Disappointing results with E6300 + 965P-S3. Please help!

C

Crooked Paul

Weaksauce
Joined
Dec 6, 2006
Messages
76
First lemme say that this is by far my favorite OCers' forum. I know in my heart that you [H]ard mofos can help me reach my goal of 3.3 Ghz. Let's get into it! Here's my system:

Core 2 Duo E6300
Gigabyte GA-965P-S3 (rev 1.0, C1, BIOS F7)
2 x 1GB G.Skill DDR2-800 (Newegg link for clarity's sake)
eVGA GeForce 7600GT KO
+ 2 SATA hard disks and an IDE DVD burner

I had been running a mild overclock of 400 x 7 for a while, and it was rock-stable (overnight in Orthos, plenty of Half-Life 2 and Oblivion). Today I finally got around to disassembling the whole rig, lapping my Arctic Freezer Pro 7 and reinstalling, and swapping out the stock northbridge heatsink for a Zalman ZM-NBF47. (Used Arctic Silver MX-1 compound for both.) Now that I'm not worried about heat anymore (CPU runs at ~40C idle / 54C full load in TAT), I decided to quit screwing around and really overclock.

But I can't get it to boot any higher than 420 FSB! :mad: This is extremely disappointing. The first iteration of this system -- before the mobo inexplicably gave up the ghost and I had to RMA -- ran totally stable at 450 FSB, and that was with the stock Intel HSF, stock NB HS, and a less mature BIOS (F5 or F6, not sure). With the upgrades I was hoping to hit at least 470....

A little more info that might help. Right now my BIOS settings look like this:
Vcore: 1.35v (1.325 stock)
Vdimm: +.3 (=2.1v)
Vfsb: +.1
Vmch: +.1
PCIe: Auto (also tried locking to 100 MHz)

I have the memory timings set manually to 5-5-5-15 and Auto thereafter. (My RAM is rated to run at DDR2-800MHz 4-4-4-12.)

When I put any setting in the BIOS that the system doesn't like, it restarts itself during POST after the CPU/memory check but before the drives are detected. Sometimes it restarts more than once, then goes back to stock settings and boots fine. A couple of times when I chose Save & Exit in the BIOS, it rebooted but wasn't sending a signal to the monitor. Hitting reset or powering down and then back up fixes this, albeit back at stock settings.

I guess that's all the info I've got. Wondering where to go from here. From doing some forum searches and a lot of reading, here's what I'm considering, in no particular order:

Reset CMOS and start over.
Bump voltage(s) (but which ones, and how much?).
Change BIOS versions.
Read up on the more esoteric RAM timings and see if changing those might help.

So..... what do you guys think? And thanks to everyone who read this far. :)
 
You need to actively cool the northbridge on this motherboard at speeds > 400FSB. Gluing a 50mm fan to the wimpy passive sink helped my overclock immensely:
nb.jpg

I would not be surprised if the explanation for your "inexplicable" failure was that you burned the board out for lack of proper cooling!

Try running the CPU at 1.4125 and the DDR at +.4v (2.2v). Leave the FSB and MCH at +.1v. Also use F5 BIOS because F7 brings more problems than it solves.

If that doesn't help you then you probably got a dud clocker. It happens :(
 
Thanks for the suggestions.

I set Vcore to 1.4v, Vdimm to +.4 (2.2v), left vFSB and vMCH at +.1, and set FSB to 450 MHz -- and it wouldn't POST. No signal to the monitor. So I hit reset, and as usual it had changed most of the settings back to defaults (266x7, most things set to Auto) except the voltage settings, which it left alone. Interestingly, the red flashing "System Voltage Not Optimized OMFG!!!111!!" (emphasis added ;) ) was a nice solid green "System Voltage Optimized," even though it had left all my voltage settings the same, as described above. Not sure what that means, if anything.

I tried the same settings at 440 MHz FSB with the same result.

I read a post last night where someone had tested the actual voltage supplied to FSB and MCH with a meter, and on his DS3 (which is mostly the same as my S3), when the system is overclocked and FSB or MCH voltage are set to +.1 or +.2, it actually supplies LESS voltage than if they're set to Normal. Apparently Gigabyte is second-guessing us behind our backs.

