Sign in with
Sign up | Sign in
Your question

e8400 quasi stable @ 4ghz need help on next move

Tags:
  • CPUs
  • Overclocking
Last response: in Overclocking
Share
October 7, 2008 2:48:23 AM

So I got a new e8400 and paired it w/a dfi dk x38 t2rb mobo and am trying to hit that golden 4.0ghz plus. After no brainer OCs to 3.6 and then 3.8 that were prime stable for 2+ hours (400x9 and 420x9 respectively) I decided to try for the big 4.0ghz and went 450x9. Almost immediately after booting into windows my computer froze. That's right, no BSOD, no reboot, it just completely froze. I haven't seen windows freeze since the days of 3.1 when it happened ALL THE TIME... well I guess it happened a few times in 95 too but that's beside the point. I'm running xp pro w/sp3 and just installed it a week ago. Even when I OC'd my AMD x2 3800 I got prime errors or restrats/bsods, not freezes. Anyway, I backed off on the fsb to 444 and got it to pass prime 95 large fft torture test for 4 hours. Next I tried OCCT and the system froze every time I ran it.

Now to back up a sec and mention ram issues. The patriot ddr2 800 sticks I have show up in the cpu-z spd as being jedec @ 400mhz 5-5-5-16 1.8v | EPP1 400mhz 4-4-4-12 trc 21 2t 2.2v | EPP2 500mhz 5-5-5-15 35 tRC 2t 2.3v I had these sticks in an amd system on a gigbyte m55sli-s4 board and was able to get them to run at up to 475mhz at 5 cas (the board had no ram timing tweak other than CAS) and 2.1v (max I could select on the gigabyte board). When I stuck them in this new system on the dfi board w/the e8400 they booted up fine and I could get them to 4-4-4-12 400mhz np. But they wouldn't post past 410mhz, even when I set the timings to 5-5-5-15 and gave em plenty of juice. Turns out, I had to run em at 5-5-5-16 and they work now. Been wondering if there's a way to get em to run at the EPP standard up at 500mhz, well that's a different story.

Anyhow, once I figured that out, we're back at the beginning: 450mhz fsb freezes, 444 passes prime95 large fft for 4 hours but OCCT freezes the comp. So my next try (after reading the great sticky on OCing a c2d/c2q cpu; that's where I also learned that the small fft is the better test. I documented my results and need some help interpreting the data) was to push the vcore.

Series 1: 444x9 | strap at 1:1 (400:800) | 5-5-5-16 | Dram 2.175v | all CPU enhancements turned off except Thermal Management and multi core processing | all other voltages stock except cpu core voltage which is my variable.

1) +12.5mv vcore special add (cpu-z shows 1.248v for cpu @idle). Prime small fft fails after 25min

2) +25.0mv vcore special add (1.264v droop to 1.248). Prime small fft stable for 6hrs 18min when I halted it. Tcase temps hovered around 52-53c (when I callibrated speed fan, my thermometer was not that reliable, I'd give it an accuracy of +- 3c more likely it reads a little high.) I'm using a Thermaltake big typhoon (the newer one) and AS5.

At this point I figured that I probably have plenty of voltage and that the problem with freezing or whatever else lies somewhere else. On to series 2: testing Ram voltage. This time I used memtest to test the memory stability

Series 2: same as series 1 but with +25mv vcore leaving cpu @ 1.248 after vdroop at load | variable is ram voltage

1) 2.175v Memtest: system freeze w/in a few min of beginning the test. (same freeze as when I did the OCCT test)

2) 2.225v Memtest: system freeze w/in 5min

3) 2.25v memtest: freeze at 60% coverage (guesstimate about 10min in)

4) 2.275v memtest: freeze @ 180% coverage (guesstimate about 30min in)

5) At this point I thought that a big problem must have been ram voltage, but I wanted to test and see what affect NB voltage had, esp since I didn't want to push the ram juice much higher. So I put ram voltage back to 2.175 and bumped the NB up 1 notch to 1.278v (bios setting, I have no idea what it really is). Memtest caused system freeze almost immediately.

6) Then I Put the ram back to 2.275v and left the NB at 1.278 and memtest lasted until 98% coverage before freezing.

7) my last test was to see if the cpu clock speed had any affect. So left everything as it was and dropped the multi to 7x which gave me 3108mhz cpu. This time memtest lasted 1350% coverage before I stopped it. I'm guessing it was somewhere between 1-2hrs of testing.

