Trying to Overclock E6400 (2.1ghz) to 3.2Ghz , Help needed !

Toggle sidebar Toggle sidebar
louno

louno

Distinguished
Sep 7, 2009
146
0
18,680
Alright,
Well this is my first time overclocking a cpu, here are my system specs :

MY SYSTEM SPECS :
Motherboard : ASUS P5W DH Deluxe
Power Supply : Antec Earthwatts 500
CPU : Intel Core 2 Duo 6400 ( 2.13Ghz ) stepping 6, revision B2
Heatsink : Xigmatek Dark Knight S1283V
Ram : Gskill F2-6400CL4D-4GBPK DDR2-800 / CL4-4-4-12 / 2.0v ~ 2.1v
Graphic Card : NVIDIA GeForce 9600GT
Hard Drives :
1 x WesterDigital WD2500JS ATA 250GB 7200 RPM 8MB Cache SATA 3.0Gb
1 x WesterDigital WD2500KS ATA 250GB 7200 RPM 16MB Cache SATA 3.0Gb
Power Supply : Antec Smart Power 2.0 SP500 ATX P4 500W 20/24Pin
Case : Antec NSK6500 Mid Tour ( 1 small intake fan in front and 1 big exhaust in back )
OS : Vista 32bit


RESULTS SO FAR :
Alright so my goal here is to reach 3.2ghz which would allow me to run my ram in synchronious mode ( ram/fsb ratio 1:1 ) ... with stock voltage i was able to reach 2.8ghz stable running orthos for at least 30min without errors. From there what i did was put the fsb to 400 ( 400 x 8 = 3.2ghz ) and raised the voltage of my ram to 2.1v and had to raise voltage of my cpu (vcore) until i was able to enter windows ( initially i wouldnt even get to the POST ) . After trying to run orthos blend a few times it keeps failing, i slowly raised voltage (vcore) and am now at 1.45 and still orthos fails ( although its better now, longest ive run it was 6min without error, but sometimes it can fail within 1 min )

Im not quite sure what to do next... would it be safe to raise voltage again considering im at 1.45 ? are there other things i can tweak to avoid raising voltage without lowering the FSB ?

Using Core Temp I can see my temps at idle are 41-42C , they reached 56C after 6 min of orthos they were at 56C . Note that while in bios the cpu temp being reported are lower, bios says its in the 25-30C range.


SIDE NOTES :
CPU FAN ISSUES :
Before I started overclocking right after installing the Heatsink i checked my temps with the core temp program, they were at 35-40. Then, again before starting to overclock, i decided to update my bios because i had never done so in the last 3 years, I went from version 1001 to 3001 but apparently that was a bad ideal, my idle temps didnt improve BUT now, unlike before, my Xigmatek fan is running at 100% speed at all time which makes it a little bit noisy, it was really silent before... i guess i can live with the noise but im not sure if this is an issue for the fan though. There are a few options in the bios to manage fan speeds (Q-fan for instance) however after trying a bunch of different settings none of them changed anything... Finaly i decided i'd just go back to an older bios version because asus keeps all their old bios patch history online but i was unable to complete the bios flash because i get a message saying that the versions I try to install are older than the one i have and the installation stops there... i didnt make a backup of my 1001 because i knew it was available online so im stuck with the latest version...

 
Sort by date Sort by votes
I'm running E6400 @ 3.2 (almost 3 yrs). w/4gigs of Ram I had to raise FSB voltage and (G)MCH by +0.1 for stability. Dram voltage was raised to specs, All other voltages including vcore are are stock. I switched RAM and was able to lower FSB voltage and (G)MCH to 0.05 above stock.

As Sportfanboy stated start with a lower multiplier, and raise. (W/ FSB Voltage and (G)MCH raised by 0.1 V. If get stability @ 8X (3.2 GHz) then try reducing FSB and (G)MCH to alonly +0.05 above stock - One at a time.
 
Upvote 0 Downvote
Yes to keep 400mhz, lowering the multiplier would be an option i guess, but i'd like to avoid it since it is a big drop from 3.2ghz to 2.8ghz.

RetiredChief,
Wow that is impressive, 3.2 at stock voltage !?

I have my FSB Termination Voltage and MCH Chipset Voltage set to [Auto] in bios, FSB options goes from 1.20V to 1.50V in 0.10V increments while MCH Voltage goes from 1.55V to 1.85V again in 0.10V increments, there are no 0.05 increments... Also I do not know what are the "default" values that are used with Auto mode so im not sure which options would result in a 0.1V increase over default value...