With that in mind, I left all voltage settings at stock (including Vcore) except Vdimm, which I set to +.4 (2.2v). Then I set the memory timings to Auto. With these settings I could boot successfully at 420 FSB, 430 FSB, and 440 FSB (which I'm running now as I type this). I'm going to reboot it one more time and try to hit 450, then do some stability testing.

I'm still using F7 BIOS, but I downloaded F5, F6, and F8 (from GB's .tw site -- it's not posted on the US site yet for some reason). If I can't get past 450 FSB with the current BIOS, I'll do some testing with the others.

Any other ideas in the meantime?

@ Zinn: It seems unlikely -- but possible -- that I fried the old board with too much heat on the NB, because mine wasn't running nearly as hot as some people reported. I could keep my finger on the stock passive heatsink indefinitely, even while doing CPU/memory intensive stuff like encoding video, but others reported it would literally burn them under load. Anyway, now that I have a much better aftermarket heatsink on there with better compound, it barely gets warm at all. So it might still be the MCH that's holding me back, but I doubt it's because of heat. Thanks for your help!
 
Just booted at 450 x 7 with all the same settings as above. The only voltage change is still +.4 to Vdimm, otherwise stock.

I suspect the key change that allowed me to boot at 450 (where 420 seemed to be a wall before) is that I have memory timings set to Auto rather than 5-5-5-15. CPU-Z reports it's running at 5-7-7-19 right now. What do you guys think of this theory?

I'm going to run Orthos for a while now, and assuming I get no errors, I'll test my theory by manually setting the mem timings a little tighter without changing anything else... stay tuned, and thanks again for helping me out.
 
Okay, I think I getting there. I have a bunch of new info which I hope will help you guys help me finish this mofo off. :)

I booted to XP at 450 FSB, all voltages stock except +.3 to vDimm, but Orthos errored on me. I bumped vCore to 1.4 and got the same result, which I figured meant the CPU was fine but the RAM wasn't happy. Also my core temps went up significantly with the voltage change, so I backed that down to 1.375 (.05v above stock) and set vDimm +.4. It was very happy at those settings.

So I set FSB to 460 and left everything else as it was. That made Orthos throw another rounding error after about 25 minutes. In order to get more info, I took Orthos off "Blend" and set it to stress the CPU, and it ran like that for 4+ hours. Then I set it to stress RAM, and got an error almost immediately.

I figured that meant the CPU is happy with its slight voltage bump. Do you guys agree?

Anyway, I decided to keep tweaking... I set vMCH to +.1 with basically the same results, but when I set vMCH to +.2, it was totally stable at 460 FSB. (This is how it's running now.)

I went up to 470 x 7 and it booted into XP but Orthos stopped immediately. Since vFSB was still at stock, I decided to bump that and see if it helped. With vFSB at +.1 got the same results. At vFSB +.2 Orthos froze the system.

That's when I decided to bring it back down to the last known stable config and share my results with you guys and try to get more advice. So, to recap, here's how it's running right now:

FSB: 460
Memory: 1:1 (920 MHz @ 5-7-7-19) (I'll tighten the timings at the last step)
vDimm: +.4 (2.2v)
vMCH: +.2
vFSB: Normal

So what would you guys suggest? I'm really in the home stretch here... I only need 472 FSB to hit my 3.3 Ghz goal, and I think I've got the headroom to do it. TAT reports temps of 54C at full load, and my new northbridge heatsink is barely warm to the touch, even while stressing the system with TAT, Half-Life 2, and video encoding... I feel like I just need to get the right FSB/MCH voltage settings and I'm golden. Please help!

And thanks!
 
What I've found is that if you leave vmch (not sure about vfsb) at normal, it'll automatically scale it up as you raise fsb.

Also, get speedfan and look at vcore2. It's actually vdimm. I've owned two ds3's and one is 0.1v higher than set while the other is 0.1v lower. I've checked with both vcore2 and an actual dmm.

As for cpu, what's the actual vcore when under load? You could be experiencing droop and thus have a lower vcore than what you believe.
 