At this point I'm a bit stumped as to what my next move should be. My gut says that the memory just isn't getting enough voltage at that clock speed, but I'm real skiddish about giving it much more b/c the bios voltage readout for the ram (only way I know to read it) was showing darn near 2.3v (if not slightly more) at the 2.275v setting. I noticed that as I gave it more juice it went from being lower than the setting to actually being higher. On the other hand, I don't know enough about the NB to say whether it has any real affect on the freezing, or whether the VTT might have an affect. Course, the fact that memtest ran stable for a fairly decent length of time with the cpu at a slower clock speed while the ram was still at the same speed is the most puzzling to me. That seems to imply that I may still need some more juice to the cpu. And I'm well within the CPU's safe vcore setting.

I hope someone can help me make sense out of this Ultimately I'd like to be able to hit 4.25ghz or better but will be very happy w/a rock solid 4.0ghz. :sol: 

Oh 2 side note questions, I'm wondering if I should reseat the heatsink since I would have expected the TTake typhoon to be much more effective at cooling one of the new wolfdales, esp since I hardly added any juice to it and I got the e0 stepping.

Also, I'm wondering if I ought to get some 1066 instead of the 800 ram since the ram seems to be holding me back from a higher cpu speed and unfortunately I can't run the 800 any slower than 1:1 when the fsb is at 400mhz. I am definitely getting more ram (my current sticks have to go with my old comp) but would like to be able to get away w/a high oc and use 800 ram since it's a fair bit cheaper.

More about : e8400 quasi stable 4ghz move

October 7, 2008 3:06:58 AM

Bump.

Eager to hear what the braintrust has to say....
a b K Overclocking
October 7, 2008 10:54:04 AM

well your useing very low voltages so you can use 1.4 vcore 1.4 mch(north bridge), well id swap ram out, but dont go over 2.3v, your best bet is just to try higher voltages and then lower them individually to see which one is makeing it unstable. im running my e8400 at 1.53vcore to get prime stable at 520fsb for 4.44ghz. so dont expect the voltage increases to be linear
Related resources
October 8, 2008 1:32:02 AM

Ok first off, set your timings to 6-6-6-18, you can work on ram stability after getting your CPU overclock.

Next, bump your vcore up to 1.33 (up to 1.36) (its ok, your running REALLY low voltages right now, thats why you have the problem)

Question; Do you have LLC (loadline calibration?)

Alright, after the vore, set your VTT to something like 1.24
Set the PLL voltage to 1.5
Set the NB voltage to 1.3

Test for stability @4.0ghz, if you've got it, then go to 4.2, 4.4 etc till you run out of stability. Once you do, up the Vcore to 1.365 if you haven't already. If it still fails then go to BIOS, and set this setting

GTL 0/2: .635

If you have more stability, then find your max ghz
If you don't go back and set these

GTL 0/2: .630
GTL 1/3: .675


If you have stability, then find your max ghz
If you don't, then determine if you had better stability with GTL 0/2: .635 or GTL 0/2: .630
GTL 1/3: .675

Then go back, set it to the most stable of those, and increase your VTT from 1.24 to 1.32

If you have more stability, then find your max ghz
If you don't go back and set this:

VTT: 1.24

And then just stick with whatever OC you have
You could try increasing the PLL to 1.54 or 1.56 at this point if you want
Bumping the Vcore would also work, however this would put your voltage out of intel's "safety spec" although with Vdroop it's technically ok.

Anyway, i hope that hepled, be sure to answer my question about LLC, and if you're confused, be sure to ask.
October 8, 2008 2:26:07 AM

K quick answer, haven't tried a lot of that yet, but I did try the 6-6-6-18 settings and board wouldn't post (c1 error on the mobo lcd display which is a memory stop error). Didn't see anything resembling LLC in the genie bios section (the OC area), though I just realized that I didn't look in the other standard areas. I'm gonna check the DFI bios guide on clunk forums to see if there's anything like that there.


Otherwise, before I read your post I made 2 more tweaking attempts:

444x9 | 1:1 400:800 5-5-5-16 2.3v (reading in bios, 2.275 setting) | 1.248 vcore (after vdroop) | everything else stock. Variable was the NB voltage:

1) showing 1.26v mem test freeze at 198% coverage

2) NB (jumped 2 levels) to 1.317 bios setting, 1.29 showing. Mem test passed through 520%+ before I stopped it. (not quite 2 hours).

2b) Same settings this time using the 1hr Auto mixed test in OCCT. It passed with flying colors.

At this point, considering that I already passed Prime small fft for 6+ hours I think my system is pretty stable at 4.0ghz. Though I will need to do another prime test.

