Begining To Oc

[Encrypted_God]

dont get me wrong, i understand him as well, but i was just trying to say that i was not just picking random settings and changing random settings to achieve a higher oc, i actually sorta had a method.

I'm sorry. I didn't word that correctly My fault. What I meant (and was directed to Verran) was; "I understand you Verran"

 

AND

 

......IMO you need to start over with your OC

[gotdamojo06]

ok i started over and ran super pi afer every time that i push the bus speed up (5mhz at a time) and what i have been looking at is cpu load temp, idle temp, and the super pi 1 and 4 M tests. to tell if it is stable, what should i be looking for? everytime i run the super pi, so far everything has been either the same or a little bit faster times, also the temps are very close to eachother. earlier when i was giving temps, they were wrong, i forgot i was looking at my own probe that i had placed in the actual heatsink for the cpu, but now im using the temps from speed fan, any other suggestions?

 

so far i have pushed the bus speed up to 215 mhz, and been runing a idle temp of 48C and load of 53C, im up to 2.37 GHz so far. i have not removed the duct and redone the air setup that was suggested earlier, but i will be doing that soon.

 

<---EDIT--->

 

ok just did the fan re arrangment, and it made a big differentce, now the load temp is 48C and the idle is around 43C. Thanks for the idea of taking the duct away. now time to keep pushing it...hope i can get it up there

[Lancer33]

Wow...I have never taken mine above 1.475v on air.

[hardnrg]

i ran my 3500 venice at 1.568v... maybe 1.600v for extreme benching... but that was with an XP-120, no ihs, a Panaflo 120x38 H1A @ 12V, and a hat and coat on in my room because it was that cold haha

 

personally, i think 1.53-1.58v is the sweet spot on air for most venices... generally, sweeping statements and generalisation there lol... some cpus do like more voltage... but you'll probably find that higher voltages actually have a detrimental effect... feel free to experiment to see if the voltage scales, but i'd use caution, and wouldn't use stock air unless i was some kinda madman that lived in a meatlocker or something...

[gotdamojo06]

lol, last time i looked around, it didnt look like it was a meatlocker, but you never know what it is around here....dorm room, sorta like meatlocker

[DECwakeboarder]

ok i started over and ran super pi afer every time that i push the bus speed up (5mhz at a time) and what i have been looking at is cpu load temp, idle temp, and the super pi 1 and 4 M tests. to tell if it is stable, what should i be looking for?

First off...super pi 1M and 4M are not stress tests, they are benchmarks. If you want to use super pi, then you need to at least be doing 32M. I would recommend another program though (OCCT if it is still around). No need to do this every 5mhz, I'd say every 10. If it fails, try to bump the volts up (I would stay at least under 1.6v, especially on stock cooling) if the extra volts don't help, drop it down 5mhz or so and find the highest stable clock speed you can get.

 

Note - all of the above should be done with a very large RAM divider...once you find the CPU's max speed, then start lowering the divider and lowering timings until it becomes unstable, then go back to the lowest stable setting you managed. (get the divider as close to 1:1 as you can, then start lowering timings)

[gotdamojo06]

Ok i have managed to get it to oc to 2.42 GHz while being stable with a 49C boot temp and a 46C idle temp. I think that is pretty good with the stock cooler. The voltage is set at cpu default, and i used OCCT to tell if it was stable or not.

 

thank you all for all of the help and suggestions

[DECwakeboarder]

Check the temperature using the same program with 0% load on the CPU and 100% load...I doubt that a stock cooler can keep a 3* delta, considering that even the best watercooling can't do that.

[cchalogamer]

No way in hell your temps are right. I can hit an easy 50C load on the stock cooler of my 3000+ before I top 2.4 with 1.45V...Just what are you using to read temps?

 

Personally my Venice doesnt see ANY improvement over 1.5V even above 2.8ghz, but that does varry by CPU. Also how are you checking these loads temps? It should be run doing a heavy task for atleast 10 min solid and checked BEFORE that task is complete (for example if you play a game to give it load then checking after closing the game or alt-tabing will usually cause a quick drop in temp, so it's best to log temps with some app to have the best results).

 

Just to see what it says the temp is try either:

 

SpeedFan (new temp added called "core" that does basically the same as the app below)

 

or

 

Core Temp

 

A little info on Core Temp for thoes who dont click the link:

This is a new program that lets you monitor Intel's "Core", "Core 2" and all AMD K8 chips' die temperature. The temperature readings are very accurate as the data is collected from a Digital Thermal Sensor (or DTS) which is located in each individual processing core, near the hottest part.

The way this program works is as follows: Intel defines a certain Tjunction/TCaseMax temperature for the processor. In the case of Yonah it is 85C° or 100C°. First of all the program reads from an MSR, detects the Tjunction/TCaseMax temperature. A different MSR contains the temperature data, the data is represented as Delta in C° between current temperature and Tjunction/TCaseMax.

So the actual temperature is calculated like this 'Core Temp = Tjunction/TCaseMax - Delta'

The size of the data field is 7 bits. This means a Delta of 0 - 127C° can be reported in theory. But from preliminary tests, the reported temperature doesn't go below 0C°, no matter what kind of cooling was used.

