Overclocking gone wrong?

outsid3r

Weaksauce
Joined
Jan 28, 2007
Messages
106
I'm still fairly new with overclocking. I set it to 225x9 System Memory Multiplier is on auto
It won't boot on that...

I'm not sure how to use the (SPD)

I folowed an overclocking guide and with my current specs

It was said to do a small over clock by putting it to 9 - 3.0 - 266. When I do that my computer won't boot.

Am I not able to overclock at all? This board and proc was said to be the best overclocker for the price.
 
Something is horribly wrong,

Please download and install coretemp and tell us what it reads as your core temps at stock or at whatever you can boot at .

9 - 3.0 - 266. the 9 I understand (cpu multi) the 266 must be FSB what is the 3.0 ? if memory mulitplier that is your problem it should be 2.0.

If you used my guide please go double check, if I put 3.0 in I made a mistake and want to fix it.
http://www.hardforum.com/showthread.php?p=1030795318
 
Both cores are at 53 - 55ish Celcius. When I go into bios it reads like 37 to 42c.

When I look at defaults why is the mem multi at 4?

With it set to mem multi 4 I was just able to get it to 220FSB which was really nothing.

And yes it says 3.0 in your guide.
 
OK 50ish at idle with coretemp - way way too high. If using the stock heatsink I would guess one or more of the pushpins is not fully seated. It takes a lot of force to seat them, so much it will hurt your thumb and I do not like doing it with the board in the case as the board gets flexed a lot due to the pressure you need to push with.

If its not a stock cooler it needs to be remounted.

You probally have EIST enabled which when the machine is not under load will reduce the voltage and mulitplier to the cpu to run it slower to save energy.

PC2-6400 (800Mhz)_______ 293______( 880) _________ 1.76GHz __ use 3.0 multi

The above is part of a memory testing stragety, we are not there yet.

Get you idle temps down into the 30s and we will go from there.
 
Alright going to pull out my mobo right now and check. I'm almost positive though that all 4 clicked when I put it in the first time. I will get back at you in 30 mins or so. Thanks for all your help so far I really really appreciate it.
 
np I watching a movie and will pop back in a few.

If you can get your idle temps down try this.

cpu mulit 8
fsb 300
memory divider 2.0
memory set to [Auto]

Vcore 1.375V
memory voltage to what the manuf recommends, if they give a range 1.8-2.0 v put it in the middle or higher value. default is 1.8v so for 2.0v you would put in +.2V for example.

I will check back in.
 
Alright well I took everything apart took off the heatsink and I dunno if it has to do with anything but the paste that's on it is really light not sure how much is supposed to be on. But I guess there is enough to cover the whole top of the proc. but still able to lightly see the metal. Anyways with the computer completely on idle except for this website open it's temps are at 48C. I don't think I really did anything. I made extreme sure that the heatsink was pushed in untill all 4 clamps clicked and than more. :p


1.8v - 2.0v is manuf on Ram
 
it must be hot in your room is all I can figure Leave the side panel of the computer off that might help.

Try the settings I mentioned and let us know what happened.

Try the ram at +.1 (which will be 1.9V) we are not going to push it at first lets see what we can do about getting the cpu going.
 
Actually it's freezing in my room lol. I live in a campus appartment and aircondtioner is set to 70 we don't have to pay for electrical or water bills. What I really think is core temp isn't reading it properly. I mean I go boot into the bios and it's under 40c.

Ok I tried the program "SpeedFan" Here are the Temps:

Temp1: 36C
Temp2: 39C
Temp3: -2C
HD0: 30C
 
Just a thought it sounds like you used a paste but I am not sure. The stock heatsink comes with a gray material pre-applied. Did you replace that or are we still using it ?
 
Yes using the stuff that was already on it. Also I was in a game when I had Temp2 at 39C now it's at 32C with the game turned off. According to "SpeedFan" Program
 
Bah site went down. I am concerned you are using the thermal materail that came with the heatsink. Its not really made for reuse. If you can get some thermal grease like Artic Silver 5 or any other good brand and remove the original Intel stuff with alocohol and use and apply a very thin layer of new grease it should help a lot.

The temps listed as core1 core2 is what we are really concerned about and want to watch, those are the reading from internal to the cpu digital sensors. The others are temp sensors on the board somewhere and are not nearly as important.