Questions: what do LLC, VTT voltage, and PLL voltage do?
October 8, 2008 2:46:10 AM

Ok so a quick google search on LLC and I know what it is/does. Yes my bios has what's called "Vdroop conrol" and that option is/has been enabled. Prime small fft causes a vcore vdroop from 1.264v @ idle to 1.248v @ load.
October 8, 2008 3:30:37 AM

Ok, then start with 1.33, rather than 1.36 (i wasn't sure if you had LLC)
Other than that, follow my instructions

Also "stable" is prime95 small ffts 6hours+
October 8, 2008 3:32:00 AM

Edit: Sorry doublepost
October 8, 2008 3:41:50 AM

yeah I've already run prime stable for 6+hrs w/vcore @ 1.248 (after droop), mem @ 2.175v, NB @ stock, 444x9. It was only when I did some test that involved memory that I got the system freezes. I think that since then bumping ram to 2.3v and NB to 1.317 (1.29 showing in bios), it has passed several tests that use memory. Though I haven't run Prime95 again. I doubt that the extra MCH and Ram voltage would cause the cpu to lose stability.
October 8, 2008 3:44:21 AM

Like i said, don't deal with the RAM till you've got your CPU stable. That way it's less variables
October 8, 2008 3:48:55 AM

yeah and as per my earlier test, my cpu was stable in prime. But I had to figure out why I was getting system freezes when I booted into windows at 450fsb. And I think it was due to ram voltage and NB voltage being too low.

So what is the VTT and PLL for? I remember reading something about VTT that said CPUs are much more susceptible to getting fried when you start messing w/VTT. And while I don't doubt your instructions, I'd like to know what it is I'm doing before I go risking my brand new system getting fried.
October 8, 2008 3:57:09 AM

Well you want above 450 FSB right? That's what i though you said, thus i told you how (its not a ram voltage problem, the problem is your vcore). If you just want 444x9 stable and ram stable and leave it at that, don't heed my instructions. If you want something like 475x9 or above, do heed my instructions, and as i said, set your RAM timings to 6-6-6-18 to make sure that any problems are due to the CPU, not RAM (as overclocking FSB overclocks the RAM also).

VTT is the FSB termination voltage. All you need to know that increasing it increases stability. However, since this setting does have an effect on the CPU, i.e its the voltage on a part of the CPU, you shouldn't set it above 1.4V (basically the max normal setting for the Vcore anyway). A general rule of thumb is to keep it below your Vcore.

Edit: You don't have to take my word for it, heres intel's spec sheet. Look at section 2.6.1 and 2.6.2
http://download.intel.com/design/processor/datashts/318...



PLL is the clock generator voltage. It's not going to fry your CPU, it starts at 1.50V normally, so 1.54 isn't a big increase, don't worry about it

GTL voltages (you didn't ask but just in case) are voltages that tell your CPU when it's a 0, or a 1, i believe per the voltage supplied, thus it increases accuracy, and thus it fails prime left. This cannot fry your CPU.
October 8, 2008 4:08:23 AM

I wish I could set the ram timings to 6-6-6-18. But I tried that and the board wouldn't post. :-(

Thanks for the info. Am gonna try now :-)
October 8, 2008 4:33:47 AM

Aight keep em at whatever you have them at, just try to keep the mhz as low as possible.
October 8, 2008 4:36:30 AM

ok, I'm now booted in windows @ 466x9 and running prime small fft. After ten minutes temps are: 52c tcase | 38c NB | Cores show 58c/56c

vcore: special add +62,5mv | cpuz: 1.296v idle, 1.280v at prime load
VTT: 1.24v setting | bios reading showed 1.23v, DFI's Smart Guardian says 1.21v
NB: 1.33v setting | 1.31v actual
DRAM 2.275v setting, 2.3v actual

Gotta say this is a far cry from my last OC attempts w/a X2 3800+ and Gigabyte Nforce 4 mobo. Max I got on there was 2.48 ghz 248x10. Though I was able to get the FSB on that sucker up to the 260s....


Gah, prime failed at 15min. Gonna go up the vcore a little more.
October 8, 2008 4:54:07 AM

Back up w/+75mv aka 1.312v idle, and 1.296 under load. We'll see if prime holds or not. *crosses figers* I looked at PLL and the bottom option is 1.51v in bios. Next possible option is 1.64 not sure if I should mess with that or not. That seems like a bit of a bump.
October 8, 2008 2:37:05 PM

Oh i didn't realize your mobo only lets you select numbers. Ok, well don't go to 1.65 PLL for sure
October 8, 2008 2:59:33 PM

K i'll leave PLL alone then. So far I'm 10 hours Prime stable 466x9 w/the 1.296vcore including vdroop. Cpuz says I'm actually at 465.4x9 so 4188mhz. 40% overclock next stop 475x9 :sol: 
October 8, 2008 6:55:32 PM

Nice, seems to be doing well

You haven't raised anything but the Vcore yet? or have you

Just want to know what step you're at
October 9, 2008 2:32:25 AM

Yeah I got it stable at 466x9 for 10hours last night. All my attempts at or around 475 fsb caused system freezes (similar to when I first started messing w/450+ fsb before). Here were my settings:

| 5-5-5-16 | 400:800 1:1 ram strap |
same voltage settings as my successful 466x9 (Dram 2.3v | vcore 1.312v/1.296v after vdroop @ load | VTT bios set 1.24v showing 1.23v | NB bios set 1.33 showing 1.31 |

1) 475x9 system freeze @ windows load screen

2) bumped vcore to +87.5mv (1.312v idle) VTT to 1.245 bios setting. Not sure what VTT actually was. This caused a system freeze 1 min into prime testing.

3) vcore +100mv (1.328v idle, 1.312 @ load) NB to 1.343 freeze before windows finished loading

After that, I dropped back to 466x9 to try seeing if I could get it stable w/C1E on. 2 tries w/vcore back around 1.3 and I noticed a possible pattern that suggests the C1E may not like to give the cpu more than 1.296v. In any case, neither try was stable.

Now, I'm going to go back to see if I can get 475x9 stable.

My feeling, is that I'm likely hitting a wall w/the ram. And am planning to get some DDR2 1066. I was thinking 2 2gb sticks. Any suggestions to a particular brand/model? I don't really want to go Gskill cause I had 2 sticks of their supposed 4-4-4-12 ddr2 800 ram that according to spd was 5-5-5-15 at 400mhz and 4-4-4-12 at a slower clock speed. On the other hand I've been really happy with my patriot sticks, as I got them to 475mhz back on my amd board and the SPD shows an EPP rating of 500mhz, 5-5-5-15 2t and 35tRC @ 2.3v

Guess I'll go try somethign else to see if i can figure out what's causing the system freezes. esp, since I still have some room w/the vcore and haven't messed w/the GTL yet.
October 9, 2008 2:34:34 AM

oh 1 other thing, I have no idea how to set my command rate or the tRC value in this DFI bios. Considering all the ram settings, there's gotta be a way, but I've consulted dfi's bios guide and Clunk's translation: http://www.clunk.org.uk/forums/hardware/3203-dfi-bios-g... and can't find anything that resembles tRC or command rate.
October 9, 2008 2:45:42 AM

Sorry don't know anything about that

No idea about DFI boards, i have a P5Q-E myself.
October 9, 2008 3:37:30 AM

I think I may have discovered the cause of the system freezes... *crosses fingers* After half dozen different tries at 474x9 tonight w/various voltage tweaks and always the same thing: system freeze either at the very beginning of starting prime, or sometimes even sooner, I am currently 15 minutes into a prime test. The main change..... I switched my ram bootsrap from 400:800 to 333:667. Still 1:1.

Current voltages: vcore 1.344 vdroop at load (1.360 at idle),
According to Smart guardian (DFI monitoring program) VTT 1.23v, NB 1.32v, VDIMM 2.27v

Also, a little disconcerting is the fact my Tcase temp is sitting at 62c (w/the case open and a fan blowing on it. Ambient, I can only guess at between 65-70F)
October 9, 2008 4:36:32 AM

Yep I was right. Prime was stable for 1hr. Then I stopped it so I could test 1 more higher OC before bed. (I'll do the long tset for 474x9 tomorrow while I'm at work :D 
So, gonna test 484x9. :sol: 
October 9, 2008 4:59:01 AM

Looks like I'm about ready to start futzing w/the GTL. 484 fsb caused a system freeze 5min into Prime small fft.

I checked my board, and my options are for the GTL 0/2 and 1/3: .667x (default and current setting), .648x, .636x, .620x. NB GTL can go between .667x (default and current setting) and .610x (only other option).

Rats... mem test just showed an error at 23% coverage :(  474x9. Guess I've hit a memory mhz wall.
October 9, 2008 2:07:08 PM

Set your memory to a lower speed.
I.e. if you have 500FSB, you're RAM would run at 1000mhz at a 1:1 divider, but you can set the divider differenly so it runs below 1000mhz

Do this, the option should be where your RAM speed is shown
October 9, 2008 2:14:16 PM

I don't know the proper etiquete here, but I have an e8400 stuck at 3.85 GHz. I don't want to hijack the thread and all, but we are talking the same issues, albeit with different system specs. Vinegar and Dope--is that OK, or should I start a new thread?