 
Upvote 0 Downvote
Sorry, I can not check my E6400 system, currently on vacation in Calif. (LA)
Also I have a gigabyte 965-DQ6 MB and it does not list the voltage in the options, it list amount to increase.

You might do a search for old post on the E6400 overclock on tom's.
Also google nominal FSB Voltage

jsc, while you may be 100% correct, back when the E6300/E6400 first came out a good percentage were reaching the 3.2 on the E6400.
 
Upvote 0 Downvote
alright well, i increased vcore voltage from 1.45 to 1.46 and i was able to run orthos for a little longer... Idle temps stayed pretty much the same, as im typing this i have a 39C temp in core temp ...

What would be the safe maximum voltage i could use ? My computer is pretty much open 24/7...

My ram is supposed to be running at 400mhz so im assuming that the problem is the cpu, but since temperature seems fine, the only thing i could do is raise vcore... unless there is something else ( which im not aware of... )

Thanks,


Also should i add voltage to MCH ?
 
Upvote 0 Downvote
Anyone ?
im still at 1.46 vcore and 3.2Ghz ( 8 x 400mhz ), i've been running orthos "Small FFTs" test for over 45min with no errors and temps never went higher than 54C...

It seems that the blend test is more prone to errors, what would that mean ? Should I also try to use the "large in place FFTs" ???

Would my system be stable enough if it doesnt fail the small FFTs but fails the blend test ?
 
Upvote 0 Downvote
Small FFts - test CPU
Blend - both memory and CPU (But not as good as S FFTs for CPU
Large - From what I gather The best for STRESS Test system.

Your problem sounds more like memory and you could proably lower Vcore. Anyway I would try getting your memory stable, then start lowering your Vcore.

And Yes to raising your MCH voltage.

Run CPUZ and look at memory and make sure your Settings (Voltage and CL ratings) match the profile for your memory

ADDED:
(1) You might try setting your bus to 398, 399, 401, 402. Some MBs have a problem at a specific Bus speed and will work Fine @ slightly less/more.
(2) Your temp (39 C as I'm typing this) is not the temp you need to look at. This is basicly the idle temp and depending on your ambient (room) temp may be High ( my Idle temps for both cores is about 30 C this time of year. You need to monitor temp when CPU is under a heavy (max) load. ie Monitor with prime95 running or what I use to really check the CPU is Intels TAT program
http://anonforums.com/builds/TAT.zip

(3) From below listed thread
FSB nominal 1.2
MCH nominal 1.45
Also See:
http://www.tomshardware.com/forum/210280-29-overclocking-e6400-stuck
 
Upvote 0 Downvote


See the bold above... I would move this back down to 2.0, you are not overclocking the RAM if you intend to keep the ratio @ 1:1, you're simply pumping more juice to the RAM and may be the reason for the crashes, also, maybe loosen the RAM timings and try 5-5-5-12, this may help as well.
 
Upvote 0 Downvote
disagree with jonpaul37 on voltage - concur with trying looser timings BUT it Should work at rated values.

The manuf specs are 4-4-4 @ 2.0 -> 2.1 Volts. Reading the reviews a number of individuals had to go to 2.2 / 2.25 Volts to get the memory to work. Noted one reviewers system, E6400 @ 3.2 GHz, but didn't specify Voltage.

I would leave memory @ 2.1 V and Set the CL ratings in bios. If that does not work, then try the looser timings, Next bump the voltage to 2.2. If it still flunks Blend test, I think I would email G-skill

My system E6400 @3.2 worked with Corsair 4 x 1 Gig memory Ram V = 2.1, FSB +0.1, (G)MCH +0.1, Vcore stock), which I swapped out for 2 x 2 Gig Mushkins which Vram = 2.05V and Decreased FSB & MCH to +0.05 V above nominal value

NOTE: My FSB and (G)MCH are +x.xx ABOVE default values.
 
Upvote 0 Downvote
galex, you are not getting my question. I am asking why he needs to up the RAM voltage if he is not overclocking the RAM.

Scenario # 1:

no OC of CPU or RAM, would not require a voltage increase of either. RAM runs @ 400mhz - 2.0v, CPU runs @ 266.6 which = 1.5:2 ratio

Scenario # 2:

OC the CPU with which would require a voltage increase, but the RAM is still running @ 400 mhz while the CPU is now running @ 400mhz which = 1:1 ratio

Again, if the RAM is running at the same speed in BOTH situations, why the voltage increase of the RAM?
 
