11&sid=b0f6032324c3af7a0340c7f8abb8880f

Originally posted by rdgrimes
[B]so is your DDR running in synch or asynch? Sounds like asynch, which will be about the same or even slower than running it in synch at 133 FSB. Keep playing with the timings and try to get in in synch at 166.

http://www.amdforums.com/forumdisplay.php?s=&forumid=12
http://www.nforcershq.com/forum/viewforum.php?f=11&sid=b0f6032324c3af7a0340c7f8abb8880f [/B]

how can i check if it’s in asynch or synch mode ?

i just forced it to go back to it’s SPD settings , making it 2 x 133Mhz

The 2600 is a 166/333 FSB cpu, and the Ram is running at 133/266 speed.

Therefore not in synch.

In synch, your ram needs to run at 166/333 speed, same as the CPU. :slight_smile:

Synch = FSB-166, DDR-166
Asynch = FSB-166, DDR-133 (or visa versa)
Asynch creates latency in the timing which will result in lower benchmarks on the DDR than if you ran it in synch. Obviously running it at 133 is slower than at 166 anyway, but the asynch timing is even slower. The DDR should be able to run at 166, but you will need to adjust the timings up to something like 7-4-4-2.5 to get it to be stable. Anyway that’s a good starting point. I don’t know what your BIOS allows, so I refer you to the forums for timings. Leaving it at “by SPD” will be more stable but less performance. If you have a DDR volts setting, bumping that up a notch may help also.
I love my ASUS because it has settings marked “aggressive”. I’ll use anything marked “aggressive”. :bigsmile:
Be sure to run Memtest86 anytime you make changes, it’ll pick up any stress in the DDR.

Understood.

Well , tested it… it will not boot with every setting possible. It will come to the login screen with a setting of 15:9:9:3.0 , but after that it’s BSOD paradise again.

Time to shove those memory banks inside the vendors throat again ?

Hmmm seems to me that if the memory has a standard speed of 133Mhz (because that’s what it reads from the SPD) you actually bought PC2100 instead of PC2700 memory. Mabye that cranking up the voltage of the memory will help a little, but I don’t think it’ll be really stable ever.

I’d take it back and get some “real” PC2700 memory. See if you can get some PQI memory, if you want cheap but good memory (at least, my PC2700 memory works perfect ;))…