BTW, Dope: funny system specs. I just noticed that you "upgraded" imaginary CPU specs...
October 9, 2008 3:44:16 PM

Dope, wish I had that option in my bios. :(  Slowest I can set the ram is 1:1 and I have 2 options for that 400:800 or 333:667. This is why I mentioned that I think I'll have to get 1066 ram to be able to take advantage of maximum OC here.

Husky, I don't mind, though you may get more response if you start your own thread. If you haven't gathered from the above, I reached 3.8 with out any voltage push. Once I started going for 4.0 and above I need to start increasing vcore to keep things stable. I found that a slight NB increase helped for Ram stability as well, since I was already pushing my ram to 2.3v. But a big thing, is making sure your ram is ok (like Dope says, keep it slow and the timings loose). Also, I found that I could keep C1E enabled up to 4.0ghz. Above that and I lost stability. 4.2 was where I had to start pushing a lot more voltage to achieve stability (across the board, though it's possible that I may be able to lower some of those values). And sadly, I haven't gotten it stable above that yet.
October 9, 2008 3:46:09 PM

oh yeah, do you have c0 or e0 stepping husky? That will make a difference in heat and vcore amount as the e0 are more efficient and generally require less vcore. I have an e0
October 9, 2008 4:01:21 PM

Hi Joe:

Thanks for the note. Mine is C0, so I have to push more voltage through it. 1.35 or so right now. Can't seem to do C1E or any power savings without loss of stability. My limit may simply be a function of maxxing out a P35 board. Really, I can't complain as 3.85 is really overkill for gaming at the moment anyways. But it would be fun to get it over 4...

I am traveling for the next few days, and won't be able to tinker. Thanks for the help!
October 9, 2008 4:38:00 PM

Ok, well vinegar, as far as i can tell, you've either hit the max CPU speed you can have, or your RAM just can't run that fast.

You have 4 sticks right? Like i said earlier, take out two, and you should be able to run them faster (becuase with 4 sticks, you have to have looser timings and lower mhz to make sure they are all stable)

If it's your CPU, then i dunno, your stuck, it might be the C1 stepping (E0 overclocks much better), but i don't have any idea really.

If it turns out that RAM is your problem, running 2 sticks should let you get 4.0ghz, and then you'll just have to buy a 2x2gb kit (always do that in the future, it's MUCH more stable)

@husky --> =P i see you noticed my sig
Post all your specs, or a screeny of CPU-z
What's your Bios vcore, and whats the actual vcore under load and idle (via CPUZ)
Do you have LLC?
List these settings from your bios (if their auto, just say auto)

CPU vcore:
FSB:
RAM spd:
vRAM:
VTT:
PLL:
NB volt:
GTL 0/2:
GTL 1/3:

Also what are your temps (both of you guys) and what is your motherboard husky, do you have 4 sticks of RAM, and what are they

Also, again both of you, what PSU's do you have, (mostly likely not a problem, but just in case)
October 11, 2008 5:56:32 PM

my temps were starting to hit 60-62c (tcase) when I was stressing 474x9 w/1.344 vcore (after droop at load). I am only using 2 1gb sticks of ddr2 800 ram atm. And the slowest I can set the ram is 1:1 ratio (400:800, or 333:667). But I'm ordering 4gb (2x2gb) of some 1066 soon. That should give me a bit more room. At 4.0 and 4.2ghz I was seeing 50-55c load temps (tcase).
a b K Overclocking
October 12, 2008 1:52:55 PM

go on joe take it 1.5vcore and see if you can hit 4.5ghz :D  when you get you new ram.
October 12, 2008 2:31:34 PM

richardscott said:
go on joe take it 1.5vcore and see if you can hit 4.5ghz :D  when you get you new ram.


:pt1cable:  LOLs..... uh.... no. Unlike some peeps, I don't have the cash for extreme cooling and would like to keep my chip for longer than 30 seconds :kaola: 
a b K Overclocking
October 12, 2008 7:31:04 PM

hehe yea i dont blame you, im planning on buying a 45nm e5200 just to see how long it lasts at 1.6v :D , see if i can get a 100% overclock on it.
October 12, 2008 8:18:14 PM

For a true test, you should do that with no heatsink.
a b K Overclocking
October 13, 2008 5:51:17 AM

lol it wont even post :p  wont break it either just wont work.
October 13, 2008 3:42:41 PM

aww... I was hoping for fireworks
!