Need help stabilizing E8400 on IP35-E @ 4GHz

I have the following configuration and am having difficulty getting my o/c to be completely stable.
E8400 Q746A483 @ 9x445, VCore set to 1.465 in BIOS, actual voltage according to CPU-Z and Abit EQ 1.392 @ idle and 1.360-1.376 @ load
NB @ 1.33V, VTT @ 1.2V in BIOS but reading 1.1V under load according to Abit EQ, GTLREF @ 63%
Abit IP35-E BIOS 15 (latest official)
Zerotherm Nirvana 120 (fan @ full RPM)
Geil 2x2GB DDR2 800 o/c'd to 890MHz using 1:1 (5-6-6-15-2T @ 2.1V)
rear 92MM exhaust fan
front 80MM intake fan
open case (mid-tower)
XP Pro

My system fails memtest test #7. I have tried bumping the voltage of both the RAM and the northbridge, but have seen no difference. All other tests pass without issue. My HSF, NB HS, and RAM heatspreaders are cool to the touch during memtest so heat isn't the problem.

This is my first endeavor at overclocking an Intel platform, so having to tweak all these timings and voltages is quite new to me. I'm thinking I need to either throw more volts at this or relax timings even further. I manually set TRP to 6, RAS to CAS to 6, TRAS to 18, and TRFC to 52.

Here's a screenshot of the "usual suspects":

http://img.photobucket.com/albums/v2...ran/c97745.jpg

Does anything jump out at anyone? Temps seem fine to me, all voltages are well within "acceptable limits" from what I've gathered reading this forum, RAM timings are severely relaxed from stock, I just don't get it...

Do I just need to buy better memory or what? I was thinking of getting G.Skill's 2x2GB PC8000 kit, so if that's what it's going to take then that's what I'll do.

Thanks in advance.

I don't have your mobo and am not that familiar with overclocking 2gb sticks but it seems to me that your northbridge is a bit low (most run 1.4-1.55) and definitely your ram - ddr2 ram at least for the 1gb sticks need 2.2v usually for that fsb. Of course lock pci-e to 100 and turn off eist/speedstep, etc.

Quote:

---

Originally Posted by LagunaX

I don't have your mobo and am not that familiar with overclocking 2gb sticks but it seems to me that your northbridge is a bit low (most run 1.4-1.55) and definitely your ram - ddr2 ram at least for the 1gb sticks need 2.2v usually for that fsb. Of course lock pci-e to 100 and turn off eist/speedstep, etc.

---

Thanks, I'll try upping the DDR & NB volts a bit to see if it helps. PCIe clock is already locked @ 100 and all power-saving CPU features have been disabled.

Edit: no go. Up'd DDR to 2.2 and NB to 1.45, if anything it actually failed the tests quicker (only took 2 minutes on test 7, usually takes a full pass). Also, I was able to run small FFTs in Orthos for about an hour and a half without issue, but 30 seconds of SMP FAH caused the system to reboot. The system is stable under heavy CPU load, but not memory-intensive tasks. I know my CPU can take it, so that means either my memory isn't up to the task or the board isn't.

Looks like voltage isn't the answer. Any thoughts on the timings I'm running? Have I over-looked a particular setting that may give me stability?

try matching vtt with vcore.

How do we match VTT and Vcore?

E8400 Q746A483 @ 9x445, VCore set to 1.465 in BIOS

this is way too much vcore....keep it 1.4v BIOS.

Quote:

---

Originally Posted by nzbleach

try matching vtt with vcore.

---

Quote:

---

Originally Posted by aoch88

How do we match VTT and Vcore?

---

Yes, how do we match vtt and vcore? AFAICT my board only allows VTT up to 1.2375, and from what I've read these chips do not work well with high VTT.

Quote:

---

Originally Posted by TEDY

E8400 Q746A483 @ 9x445, VCore set to 1.465 in BIOS

this is way too much vcore....keep it 1.4v BIOS.

