Bump voltage or RMA?

[DarkRogue]

Hello,

I finally finished putting together my new rig with the E8400 and 4GB of ram.
Got Windows and everything installed and now I'm trying to OC it a bit.

I'm still dealing with some random reboots here and there but that *might* be my PSU, I'm not sure. More on that later.

I'm running Orthos to do some stress testing (only that for now as I don't have too much time to tinker with it.) At STOCK speeds (333x9 1.225v) Orthos runs for hours on end perfectly fine with Blend (until the PC reboots, may be unrelated.)

Next, I set it to 400x9 1.225v for 3.6GHz at stock voltage. I re-run Orthos and this is where things get sticky. If I run the CPU-intensive Orthos tests (small FFTs) it will run forever (or until a reboot, again possibly unrelated.) However, as soon as I hit Blend or the Large FFTs, Orthos will fail in under a minute. This leads me to believe there may be an issue with the RAM.

I have 4GB (2x2GB) of the Corsair TWIN2X4096-6400C5's. They're rated at 1.9v and it's currently (auto) set to 1.9v. They're also at default/stock timings, I haven't touched it (5-5-5-18.) I ran Memtest86+ on STOCK CPU speeds (333x9 1.225v) and it ran for 11 hours (18 passes) with no errors. As soon as I upped CPU to 400x9 1.225v though, Memtest86+ detected 110 errors throughout Pass #1. I'm not entirely sure what this means, do I need to boost its voltage a bit or RMA, or what's going on?

As far as the rebooting issue goes, it seems to have been carried over from my old PC or something as I was getting random reboots there too recently before I built this rig. There's no pattern to its rebooting, it'll reboot whenever no matter what I'm doing to replicate it. The only things I've re-used from the old PC are the PSU, Raptor74 (admittedly with some bad sectors, but it's not my current boot drive), a WD 500GB drive and the Sound Blaster Audigy. Everything else is new, but I'm thinking it may be my (relatively new) PSU being ddefective or something.. I can't exactly prove it though so I don't know if Newegg will replace it.

Anyway, sorry for the long post, but any help is appreciated, thanks in advance.

Edit: Typos.

 

[BlueWeasel]

Bump the RAM voltage to 2.00v, rerun Memtest, and see if you get the same errors. You could also test each 2GB stick individually.

 

[DarkRogue]

Oops, I forgot to mention the motherboard lol, sorry.

I am using a Gigabyte GA-EP35-DS3P Rev 2.1 with the F2 BIOS.
MIT settings..
Robust Graphics Booster - Auto (What does this do?)
CPU Clock Ratio - 9X
CPU Host Clock Control - Enabled
CPU Host Frequency - 400
PCI Express Frequency - 100 MHz
CIA 2 - Disabled
Performance Enhance - Extreme
System Memory Multiplier - 2.00 (1:1)
DRAM Timing Selectable - Auto
System Voltage Control - Manual
DDR2 Overvoltage Control - Normal
PCI-E Overvoltage Control - Normal
FSB Overvoltage Control - Normal
(G)MCH Overvoltage Control - Normal (If this is the NB, what is the FSB one for?)
Loadline Calibration - Enabled (Not entirely sure about this one either)
CPU Voltage Control - 1.2250v
Normal CPU Vcore - 1.2250v

Also, EIST and C1E are disabled until I can get the overclocks stable.

@BlueWeasel - 2.00v seems quite a jump for the RAM.. shouldn't I try 1.95v first?
I'm not sure if testing the sticks individually will do anything, Memtest detected errors from the 128K - 2048 range, then more from there until the 4096 range, so it appears that errors could exist on both sticks. Both were errors on test #7, the random number test I think.

 

[DarkRogue]

Okay - this looks promising so far.

I didn't want to overvolt anything so I tried to increase the DDR2 voltage by +0.05v. Unfortunately, this resulted in the PC refusing to boot (said there was some 'hard error' with ntld.dll or something similar, can't remember off the top of my head.

I decided to undo that, and instead increase the (G)MCH (northbridge, says in the manual) voltage as small as I could, +0.025v I think. This allowed me to boot into windows, and a quick Orthos run with Large FFT's for 8 minutes and it hasn't failed yet. I'm going to reboot into Memtest 86+ and redo that and see if the NB voltage bump did the trick.