I will be watching thread.
 
Bios temperature readings are not always accurate. I keep this picture handy to prove the point.

biostempap5.jpg


A CPU at 13C with a room temperature over 20C is impossible. Some days it looks more believable but if it can produce nonsense like 13C then it can't be trusted.

The problem is you have an E4300 processor which CoreTemp 0.95 assumes has a maximum Tjunction of 100C. That 100C number is not documented by Intel and it is likely only 85C like all of the other Core 2 desktop processors. What this means is that your core temperatures are likely being reported 15C higher than what they actually are by CoreTemp.

The only Intel temperature information for a Core 2 Duo is the amount of head room you have until thermal throttling begins. Set CoreTemp 0.95 up like the following picture to read the on chip digital thermal sensors directly.

coretempdtsdc4.png


50C of head room at idle until your processor reaches it's maximum Tjunction temperature is a NORMAL Core 2 Duo idle temperature and you should be pretty close to that number.
No need to buy super paste or R&R your heatsink 101 more times.

Drop your memory multiplier down to 2X and raise the memory voltage to 2.1 volts so it doesn't hold back your overclocking fun. Set the memory timings to CL 5-5-5-15. Once you reach a suitable CPU overclock then you can adjust your memory to see if any additional performance can be gained without killing stability or not being able to boot up.
 
Bios temperature readings are not always accurate. I keep this picture handy to prove the point.

biostempap5.jpg


A CPU at 13C with a room temperature over 20C is impossible. Some days it looks more believable but if it can produce nonsense like 13C then it can't be trusted.

The problem is you have an E4300 processor which CoreTemp 0.95 assumes has a maximum Tjunction of 100C. That 100C number is not documented by Intel and it is likely only 85C like all of the other Core 2 desktop processors. What this means is that your core temperatures are likely being reported 15C higher than what they actually are by CoreTemp.

The only Intel temperature information for a Core 2 Duo is the amount of head room you have until thermal throttling begins. Set CoreTemp 0.95 up like the following picture to read the on chip digital thermal sensors directly.

coretempdtsdc4.png


50C of head room at idle until your processor reaches it's maximum Tjunction temperature is a NORMAL Core 2 Duo idle temperature and you should be pretty close to that number.
No need to buy super paste or R&R your heatsink 101 more times.

Drop your memory multiplier down to 2X and raise the memory voltage to 2.1 volts so it doesn't hold back your overclocking fun. Set the memory timings to CL 5-5-5-15. Once you reach a suitable CPU overclock then you can adjust your memory to see if any additional performance can be gained without killing stability or not being able to boot up.

Ok I have those settings set and both cores are at

Core1: 50 to Tjunction remaining
Core2: 50 to Tjunction remaining
 
excellent that makes a lot more sense, I was unaware of the Tjuction issue. If you Tjunction is 85C, and I beleive Unclewebb he has done extensive investigation into the issue, instead of the reported 100, 50C to Tjunction means your cores are (85-50) running about 35C which is what I was looking for. There is no point in trying to OC if temp are high to start with, but we are ok.

what that is telling us is that we have about 40C of temp headroom to play with.

Download this if you have already not done so.
http://sp2004.fre3.com/beta/beta2.htm
You want the big link in the middle of the page.

Be on the lookout for some good quality thermal paste, we will need to replace that Intel stuff sooner or later, see if someone else into computers in the dorm has some Artic Silver paste and will let you have a drop, thats all it takes, so people that buy a tube of it normally have plently extra.

Keeping coretemp (or speedfan) up, run orthos, just leave whatever is set, set and hit the start buttons. what we are doing is heating up your cpu in an attempt to get that intel thermal pad to melt and make good contact between the cpu and the heatsink. With your previous issues I an not sure the computer has ever been used hard enough to do that and removing the heatsink to check the pins means we need to try and get that stuff heated up so it will melt and flow some for good contact. Let Orthos run for 1/2 hour, temps just jump up, as long as you are 10C from Tjunction or more, no worries.