---

Abit IP35 boards undervolt. A lot. Actual vcore under load (Orthos small FFTs) is 1.36-1.376. Individual core temps do not reach 60 degrees even after 90 minutes of this.

There's a new beta BIOS out for the board which some other abit P35 + Wolfdale owners are reporting reduces the drop by quite a bit. I should be able to set vcore down to around 1.4 now for the same values (about 1.36V under load) in Windows. I'm going to try that tonight, as well as perhaps upping VTT and see if I can get stability.

Raise your tWTR to 15 or higher.

And I don't see any reason to use 5-6-6-6-15-2T , if 5-5-5-5-15-2T doesn't work @ that speed, throw those chips away.

Thanks Spacehead! I'll raise tWTR and reset all other timings (except TRFC) back to default.

Upping VTT is not necessary for this OC - and I would go as far to say that increasing it all, given your FSB is just terrible advice. For your FSB, NB voltage above 1.33 (stock) is not necessary. I left both of those at stock on my OC, and from my notes, nothing but vcore was required to scale up in clock rate, especially since you're on the 9x multiplier.

I would definitely change the RAM out. To be blunt, your RAM should never fail memtest - and if it does, either add the necessary voltage or get some better sticks. It goes without saying that your RAM should be passing with flying colors before doing anything else.

Thanks for the advice Mav. Guess I'll just try increasing the tWTR setting and maybe applying more volts to RAM. It does make sense that none of the other voltages need to be changed, as I haven't seen them have any impact thus far.

Let me pose another question: if I do get some better RAM (say G.Skill PC8500 4GB kit) and shoot for a matched FSB (in this case 533MHz) is it likely that I will then have to increase any of the other voltages?

Thanks

If you go for a 533 fsb then you will most likely have to up other voltages, if you run the ram at a 5:6 timing then your FSB would be about 444 which shouldn't need much of a boost on the NB if any at all.

When I started with my CPU testing to see how high I could go with it I set my ram to 2.2v and the timings at 6-6-6-6-20 just to make sure the ram wasn't affecting my testing at all. But then again I've got cheap ram and I knew it going in.

Not sure about that ram that you're using. You tested it known good in another platform to know it can hit those speeds? As mentioned, try slacking timings and subtimings as much as you can at first to see if anything happens. I had a ip35e as well and found it to be incredibly easy to overclock. I struck 500fsb on stock motherboard volts with a e6400.

I usually set pcie a few higher. 104 or so. Figure it might bring me that slight boost ;)

Well, flashed to a new beta BIOS last night that has better Wolfdale support, also tried all the advice given thus far but I still can't pass memtest #7 at these speeds. Looks like it's new memory time. Now the question is: G.Skill 8000 or 8500 4GB, or another brand?

Never mind, I see you've already tried that.

Try testing with plain old Prime 95 instead of Orthos.

ballistix 8500

Quote:

---

Originally Posted by d50man

ballistix 8500

---

Thanks, but no thanks. I'm not touching Crucial Ballistix with a 10 foot pole. Last time I used Ballistix I had to RMA it 5 times before I got a set that worked. That was DDR but I still hear plenty of reports of their DDR2 Ballistix being just as bad.

Also I need 4GB (2x2) kit.

Wow that is some extraordinarily bad luck. I too had to RMA my DDR, but they actually refunded shipping too (which nobody does these days). So I gave them another chance for DDR2 and here we are :D

*shrug*

You never mentioned wether your ram passes memtest at stock. Does it? And if increasing your Vmem made you fail FASTER.. then try active cooling on ram and NB if you dont already have it, as failing faster with a bump in V. sounds like a possible heat issue.

Quote:

---

Originally Posted by little_scrapper

You never mentioned wether your ram passes memtest at stock. Does it? And if increasing your Vmem made you fail FASTER.. then try active cooling on ram and NB if you dont already have it, as failing faster with a bump in V. sounds like a possible heat issue.

---

Hmm, thought I did... Yes, it passes @ stock settings no problem. 5-4-4-12 timings (I thought it was 5-5-5-15 for some reason) @ 2.0V and 400MHz I can throw memtest at it all day long, even just test #7.