 

[Martimus]

If you really do have PSU issues, it is best ot replace it now before it kills off your brand new components. What PSU do you have, by the way?

 

[OCGuy]

I had to set my Vcore to ~ 1.285V in BIOS to run 3.6ghz (Shows as 1.256 in CPU-Z).


The Auto feature on my MB had it set at 1.3V, showing 1.272 in CPU-Z.


I guess my point is not all E8400s can run 3.6ghz at stock voltage and be stable. Hardly a con IMO, though.

 

[DarkRogue]

I'm not entirely sure it is an issue with the power supply as I can't find a way to reliably recreate the reboot condition, so I have no way of testing it. But for the record I am using a Seasonic S12II 380GB 380w PSU. It's not an issue with power draw as I am nowhere near its limit. My Kill-A-Watt reports 170w LOAD at the overclocked rate of 3.6GHz. And this is from the wall, add in the 80-85% efficiency, and system draw is a little lower than that.

My E8400 is part of batch Q745A which was reported to be nice to overclock with, but of course that doesn't mean it's necessarily true. However, as of right now I have it running at 3.6 with the smallest bump in north bridge voltage and it seems promising. It passed 2 passes of Memtest86+ with no errors, and I'm running it through Orthos Blend right now.

 

[DarkRogue]

Okay, Orthos Blend ran for 1 hour and 15 minutes and failed again.
It says consult stress.txt but I have no idea where that is.

Anyway I'm not sure what to do now. :/

 

[OCGuy]

Increase Vcore. I dont know why you are stuck wanting to keep stock voltage....wierd. But to each his/her own I guess.

 

[TheJian]

Originally posted by: DarkRogue
Okay, Orthos Blend ran for 1 hour and 15 minutes and failed again.
It says consult stress.txt but I have no idea where that is.

Anyway I'm not sure what to do now. :/

Since you barely went up on MCH and DDR2 up both one more notch.

Remember the board is not meant to run at 1600fsb. Yes it's a great board for getting there but not at default volts probably. UP the FSB one notch also. This will probably end everything, then drop each one at a time to see which is causing the problem. You already saw you went from minutes to over an hour with one notch on mch/ddr2. That means you're getting closer to stable. The question is will you need so much it's not worth what you get?

Regarding the cpu since I see you're 1.225: I've had a cpu where i'd boot at default (e2160 1.8ghz) at 3ghz on 1333fsb. I couldn't get near windows. Add a notch on cpu and get to logon, add another and could get into windows at the desktop but prime crashed or worse couldn't even get it to start...LOL. One more notch all problems went away at like 1.36v. Note the FSB was not a problem and I was using 1.8v memory 5-5-5-15 which I already set at 1.9 out of the box just to not have to deal with it during my OCing testing. Once solid I backed off the mem volt. I was just ensuring it wasn't a factor. Yours got better with mch, so that's part of the issue. Do you need more? Not sure it could also be cpu. All of these can affect you. Going up one notch on any of these will not light your system on fire. That board is built for this. The memory should handle anything 2.1v or below and last for a long time, all memory should. I prefer less, sure, but it won't kill it (some run up to 2.4, no I wouldn't). Above that...roll those dice baby Depends on the memory, mileage may vary.

That bios tells you which voltages are safe in their minds, note the red where they tell you UH OH BACK THE TRUCK UP flashing at you...LOL. I have the DS3R board. With your cpu at 1.225 yeah I'd probably just go straight to that and up it to 1.3 and work my way down if solid ignoring other voltages. Intel specs that chip up to 1.36 if memory serves so you won't blow it up. The server version E3110 is up to 1.225 max default and comes with a puny heastink for servers so you have a good chip at that volt. Also note that the bios probably tells you its actually LOWER (go into PC Health page and you're probably 1.2v). I've noticed all models of that board are .02 less than I set. Like if I set at 1.36 it's 1.34 on pc health page. Just knowing Intel can ship you one out of the box at 1.36 means they think it will make it there for 3years server OR desktop model. My E3110 came today. I hope it's like this one:
http://www.circuitremix.com/in...?q=node/122&page=0%2C0

4gigglehurtz at 1.292v! Check out the undervolt at 1.197! 3.7ghz!