 

For AMD the value the CPU reports is read and converted into an actual temperature. AMD chips can report temperatures between -49C and 206C

 

I personally want to have a better hold on what the temps REALLY are before helping you go higher.

[LightningJR]

Ok, personally I would use 10Mhz steps because it could take quite a while to get to your max and 5Mhz increments can be overkill. SuperPI isn't the best way to find the stability of your PC. You need to stress all you components together, especially your GPU. It will increase your case temps and therefore increase your CPU temp. You should be using 3DMark and doing batch tests and a CPU test at once, this will stress your full PC and you will get the highest temps possible. You should be at least doing this tests for 20Min to get best results for the highest temps and even higher for stability testing. You need to make sure that you hgave rulled out everything but the CPU when you are trying to overclock. Bring ram down as low as you can go and HTT and make sure the PC/PCIE etc speeds are locked or are their right speeds. Idle temps are the temps when your PC is at idle for 20Mins. Increase 10Mhz and stress test and record the temps and show see if it is stable. Do again and record the temps. (log temps because like cchalogamer said, they can change as soon as you stop the stress testing or press a key.) Keep going until the stress testing fails. Then you increase the voltage (small increments) and test again, keeping a close eye on temps. If its still not stable increase again, always keep an eye on temps. Like most people said 1.7V is very high and most people wouldn't recommend higher than 1.6V and that's still a bit high. If your CPU can handle it than do what you like but if you don't see any change in stability then it's best to lower to a less extreme level. If the voltage increase makes the PC stable then do more stress testing and keep increasing the CPU speed until it starts to become unstable again, and then keep increasing the voltage, always keep a close eye on the temps. When you reach your ceiling then you have to decrease the bus and test again, and keep decreasing it until you can find the stable setting. When you have found the sweet spot you need to stress test for absolute assurance. You will need to stress for about 20hrs, most people like to say that 10hrs is enough but ive read that some people got instabilities within 18hrs, but it's up to you how stable you want your pc. I know I would want 100% stability. When you get that done and all is good you can increase your ram and htt, you will either find a perfect multiplyer with both or you will either need to go higher than default ram/htt or lower. You will need to test again if you go higher than default. At 250bus you can have 4:5 ram and 4x htt and the ram is at 400Mhz and the htt is at 2000Mhz which is goo, But test to make sure, it will add to the temperature. lol, lots of info, some people do it different but this way works too. It takes long but you will get the most out of your PC if you do this. Correct me if im wrong but I don't think that 3DMark06 or 05 will be enought of a stress test on your CPU. So I recommend running 3DMark and an instance of prime95 or superpi batch run also. Well that's my 2 cents, well maybe 4.... :thumbs-up:

[Verran]

@LightningJR:

There's some good points in there, but good lord man, take a breath. Put some spaces in there please. No one wants to read a wall of text like that.

 

And with regards to the SuperPI stability point:

-I- was the one who suggested he use SuperPI, and I stand by it. I -did- put a disclaimer about this at the end of my post in which I made the suggestion.

 

It's funny to see people say SuperPI isn't enough, and then say that testing at every 5 mhz is overkill. You're taking away one complication and adding another. I use SuperPI to "feel out" my chip's ceiling, and once I have a decent idea where it is, I use OCCT and Prime to find it exactly. This process works very well for me, and it's the exact one I used to get to 3.0 on my Opteron.

 

I set a divider on the ram, and then use clockgen to jump in 5 mhz intervals. I run 2 1M tests, and then 2 4M tests, and document the results (times if passed, X's if failed). If it fails, I jump the volts up and try those same speeds again. I do this all the way up. I generally have a top end voltage and temp that I'm not willing to go past, even before I start. When I get to the "ceiling" using this method, I start phase 2 of testing.

 

Let's say the top of my Pi tests left me at 300mhz HTT as my "ceiling". Now I'll drop to 290 or so and start testing with OCCT 30 minute test. I'll do these at 1 or 2 mhz intervals until it fails. Then I'll step back to the last successful speed and run a 10hr+ OCCT or Prime test.

 

I like to "narrow down" my testing in phases. Whoever says SuperPI is not a stability test is dead wrong. Even 1M can tell you a lot about your stability. It's not the end of the road in stability testing, but that doesn't mean it doesn't have its uses. I've found it to be pretty reliable that true stability is about 10mhz under 1M stability. Using SuperPI with this sort of information in mind can save you LOADS of time in the "bottom end" of your testing procedure, while still helping you gather a lot of info about your chip.

[LightningJR]

I know, I was just typing, typing, typing. I didn't say that SuperPI wasn't good it's just a CPU stress test, you need a full system stress test to check stability. The temperature increase of the hard drives, ram, video card, etc. will increase the case temperature and in turn will increase the CPU temperature which can affect your stability. If you just use SuperPI or Prime95 to stress test and your CPU is stable then you go and play a game for 3 or 4 hours and your PC locks up you wont be very happy.