As for heat - definitely not a problem. Even at the highest settings this stuff isn't even warm to the touch. My CPU fan draws air over the heatspreaders and I have an open case.

Problem in the ddr2 world is that micron has such a strong leg up over the other companies. I'm not sure exactly what IC's the geils use, but I've used a few non micron ic's that just hit an absolute wall in the mid 400's despite voltage or slacking timings.

So you're talking :banana::banana::banana::banana: about ballistix. Here's the problem. Depending on which gskill you go for, it'll be 1. more expensive, and 2. have the same problem as the ballistix if they are d9gmh/gkx. I've gone through a bunch of d9gmh myself, only once or twice exceeding the "recommended" voltage. Despite good cooling and whatnot, it's still very likely for these chips to die early. You kind of just have to live with that.

That said, there are reports of newer ballistix (at least the higher speed ballistix) using high density chips that don't overclock well at all and indeed can likely blow up within a week.

I'm not necessarily agreeing with your hatred for ballistix (I just popped in 2 ballistix a few days ago and am absolutely in love). But I do understand where you're coming from. And if you do decide on some d9's, you may be in for some upset down the road.

I'm trying to stabilize my E8400 on an ABit IP35 PRO at 500MHz, and leaving CPU for now at 6x so 3GHz. I was only able to hit 500 FSB after changing the RAM (2x1g Crucial Ballistix 8500) from Auto to Manual.

1. I've got the RAM volted at 2.2v... is that enough? It's working 1:1 with the FSB, so is it potentially overvolted?

2. I changed to manual the first four settings... 5-5-5-18... would anybody know the other subsettings that I ought to manually set aside from tWTR to 15 per earlier post?

The issue is that I'm testing 3dmark06, and it'll begin stuttering in video, then BSOD or hang. Any other voltages I should be focused on right now? I've got mch running usually around 1.52... have had to modify other voltages here and there, but haven't found a much concrete way to proceed... not sure which voltages apply directly to having a stable FSB

Thanks!

---
E8400 | Abit IP35 Pro | 2x1GB Crucial Ballistix 1066 | Antec 900 w/ ZeroTherm NV120 | EVGA 8800GTS 512MB G92 | Cooler Master Real Power 650W

mezryn - you need to figure out the difference between BIOS settings and Windows (or real) settings. Abit is massive vdroop as well (main reason I stayed away), but be sure to take notes on the differences.

E.g. say you set 2.23v for RAM in BIOS >> what does it give you in Windows? Exactly 2.23? DFI, for example, is -.05 for a Windows reading of 2.18vDIMM. Or vcore, for example. DFI - 1.28 in BIOS results in 1.296 in Windows. The stickler is that DFI has 0 vdroop during load, so obviously emulated my settings on a non-DFI board is a bad idea.

Quote:

---

Originally Posted by ziddey

Problem in the ddr2 world is that micron has such a strong leg up over the other companies. I'm not sure exactly what IC's the geils use, but I've used a few non micron ic's that just hit an absolute wall in the mid 400's despite voltage or slacking timings.

So you're talking :banana::banana::banana::banana: about ballistix. Here's the problem. Depending on which gskill you go for, it'll be 1. more expensive, and 2. have the same problem as the ballistix if they are d9gmh/gkx. I've gone through a bunch of d9gmh myself, only once or twice exceeding the "recommended" voltage. Despite good cooling and whatnot, it's still very likely for these chips to die early. You kind of just have to live with that.

That said, there are reports of newer ballistix (at least the higher speed ballistix) using high density chips that don't overclock well at all and indeed can likely blow up within a week.

I'm not necessarily agreeing with your hatred for ballistix (I just popped in 2 ballistix a few days ago and am absolutely in love). But I do understand where you're coming from. And if you do decide on some d9's, you may be in for some upset down the road.

---

Hey, thanks for coming in my thread and spreading your Ballistix fanboy crap all over. Great contribution.