Do that and let me know, of course if you can get you hands on some thermal grease quick, install that first. use alcohol (90% rubbing alcohol is handy to keep around) to help dissolve the grey gunk, and a coffee filter (no lint) to remove the gunk, you want to get every spec of it off the cpu top and the bottom of the heatsink then apply a drop of AS5 or other good grease in the middle of the cpu top. Then just put the HS back on, the pressure of the HS on the cpu will spread out the paste. If you want put a finger in a baggie and gently spread the drop of grease into a thin layer all over the top of the cpu. A layer so thin you can almost see thru it.

Try this after you run Orthos for 1/2 hour.

cpu multi 8
fsb 300
memory divider 2.0
memory set to [Auto]

Vcore 1.375V
memory voltage to what the manuf recommends, if they give a range 1.8-2.0 v put it in the middle or higher value. default is 1.8v, so for 2.0v you would put in +.2V for example.

The above should give you a 8 x 300 = 2.4 GHz OC with your memory running underclocked at 300 x 2 = 600MHz

The components you have should do that easily. Its just a place to start. If that works run Orthos for 5 min and see what your temps do, if you are still 20C or more away from Tjmax then try:

FSB = 350 leave all other stuff alone.
8x350 = 2.8GHz cpu speed and memory running underclocked at 700MHz .

Again if the board is set up right, your hardware should do that easily.
 
Ok I have those settings set and both cores are at

Core1: 50 to Tjunction remaining
Core2: 50 to Tjunction remaining

That's exactly what I thought. There's absolutely nothing wrong with your temperatures.

Now you can run Orthos and make sure your system is stable at default MHz without any overclocking.

Let us know how much temperature head rooom you have while running Orthos and if you're stable at default settings. You can't start overclocking unless it's proven that your system is stable to begin with.

What type of memory are you using and what are its specs? Adjusting the memory multiplier to keep your memory running within spec is holding you back at the moment.

Use CPUz which displays the speed your memory is running at. If you use 2.1 volts for the memory you should be able to run it reliably at 400 MHz which is known as DDR2-800 speed. You have to know what your memory is capable of before you can overclock your processor and you have to keep your memory running within its capabilities when you start overclocking your processor. These two are linked on the DS3 so you have to keep your memory running below its maximum capable speed or you will not be stable.

Don't leave your memory timings on AUTO. Manually adjust it to cas latency CL5-5-5-15. When you've done it right those numbers will show up in CPUz as well under the memory tab.

basictimingssb1.png


My memory is capable of running with CL4-4-4-12 timings at 450 MHz but you should start with less memory MHz and looser CL5 timings.
 
Unclewebb,

What's with CoreTemp 0.95 reading 15C higher than 0.94? My e4300 at 3.33GHz now idles at 45C not 29C, Orthos pushes me to 72C instead of 60C. Danger?
 
Alright well I still have one more class to go after that I will try and head over to a hardware store and see if I can find some arctic silver.

One more question it does read 50 to Tjunction remaining. But it also says Tjunction: 100C is that still fine?

Tjuntion: 100c "Compared to yours has 85C"
Core1: 50 to Tjunction remaining
Core2: 50 to Tjunction remaining

And since it does say 50c to Tjuntion what is the real temp of that? And how do you figure that?

I will get back to you guys later today.
 
ahh those pesky classes well beleive me, it beats working for a living :D

Ok what it boils down too is the programmer for coretemp made a mistake. The actual Tjuction temp (really should be called Tjmax) is the temp where the processor says "HELP I AM TOO HOT ! SPIN UP THE FANS" really, not kidding. Trouble is we have no way of knowing what the temp is, its set at the factory and we cannot read it so the programmer took a guess and guessed wrong. BTw there is a second trip point set slightly higher, and again we cannot determine what that is, where the processor will just say " Screw it, I am too hot, bye bye" and shut itself down.

What we can read off of the cpu and what is documented and known for a fact is a software register that give a "number of deg C until the processor hits that I am too hot trip point. So that number, a delta until trouble starts is a good number.

Here is more than you want to know, Unclewebb did some bang up testing to determine what was going on. In short we are pretty certain your Tjunction should really be 85C or close to it, if you then subtract the delta to Tjunction you get the actual temp in C your processor is running at. Tjunction of 100 minus the delta gives one number and we are pretty sure that is wrong. Tjunction of 85 minus the detla gives us a different number (15C different) and we think that would be correct. So due to a programming assumption that we think is wrong your coretemp "normal" readings are 15C high and when you reported them I was very concerned about heatsink mounting etc, until Unclewebb reminded me your processor is one of the ones coretemp does not read correctly, there are a bunch of Core2duos out now and I cannot rememer all this stuff.