Hey ziddey, thanks for your reply. I checked it out and here's what I found:

Vcore is set to 1.375 in bios. The CPU-Z and SpeedFan values agree, but CPU-Z has an extra digit of precision, so I'll use those numbers. In XP at idle it's 1.344 and under full load it seems to fluctuate between 1.328 and 1.315. (Actually it seems to switch back and forth between these values; it never displays a number in-between.)

Vdimm is set to +.4v in bios, so it should be 2.2v. Sure enough, Speedfan reports a steady 2.19v.

As for Vmch, I'm aware of the issue you mention. Currently I'm running 460 x 7 with it set to +.2v, and I can run Orthos for 6+ hours no problem. If I try the all the same settings but with Vmch at "Normal," XP boots fine but Orthos kicks out an error, usually in the first 30 minutes. Is it possible that +.2v is actually slightly undervolting MCH, and it's more stable there?
 
Crooked Paul said:
Hey ziddey, thanks for your reply. I checked it out and here's what I found:

Vcore is set to 1.375 in bios. The CPU-Z and SpeedFan values agree, but CPU-Z has an extra digit of precision, so I'll use those numbers. In XP at idle it's 1.344 and under full load it seems to fluctuate between 1.328 and 1.315. (Actually it seems to switch back and forth between these values; it never displays a number in-between.)

Vdimm is set to +.4v in bios, so it should be 2.2v. Sure enough, Speedfan reports a steady 2.19v.

As for Vmch, I'm aware of the issue you mention. Currently I'm running 460 x 7 with it set to +.2v, and I can run Orthos for 6+ hours no problem. If I try the all the same settings but with Vmch at "Normal," XP boots fine but Orthos kicks out an error, usually in the first 30 minutes. Is it possible that +.2v is actually slightly undervolting MCH, and it's more stable there?
Click to expand...

Hmm, it's hard to say unless you take a dmm and actually measure vmch at each of the set values. On my rev1 board, the values were crazy high. I understand baseline should be 1.2v, but with normal, or +0.2, +0.3 etc, I was seeing upwards of 1.8 pretty frequently. Since when was 1.2+0.3=1.8?! But as of such, it's quite possible the northbridge is just putting out too much heat. Maybe you could try changing the thermal compound to as5 or something and putting a 50mm fan blowing down, or swap out for a hr05 or something to cool the nb.
 
ziddey said:
Maybe you could try changing the thermal compound to as5 or something and putting a 50mm fan blowing down, or swap out for a hr05 or something to cool the nb.
Click to expand...

I'm way ahead of you, brother. :) Having had to RMA the first mobo I got, and not being really sure if that was my fault or not, I was careful to the point of paranoia with this one. Before I dared to push the system past 300 FSB, I removed the stock NB heatsink and replaced it with a lapped Zalman NBF-47 ("flower" design). I used MX-1 thermal compound. It barely gets warm no matter how I stress the system. So I'm pretty sure chipset heat is not the issue.

If I had a DMM I'd check the voltages myself, but I don't have one and can't afford one (I know they're only $15 but I'm poor as a mofo and I'm already pretty far over what I budgeted for this system).

I have a couple other ideas I'd like to run past you. So far I've had the memory timings set to Auto, figuring I'd tighten them up once I hit my OC goal. The memory is rated DDR2-800 4-4-4-12, but with Auto set in bios it's running at 5-7-7-19. Do you think maybe those wonky timings could be the problem, or part of it?

Also I still have the F7 bios on the board right now. Gigabyte has just posted F8 (which I believe is the equivalent of the DS3 rev1's F10), but I was waiting to see some positive feedback on it/them before I flash it, and so far the reviews have been sparse and mixed.

Finally I'm thinking it's possible that my PSU is holding me back. When I built this system I didn't understand how important steady power was to a stable overclock; I just went by the rated wattage. So I have a Rosewill 450W, which I guess is regarded as "decent" but not "good." Still, that should be plenty for my rig. I have the components in my sig plus 2 HDDs, 1 DVD burner, a wireless NIC, 2 120mm case fans, and an Arctic Freezer 7.

Does this new information suggest anything to you? Thanks!
 
Yikes, I missed that part of the first post.

I've had the most success personally with ds3 f7 (s3 f6). However, currently I'm on f10 (s3 f8 as you've said). Maybe you could try giving both those a whirl.

SPD for my ram at 450fsb makes it goto 5-8-8-20, and it's not stable. But if I manually set 5-8-8-20 it appears stable. I used to be able to run my ram at really tight timings, but am not sure what happened. They still do 5-5-5-15 stably, but I'm doing 5-8-8-20 currently to see where I can push the cpu first.

I think you might have already tried these settings, but just to make sure:

460fsb, 7x multi, pcie locked at 100 (I usually do 110, so am not sure about 100).
vcore set to 1.425v just for sake of trying. vdimm set to 2.3v for the time being as well (shouldn't be too much voltage, although I recall some non-micron IC's don't respond well to voltage so I'm not sure if this would be detrimental to success).

set timings to 5-8-8-20 and leave subtimings to auto, and trfc to 0 (which I think defaults it to 53).

Make sure memory performance enhance is normal.

Leave all the vmch, vfsb, vpcie at normal and see how it does with autoscaling.

Disable c1e and eist for cpu.

Some people reported on the biostar motherboard that enabling the cpu temperature warning caused strange results in overclocking. I've never heard similar for gigabytes but I guess it doesn't hurt to turn it off for the time being.

If these settings don't work, you could try getting real aggressive to see if it even posts / primes for more than a minute by turning vcore way up to say 1.5v or something. It gets real toasty up here and iirc the arctic freezer isn't really able to keep it tame, but temps shouldn't go above 80'c, and for the sake of testing, I suppose you could try it. It could be an unlucky cpu. Or maybe unlucky northbridge, since you said you have c1, which I also have had, and got 450fsb.
 
I don't know if this is helpful or not, but I have a 6300 with a Gigabyte DS3 motherboard and it won't go over 400 x 7 no matter what I do. I may be held back by the memory I'm using which is DDR2-667 memory that is currently running at 800. It is stable as a rock at 400 x 7 though. I get temps of around 57C using TAT running folding@Home on each core constantly. I am also using the stock heatsink/fan and have made no modifications to the northbridge or southbridge cooling. I tried the f9 bios for the DS3, but it wouldn't let me overclock then and I didn't want to fool with it, so I flashed back to f7 and now my overclock is working great. F7 supports the quad core processors for the DS3, which is really what I want. I may flash up to the newest bios and buy faster memory once I get a quad core though.

Like I said, I don't know if this is helpful or not, but maybe some chips just won't go really high. I would love to be at 3GHz+, but with my current memory I just don't think it is going to do it. The other thing I forget to mention is that the only voltage I've made is to add .3volts to my memory. Everything else is at stock and I'm running the memory at 4-4-4-12.

Good luck to you.
 
okay well one thing that i never heard was you running memtest to see if you mem was passing at 450 with any timings or what ever vmem. so start testing your RAM. at stock speeds and timings and then OC your RAM and make sure it passes 1 whole round of memtest before you boot into windows.... it sounds like it could be a RAM or memory controller issue...
 
Thanks everyone for additional suggestions... I spent a few hours tonight running memtest w/ various BIOS settings. Long story short: no great breakthroughs, but possibly some more useful info. I'll post a complete rundown after I've had some sleep; if I do it now I know I'll garble it and that won't help anybody. I'm sooo tired... :eek: zzzzz
 
Okay y'all, here's the news. I've spent god-only-knows how many hours testing 23 unique configurations since last I posted, so I'll try to give the Cliff's Notes version.

Here's what my settings were at the beginning of this epic journey:
460 FSB x 7 multi (3.22GHz)
vDimm +.4 (2.2v)
vMCH +.2 (?.?v -- dang Gigabyte and their nonexistent documentation)
vCore 1.375v
mem timings Auto (5-7-7-19)

That configuration ran Orthos stably for ~7hrs. But when I took Adidas's suggestion and ran memtest on it, I got hella errors on test 7. We're talking like 110+ errors per pass, and they were all over the address space. Quite odd, since none of the other tests produced any errors at all after 4 or 5 loops.

I hatched a hypothesis that this odd behavior was caused not by bad RAM, but by some idiosyncracy in the MCH. (Please check my work on this one -- does that make sense to you guys?) I should mention that when I first got this DDR2 set I ran memtest error-free on it for 6+ hours at mostly-stock (266 FSB, 2:3 divider, 400MHz DDR2, 2.0v) with slightly loosened timings (5-5-5-15). Back in the present, having just seen all those errors on test 7, just to be sure I reset everything back to those settings and tested it again for ~2hrs and it was fine. Then I switched it up to 400FSB, 1:1, 5-5-5-15 and that was fine. Just for fun I tightened the timings to the modules' official rating of 4-4-4-12 and that passed memtest too.

So I know the memory will pass memtest at stock settings. As for overclocking? To make a long story short: according to memtest, this ram is aces all the way up to 465Mhz w/ 2.1v, FSB and MCH at stock, 5-6-6-18 timings. But if I boot to XP with anything over 450, Orthos spits errors within the first 5 minutes. (Frequently it can run the "stress CPU" test indefinitely, but "Blend" or "stress RAM" consistently cause errors in the first 5 minutes.) Note that there's a "no man's land" between 450 and 470 where Memtest and Orthos disagree. With different voltages set, I can get either Memtest or Orthos to run without errors for ~1hr+, but never both with the same voltage settings. So WTF there?

At this point I kinda decided that the only way I was going to be really stable above 450 FSB was to try a different BIOS, so I went ahead and flashed F8 on the board.

Once again to make a long story short (skipping over about 12 configurations here), the best settings that are error-free in both Orthos and memtest are these:

450 x 7
vDimm +.3 (2.1v)
vCore 1.35 (1.375 works too, but isn't any stabler and produces higher temps)
FSB and MCH Normal
timings 5-6-6-18

Even with the F8 BIOS, any FSB higher than 450 means I can get either Memtest or Orthos to run stably by tweaking various voltages, but never both with the same settings. Ack!

I think the memory is pretty happy at 2.1v. I tested several configurations with all the same settings except vDimm 2.1v versus 2.2v, and the 2.1's would pass Orthos while the 2.2's would fail (both passed memtest). At this point I'm kinda reluctant to go up to 2.2v since it seems to hurt rather than help. But that leaves me rather at a loss of what to do next. I've tried the same config as above with FSB/MCH set to +.1/stock, stock/+.1, and +.1/+.1, and all three fail Orthos within 15 minutes. I haven't tried any higher settings on them yet. Would it be at all worthwhile to throw some +.2's or 3's into the mix, given that any .1v bump destabilizes a formerly-good config? Seems doubtful.

I feel pretty confident that something on the mobo or in the BIOS is holding me back. The CPU doesn't run very hot, and it will pass Orthos's "stress CPU" tests even when Blend and RAM tests error. This has happened pretty consistently across multiple configurations on both F7 and F8 BIOSes. Similarly, I feel like my RAM sticks are probably good at least to 465MHz, since they pass 1.5-2hrs of memtest in at least three different configs at that speed.

What do you guys think? Is this making any more sense to you than it is to me? Maybe I should flash the F6 BIOS or maybe - maybe - the DS3's F10. Is that a good idea? Sorry for the long post; I really tried to keep it pithy. As always thanks for any suggestions or comments. (Also ignore my sig for the time being... the last-known-good config in this post is how it's running presently.)
 
I greatly increased the stability of my DS3 over 400 FSB by putting a small 50 mm fan of the northbridge.

Also, I noticed that my system is much more stable when i leave vMCH to AUTO because the BIOS increases the voltage according to the FSB you set.
 
Thanks for your comments, blurp. I've had similar results up to 450, but past that I'm running into real trouble. If I had a E6400 I'd be running at a sweet 3.6 (potentially, depending on how much heat/voltage the chip could take), but with the E6300's 7x multiplier it looks like I'm stuck at 3.15GHz for the time being.

I guess I'll have to wait for an improved BIOS release. Or maybe in a couple weeks I'll be able to get a better power supply and see if that helps. Anyone have any other suggestions?
 