Did you not read where I said I had to RMA a set of Ballistix 5 ****ing times? Would you go back to a company that consistently screwed up over and over and over and over? I sure as hell won't.

As for your down-playing of G.Skill - I've already mentioned which G.Skill memory I'm looking at (4GB 8000 & 8500 kits) and they use powerchips and are quite affordable.

So I'm still trying to stabilize this thing, as even at stock settings it'll reboot randomly under load. I run folding @ home SMP 24x7 and I can count on the system rebooting on me at least once per day (usually overnight). However, I can run OCCT, Super Pi 32M, Orthos blend/small fft, 3dmark 06, play any game I want and everything runs fine.

I pumped all my voltages up just to be sure that any under-volting by the board isn't causing instability, CPU is @ 1.36 under load, MCH is @ 1.57, VTT @ 1.2375 and GTLREF @ 63%. There are no hotspots in my system. Great ventilation and direct heat dissipation on all components so I really don't think it's a heat problem.

At this point I'm about ready to just scrap the whole :banana::banana::banana::banana:ing thing and rebuild with all new components.

Oh yeah, forgot to mention that when it reboots - it's instantaneous. There is no BSOD (and thus no error logged in event viewer).

I had similar problem on Abit IP35
My settings in BIOS was:
vcore 1.485v (1.368v CPUz under Load)
VTT 1.3v (1.22v CPUz)
NB 1.45v
On those settings was fine.

there is a beta 16 bios that support e8400 series better on abit ip35-e.

445fsb shouldnt be that stressful.

I had some issues with the board I had with 500fsb, couldnt get the system to be stable at 8x500fsb using a e8200.
Lower was fine but I knew the cpu could handle more.
I bought a gigabyte cheap p35 ds3 and got it working at 4ghz.
Had some issues with that to until I had a lower vcore to fsb which worked well.
mch however got some juice.

I've got the latest beta BIOS for IP35-e, vdrop is much lower (well under control now). It's only about .01V @ idle and .025 under load.

What concerns me here is that even at stock settings I have this rebooting issue.

Did you end up trying new sticks or RAM?
I'm re-reading the 1st page, but I don't think you mention the PSU that you're using.

Haven't tried new RAM yet. I've got some old 533 stuff I could use, but this Geil 4GB 800 kit passes memtest at the settings I'm currently running (840MHz 5-4-4-12-2T 2.1V). If I put the 533 stuff in I wouldn't be able to o/c at all. In fact, I'd be lucky to run stock settings with a 1333FSB chip.

I have a Corsair550VX PSU. ATX 12V is rock-steady @ 11.97 under full load according to abit EQ.

Is that GTLRef setting stock? Mine's basically stock for DFI, which is 110 I believe. I guess see if changing it back to stock helps (or not).

Quote:

---

Originally Posted by Mav451

Is that GTLRef setting stock? Mine's basically stock for DFI, which is 110 I believe. I guess see if changing it back to stock helps (or not).

---

63% is stock setting for GTLREF. I think I'm going to reload Windows and see if that helps. If not, I'll try again on this spare nf7100 board I have. I still plan on getting new memory anyway because the stuff I have now just doesn't o/c for :banana::banana::banana::banana: but at least it passes memtest so I doubt memory is the issue.

Quote:

---

Originally Posted by techguy2k5

63% is stock setting for GTLREF. I think I'm going to reload Windows and see if that helps. If not, I'll try again on this spare nf7100 board I have. I still plan on getting new memory anyway because the stuff I have now just doesn't o/c for :banana::banana::banana::banana: but at least it passes memtest so I doubt memory is the issue.