Beleive me when I say Unclewebb and myself have looked and looked at this and I would be amazed if we were mistaken, Unclewebb has even run his processor without a heatsink to verify by experimentation what the Intel documentaion says and verified the error in coretemp.

If you would like to really get into it
http://www.intel.com/design/processor/designex/313685.htm
(The thermal and mechanical guidelines for C2D from Intel)
( section 4.2 has the good stuff)


Dont get too hung up on it all, as Unclewebb says, as long as we keep an eye on the "distance/delta to Tjunction" that tells us exactly how much thermal headroom we have for overclocking. In theory we could run right up almost to it, I would like to stay at least 10C under for long term reliability etc. So making the heatsink work as well as we can will help increase the ability of the machine to overclock but still stay way under the "I am too hot spin up the fans " trip point.
 
Alright guys I couldn't get my hands on any arctic silver so I went ahead and bought "DYNEX Silver Thermal Compund" I was ripped off by BestBuy cost $11. But that was the only way I could get some today. Anyways I will go ahead and continue to try this stuff out I dout it will really do anything but it's worth a try. I got some 91% Alchohol and car "Micro Fiber Towel" to whipe the stuff off. After that I'll test out Orthos for a good 30 mins and get back to you guys with the temps.
 
it will help, the stock intel pad must be replaced.

the new stuff will not need to be "melted" in, that was in case you could not find anything and we had to go with what was left of the original intel pad.

10 min of othos will be plenty to see what your new "distance to Tjuction" is it wont be a huge difference but now we are sure we have the heatsink in good order.

Please relax with the "negative waves" you have good equipment and I know that board backwards and forwards. It might take a bit because we have to communicate through the forum but if you hang in there, if something is not defective, I will get you the best OC possible with what you have. No worries

You want trouble, try replacing the drive belt on a riding lawnmower LOL, back in a bit.
 
What's with CoreTemp 0.95 reading 15C higher than 0.94?

Here's the simple explanation.

Reported Core Temperature = TjMax - Digital Thermal Sensor (DTS) reading

As the core temperature increases, the reading from the DTS decreases. It tells you how much head room you have before the processor will begin to slow down by throttling itself. At DTS=0, the processor will keep running but it will reduce the MHz by cycling the multiplier to 6 and it will also try to reduce the core voltage to control any overheating condition.

The DTS is a positive number and your processor is documented by Intel to be able to run safely right up until DTS=0 but for practical purposes when you are overclocking you will need to leave some head room so that you maintain stability and don't ever start to thermal throttle you processor which will slow things down.

TjMax is the maximum Tjunction temperature that your processor can safely operate at. Unfortunately it is not documented by Intel for the Desktop Core 2 Duo processors so ALL software can only guess at that value. It is generally believed to be either 85C or 100C but in my opinion, all Core 2 Duo desktop processors are rated at the same 85C TjMax temperature. All Core 2 Duo mobile processors have an Intel documented TjMax = 100C.

When the E4300 came out, the programmer of CoreTemp decided that 85C wasn't right so with version 0.95 he changed it so that it now guesses at a TjMax of 100C for the E4300 as well as the revision L2 E6300 and E6400. If you look at the above formula then it is pretty obvious that guessing wrong at the TjMax by 15C is going to result in the Reported Temperature being reported too high by 15C.

That's why I recommend setting CoreTemp 0.95 to report "Show Delta to Tjunction temp." This eliminates the possible temperature error if CoreTemp is indeed guessing wrong.
 
Alright guys sorry I took so long. It was my first time putting on Thermal Paste so I took my time and it was actually exciting and fun! So here are the results:


10mins into Orthos the lowest to, "Tjuntion Remaining was 32C."

At Idle it was 52C.

Now lets get to the fun stuff! Haha this is already fun :p
 
Here's the simple explanation.

Reported Core Temperature = TjMax - Digital Thermal Sensor (DTS) reading

As the core temperature increases, the reading from the DTS decreases. It tells you how much head room you have before the processor will begin to slow down by throttling itself. At DTS=0, the processor will keep running but it will reduce the MHz by cycling the multiplier to 6 and it will also try to reduce the core voltage to control any overheating condition.

The DTS is a positive number and your processor is documented by Intel to be able to run safely right up until DTS=0 but for practical purposes when you are overclocking you will need to leave some head room so that you maintain stability and don't ever start to thermal throttle you processor which will slow things down.

TjMax is the maximum Tjunction temperature that your processor can safely operate at. Unfortunately it is not documented by Intel for the Desktop Core 2 Duo processors so ALL software can only guess at that value. It is generally believed to be either 85C or 100C but in my opinion, all Core 2 Duo desktop processors are rated at the same 85C TjMax temperature. All Core 2 Duo mobile processors have an Intel documented TjMax = 100C.

When the E4300 came out, the programmer of CoreTemp decided that 85C wasn't right so with version 0.95 he changed it so that it now guesses at a TjMax of 100C for the E4300 as well as the revision L2 E6300 and E6400. If you look at the above formula then it is pretty obvious that guessing wrong at the TjMax by 15C is going to result in the Reported Temperature being reported too high by 15C.

That's why I recommend setting CoreTemp 0.95 to report "Show Delta to Tjunction temp." This eliminates the possible temperature error if CoreTemp is indeed guessing wrong.

Thanks for that! Will reset the TjMax to 85 and have it show Delta. Appreciate it.
 
Thanks for that!
No problem. SpeedFan 4.32 uses TjMax=85C for all of the Core 2 Duo desktop processors so you might want to use that. It also draws some nice temperature graphs which is good when you are gaming then you can go back and have a look at how hot it was running during the previous 13 minutes.
 
outsid3r : Your temperatures are good to go. Lots of head room so set your memory speed properly and start turning up the CPU MHz. The DS3 is simple to overclock and BIllParrish has plenty of experience with them! :D

I avoid going up 10 MHz at a time and like going for a home run immediately. Some motherboards have holes where they simply don't like to run but if you go higher they will run fine.

333 MHz x 9.0 = 3000 MHz

With a memory multiplier of 2 your memory will run at DDR2-666 speed which ANY DDR2 memory can do, even at default voltage. You're going to be surprised how easy this really is.
 
yep thats a good target

cpu multi 9
fsb 333
memory divider 2.0
memory set to [Auto]

Vcore 1.375V
memory voltage to what the manuf recommends, if they give a range 1.8-2.0 v put it in the middle or higher value. default is 1.8v, so for 2.0v you would put in +.2V for example.
 
Ok quick questions.

1. What is the proper quick way to know how to set the ram settings. I mean for example my default ram timings I believe are 5 - 5 - 5 - 15. Do I just immediatly set it to 4 - 4 - 4 - 14? And if it was succesful than 3 - 3 - 3 - 13? I've read the guide I just want a quick answer to that. I know that I will set the volt for the ram at 2.0v

2. Now how will I know the ram is good where it's at when I'm done? I know that if it doesn't boot there is something wrong but, I mean over a period of time. I know in the guide it mentioned some software ramtest but it required a floppy I believe is there a way of using a thumbdrive? If so how?

3. Should I just work with the ram first or do all the settings at once (Cpu, Multi, and Ram)
 
Leave your memory at CL5-5-5-15 and set everything at once and let 'er rip. It should boot up no problem as long as the memory divider is set properly. Post the part number and brand of your memory but it's likely that you will get your best overall results by running it at CL5 and concentrating on overclocking the CPU.

If your memory is causing problems you will know when you try to run Orthos and it reports errors.
 
Ok quick questions.

1. What is the proper quick way to know how to set the ram settings. I mean for example my default ram timings I believe are 5 - 5 - 5 - 15. Do I just immediatly set it to 4 - 4 - 4 - 14? And if it was succesful than 3 - 3 - 3 - 13? I've read the guide I just want a quick answer to that. I know that I will set the volt for the ram at 2.0v

2. Now how will I know the ram is good where it's at when I'm done? I know that if it doesn't boot there is something wrong but, I mean over a period of time. I know in the guide it mentioned some software ramtest but it required a floppy I believe is there a way of using a thumbdrive? If so how?