Upvote 0 Downvote
jonpaul37, He is not overvolting his ram @ 2.1 Volts. Manuf specs are 2.0 -> 2.1 which implies SOME of their memory may work @ 2.0, But some (I would bet more than some will require the 2.1). Reading the reviews at newegg, at least 5 indicate they need > 2.1.

Agree 400 MHz w/1 to 1 ratio (which I'm using) is with in specs, It's just a question of at What voltage. Also this is somewhat dependent on MB. Some mB seem to favor specific brands (Recommended Brand by MB manuf are normally at stock setting.

While the Ram is not being clocked higher than Manuf specs, He is overclocking the system ie Bus clock to 400 MHz. When you raise this frequency, it will tend to reduce amplitute of signal (MB dependent). This is why when OC you may to make "Small" changes, Increases, other voltages.

For My 965-DQ6, the Mushkins DDR2-800 had more favable comments

jon, hope this helps explain. forget my spelling errors as I did this in a hurry (And Spelling was NEVER my strong point.
 
Upvote 0 Downvote
thanks for the great input guys, i really appreciate it.

Alright so according to what im hearing here, sounds like this is a memory problem.

Here is what i did so far to look into this :

First, I knew I had already set manually in bios my memory timing to run according to specs ( 4 / 4 / 4 / 12 ) , but i opened cpuz to take a look and make sure that everything was matching. This is what I was running at :

CAS# Latency ( CL ) : 4.0 clocks
RAS# to CAS# Delay (tRCD) : 4 clocks
RAS# Precharge (tRP) :4.0 clocks
Cycle Time (tRAS) : 12 clocks
Bank Cycle Time (tRC) 16 clocks

Everything was matching except i noticed something in CPUZ's "SPD" tab... At the bottom of that tab there is a table called the "timing Table" which lists specs for 3 different profiles : JEDEC#1 JEDEC#2 and EPP#1 , I can tell/guess that EPP#1 should be the profile i use because it had the tightest timings ( 4/4/4/12 ) compared to the other profiles which are slower. These are the specs listed for EPP#1 :

CAS# Latency ( CL ) : 4.0 clocks
RAS# to CAS# Delay (tRCD) : 4 clocks
RAS# Precharge (tRP) :4.0 clocks
Cycle Time (tRAS) : 12 clocks
Bank Cycle Time (tRC) 36 clocks
Command Rate : 2T
Voltage 2.100V

This EPP#1 profile is a perfect match to the setting i had set in bios except for 1 thing : the tRAS ( Bank Cycle Time ) is supposed to be 36 clocks, and my ram was running tRAS at 16. I thought that was it, that was the solution, I simply had to change that to 36 so I went into BIOS and tried to find the option where i could change tRAS from16 to 36 but i was not able to find any settings for that... Doing a google search didnt help either, so instead I changed the bios's manual settings for my ram from manual to "by SPD" into bios.

I knew the "auto/by spd" setting would result in looser timings so if i can run at looser timings without stability issue I know for sure the problem is with the ram. So now I am back into windows, i opened cpuz again to take a look at what my memory timing are, as expected they are looser :

CAS# Latency ( CL ) : 5.0 clocks
RAS# to CAS# Delay (tRCD) : 6 clocks
RAS# Precharge (tRP) : 6.0 clocks
Cycle Time (tRAS) : 18 clocks
Bank Cycle Time (tRC) 24 clocks

If i look in the SPD tab's timing table, i see that the profile that is closest to this is the JEDEC#2 profile which should uses these timings :

CAS# Latency ( CL ) : 5.0 clocks
RAS# to CAS# Delay (tRCD) : 5 clocks
RAS# Precharge (tRP) : 5.0 clocks
Cycle Time (tRAS) : 15 clocks
Bank Cycle Time (tRC) 24 clocks

As you can see, the settings specified in JEDEC#2 are TIGHTER than the timings my ram is running at so there really shouldnt be any problem running orthos blend test... WRONG ... after 2 minute the test failed once, i tried a second time and it failed after 6 minute.

I am not sure what to do next...

I tried loosening the memory timing and that didnt work, although i didnt lower voltage when i did that i dont think it should be a problem, my ram is supposed to run at 2.0 ~ 2.1 volt according to G.Skill so that's what im using.

