Newbie Oc Questions

[mike_brie]

So I read graysky's guide to oc'ing c2d and c2q, and I'd like to try this out myself, but I'm new to oc'ing. I found his post on anandtech originally and tried asking a question there, but it's stickied and I imagine I won't get a response anytime soon, so I thought this board might be better.

 

First, I'm wondering if the general approach to overclocking a multiplier-locked CPU is simply to increase the FSB, ensure your system is stable, lower the vcore, and find a sweet spot between performance, stability, and voltage. Is this correct, or are there other things I should be playing with besides FSB to increase performance?

 

Second, I wonder if I can make big jumps on the FSB such as from 333 to 400, or if I always need to move in 5-10 MHz increments. I've read posts on some sites indicating that you can simply set the FSB to 400 and you'll be fine, but at other sites, they insist that you take small steps and test each step.

 

It seems like the best approach would be to do a "binary search" to find the optimal FSB speed, so long as there's no risk of damage by starting near the top of the possible FSB speeds. In other words, start at the max, if failure, then test an FSB halfway between the last known good speed and the last failed speed. If failure, repeat that process, if success, then test at an FSB halfway between the new good speed and the last failed speed. Is this how you guys and gals do it?

 

Finally, how long can I expect this process to take? An hour, a day, a week, a month?

 

Thanks!

 

----------

My specs:

Intel Q9550 Core 2 Quad @ 2.83 Ghz (8.5 x 333)

Gigabyte P45 EP45-DS3R Mobo

MSI 512MB Radeon HD 4850

WD VelociRaptor 150GB

G.Skill 8GB DDR2 800 RAM

Corsair 650TX 650W Power

Xigmatek HDT-S1283 Cooler

Vista 64-bit Home Premium

Dell 30" 3007WFP Monitor

Antec P182 ATX Case

[The Smith]

Hi,

 

You will need to lower the FSB and also the RAM divider to keep it at around 400Mhz.(800Mhz Dual-data rate). This way, you will be overclocking your processor, but not the RAM. It is much simpler than testing for stability on both.

 

Also, as you will be raising the FSB, you will come to a certain point where you will need to raise voltage, as you mentionned. This point can come very quickly, depending if your processor is stubborn or not.

 

Don't forget to watch your temperatures.

 

As for increments on FSB, since you're a beginner, I suggest you to begin with a small overclock. Test for stability and take the time to understand all what you do. There is no danger to raise the FSB by big increments, but just be sure of what you do. If you consider only FSB, the worst that can happen is that you need to do a hard restart because it has freezed, or blue-screened.

 

Now talking of voltage, be aware that the more you raise it, the more it's dangerous for your chip. This time the worst that can happen is that you kill your Q9550.

 

Good luck

[Turophiliac]

i like the way you ask your questions, but anyway, i'll answer what i can how i know.

 

with overclocking, everything imaginable is pretty much up to personal preference. so everybody will do it differently.

 

with overclocking my e8400, i pretty much left everything at auto (including voltage) and raised the FSB to a level that i've read most any chip similar to mine should reach. in my case it was 9.0 x 400. i'm not too sure about the quads, i've heard they can be a little harder to O/C, but you could take a similar approach as I and overclock that way. but after i did that, i made sure that i checked what the voltage was on auto and set it manually, as to prevent any weird changes when i messed with it further.

 

my approach doesn't always work, however, and many veterans will tell you not to guess, because it usually just wastes time getting to where you want.

 

so try what you feel comfortable with and make sure to test every time to change anything.

 

EDIT: The Smith beat me to it, but i agree with what he says, it's good advice.

Edited September 25, 2008 by turophiliac

[mike_brie]

Great advice, thanks to you both!

 

I have read that the Q9550 can easily handle 8.5 x 400, so maybe I will start there, keep everything on auto, ensure it works, record the voltage settings, and move upwards from there. Can't wait to try ito ut!

 

I'm paranoid of trashing my machine so I will DEFINITELY watch my temps along the way.

 

Do you guys run a 24h test each time you change a setting? Or do you run Prime95 for say, an hour or so, and then step things forward a bit if it passes that test, finally running a 24h test when you're at a setting that seems close to ideal?