3. Should I just work with the ram first or do all the settings at once (Cpu, Multi, and Ram)

My sig has the link to Bill's OC guide, perfect step-by-step....
 
Do not worry about your ram, one thing at a time, There is a plan to this. If you want to flop around and change a bunch of stuff and not know wth you did fine.

With a 2.0 memory divider you don't need to even think about your ram until we see what and were the FSB ends up when we hit the temp wall with the CPU.

your ram is rated 5 5 5 15 anyway, you are not going to be able to tighten it up much if at all. so auto and your rams specs end up being the same thing, set your memory voltage to +.2V if yo have not already done so.
 
Alright guys I have some news!

Here is what I had gotten it up too:

DDR2 - +0.2v
PCI-E - Norm
FSB - Norm
MCH - Norm
VCore - 1.350

333Mhz x 9.0
Mem Div - 2.0

Booted succesfully with an idle of 50C to Tjuntion remaining. ran Orthos for 10 mins and the lowest it hit was 22C to Tjuntion remaining. So basically a 10C drop at load before it was over clocked.
 
Alright guys I have some news!

Here is what I had gotten it up too:

DDR2 - +0.2v
PCI-E - Norm
FSB - Norm
MCH - Norm
VCore - 1.350

333Mhz x 9.0
Mem Div - 2.0

Booted succesfully with an idle of 50C to Tjuntion remaining. ran Orthos for 10 mins and the lowest it hit was 22C to Tjuntion remaining. So basically a 10C drop at load before it was over clocked.

excellent, increase FSB by 10MHz a pop until it will not boot or when running orthos for 5 min, your distance to Tjunction get (lowers) to about 10C.
 
Ok well remember I'm not using any special cooling everythings stock so is it still safe to go further than 333?
 
absolutely thats why we are testing with Orthos and watching the temps. However it should be running much much better than stock so if you are happy stopping here , try your game, then thats good too.
 
Once we find the "happy spot" for the CPU we are going to play with the CPU mulitplier and the FSB to get your memory running closer to its rated specs.

for instance you are at 9 x 333 = 3.0Ghz CPU speed but your memory is under clocked at 667MHz.

If we change the mulitplier to 8 but raise the FSB to 375 your CPU will still be running at 3.0GHz but your memory will now be running at 750MHz, faster and closer to its rated speed but still a little under clocked. that would probally be a good spot to stop. Anything more requires a lot of time tweaking and testing and frankly will bring small returns. You have gotten the lions share of performance increase already by getting the CPU speed up around 3.0GHz.

The perfect settings for you would be a cpu mulit of 8x and an FSB of 400 running the cpu at 3.2GHz with a memory divider of 2.0 so your memory is running exactly on spec at 800MHz . You will probally need a better heatsink to do that, artic 7 freezer pro is a nice HS for the money and should do the job.
You might have to raise vcore a tad to get to 3.2 and MCH voltage a bit too, but I am getting to far ahead.
 
Ok Um I got it to work fine at 343 ran orthos for 5 mins and it ran fine so I decided to take it to 363 I know went up 20 lol anyways it booted and ran orthos for 2 mins and it got:

Test 2, 4000 Lucas-Lehmer iterations of M19922943 using 1024K FFT length.
FATAL ERROR: Rounding was 0.5, expected less than 0.4
Hardware failure detected, consult stress.txt file.
Torture Test ran 2 minutes 52 seconds - 1 errors, 0 warnings.
Execution halted.

So I took it down to 353 and it wouldn't boot very weird so I'm happy with 343 untill I get a better cooling.

Now you said put the vcore to 1.375 I thought it was wrong because your guide said 1.350 so I set it at 1.350 I don't know if that will have anything to do with it. The default was 1.300 or 1.325 can't remember.
 
Basically as you make the cpu go faster you need to give it more juice. 1.325 is the typcally default.

the trick it is to run as fast as you can with as little Vcore as needed for stability. If it will boot and run orthos at 343 with 1.350 good, no need to go higher as higher causes more heat.

So 343 aint bad, but your memory is way underclocked.

343 x 9 = 3,087GHz

try 8 x 385 = 3.087 GHz it will boost your memory speed, give it +.1 to the MCH voltage.

If it will do that, its a good place to stop and play the hello out of your game to make sure everything is stable.
 
Back
Top