Im in a similar boat as you, except I have an E6400. I can't get it rock stable in the 3.5-3.6GHz range, but it also bugs me that it goes over 60c with my watercooling. I'm currently running this, until my new PSU shows up.

FSB 420x8
Ram 5-5-5-12 ~525MHz
1.45v cpu
2.1v mem (just ordered a 700w GameXStream because 2.2v fails with my current 420w Enermax)
Auto on FSB and MCH voltages, I found auto to be much more stable.


I'll get back to you and let you know if the new PSU is more stable. My temps at the above settings are 43c idle and 57c load, but its also 80F here right now so that ain't helping. Thanks for the informative thread, you inspired me to push mine more :D
 
Tahoe, since your temps are bugging you, have you tried lowering your vCore? Does it really need 1.45 to be stable? When I went from 1.35v to 1.4v for testing purposes, my temps went up like 6C. Right now I'm running the chip at 3.15GHz (450x7) at stock voltage of 1.325v, with temps of 40C idle / ~54C load.

Anyway, it's not too shabby to be "stuck" at 3.15GHz. I just finished tightening up my memtimings, and they went lower than I would have guessed running the memory at 900MHz. Got 'em down to 4-4-5-12, which is a pretty sweet consolation prize.

Thanks to everyone who took the time to read & post advice. You guys are teh roxors.
 
Tahoe, I thought of another thing. Maybe you're stuck at 420 because of the dreaded "FSB hole" that most boards have in the 420-449 range. As I understand it this is due to the MCH being massively overclocked until you get up to 450 and a new strap setting kicks in. You do have a 965-based mobo, right? I couldn't find it in your sig.

If you can get your temps down a little more (maybe try lowering vCore like I said), I'd say skip that whole range and go for the juice! Just set it to 450 or a little above and see if it works. With water cooling I think you might could run the chip that fast.
 
Yes I have the same board as you. This chip seems to need a lot of voltage to run like bull. Buuut I also noticed a fairly large voltage droop, and I think my 420w PSU is starting to hernia from my 4 HDs, 7900GS OC'd, 2 DVDrws and the duo, mobo, fans, etc.

Once the PSU comes in I'm going to go for the gusto and set it to 450FSB, which I have done already and its posts and loads windows but craps out of ORTHOS after a short while. I'll update you when I get it in a couple days.
 
Ohhhhh baby, you have hella drives in that beast. Well, I didn't realize that. I totally agree with your reasoning and I wish you good luck!
 
Since we have the same ram I'm going to try some new settings once I lap my cpu and ninja.

Thanks for the boost in confidence. :)
 
Yeah this G.skill memory is the balls! Mine's at 900MHz with timings 4-4-5-11 at 2.1v (only .1 over stock). Can't argue with that. I'm almost positive I could hit 1000 with it if I loosened the timings and my mobo could go higher than 450 reliably...
 
I'm going to go out on a limb here and ask why the hell you aren't pushing the vcore any higher but expecting magical results

Push that bitch to 1.5v and see what she can do

that is unless you are running the stock cooler
 
Totally valid question. For one thing I don't think the chip is the limiting reagent in my current setup. I think it's either a finicky MCH or my power supply isn't up to the task of running it faster.

The reason I say this is that during my stability testing, Orthos would routinely pass the "small FFTs" CPU stress test but barf 5 minutes into either blend or RAM tests. If I bump the vDimm it's even worse (and my memory timings were a damn sight looser then).

Also, any push to the vCore sends temps WAY up. Like at 1.4v my load temps are 62C-ish. I know that's not dangerous, but I'm not going to run it like that 24/7 nohow.

For cooling I've got a lapped (but not obsessively so) Arctic Freezer Pro 7 with MX-1 compound (which is the balls), and I got a lapped Zalman on the NB. CPU temps are about 53 at load.

I got a CPU-Z validation link for my current setup. I'm happy with it for the time being. I'll try again when a new BIOS drops or I can afford a really good PSU.

http://valid.x86-secret.com/show_oc?id=161770
 
You must log in or register to reply here.
Back
Top