Originally Posted by Smo
Nice dude - any idea what sort of turbo those things have as standard? A 100% bhp increase is almost unheard of, especially for something that comes out the factory on a production car.
Love the bike by the way, great excuse for getting yourself killed! I want one
Thanks for posting that - I've just had a read through and I must say, I'm stunned. To think that to potentially utilise the full potential of my out-of-the-box NVIDIA hardware I would have to overclock a quad-core CPU to 4.8GHz is mad.
Good job I'm up for trying it then
I have an Antec Kuhler H2O 920 CPU Cooler so temperatures shouldn't be a problem. I'm currently at 4GHz, as last night I downclocked to look into my voltage. What I'll do is run some Benchmarks as is (with my 590 @ 670Mhz) then do a quick forum search for upping the 2500k to 4.8GHz.
I'll work out an overclock that will hopefully prove Prime95 stable and re-run my benchmarks, see what happens. This is also my first in-depth overclock so you're popping my cherry here Cain!
Thats how most of us do it, we raise clocks, estimate the voltage needed, then we run Prime95 to test it out.
If it goes for more than 3~24 hours right off the bat good chance you got lucky and may even be able to lower voltage. I like to "trim the fat" on voltage, and use one increment higher than what took me to be stable.
So for 4.4 GHz (what I run 24/7) it took me exactly 1.389v to stabilize. This is just under my maximum safe voltage for my CPU. I went ahead and just put it as 1.4v, even though its Prime95 stable, just as a rule of habbit. I have been stable on Prime95 and IBT and every once in a while I will have a game crash for no apparent reason... I consider it to be some type of dip in voltage and usually just small nudge up ensures complete and total stability.
A lot of these errors can occur if you have too much voltage as well, going to overclocks beyond what your chip is capable requires more voltage, but more voltage does not guarantee more frequency or stability. So if you are starting low which is how I overclock, this may help you figure out how much voltage you need to add/adjust
Here is the order of instability when overclocking just the CPU I go by, but its not all encompassing.
Usually Needs Large amounts of voltage increase
Needs moderate amounts of voltage increase
- Computer fails to enter post: Really Unstable, CPU Voltage Too Low
- Computer freezes during POST or in BIOS: Really Unstable, Voltage Too Low
- Computer freezes after POST but before entering Windows: Really Unstable, Voltage Too Low
Needs fine tuning via voltage increases
- Computer gets through POST and Windows Boot, but freezes on Desktop: Needs moar voltage... still pretty unstable.
- Computer POSTS/Windows Loads, weird application errors come up: Needs moar voltage... still pretty unstable.
- Computer boots to desktop, can't get a specific program to load: Needs moar voltage... still pretty unstable.
- Computer boots to desktop, every thing looks okay, applications seem sluggish to load, fail to even load period but not necessarily errors: Needs moar voltage... still pretty unstable.
- Computer boots to desktop, every thing looks okay, programs load, but there are a few strange errors in EventViewer Logs (drivers & media applications seem to have a lot of warnings/errors.) Needs moar voltage... still pretty unstable.
- Everything boots up, and everything loads, no visible signs of errors, but I randomly get BSOD involving error codes that involve drivers for example: 0x0000037 I believe is an nVidia driver (people think its the graphics card and not the CPU) / or CPU errors here 120~129, and or IRQ_NOT_LESS_THAN_OR_EQUAL errors. Any form of crash that is unexpected is unstable. Up ze voltagez!
- Everything boots up, and everything loads, no visible signs of errors, but video games are not playable (crash): Any form of crash that is unexpected is unstable. Up ze voltagez!
Everything boots up, and everything loads, no visible signs of errors, maybe even video games are playable: This is where most people STOP upping the voltage
. Because X Y Z game is playable. In reality, not la la land, its STRESS TESTING TIME!
Needs fine tuning via voltage increases
- Prime95 runs for about 5 minutes, and the whole program crashes (not Windows). This is slightly unstable, but not enough to bring down the system. You have made progress, but still need to up voltages.
- Prime95 (SmallFFTS Test) runs for about 1 minute, and an error popups. This is slightly more stable than the program not crashing (no matter how long Prime95 ran before crashing), but its even less likely to bring down the system. You have made progress, but still need to up more voltage.
- Prime95 (SmallFFTs Test) runs for about 30 minute, and an error popups. This is much more stable, and in fact you have just entered the upper levels of stability even though you have had an error, but its even less likely to bring down the system. You have made progress, but still need to more voltages.
- Prime95 (SmallFFTs Test) runs for about 3 hours, and an error popups. This is very stable, and in fact you have just entered the what some considered the highest levels of stability. Even though you have had an error, but its very less likely to bring down the system. You have made progress, but still in all honesty, need to up more voltages slightly.
- Prime95 (SmallFFTs Test) runs for more than 8 hours, no errors. This is enough for me to claim stability, BUT its not as long as some people do run it, especially people who Fold/BOINC and need 100% accuracy on their CPU cycles. I encourage letting it run for at least 24 hours, but at this point, I use my own logic/experience. I let it run for 8 hours, and then I know I am going to increase voltage by one or two increments after its done regadless. To account for the unknown stability. It achieves essentially the same stability in less time, for least amount of voltage required (efficient.)
Once the CPU is "stable" then you switch testing to Prime95 Blend if you are about 4GB of RAM, or run CUSTOM for having large amounts of ram 8/12/16/24 GB of RAM (like me) and just increase memory usage to make it 80~90% of your system RAM. This ensures CPU / Communication / System RAM are also error free. I just like Prime95 because I am used to it and its been around almost as long as I have been playing with computers.
This is how I think others should test for stability. I will tell you though A LOT of people on this forum don't do this... I think primarily for either laziness, heat, or just because they don't understand how it works. I also liked to overclock, but I NEED a stable system. Thats whats very important to me. I think that overclocking has gotten so easy that anyone can now go in and just up multipliers without having to struggle with it etc, so there is nothing learned nor anything risked really.
A lot of people also use the excuse "nothing will run your hardware that hard like Prime95 or IBT" and that is not what the programs do. All those programs do is find an error in mathematics. An error or instability can occur under ANY difficulty or stress on the CPU. They illogically think that only stress produces error. This is simply incorrect, and misinformed. It also allows them to continue being lazy, and encourage others to do the same. It actually pisses me off severely. I don't mind people doing what they want with their hardware, but then they are essentially spreading lies to the new guys, who then learn it this way, AND show up months later and their entire computer doesn't boot. Argh... anyways, enough ranting.
I guarantee 50% of the errors on this forum in form of a game crash, driver crash, media crash, or x y z problem occured, is because a lack of proper overclocking testing. One other extremely useful tool is Window's EventViewer, which you can get to by going to Control Panel -> Small Icons (top right) -> Administrator's Tools -> Event Viewer and check out Windows Logs, specifically Applications, Windows, and Setup (if you are installing a program that crashes.)
It gives you the equivalent of the information found in a BSOD for each application crash. What happens though on an unstable system many background services/tasks/file transfers/updates are corrupted by instability giving you absolutely no notice of it. This is how eventually the System goes down. You get enough of these and its bye bye booting. Don't panic if you see errors though, you have to analyze what type of error, if its the same device/same application it maybe a faulty application. Getting errors here though is a good way to know something is not working the way it should, and many average users don't ever peak in here.Edited by RagingCain - 8/8/11 at 12:13pm