 

[DarkRogue]

I'd like to avoid raising the vcore if at all possible, so I'm trying to exhaust other options first, maybe I'm an idiot for doing so, but hey, I'd never know unless I try it.

Jian, awesome write-up. Small correction though, I reset the DDR2 voltage, because adding 0.05v caused it to be unable to boot (hard lock error) so I set it back down to normal and increased mch by one notch. Then the aforementioned Orthos fail at 1hr 15min but I'm confused right now. I re-ran Orthos Blend while keeping those same settings to see if I could repeat the fail, but it's been going for 2hrs 15min now and still going.

I'm actually afraid to 'test' the limits of increasing voltage because I've never done it before, so I want to play it safe. What's the deal with increasing FSB voltage though? That one I can't understand. I may try upping mch once more if I get another fail because some people say you need to put more juice into mch when running 'higher' sticks (ie; 4GB or running 4 DIMMs.) I don't know about upping DDR2 voltage since +0.05v didn't work at all.

As far as the reporting 0.02v less thing, I believe I read somewhere that it had to do with Gigabyte's Loadline calibration but I get conflicting reports of whether it's better to leave it on or off. I currently have it on auto so who knows what the heck it's doing.

 

[TheJian]

Missed the reset part, but still you're safe with even 2.0v on memory. I really don't think raising the voltage on memory will ever be the cause of a crash at 2.1v or below. They just don't get hot enough to crash usually at that volt. Above that heat can circle round and bite you in the butt. In the case of you raising mem and not getting a good reaction I think it was just the real problem STILL biting you so to speak.

In an effort to shorten the post I left out some stuff with the cpu volt story. When I started prime I was blown out on one occasion and another reset/rebooted. The blow out came first which I considered a tame result. But I wasn't deterred with the next increase that caused an actual reboot which I considered a worse result. The point I'm trying to make is it won't always react the same. You could leave everything at one setting and get different versions of crashing if you repeatedly tried. Maybe a voltage fluctuation caused the first way to be a bit higher and more forgiving. PSU's fluctuate between 1-5% on most brands with crappy PSU's being even worse. So it could just be that.

Also to kill your cpu fears:
http://www.anandtech.com/cpuch...howdoc.aspx?i=3251&p=2
Note the part where they say running past 1.45 would worry them. Intel confirmed this. Meaning you're perfectly fine at 1.3 that I suggested. For that matter I'd have NO fear to Intel's own 1.36v barring flying high temps. IF they're willing to ship a core at 1.36v on that model I'm not worried about putting mine there. Even if you have a totally crap chip that heats up it will just kick in the brakes at 83c temp. So you won't be there for more than a bit and Intel will save you... I completely understand your fear as a new overclocker though. But if Intel can do 1.36v on E8400 and think they have no warranty issues and won't see you RMAing it in 3 years I have no fear at anything less than 1.36v.

BTW, I didn't catch what heatsink you're using. Is that the Stock one or something better? Personally I have no fear at 1.36 on stock. I'm running there now on my e4300 without more than 60c load temps (though the bios says 1.34). My sisters only took 1.29v to get to the same 3ghz but it's a E2160 and much later date.

Oh, loadline: http://www.anandtech.com/mb/showdoc.aspx?i=3208&p=2
Anandtech says leave it disabled. I can't remember what I have on mine...LOL. If I'm within Intel volt limits and temps are ok I never really care. But if I was nearing scary temps and desired volts to be lower to solve the problem I'd disable it. But yeah, reports vary. Good luck, and report back I'll hopefully get to my e3110 today, but I have a server of my own to play with first. Then my dad's should be here Friday...LOL. Xmas was definitely early this year...ROFL So many toys and so little time

 

[Idontcare]

Originally posted by: Ocguy31
Increase Vcore. I dont know why you are stuck wanting to keep stock voltage....wierd. But to each his/her own I guess.

You ever notice people's screen names rarely seem to reflect their true character? I use to think it was the opposite, people used their screen names to reflect how they really view themselves...but it seems to be the opposite on so many occasions.

Even my own screen name falls into this observation.

