SOLVED Odd Issue

All of my system specs are in my sig.

A few months ago, I overclocked my rig. It passed P95 overnight with no issues and good temps. A couple of days ago, I decided to see what my load temps were like with this cooler weather we've been having, so I fired up P95 again, and 10 minutes in, I had a freeze up. From there, I've been searching forums and looking at things all over the place. I found out that the PCI power plug I was using for my GTX card was the one on the same +12v rail as the CPU, so I swapped them. I tried lowing my overclock, I cleaned my case out good and disconnected and reconnected all of the plugs, and I'm still freezing up in prime.

Now here's the question: When I got my new Motherboard, I didn't reinstall Windows. All I did was uninstall all of the drivers and software associated with the old motherboard and install the new drivers for the new one. With that, I got 4080mhz stable overnight. Is it possible that the Windows install has deteriorated or something? Or is my CPU starting to fail? Any advice or knowledge given would be appreciated, but I probably won't be back on until tomorrow morning. Lot's of things to do on a Friday night, ya know.

Yeah, I think I would:
1. Post back and attach pics of CPU-z tabs: CPU, Memory and SPD
2. Check core and CPU temps with HWMonitor while running Prime for however long it will run. Post back with a report of those temps.

We'll see if we can spot something amiss for you from that info.

Here are the screenies. I stopped P95 about 5 minutes in, as it usually freezes pretty soon there after. Temps never get higher than 65 on the socket and the highest I've seen core temps at is about 48.

1. At 1.408 your vcore seems awfully low. It takes me 1.475 or so to be stable at 4.0 on my 960T.
2. What is your CPUNB voltage set to? Since you are overclocking the CPUNB you would need to supplement the stock voltage.
3. I would lower you HT Link speed to between 1800 and 2000 mhz. There's no advantage performance wise to overclocking it and despite the popular notion that the CPUNB and HT Link should be kept in lockstep we have found that not all Thubans like that.
4. Your socket temp at 65c is getting pretty warm. I would lay a fan in that area to move some air across it.

I can't explain exactly why but I have found that generally I have to go back and make adjustments a few weeks to a few months later after my initial overclock that was Prime95 stable originally after a new build. For some reason, whether it's deterioration of components or whatever, it becomes unstable.

If none of these suggestions help I would suggest testing the ram.

trents said:

Yeah, I think I would:
1. Post back and attach pics of CPU-z tabs: CPU, Memory and SPD
2. Check core and CPU temps with HWMonitor while running Prime for however long it will run. Post back with a report of those temps.

We'll see if we can spot something amiss for you from that info.

Click to expand...

I know you guys ask for the SPD tab of CPU-Z but it is rather useless outside of manufacturer, model and manufacture date. Sometimes the model isn't even right or the manufacturer is blank but filled in in the model spot instead.
Most of the time the SPD table is useless...it never does what it's supposed to and reads out trash.
Furthermore, I don't know if you all know what XMP Profiles are but when they show up in the SPD Table they are correct, and not only are they correct but they are what the memory is designed to run at by the manufacturer, not the last JEDEC SPD.

I see this on every thread and I don't mean to be rude, it's just something that bugs me a bit.

Anyway...

Dlaw, generally core temps + 15c are accurate on Phenom II "Thuban" or in this case "Zosma" CPUs. If you are peaking at 48c then you are up around 60-65c just like the motherboard reads out.
Before you try to push further I would try to get cooling in check.

In case it helps, you can try increasing vcore an extra .01v but your CPU is running rather hot.

Contrary to what trents suggested, to keep heat down I would actually lower the CPU-NB multiplier and actually undervolt the CPU-NB or overclock the CPU with only CPU Multiplier.

@ trents: The 1.408 was at idle. Under load, LLC pushes it to about 1.45. CPU-NB is set at 1.1v. So it's normal to have to tweak an overclock a month or so after it's initially set? I'll try dropping the HT link back to stock and add another increment of voltage to CPU-NB. After I cleaned my case out, all of my temps seem about 4 degrees cooler, so now I think my socket temp is closer to 61

@Beep: I said the highest I've seen it was 48, and now that I think about it, I want to say that was with my GeminII. 35-38 is about normal under load for this cooler.
I can't get over 3.6 with just the multiplier. For whatever reason, I can't get stable with even 1.5v at 3.7 using just the multiplier.

Okay, so I restored my overclock to 4080MHz, and tweaked a little. Here's where I'm at right now:
CPU Freq.: 240 Multi: 17 = 4080
CPU Voltage: 1.375
NB Freq.: x10 = 2400
NB Voltage: 1.1125
HT Link: 2160
LLC: Auto (Options are Auto/Enabled/Disabled)

Running P95 right now. About 5 minutes in, temps are about the same as they were in the screenies. Will update with a pass or fail, and screenies if I can get them.

Here's some shots at about 45 minutes in.




How's it all look? I know the socket temp is a little high, I'll see if I can fit the 80mm fan I have off of an old heatsink in there somewhere. I'll take a picture of the inside of my case and ask for some advice as far where to put it.

That looks great! The socket temp is a little higher than I'd like to see it but not dangerously high. Do you have a spare fan you can lay in there to blow air across the area?

You might need to bump the CPUNB up to about 1.2-1.225 if you are still not stable. You have it overclocked to 2400 mhz. Notice that LLC Auto is bringing the vcore up to 1.45 under load now and that is the main reason it's more stable, that and maybe lowering the HT Link speed.

Right, If it fails I'll try bumping CPU-NB voltage first.
I do have a couple of 80mm fans around from old heatsinks, and I think I have one more 4 pin fan controller left on my mobo. I'll get a pic of the inside of my case real quick and maybe you can help with placement?

If necessary, you can put two fans on a "Y" splitter and run both off the same header pins. You can get them for $2-$3 on ebay if you can't get them local.

Here's a picture of the inside of my case:
https://www.dropbox.com/s/vmw2kgdk98ah70u/IMG_20120810_112839.jpg
Sorry for the link, the picture is too big to host here.

About an hour and a half in, and the only max temp that's changed is CPUTIN at 66. So long as the last half hour goes well, I'll look into getting that 80mm in there and call it good.

I guess I would try to position a fan between the video card and the CPU cooler base blowing at an angle downward and upward, maybe about 45 degrees. Is there room to do that?

I think that's possible. I can try it any way. So what you're saying is angle it so that the bottom of the fan is touching the gfx card and the mobo, and the top is touching the heatsink, blowing towards the CPU?

Yes.

Zip ties might be used to hold the top edge of the fan against the cooler base and a narrow strip of that double-sided adhesive tape laid along the edge of the fan making contact with the video card might secure the bottom.

Cool, I'll try that after this test (at 2hrs in 15 minutes).

Scratch that, just got recruited to go to the park with the family. I'll try that in a couple of hours.

Personally, if you can pass a two hour Prime95 blend test I would call it good. I see no point in torturing equipment for many hours. And I know many on this forum would disagree with me. If you can pass two hours of Prime I don't think you'll find any instability in running real world apps.

After 2 hours of P95:

I'll be back in a couple of hours to try the fan thing.

Boy, your core temps are great! But having said that, it is common for the Thuban core CPUs to misreport the core temp by about 10c or even a little more on the cool side. That's why it's important to look at both the core temps and the CPU socket temps, especially with some CPU models.