Suddenly lots of freeze-ups

NotAPimecone

Distinguished
Nov 18, 2006
299
0
18,780
System in sig.

I tested it quite thoroughly, many hours each of memtest, dual-prime95, and orthos. All went without a hitch. Lately, however, it's started freezing up, and I'm not sure why.

Even lowering FSB to 400MHz (so the RAM wasn't even overclocked) nothing changed.

Gradually it got worse, from a freeze only once in a blue moon, to the point now where it can't even get into Windows, and freezes during memtest at the start of test #7. Not errors, a freeze-up.

I've reset everything to stock (except for RAM - voltage and timings to the spec 2.1V, 4-4-4-12, 1:1 ratio instead of the crap SPD settings) and started memtesting again. It made it through a pass so far, so it's probably fine. Also tried stock FSB with RAM at 3:2 (for its rated 800MHz) - also passed a pass of memtest. Everything seems to run fine as long as the CPU is not overclocked.

Haven't tried out a milder OC, like 333FSB, yet, but will at some point.

So, did my CPU run out of mojo? I don't see how I could have fried it, vCore was set to 1.3, and I tried adding a little more up to 1.325 (to no effect) when the freeze-ups started happening more frequently. Temps, as measured by CoreTemp, rarely broke 50C during stress testing, and never above 45 under normal use. I think my all-time high was, with the room very warm, and TAT running 100% stress on both cores, 59C.

Any idea what the problem could be?Could it be something to do with my PSU? A while after buying it, I heard it has more "ripple" than is desirable. But this PSU has only been in the system for about 2 months, and it's not on anywhere near 24/7.
 

mkaibear

Distinguished
Sep 5, 2006
678
0
18,990
Do you have access to a multimeter?

Problems like this always scream *POWER SUPPLY* to me - especially if you've reduced the OC and you're still getting the stability problems.

If you have access to a MM, take the PSU out, short the switch to get it started, and check the rails.
 

NotAPimecone

Distinguished
Nov 18, 2006
299
0
18,780
Yeah, I bought a multimeter a while ago for shits and giggles. Haven't used it much yet, as everything was working just fine until now.

I'll try it out some time, probably fairly late tonight or tomorrow evening. I'd do it sonner, but I have work and other stuff to do.

Whee, I get to take my computer apart again! At least it's not a full tear-down, though. Already done that a couple of times (had to RMA mobo - man, did that ever suck)
 

mkaibear

Distinguished
Sep 5, 2006
678
0
18,990
Last time I did a full tear-down was just after I built an HTPC - I had a whole pile of cable ties left over (bought a HUGE box :)) so I decided to tidy up my main PC as well.

Oops.

Let's just say that certain front-panel connections have never been the same since. *sigh* :)
 

NotAPimecone

Distinguished
Nov 18, 2006
299
0
18,780
Heh, tidy? What's that? I just try to cram my (multitude of) extra cables out of the way as much as possible. As long as airflow seems ok, I'm happy. No window on my case, what I don't see won't hurt me :wink:

But dammit, I thought I was done messing with this thing's guts. My old (non-overclocked [incapable of being overclocked] P4) never gave me this kind of trouble.

Then again, part of the reason I built this one (aside from just plain wanting a faster computer) was to have a spiffy PC I could play with and overclock and learn. I guess it'd be pretty lame of me to bitch too much about having to get my hands dirty to keep it running :)

It just sucks when it freezes up 90% through encoding a video :p
 

cb62fcni

Distinguished
Jul 15, 2006
921
0
18,980
It's not going to hurt your performance any to loosen your memory timings further, to say, 5-5-5-15. Intel systems aren't as reliant on memory timings. I agree that it sounds like a PSU problem though, which is strange, because OCZ's are pretty darn good. Another problem could be the VRM's on the motherboard. Watch your Vcore under load and see if it fluctuates.
 

NotAPimecone

Distinguished
Nov 18, 2006
299
0
18,780
That's not really the point, though. This RAM is rated for 4-4-4-12 at 800MHz and 2.1V, it should damn well run at 4-4-4-12, 800MHz, and 2.1V. That's what I paid for.

And it does run at those settings - passed memtest with the CPU at stock speed. It only failed when the CPU was overclocked and it didn't exactly fail (no red errors), it just completely froze up. I don't think my RAM is the problem here.