Originally posted by: Extreme!!!Overclocker!!!FTW!!!
I'm not really comfortable running my system at anything but stock settings...

 

[Owls]

My screenname is extremely indicative of my overclocking prowess

 

[Idontcare]

Originally posted by: Owls
My screenname is extremely indicative of my overclocking prowess

I don't care!

... or do I?

 

[CTho9305]

I have 4GB (2x2GB) of the Corsair TWIN2X4096-6400C5's. They're rated at 1.9v and it's currently (auto) set to 1.9v. They're also at default/stock timings, I haven't touched it (5-5-5-18.) I ran Memtest86+ on STOCK CPU speeds (333x9 1.225v) and it ran for 11 hours (18 passes) with no errors. As soon as I upped CPU to 400x9 1.225v though, Memtest86+ detected 110 errors throughout Pass #1. I'm not entirely sure what this means, do I need to boost its voltage a bit or RMA, or what's going on?

You're considering RMAing hardware that works fine at the rated specifications? Is that like buying 20lb fishing line and returning it because it breaks when used with 25lb loads?

How do you know it's not the chipset or CPU I/O logic that's unable to reliably handle the higher bus speed?

 

[DarkRogue]

Thanks for the reply guys.

CTho, I thought the memory may have been bad because it started failing when I raised something OTHER than the memory. Memory has always been left at stock, I only changed the CPU speed to match the memory's FSB so I thought I would take the memory out of the equation.

lol @ the humor though, Idontcare

Jian, thanks for the clarifications. I suppose I will give DDR voltage a +1.0 bump and seew hat happens if it errors again, seeing that hard lock at +0.5 kind of deterred me lol.
I'm fairly certain my CPU is rock solid and it's something to do with my memory though.
Orthos ran for 10 hours this time before it failed. A look at the test history shows it was doing small FFT's the entire time but when it went to the large FFT's it failed. Memtest86+ doesn't seem to find any errors though, so I just got HCI Memtest and I'm going to try that. I set mch voltage one notch higher though, at +0.05v now (and currently the only voltage change) and I'll see if that helps any.

As far as heatsink, no I am not using the stock heatsink (got an OEM chip anyway.) I'm using a Scythe Ninja Plus Rev B mounted with the Thermalright LGA775 Bolt-Thru Kit, using AS Ceramique. (I could've used something else but that would cost more money and I had tons of this stuff lying around lol.) I don't know which temps to trust and TAT won't run for some reason, but Coretemp reports load temps of 65c, while realtemp reports loads of 55c.

As far as Loadline is concerned, yeah I know Anandtech recommended it off, but I dunno, some people said it was good. I suppose I can try disabling it as well. Just sucks because testing takes so long and I gotta test one thing at a time.

 

[TheJian]

Especially with OEM. Test that good! Is that a 30day warranty or does the place you bought from have better than that (since it is up to them on OEM)?

Here's some places to look for more help and see what others are comfortable with.
http://www.ocforums.com/showthread.php?t=549725
E3110 (same as E8400 just generally better volts/heat, same mhz roughly)

http://www.ocforums.com/showthread.php?t=544848
That's the E8400/8500 thread.

Seeing that you have oem, I wouldn't go over 1.3v. Sure Intel might ship one with 1.35v and tell you 1.45 is fine (Anand says intel engineers confirmed this) but with OEM you're SOL if your chip dies. That's a SUPER conservative statement with regards to volts but I don't mind being called paranoid over $230 (more? I paid $209 (mine)/$228 (dads) for my two). God forbid you paid $260 or something.

 

[DarkRogue]

I believe they said warranty on the OEMs was 1 year from Intel. Dunno though.
I paid $235 shipped for it so yeah quite a chunk of change so I don't wanna kill it.

Anyway, I'm completely out of ideas. DDR voltage just won't play ball.
Default settings for all my runs are 400x9, PCI-e @ 100, 1:1 ram ratio and 1.225v vcore along with Loadline @ Auto and MCH @ +0.05. It boots but it will fail orthos with large FFTs (I set it custom to 768-32768K FFTs.) Oddly though, HCI Memtest did not show any errors, nor does Memtest86+.

Each of these were done independently so I wouldn't have 5 different voltage changes as variables. IE; I reset everything to the above 'default' settings before doing each one.