[Turophiliac]

you know, i've never seen the purpose of a 24 hour test... i feel that if my processor can handle over an hour of ORTHOS or Prime95 stressing at 100% without error, it can handle anything i could ever throw at it... the longest i've ever run my cpu at 100% (minus testing) was about 2 minutes... so i think an hour of stressing at 100% should be fine, but like i said before, overclocking is different with everybody, so if you want your system to be 24 hours prime stable, then by all means, go for it. but i say that you only really need about an hour for real world application.

[mike_brie]

Cool I definitely won't bother with long tests for each new setting then, sounds like a good approach that will save me TONS of time

[crazy8s846]

Cool I definitely won't bother with long tests for each new setting then, sounds like a good approach that will save me TONS of time

Run occt stress test, and you will know way before an hour, if you can run it for an hour your doing good.

[The Smith]

Run occt stress test, and you will know way before an hour, if you can run it for an hour your doing good.

As for me, I suggest once you achieved your final OC, you run it at least 8h for the final test. But it depends if you're afraid of blue-screening...

[mike_brie]

OK, so I tried it out, and a 400 FSB was no problem at all. I incrementally raised it to 445 MHz, still no problems. Under stress, the CPU core temp reached a max of 73. I assume this temp is still considered safe. Prime95 passed (haven't run memtest86 yet). Sweet, a 3.78GHz CPU, that seems too easy!

 

A question about memory I have DDR2800 memory. My GB motherboard has a multiplier rather than a divider, so I chose the lowest multiplier, 2.0D. This runs my memory at 890 MHz. So this means I'm overclocking my memory, right?

 

And I notice that timings are automatically adjusted from 5-5-5-15 to 5-6-6-17 after moving up this far. I wonder if that's actually worse off than running at a 400FSB? I guess benchmarking will be the best way to tell, which I'm currently doing. But if people have advice, or notice that I'm doing something dreadfully wrong, let me know, thanks!

[Turophiliac]

I'm doing something dreadfully wrong

 

73C is WAY too high for your processor, even if it's just a max temp, i would strongly suggest that you keep your maximum load temp under 60C, i try and keep mine below 55 at all times. if you are using the stock cooler, i would strongly suggest that you invest in an aftermarket one, it's well worth the 50 bucks or so to keep your processor safe from harm...

[mike_brie]

Thanks for the info turophiliac, I'll aim for a lower max temp. I've read elsewhere that a max temp of 65 is reasonable or even conservative for the quad-cores. Do you think this is still too high?

 

I am using an aftermarket cooler -- the Xigmatek HDT-S1283 -- and some arctic silver 5. CPU fan speed reached about 1400 RPM; I have it on auto.

 

I suppose it's time to find the temp/performance sweet spot, and I have two routes to lowering the temp: 1) lower the vcore voltage (which is currently at stock: 1.2375) and test for stability, or 2) lower the FSB.

 

I'll try both and see where it gets me. Well actually, I'll focus on lowering vcore voltage. I already know idle/core temps at lower FSB rates during previous testing.

 

I'm curious to know the reasoning behind aiming for 55 as a max temp, vs. say, 65. Where do people come up with good max temp numbers?

 

I've read this is a guessing game, since Intel doesn't release official numbers. They do say that the thermal specification temp is 71.4 for the Q9550/C1 stepping (http://processorfinder.intel.com/details.aspx?sSpec=SLAWQ#), but that is the tcase temp, and I'll admit I'm not clear on how that relates to core temp.

Edited September 27, 2008 by mike_brie

[Turophiliac]

hmm, your cooler should be cooling that quad core way better than it is, i know my dual core vcore is 1.375v and at load it runs only about 51C (granted it's right next to the window and i live near the coast) but at 1.23v vcore, you should at the most be getting 55C max temp. i would suggest reseating that xigmatek. with that quality of heatsink, i would never expect temps that high, unless you were like in the desert somewhere and it was over 100F in your computer room...

 

as for my reasoning behind the 55C max. it's just what i'm comfortable with. the general consensus is that 60C is what most people are comfortable running as a max temp, as far as i've read, and i just liked to be a little lower than that. but 65C as a max temp isn't going to hurt anything, but i'd be wary of anything higher.

 

as for the thermal specification temp at 71C, it really doesn't matter all that much, all that really matters is the core temp, because that is what does the work in your CPU. the tcase temp is the temperature of the case your processor is in, but all that is is a piece of metal around the core. since you are using coretemp, i would suggest making sure that it is the correct program, i know my chip's temperatures are reported more accurately with realtemp, but that may not be the case with yours.

 

i hope this answers your questions.