If it's the motherboard again, then I am going to kill Gigabyte - I already had to RMA once when it suddenly died one day. Isn't the "D" in DS3 for "Durable"? Unless, of course, the PSU is killing the motherboard. :roll:
 

martyjs

Distinguished
Jun 25, 2006
367
0
18,780
Have you check the basics? Ran MEMTEST86 and tested the Hard drive using the manufactures test tools?
Some times it could be a software problem, or a failing video card.
 

martyjs

Distinguished
Jun 25, 2006
367
0
18,780
Don't mean to pick. If your over clocking which can cause your parts to fail. I would assume you have voided your warranty (Can't blame Gigabyte for that) I know they provide the software to do it, but if you read the fine print in the warranty, it's not covered) :(

If people choose to over clock and that causes hardware failures, why should the rest of us pay the extra cost of your hobby? :(
 

cb62fcni

Distinguished
Jul 15, 2006
921
0
18,980
Well, a freeze is pretty much a failure, red or no. Just for grins and giggles loosen those timings up and see what happens. Or bump the RAM voltage to 2.2 and leave the timings. The D for durable refers to the caps, not the VRM.
 

NotAPimecone

Distinguished
Nov 18, 2006
299
0
18,780
Don't mean to pick. If your over clocking which can cause your parts to fail. I would assume you have voided your warranty (Can't blame Gigabyte for that) I know they provide the software to do it, but if you read the fine print in the warranty, it's not covered) :(

If people choose to over clock and that causes hardware failures, why should the rest of us pay the extra cost of your hobby? :(
The settings are not outside what the board should be able to handle. And, if by "they provide the software" you mean EasyTune, I don't touch that shit. Only OC through the BIOS.

Are you saying that running a higher FSB than what's stock for my CPU can hurt my board? So, if I put a CPU that runs a default 200MHz FSB in the board, then what? I can't raise the FSB, say, to 266MHz without voiding the warranty on the board, because the board "can't" handle it? Even though that's the stock FSB for another CPU that also runs in this board? Now, if I have to raise the FSB voltage to do that, maybe there's a case for voiding my warranty. But I haven't done that.

Similarly, different CPUs require different voltages. I've seen Core2s where the stock vCore was around the highest I ever ran. Hell, I've got a 775 Prescott (which should run in the 965P-DS3 if I wanted to put it in there) with a stock vCore of 1.370V, much more than I was giving my Core2 (1.300V-1.325V) - should 1.325V vCore then void the warranty on the board? I should hope not.

If the warranty on anything would be void, it would be the CPU. It's the only component for which I've raised the voltage from stock, and even then, not by much.

The only other thing running higher than spec was the RAM, by a mere 50MHz, with NO more voltage than specified by the manufacturer - warranty probably not voided.
 

NotAPimecone

Distinguished
Nov 18, 2006
299
0
18,780
Well, a freeze is pretty much a failure, red or no. Just for grins and giggles loosen those timings up and see what happens. Or bump the RAM voltage to 2.2 and leave the timings. The D for durable refers to the caps, not the VRM.
Why do you think it's the RAM? Unless 800MHz, 4-4-4-12, 2.1V at 400FSB (1:1) (failed) is that much different from 800MHz, 4-4-4-12, 2.1V at 266FSB (3:2) (passed) as far as the RAM's concerned?

I'd expect a freeze to have more to do with the CPU, isn't that how it works? Memtest moves values around in memory (presumably, the CPU does this). If they don't match up to what it expects, it shows errors (whether the RAM screwed up or the CPU got confused is for the user to figure out). But to just halt? I would think that would only happen if the CPU froze up. Especially since the RAM does not fail at the same speed but with the CPU at stock settings.

But, if you really want, I'll try it later overclocked with relaxed timings. I expect it to fail again.
 

cb62fcni

Distinguished
Jul 15, 2006
921
0
18,980
Well, there's a ton of things that can cause of freeze. RAM is much more likely than the CPU as long as the CPU isn't overheating. And Yes, memory divider can be quite a factor. The reasons I suggested playing with RAM timings and voltage are because it's easy and it's a common factor in overclocking failure.

Overclocking does NOT void the warranty, so long as you don't tell them...heh heh heh. Seriously, it's their own fault for making it available, if they didn't want you to overclock, why include bios options that allow you to do so?

Finally, can I get some more information about the problem? NB temp, NB volts, CPU idle/load temp via TAT, etc? Work with me and we'll get you through this, promise.
 

martyjs

Distinguished
Jun 25, 2006
367
0
18,780
If you read the fine print on most warranties( I know :lol: no one reads them. I only know because I'm in charge of our Computer warranties) there pretty tight on what they will fix (Again only if people tell the truth).

I don't care if people over clock or not. I've done it a few times but only about 5 to 10%.

I'm also prepared to put my money were my mouth is. Eg: about seven years ago I dropped my brand new hard drive and even though it was damaged internally I didn't take it back. :oops: :oops:

I new my supplier well and they had been good to me, so I didn't want to shaft them.

Most suppliers have to pay freight at least one way and there's also a lot of man power to process warranty claims.

The big company's might be able to cover the cost, but in the end it's the customer who pays cause the price includes the warranty. More claims there are higher the Price. :(

Personally when a customer brings back something that I know has been used the wrong way (and believe me sometimes its easy to tell) and try and tell me "THEY DID NOTHING WRONG. " I'm not happy , cause if I can tell so can the repairers. They waste my time and my bosses money sending it back only to have the repairer say "its not covered under warranty"

I don't mind if the customers honest with me because sometimes you can't tell and we then try and help them. usually if that happens we warn them not to do it again as we can't keep doing warranty claims in that case.

