5200 Windsor lockups

[steiner666]

got my new system, started overclocking last night

set it from the stock 200x13 to 250x12, and bumped up the vcore to 1.45

set the ht to 4x

set the memory divider to 667 and upped the vdimm voltage to 2.1

ran 35mark05 to compare overclock benchmarks to my stock ones and it got a couple thousand points higher with no crashes or problems.

So I was happy with my 3.0ghz and with the fact that my memory speed and ht were all close to their stock speeds. Then I decided to loosen my ram timings and remove the memory divider and up the v a bit to 2.3. System booted just fine and once in windows I ran 3dmark again and the tests ran good. Then at the end of the very last test the last frame of the test (which I believe is CPU test 2) stayed on the screen and my cursor appeared. I could move the pointer around but couldn't alt+tab, windows button, or ctrl+alt+del or get to the desktop by any means.

I set back to the OC that was stable with the memory divider, and ran the test again and had a similar issue at the end of one of the 3dmark tests. Then, I was messin around with the settings a bit, and I dont know what I had it set for, I think I may have removed the memory divider and set the timings even looser, and in the middle of 3dmark i got a windows error about losing communication with the display driver, and my screen was on that ugly like 16-color mode but i was able to restart and it was fine.

I also noticed that when I use speedfan to monitor my temps during benchmarking that they reach these max:

Temp1: 34C
Temp2: 67C
Temp3: 72C
HD0: 25C
HD1: 29C
Temp1: 66C

I'm not sure what these are since I can't find a profile for my mobo (Abit KN9 nforce4 ultra), but they're a good bit higher than I'm used to seeing on my old s939 setup, even when I had a 5oomhz overclock on it.

Today I ran prime95 torture test on that OC with the mem divider and the vcore on 1.5 and mem or 2.3 and it hadn't found any errors or problems by the time i had to leave for for, but that was only 15-20 minutes and i noticed that in speedfan it was using exactly 50% of my CPU the whole time and I was getting the above temps.

What I'm wondering is if I have just a heat problem, if my voltages are too high, if I'm doing something wrong, or what? I was thinking that maybe something is being overclocked that shouldn't be, like the PCIe bus or something, and that might be causing the 3dmark errors and display adapter crashes. but the only place thats listed is right under the CPU fequency and multiplier and it's set on 100 and can only be set higher than it.

[nn_step]

what cooler are you using?
Run WCG for about a week in the background and it'll find any error

[steiner666]

stock. I plan on getting a Zalman or Scythe or some other good air cooler. but I used stock on my 3700 and it never had problems keeping it cool, guess these X2s just run a good bit hotter.

[mAJORD]

sounds like HT corruption, or maybe even ram to me (the vid card related error leads me to the former though)

CPU core errors don't normally result in I/O and driver type errors, but usually result in BSOD's when failing in normal apps, or an error in stress prime/orthos.

using Prime / ORTHOS may not pick up a HT instablity however, as it is really only stressing CPU functions itself via mathematical fucntions + RAM. It doesn't talk to other devices.

Try using the 13x divider + lower HTT speed with the same overclock. Leave the LDT multi on 4.

Also try raising the NB core voltage if it allows, and different drive strength settings.

[steiner666]

hey thanks for the tips, I'll try clocking it different when I get home keeping an eye on the HTT. I dont think I pushed it above its 1000mhz speed tho. But if i put it on 230x13 it WILL reduce my memory overclock a bit so I guess it would result in more stability there. the htt will be about 924 or so if my calculations are correct but that shouldnt effect things much.

I'm not sure what the other voltage adjustments are in the BIOS. I did see one for the pcie and its on 1.5v... does it need more juice maybe?

[steiner666]

heres the voltage options I have available:

http://www.geocities.com/x_carpe_noctem_x/DSC00031.JPG

anything of interest that should be changed other than the vcore and vdimm?

[mAJORD]

Yeah, it should be ok, but try it anyway..

Another thing I forgot.Looks liek your PCI-E has been locked to 100 which is good, might be worth using a program like clockgen to show you actual freq's though.

Why is your vcore back down at 1.3ish?

Bump the hypertransport voltage up a bit, say 1.22 for now

[steiner666]

oh those settings in the picture are nothing close to what they're at now, those are the stock. I have my vcore closer to 1.45-1.5v and the ram is always between 2.1-2.3v. I was doing some reading on different forums and some people said that increasing the pci-e frequency increases their max stable overclock if they put it somewhere between 110-120. I was thinking about trying that when I get home too, maybe up the pci-e voltage too along with the htt? What about that CPU VDDA voltage?

[DyJohnnY]

hi steiner i'm planning a build almost like your's [assuming i can get a x2 5200 anytime soon] and i was wondering what your progress with this was.

thanks