Up DDR voltage by 0.05v = windows won't boot (BSOD)
Up DDR voltage by 0.10v = windows won't boot (BSOD)

Up Vcore 1 notch = BIOS resets/won't POST
Up Vcore 2 notches = BIOS resets/won't POST
Up MCH to 0.075v = BIOS resets/won't POST
Up FSB 1 notch = BIOS resets/won't POST
Lower MCH 2 notches + FSB 1 notch = BIOS resets/won't POST
Lower MCH 1 notch + FSB 1 notch = BIOS resets/won't POST
Lower MCH 1 notch + Vcore 1 notch = BIOS resets/won't POST
Enabling/Disabling loadline doesn't seem to do anything although disabling it sometimes prevented POSTing as well.

ATM I'm back at my "default" but of course Orthos will fail with large FFTs and I don't know why or how to fix it, even though both other memory testing apps don't show anything wrong.

EDIT:
Made some clarifications to what I was trying to say.

 

[Xpoc]

If you are having problems set "Performance Enhance - Extreme" to "Performance Enhance - standard"

 

[DarkRogue]

What exactly is it supposed to do? I remember reading somewhere that it was recommended to be set to extreme so that's what I did, but I'm not sure what it's supposed to do.

 

[Xpoc]

It is not recommended when overclocking. Try with it set to standard.

It's basically an auto tweak to make system run better, but when you are manually adjusting settings it tends to conflict.

 

[TheJian]

Strangest thing I've seen with these. You're the winner...ROFL.

No Intel warranty as far as I know. AS a reseller I got 30days but that was from my supplier which was THEIR warranty not Intel. They eventually completely halted selling anything but BOXed models for both AMD/Intel. This is ASI I'm talking about. They were pretty much my main source for cpus unless out of stock and I had an emergency. I'm pretty sure their rules were standard in the industry. Most places don't give any more than that for the same reason. Like when I sold OEM I only gave 30days because I was only getting 30 myself Needless to say I avoid them at all costs because of this. But I understand you wanting to get your grubby hands on one ASAP...LOL.
http://www.intel.com/support/p...ssors/sb/cs-020033.htm (warranty info here!)

At this point I say RMA if you can, tell them you couldn't get it to work properly in your board/setup...

Then buy a 3yr box.
excaliberpc.com $229 in stock
Allstarshop.com $239

J/K

I'm wondering if your PSU just sucks. If you have a fry's nearby you could pick up a monster PSU just to test and RMA it the next week. Get creative Or keep it since it solved your problem Before someone says "You're an ASS". Shut up, IDONTCARE :evil: about your issues with my morals... :laugh::evil:

I love that guys nick...LOL.

I'd still like to hear what happens at 1.3v. Or even 1.28+. 1.22 is really low. Hell 1.3v is pretty low.
http://processorfinder.intel.c...tails.aspx?sSpec=SLAPL
SLAPL is probably on your chip. Which means it's safe up to 1.3625v Straight from Intel.

 

[TheJian]

LOL...I didn't see that in his previous post. You're right fix that setting.

[edit] mine is on standard.

If my downloads get done soon I'll reboot and check out my settings. But that looks like some hours to go.

 

[DarkRogue]

I just did, set Performance enhance back to standard as the only change (rest at "default")

Going to run Orthos again now..

Edit:
What's the "In place" setting in orthos do?

Anyway, in regards to your other post, I guess I won't have warranty, but I don't know if Intel would service/honor it after it died trying to go to 400 FSB lol.

My PSU is supposed to be fairly good, Seasonic manufactures Corsair's PSU's I believe. Maybe it could be defective but I have no way of testing it short of buying another one which I can't really afford as I spent most of my budget on this rig as it is, going to be a while before I get more spendable cash on hand.

As far as voltages go, I'm sure it can handle 1.3v fine but I want to run it as low as possible. I don't know if I'm reading the results incorrectly or not, but the CPU seems rock solid, it's only when orthos starts doing large FFTs that it fails. Small FFT's to stress only CPU, it passes. Memtest86+/HCI Memtest to stress only memory, it passes. Orthos both of them and we have a problem.

Edit2:
Well poo, it failed again. I guess I will try Turbo?