Now, I could :

A) change the fsb from 400mhz to something else (+/- 1 ), but really that is what i'd like to avoid the most... I could run at 2.8 / 2.9 ghz stable before with my old ram, i bought this 800mhz ram just because i wanted to overclock higher and get a 1:1 ratio, so if i change the 400mhz fsb it kinda defeat the purpose of me buying that ram.

B) Increase MCH/FSB Voltage. RetiredChief, this is something you seem to suggest however not that I am doubting you but after reading a little bit about MCH voltage increases it seems like the point of doing so is mostly useful in situation where the ram is overclocked, in my situation the FSB is only 400mhz which is the ram's normal speed. I'd be willing to give it a try though, I know you said to try these :
FSB nominal 1.2
MCH nominal 1.45
But i didnt see, in the link you provided, where those voltage were suggested ? also what do you mean by "nominal" ?

C) Increase Memory voltage. I dont think this would work though, if the memory that is rated for 4/4/4/12 @ 2.1 volt fails orthos using loose 5/6/6/18 timings with 2.1volt why would it work with higher voltage still ? In fact when using JEDEC#1 or #2 profile, the voltage suggested is 1.8V ... unless using too much voltage on loose timing can also result in instability ?

D) could switching the ram position on the motherboard help ?

E) maybe the problem is somewhere else ? As i stated before, while i was using the "tight" 4/4/4/12 memory timing and trying to overclock the cpu, when i raised the cpu vcore voltage i was able to run the orthos blend test for longer, If this is indeed a memory stability issue, then raising vcore should have had no effect ? maybe it is a coincidence... ?

Aww man, im sick of this, I have the worst luck with computers :)
anyone has any suggestions ?
 
Upvote 0 Downvote
Time to fall back and regroup.

With every thing at stock ( clear CMOS and load optium values) You should be able to run Prime 95 (blend) stable, if you can not, then run one stick at a time (You could very well have some BAD memory). If both sticks can pass prime 95 one at a time Min 2 hr test) then put both sticks in one set of sockets, then the other set.

ONCE y
 
Upvote 0 Downvote
Quick update :
I tried changing a few things, I put my memory voltage at 2.2, then changed the MCH setting from auto to the lowest and second lowest voltages and did the same for FSB voltage. Nothing worked, still was failing orthos within 6-7 mins.

Then I tried to up the vcore voltage again to 1.47, i put the FSB and MCH voltages back to AUTO. Doing that allowed me to run orthos blend for a little over 1 hour without any errors, i then stopped the test myself and went back to bios to lower the memory voltage back to 2.1 ( I had left it at 2.2 ) then I ran orthos again, I got an error at 56 minute this time... Not sure if it is related but the error came up at about the same time the screensaver started.

During those long test periods my cpu temp never went above 55C, it usually was at 53C and idle temp is 41C.

So one thing is for sure, upping vcore really seems to help with stability in blend test.

Retired Chief, I will try to do as you say and restart from scratch but first :
First I will try to find another bios version, you said I had to clear the CMOS , that means i have to take out the battery on the motherboard and wait a few mins right ? Also what do you mean by Optimum values... ?
Finally when you say to run prime 95 blend, does that mean Orthos ? i thought orthos was sort of like prime but better for dual core cpus ? And is 1 hour enough ? I really wasted TONS of hours already on this :) Basically im not sure if I am seing this the right way but to me being able to run blend for 1 hour or so is pretty stable for real world usage because its extremely unlikely that i'll ever be doing anything on my computer that requires 100% cpu for one hour ? but maybe that is not the way it works and im missing the point ?

I know for sure that before I had the new ram when I had my cpu overclocked at 2.8 - 2.7 ghz i could run orthos blend for many hours.

So it seems the problem is related to the ram but why is it being affected by the vcore ? Do you think 1.47 is too much ? if all else fail can i say to myself AT LEAST that i know i can safely go back to this current oc setup ?

Thanks
 
Upvote 0 Downvote
(1) To Clear CMOS, Check your manual. Should be able to (WITH Power off) there should be a jumper that you can short.

(2) In BIOS there should be a Fx to set all parameters to "Fail Safe" and a 2nd Fx to use optimium values.

Ref voltages quote form link I posted
http://www.tomshardware.com/forum/210280-29-overclocking-e6400-stuck
vCore: 1.300 (and I might be able to back this off further, just haven't tried lower yet)
DDR voltage: +0.3 (2.1)
FSB voltage: +0.2 (1.4)
MCH voltage: +0.1 (1.55)
 
Upvote 0 Downvote
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom