Need some help - GA-P35-DS4 and Q6600

Snarkasm

Hey guys -

I need some help. I'm desperately trying to OC this system, and I just can't seem to figure out my settings. I know I'm doing some of it right, but then I reboot and CPU-Z shows me that I'm actually only throwing 2GHz (a drop from 2.4) because for some reason my multiplier gets lowered to 6. First, specs:

GA-P35-DS4
Q6600
Scythe Infinity heatsink with a pair of 120mm 70cfm fans on it
4 gigs OCZ Platinum
BFG GeForce 8800GTS OC

And my bios overclocking options are: Option: [setting]

Robust Graphics Booster: [Auto]
CPU Clock Ratio: [9x]
CPU Host Clock Control: [Enabled]
CPU Host Frequency(MHz): [333]
PCI Express Frequency(MHz): [100]
C.I.A.2: [Disabled]
Performance Enhance: [Standard]
System Memory Multiplier(SPD): [2.00] (which gets me to 667 and a FSB:DRAM of 1:1)
High Speed DRAM DLL Settings: [Option 1]
System Voltage Control: [Manual]
DDR2 OverVoltage Control: [+0.25V]
PCI-E OverVoltage Control: [Normal]
FSB OverVoltage Control: [Normal]
(G)MCH OverVoltage Control: [Normal]
CPU Voltage Control: [1.36875V] (which I picked as it's just below the upper limit for Q6600s as a good point to start testing)
Normal CPU Vcore: 1.27500V (can't change, that's just a notifier

With all of these settings, the BIOS blinks at me (right in between High Speed DRAM DLL Settings and System Voltage Control) the message "System Voltage NOT Optimized!!"

So as you can tell, I'm a freaking newb, and I understand some things (like that I want to get my FSB to 1333, which is 333x9 and that I want an FSB:DRAM of 1:1, but I pretty much don't understand any of the other options and I don't know why, when I boot and start up CPU-Z, it flashes 3GHz before then going to 2GHz and a multi of 6. It seems to me like that's a voltage problem, but like I said, I know nothing, so I defer to the masters here.

Doesn't help that everybody's bios is different, but so it goes. Please help, guys! Thanks. I'll be here all day working on this, so I can test things quickly and get back to you if need be.

Thrax

Did you leave EIST, SpeedStep or Extended CPUID functions active in the BIOS? If so, turn them off. Your settings are fine.

Snarkasm

Hmmm... EIST is disabled, but CPU Thermal Monitor2 is enabled as is CPU Enhanced Half. The thermal monitor says the core clock ratio and voltage will be reduced, which looks like what happened...

I just raised the alarm temperature to 70 C from 60, maybe that will do something.

Snarkasm

Now I'm watching it go between x9 and x6 in CPU-Z. CPU-Z's also reporting still 1.8V for my memory voltage, though I overvolted it .25 based off of that 1.8... which makes me think I did something I shouldn't have...

Using PCWizard, I'm seeing 1.33V on the CPU (which matches CPU-Z) and it seems to stay there regardless of the multiplier status. My DIMM voltage is reporting 3.3V (which I think is bad) and my +3.3V rail is only reporting 2.03V.

What the nuts is going on?

Edit: But when I run an instance of Prime95, it seems to stay at x9 (and thus 3GHz) solid, so maybe it was a power-saving feature? Still concerned about my memory voltages though.

Thrax

1. You have speedstep on. Turn speedstep off.
2. CPU-Z doesn't report memory voltage as far as I remember.
3. You need to run 2.2v or less. Your board may be overvolting. Increase VDIMM by .15 or .2v.
4. Your 3.3v rail sucks, and you need a better PSU.

Snarkasm

There's nothing called Speedstep exactly, but CPU Enhanced Half I think did the same thing (I read up on it and it said it lowered voltage and multiplier in halt states), so I turned that off and I'm getting 3GHz constant.

I readjusted my memory overvoltage (took off my +.25V) and my +3.3V rail went down to 1.87 and my DIMM voltages stayed at 3.30. Is there a chance those two values are switched? That would make sense to me, the values would be pretty much correct and I don't see how dropping my memory overvoltage would affect my 3.3v rail and completely unchange the actual memory values.

Thanks for your help, Thrax.

Thrax

Did you update your BIOS at all?

Snarkasm

Yep, first thing I did. I forgot to update here, but I worked on it more. Got it stable at 3GHz, got the memory running at the right voltage, running 4 instances of Prime95, all temps still 57-58; I'm convinced that PCWizard mixed up those values, and Gigabyte's "Easytune" program, while generally useless for anything else, confirms a correctly-volted 3.3V rail. So tomorrow I either start upping the FSB or lowering the voltage. I can't, unfortunately, find where to set my memory timings, they're stuck at 4-5-5-13. I'll have to read up more on that.

Thrax

PCWizard is garbage. The only tool that monitors voltages properly is the BIOS, really.

mas0n

MJancaitis wrote:

My DIMM voltage is reporting 3.3V (which I think is bad) and my +3.3V rail is only reporting 2.03V.

Seeing as your RAM has not burst into flames, I think it's much more likely that those voltages are somehow swapped, especially since you said your RAM was at 1.8V and you added .25V which makes for ~2.05V and that's pretty close to what is being reported.

Thrax wrote:

PCWizard is garbage. The only tool that monitors voltages properly is the BIOS, really.

I agree that PC Wizard is garbage but I would also not trust your BIOS. The only tool I trust is my multimeter.

Snarkasm

All right. After shamelessly flirting around with it, I got the proc up to ~3.5GHz (9x388), but as soon as I ran one instance of Prime95 it BSODd on me, haha. In the interest of getting a 1:1 ratio and giving my memory the 800MHz it so wanted, I bumped the multiplier down, set it at 8x400 and matched the memory and I'm sitting moderately pretty at 3.2GHz on air, full load around 58 C.

Prime's running overnight after which, if it survived, I'll set up SMP again and get back to my mad folding ways.

If anybody has any input or wants to tell me where I've gone wrong, I'll be glad to learn! Thanks for everybody's help so far, especially Thrax.

Leonardo

There's nothing called Speedstep exactly

Some BIOSes will call it C1 or C1E. Turn it off.

Qeldroma

It is called the CPU EIST Function in the BIOS where EIST is the Enhanced Intel SpeedStep Technology.

Bascially what you are trying to do is shut off all the power regulation and CPU protection features of your board. Of course, for a board that is supposed to be designed for overclocking, I can't figure out why they bother having them in the first place.

Snarkasm

What, nobody reads now?

I already turned off C1E and EIST. Thanks for explaining EIST = Speedstep, though. That makes sense.

Snarkasm

In a somewhat-related yet incredibly-different event, is there any reason an overclock would mess with SATA hard drives or my memory somehow?

I've been getting what can only be described as sporadic short (3-10 second) freezes where it sounds like a hard drive has to spin up, and it clicks at the end some times. I've got 6 hard drives in there, all SATA, so I narrowed it down to what I thought was one drive, and unplugged it. When I booted, still had the same thing. Did that a few more times before I was convinced it was a system problem and all my hard drives weren't going bad at the same time.

I thought perhaps it was a driver/BIOS issue with the SATA controller, so I contacted Gigabyte to see if they had any ideas. Their response was to test the ram, single-stick only, etc etc. Mentioning that gave me other ideas, so I checked my RAM timings and voltages, made sure everything was right; then, just for fun, I returned my Q66 to stock settings, 266x9 with 1.275V, and the lockups don't seem to be happening. This is by no means an exhaustive test, as it basically consists of me booting up, starting the SMP client and playing a game; if both the SMP client and the game are running, the game is a very easily-noticed indicator when the freeze happens when I might otherwise miss it. In the interest of full disclosure, with SMP folding not running, I do often still hear the spinups, but the system then seems to push through it with an imperceptible freeze, if any. So 100% CPU usage seems to be involved with the freezes, but not with the general cause of the issue. So my question is this:

Would my overclock have caused some kind of timing mismatch or some other kind of error with my hard drives, could this have happened with my multiplier, or does anybody have any other ideas? I'm stumped, and definitely don't want to run this chip at 2.4GHz, it's such a waste of such a good proc.

Thanks once again for any help or insight you guys can offer.


Edit: at the no-overclock, clicks still seem to be happening. Guess I'll be swapping out some memory.

Leonardo

Would my overclock have caused some kind of timing mismatch or some other kind of error with my hard drives

It could. My first thought was that yes, overclocking could be stressing the IDE (PATA & SATA) buses. I was going to suggest boosting the southbridge voltage, but the DS4 doesn't have that option. Are you running your DRAM overclocked too? If so, try dividing it back close to it's default frequency.

Edit: at the no-overclock, clicks still seem to be happening. Guess I'll be swapping out some memory.

Oh, well then, hmm. Do you have the AHCI option turned on for the SATA controllers or RAID? You might try testing those options as well.

Snarkasm

Now I'm stumped.