Sorry for going on, I don't want to be a party pooper, all I,m trying to say, think of others as well as your self and everybody will be happy.

It is very unusal to have 2 motherboards fail in a warranty period I,ve been fixing and building stock PC's for both my self and customers for over 10 years. In that time its only happened twice.

Hope you get it worked out. :)
 

NotAPimecone

Distinguished
Nov 18, 2006
299
0
18,780
OK, did some memtesting. Things are kind of interesting.

Settings:

FSB: 400MHz
CPU Multiplier: 8
RAM Multiplier: 2 (1:1)
RAM Voltage: 2.1V
CPU Voltage: 1.325V

I tried out the 5-5-5-15 timings. It passed a pass, which was a step up from what I'd observed with 4-4-4-12 (couldn't even get through 1 pass). But just to be sure, I changed the timings to 4-4-4-12 again, and it also passed a pass. I left it for a while, and it froze up about half way through pass 10.

Changed it back to 5-5-5-15, and left it overnight. It was still going this morning, after 27 passes.

So, what does this mean? My RAM never had a problem running 4-4-4-12 before. I don't care that it won't impact performance much, and if I have to run 5-5-5-15, that's fine, but it's rated for 4-4-4-12, why doesn't it run 4-4-4-12? Will I have to loosen timings again in another few months? I haven't even had this system for half a year, and I've never had the RAM voltage higher than the manufacturer specified 2.1V.
 

cb62fcni

Distinguished
Jul 15, 2006
921
0
18,980
Well, the P965 chipset is kinda weird about memory timings, 5-5-5-15 is what most recommend with that chipset. As for why it suddenly stopped working, that's a real tough one, too many factors involved. When did you update to F10 bios? Is your NB getting too hot under load? You VRMs may have degraded slightly to where they're not giving the full 2.1V to the RAM, that's why I also suggested you bump it to 2.2. The important thing is that it's stable now, from here you can work to try to find out what caused your problem.
 

NotAPimecone

Distinguished
Nov 18, 2006
299
0
18,780
Hmm. It's still not really stable, I keep getting freezes.

The northbridge cooler feels pretty hot to the touch, and the way the board is laid out, it's kind of blocked by the RAM in terms of airflow from front to back. The case does have a fan blowing in the side, but air from that might be blocked and/or sucked away by the horizontal-blowing Tuniq CPU cooler.

Maybe I should get a 40mm fan to try and improve airflow over the Northbridge. Can't hurt to try.
 

jonrem

Distinguished
Jul 7, 2006
125
0
18,690
DS3 lately = BS. Since the F10 bios I've had nothing but problems (running mostly Vista, playing SupCom.) I've had a lot of experience with the mobo since November. At times it's been great, at other times it makes me wish I waited for a 680. Lately I run into a TON of freeze-ups when SUPCOM starts to get fun. Even if I'm doing word processing the NB heats up to the point you can't touch it. I have a 50m fan blowing directly on it and it still semi-melts the plastic of the fan(!!!) I recently switched to 4 gigs from 2and I think addressing that extra mem is the straw that breaks the camel's back. Enough of me, I hope your problems get better soon.
 

cb62fcni

Distinguished
Jul 15, 2006
921
0
18,980
On my little brother's DS3 we removed the heat sink from the NB, cleaned it and the chip up, and used AS5. It seemed to help a little, temps dropped about 3C according to BIOS. Or you can pick up a nice aftermarket NB cooler from newegg for 20 or 30 bucks. Believe it or not, the NV 600 series is WAY hotter than the DS3 as far as NB go, even with their crazy big heatsinks.
 

NotAPimecone

Distinguished
Nov 18, 2006
299
0
18,780
Well, I think northbridge temp must not be the whole story.

I took off the side panel of my case and have a powerful fan blowing right at the NB cooler, and it's cooled down massively (now only barely warm to the touch under load), but I'm still getting freeze-ups.

Encoding a video (the use which was getting all the freeze-ups), temps are now:

I've clocked it all the way down to 333MHz FSB to see if the problem still happens.
 

Dave_McMahon

Distinguished
Jun 20, 2006
160
0
18,680
things to point out

Is the big heatsink stressing the MB (some cases you cant use all the mounting holes on the MB)

2, Is the memory modules clean does the problem replicate with only one module?

3, Bad BIOS Update (try reflash F10 or try F9 etc.)

4, Try a cheap PCI (or Spare PCI-e GPU)

5, Turn off the APCI stuff in BIOS (Speedstep No-execute etc

6, Turn off everything in BIOS except FLoppy to test MEMtest

Just thinking along these lines