---

i had the same exact problem as you with my memory. it would reboot randomly especially over night but would be rock solid stable during the day when benching and playing games. i was 2 weeks trying to figure this out. the last hardware i changed was infact a new kit of ballistix memory. under memtest they ran fine for 1 pass each time i tested even at stock. when i realized that the problem was over my head i dropped everything at stock settings. it worked well for a few days then it restarted to do the same thing... so i got tired and i swaped the memory with my gf's memory. the problem was gone. but the problem was on my gf's side. after few more days of benching and all NOW memtest failed and all. i got a new kit of memory and since it was all good (except that new kit just died few days ago, but its d9 chip so expect it, my first kit was also d9).

the lesson here that i've learned is that even if the memory passes memtest it doesnt mean the memory is quite 100% stable, infact that is why you can play games and bench during day and on idle during night it reboots. and eventually the memory will get worse and worse and might fail memtest. try to run memtest like 10 hours straight like over night or something. test the memory in an other setup if you can. you could also maybe test your 533memory and downclock your e8400 a little bit just to see if the problem persist.

anyways im not saying your ram is the culprit here, but your story is so similar to mine, so i thought that i could share my experience and try to help you:up:

Go Ballistix!!!!! Best Ram Ever Made!!!!!!!!!

Quote:

---

Originally Posted by finch

i had the same exact problem as you with my memory. it would reboot randomly especially over night but would be rock solid stable during the day when benching and playing games. i was 2 weeks trying to figure this out. the last hardware i changed was infact a new kit of ballistix memory. under memtest they ran fine for 1 pass each time i tested even at stock. when i realized that the problem was over my head i dropped everything at stock settings. it worked well for a few days then it restarted to do the same thing... so i got tired and i swaped the memory with my gf's memory. the problem was gone. but the problem was on my gf's side. after few more days of benching and all NOW memtest failed and all. i got a new kit of memory and since it was all good (except that new kit just died few days ago, but its d9 chip so expect it, my first kit was also d9).

the lesson here that i've learned is that even if the memory passes memtest it doesnt mean the memory is quite 100% stable, infact that is why you can play games and bench during day and on idle during night it reboots. and eventually the memory will get worse and worse and might fail memtest. try to run memtest like 10 hours straight like over night or something. test the memory in an other setup if you can. you could also maybe test your 533memory and downclock your e8400 a little bit just to see if the problem persist.

anyways im not saying your ram is the culprit here, but your story is so similar to mine, so i thought that i could share my experience and try to help you:up:

---

Well, I am planning on swapping the memory out for another 4GB kit anyway, so I certainly won't feel bad if that fixes the problem. Until then I think I'll just keep plugging along the way I have it now. No biggie if I have to restart once or twice a day.

I think I've finally got it stable @ 3.87 (9x430) until I can get some decent overclocking RAM and shoot for 4GHz (7.5x533) or better. I had to overvolt the crap out of everything to do it though. Vcore set to 1.4 in BIOS, reads 1.376 under load and 1.4 @ idle which is way higher than I would like but so far so good and temps are still only in the 50's per core under full load.

Also had to bump up the RAM voltage to 2.2, MCH to 1.57 (this could probably be brought down but it's stable so I'm leaving it for now), VTT to 1.2375, GTLREF still @ 63%, and I reduced my RAM's timings to 5-5-5-15 with tRFC at 42 and tRD at default 8.

Finally had the thing running FAH SMP for more than 12 hours so if it's still on when I get home from school tonight I'm calling it stable.

So based on the small changes I made, it may have just been vcore that did the trick. I can't believe I need such a high vcore for this chip and have yet to get 4GHz stable, although that's a separate issue for which my cheap RAM is at fault.

Wow, I don't think I've ever heard of RAM causing a system to reboot randomly at night (but selectively NOT during stress testing). I guess I'm lucky in that sense. Either the RAM was good, or it was chock full of errors. Sounds more like a power delivery issue if it's slowly degrading over time tbh.

Quote:

---

Originally Posted by Mav451

Wow, I don't think I've ever heard of RAM causing a system to reboot randomly at night (but selectively NOT during stress testing). I guess I'm lucky in that sense. Either the RAM was good, or it was chock full of errors. Sounds more like a power delivery issue if it's slowly degrading over time tbh.

---

I don't think it's necessarily degrading, I've never really had this thing completely stable since I've built it, I've just been slowly ramping up volts until it became stable.