This time, I went back and gave it a 9x360, with a 2.5 memory multiplier for 900MHz. Boot back in, and... played a level of Hitman, played two levels of Portal, nary a blip. It would be ironic if the memory were unstable at its native frequency. btw, Leo, before, it was at 400x8 with a 2x mem multiplier for its native 800MHz. That's why I was so sure there was no way it had to do with OCing the memory... until now... when it seems I need to OC it, haha.

btw, the hard drives are all in SATA/RAID mode as I'm running a pair of Raptors in RAID; AHCI is thus not selected, and they're using the legacy whatever instead of the native mode.

Thanks for your help. Any other ideas, or should I just thank the computer gods everything seems to be working and damn the explanation?

Leonardo

Strange. If you really want to know what causes what, and what stabilizes what, you'll have to start at a baseline - default everything. Record your settings and work your way up. Quite often, instabilities can be fixed merely by boosting FSB voltage and vCore. Sometimes it's merely DRAM voltage pumped a little higher than seemingly needs be.

Leonardo

I've got two packages waiting for me at the Post Office. Both are pre-owned heatsinks I purchased in private deals. One is a Zalman 9700 NT and the other is a Tuniq Tower 120. The Tuniq will go on System No. 1. The Zalman will go on one of my other machines that will soon be upgraded to a G0 6600 system. We'll see if the current B3 can knock out some more MHz with better cooling.

Snarkasm

Good luck with it. It's addictive.

I moved to 375x9 and gave it some extra voltage to work with, and a 2.4x multi gives me 900MHz for the memory, same as at the last stable point. We'll see if this holds up! btw, at these settings, with 4xPrime95, I'm punching 63 and 60C on the cores. Scythe Infinity, two 70CFM fans on either side, in an Antec 900. I hit 65 sometimes, so I don't know if I'll be leaving it here. Then again, Folding runs cooler, so we'll see.

Leonardo

375 X 9? Not bad at all! So, please tell me, that's a G0 stepping, right? I'm running the same board with a B3 and can only muster 3200 (9 X 355) at full load. That frequency is at vCore set at 1.35625 and measured 1.31 by CPU Z. Core temps are 52-58. I've tried 3300 (9X367) with vCore of 1.36875. It's stable under idle and light use, but black screens and restarts under full load. 1.3750 didn't work either. Hmm, what vCore are you running to get 9X375?

Earlier you were commenting on how your DDR voltage wasn't changing, according to CPUz. You were looking at the SPD figures for your RAM, that is, the presets that are programmed into your RAM. In other words, the default voltage for your RAM is 1.8 unless you change it. The SPD view in CPUz reads straight off the RAMs electronic data tag, so it won't change.

Snarkasm

Right, right. I figured that out. Went by PCWizard to get the accurate voltages.

It is indeed a G0. In an Antec 900 and with a Scythe Infinity, it's touching 58-60 Folding, 63-67 under 4xPrime95. It's ridiculous how different they are. But yeah, 9x375 for 3.38GHz, and I think I put it at 1.375V for that, droops to 1.306ish according to PCWizard. Not absolutely sure on that though, I'll have to check for you to be sure.

If I go any higher than 9x375, it bluescreens/core dumps at the first sign of trouble, and actually at one point told me some of my Vista boot files might be corrupted. I thought that was interesting. Backed it off and it booted just fine.

I can't decide if I want to try and be more aggressive and drop it to something like 8x450, but at 8x400, which was only there to try and help the memory, gave me tons of problems which I think was actually due to the memory. At any rate, 3.38 is pretty good, and I'll probably leave it at that due to the lack of problems here.

Don't forget, vCore can go up to 1.5V if you really want it to. I had it at 1.47 at one point to see if my 9x380 had any chance at being stable, but no. Might just be a black hole at the fsb around that point. Good luck, lemme know how it works out.

Leonardo

Hey, thanks. BTW, PC Wizard is crap, in my opinion. I recommend Core Temp for accurate cores reading. I also use SpeedFan, which is quite accurate if you use Core Temp for calibration. You might also want to check out Everest. I think though, the best picture of the PC's operations will be with a combination of Core Temp, SpeedFan, and CPU Z.

Yeah, the G0s are definitely superior to the B3s. But all in all, I'm pretty happy with this B3 of mine. Man, these quads are just bulldozers! Amazing performance. Throw anything at them and they chew it up and spit it out pulverized.