Quote:

---

Originally Posted by techguy2k5

I think I've finally got it stable @ 3.87 (9x430) until I can get some decent overclocking RAM and shoot for 4GHz (7.5x533) or better. I had to overvolt the crap out of everything to do it though. Vcore set to 1.4 in BIOS, reads 1.376 under load and 1.4 @ idle which is way higher than I would like but so far so good and temps are still only in the 50's per core under full load.

Also had to bump up the RAM voltage to 2.2, MCH to 1.57 (this could probably be brought down but it's stable so I'm leaving it for now), VTT to 1.2375, GTLREF still @ 63%, and I reduced my RAM's timings to 5-5-5-15 with tRFC at 42 and tRD at default 8.

Finally had the thing running FAH SMP for more than 12 hours so if it's still on when I get home from school tonight I'm calling it stable.

So based on the small changes I made, it may have just been vcore that did the trick. I can't believe I need such a high vcore for this chip and have yet to get 4GHz stable, although that's a separate issue for which my cheap RAM is at fault.

---

this sounds wierd generally wolfies will oc higher then that with less vcore. at least 4ghz easly. my wolfie needs 1.325v for 4ghz, 1.362vcore for 4.1ghz and 1.450vcore for 4.3ghz considering vdroop at load this makes 1.376v.

Yes, it is quite a strange situation. I haven't been able to devote a lot of time to testing either, I'm making larger changes than I should be but it's definitely stable now so until I'm ready to switch to some good RAM I'll just leave things as they are. Once I get a better 4GB kit I'll re-visit the actual voltages necessary to attain my desired o/c.

Update (as posted in the retail e8400/e8500 o/c'ing thread)

Finally got new memory in last night, G.Skill PC8000 4GB kit. Man is this stuff fast. With some additional cooling on the northbridge and RAM I'm testing the limits of stability @ 525 FSB and 4.2GHz. I can load Windows and use it normally, but have not yet stabilized serious benchmarking/stress-testing. Was able to get a quick Super Pi 1M run in though - 11.125 seconds. Everest mem speeds were 8.66GB/s copy, 8.5GB/s read, 11.1GB/s write, and 54.5ns latency. I feel like a little more tuning (and an O.S. reload - thanks to earlier corruption) could get things running perfectly at these speeds.

Unfortunately, I've got to pump some serious volts through the various components even for the relative stability I have now. 1.425 Vcore (1.4 in Windows under load), 2.1 Vdimm, 1.57 Vmch, Vich 1.55, Vtt 1.2375, GTLREF @ stock 63%. Also had to raise tRFC to 55 (all other timings auto). I don't think it's a smart idea to raise any of these voltages further on air cooling, except Vtt. I think I can probably tighten some of the sub-timings on the RAM, since I'm only o/c'ing the DIMMs by 5%.

Any suggestions for RAM timing changes or other voltages to try? Does vTT need to be increase with vMCH so high? Is it safe to go higher than 1.57 on vMCH? The next step for me in the BIOS is 1.67, gut instinct tells me that's too much for air.

Well, 4.2 isn't stable afterall, in fact it seems 8x500 is about as high as it goes with 100% stability, but that's fine by me since 4GHz was my original goal anyway. If I can tighten the RAM timings for a little kick in the pants I'll be even happier. Stability tested with Super Pi 32m, OCCT, and Folding@Home SMP (overnight).

Here's a SuperPi screenshot showing results for 1M and 32M. Some of you folders may note the sub-10 minute frame times on a 2653 WU which equates to 2500-2600 PPD. Pretty darn nice for a dual-core :D

http://img.photobucket.com/albums/v2...an/spi4ghz.jpg

Now I just have to figure out the max stable clocks for my 8800 GT SC. I can't seem to get through the last game test (Deep Freeze) with my previous high clocks of 720/1805/1940. Memory seems stable up to 2000 now with tnew DuOrb cooler on it, but even 700/1750/2000 locks up the last test now (not a hard lock, but I get the "D3D9 device lost" error message and the driver resets). Guess I have a lot to look forward to when the 9800 series comes out :p