I haven't tried any wild and crazy vCore settings yet. Not sure if the extra power consumption and heat is worth just another 100MHz. But, knowing me, I'll have to try it. 1.375 is the highest I've had it. It was stable with one FAH SMP client, highest core temp at 57C. It crashed when I launched the second FAH SMP client. BTW, I'm running a Tuniq Tower 120 that I just purchased, private trade from a guy at [H]. Awesome heatsink. I'd love to run the T'Right U 120 Extreme, but, pricey. Maybe later...when the wife isn't watching the books too closely. (nah, not a chance - she is German and doesn't miss a single detail when it comes to finances)

Oh, I should mention, my No. 1 rig is also housed in an Antec 900. One of the top five case designs of the last ten years, IMO. Just wish it was a little bit bigger.

Snarkasm

^ Agreed, I'm out of hard drive space. Plus the holes for watercooling hoses exist solely to tempt me into spending more money. But I'm glad they're there all the same.

I run CoreTemp for temp readings, I use PCWizard for voltages, that's all. I've tried Everest too, it's fine, I just don't need any of them anymore until I decide to tinker again.

Snarkasm

Update time!

After seeing my OC was horribly underwhelming according to some other sources, I decided to play around a little more. Right now I'm sitting at 8x450 for an even 3.6GHz, with the RAM at 2x multi. I had to boost my FSB voltage a little bit and the VCore another notch, and then just for fun, I upped the (G)MCH voltage, which according to the book is my NB and, surprise surprise, it seems to have fixed my random hard drive spinup issues. Obviously this is on only a little bit of testing and I'll keep watching, but I'm encouraged.

4xPrime95 was having some fun, with just one of them erroring out and the others continuing on nicely; when I restarted the stopped one, it just halted again a few minutes later. Upped the voltage a bit and tried again, and so far one halt through 120 tests. One more slight adjustment and I should be sitting happily.

Temps are 62-63ish on the hot cores under Prime; will be lower with FAH as usual.

I did have a question in all this: would the NB voltage really be related to my SATA HDD spinups? Isn't that a southbridge thing? I suppose I don't really care if it actually works, but it would be fun to know all the same.

Leonardo

Glad to see the improvement. 3.6GHz is singing sweetly.

Snarkasm

As it stands, I've got the BIOS up to 1.48750 V (1.44 or so, I think, after the droop), that appears to have fixed my Prime95 faults, so I don't know if I'll be trying any higher. I bumped the FSB voltage one notch and it wouldn't let me complete a boot, so I dunno. While I am supremely happy with a 150% overclock, wouldn't it be fun to touch 4GHz just once? Haha.

I know it's more a function of my hardware than my skills, so it's not really a compliment-worthy achievement, but I'm going to say thanks for the compliment anyhow.

Leonardo

I'm running the same board with a B3 and can only muster 3200 (9 X 355) at full load.

I dropped the multiplier to 8 and run the memory at 412.5 MHz and FSB at 1650 - total frequency 3300MHz. Certainly nothing for the record charts, but altogether dandy for a B3.

I've noticed the price on these GA-P35-DS4 motherboards has gone up steeply since I got mine a few weeks ago. I guess we all are saying nice things about them and the sellers noticed!

Snarkasm

That link somebody posted in the news section a little while ago about the Foxconn board made me look at what the DS4 could really do. It's a solid piece of equipment. Only gripe I have is that the memory is multiplier-locked to the FSB, and doesn't go any lower than x2.0. Wish they were separate, but that's all. My memory has a forced overclock, haha.

Snarkasm

Eeek. It was stable as all hell under Prime95, so I quit it and went ahead and started the 2xSMP up again. Gave me a lot of nasty errors (FILE_IO_ERROR, MISSING_WORK_FILES, 0x7B, etc), so I just deleted all the old info and just let it churn for a couple minutes. Picked up a new core, picked up new work, and now they're showing me the core successfully loaded and they're chewing on some new units... so hopefully it was just a quick shock from the changes and not something indicative of overall instability. I don't want to lose any more WUs

I love how much lower-temperature FAH runs. I drop 8-10 degrees just by running it instead of Prime95, haha.

Snarkasm

Welp, I'm not sure exactly what changed, but nothing's stable anymore. Prime is still rock solid, but F@H is screwing the pooch consistently, and on both instances. EUEs, mismatched digital signatures, FILE_IO_ERRORS, and various other states of suckitude.

I switched the HDD the F@Hs were running on, and file IOs continued, so it doesn't look like it's a failing drive, per se; anybody have any ideas on what voltages I should be boosting?

Thanks. If this continues, I may have to back it down.... or may have to get a new PSU and mobo