Overclocking stability

[vanillasnake21]

Hi, I'm pretty new to overclocking so I have a few questions:

1)How stable is an overclocked machine? Lets say I have a 3.0 ghz quad q9650, and I OC to 3.4 would there be any instability at all (even minor). The reason why it's important is because I do a lot of graphics work on my comp so I would want the safest environment possible.

2) If I overclock does the machine run overclocked all the time? So even when I'm surfing the net, checking email etc (non intensive tasks) the machine will still be running the cores at needlessly high freqs?

3) In another post someone suggested that I should get 1033mhz ram only if I overclock, otherwise 800mhz ram will suffice and I won't notice a diff with high speed ram, is that true? Up to what speed can I overclock with 800mhz ram?

Thanks, sorry if the questions sound noobish...

 

[V3NOM]

okk. well you would have found the answers to these questions pretty easily if you'd just googled them. but since i'm in a good mood (OBAMA!) i'll try and answer them anyway.

1) nothing is guaranteed. some processors (even the same type, lets just say Q9650 for the moment) will require 1.3v to get 4ghz, some will require 1.4v. thats just the way it works. IMO, if you went into core temp and wrote down the VID, then went into the BIOS and set the VID to that, you should be able to get a small oc of 3.4ghz.

the idea in ocing is to set a frequency, then stress test it. if it fails stress testing, you have two options:
1) increase the voltage to make it stable.
2) lower the OC

read some OC'ing guides to get more info on that... toms sticky'd one at the top of this forum is ok, although personally i prefer oc.net's one...

as for your second question,
Intel uses some "power saving" features called Speedstep (or EIST, Enhanced Intel Speedstep) and/or C1E. these either reduce the multiplier and/or the voltage to save power when idling. however this usually makes OC'ing unstable, so everyone just turns them off :lol:

imo, the small power savings arent worth it. if the CPU isn't doing anything (0%) then i fail to see how frequency matters.

alternatively, some motherboards come with software that allows you to overclock in windows etc. i know my MSI board came with an interesting (if useless to me) feature called "D.O.T" which astands for Dynamic Overclocking Technology... apparently it detects CPU load and accordingly sets frequencies (and the voltages necessary for them)

not really that much different to EIST. but an interesting lil program nonetheless.

3) RAM really depends on the CPU that you're overclocking. eg my e5200 will literally run with old DDR (400mhz) RAM at stock. on the flip side, an e8400 has to have a minimum of 667mhz RAM.

basically, if you divide your CPU's (QDR - Quad Pumped Data Rate, the effective frequency. just like DDR RAM's Double Data Rate effective.)FSB, for exmaple a Q9650's 1333 by four, you get your base (raw?) FSB. now since RAM runs at DDR, you will need minimum of 667mhz RAM.

1600 QDR = 400 FSB = 800MHZ RAM. be aware that this is a minimum required RAM speed for a certain FSB, called a 1:1 FSB : DRAM ratio. (400:400 raw clocks, since base ram clock is 400, multiplied by two since its DDR...)

so 800mhz will give you a maximum of 400FSB (without overclocking the RAM) while 1066mhz ram will give you 533mhz FSB. bear in mind however, that most 1066 RAM is actually high quality 800mhz RAM which has been overclocked and overvolted from the factory.

sorry for the wall of text :lol:... hope that helped. as i said, read some OCing guides, they can probably explain it better than me

 

[jsc]

I also would tell you, "Google is your friend." But it's the start of my weekend and I also am in a pretty good mood - even if Obama won.

Answer 1. It depends on what you do. The usual method of testing for stability is to use Prime95 to load all four cores. This does two things - it does a lot of precise calculations, and because the program puts a 100% load on the cores, you can see what the maximum temperature is going to be for your particular case/HSF combination. How long do you test? It's up to you. I have seen recommendations for 2, 4, 6, 8, 12, and 24 hours. I test for 24 hours. One of my setups failed at 14 and 18 hours. I ended up with a Q6600 and a TRUE HSF in an Antec 900 case that ran P95 for 72 hours.

If your talking about a Q9650, you can raise the FSB to 400 MHz. and you'll be running at 3.6 GHz. With a good HSF, you will need very little vcore increase. And at 400 MHz, good DDR2-800 DRAM will work nicely.

Answer 2. Can't add anything to what V3NOM said.

Answer 3. I do not think that 1066 RAM is worth the extra price. I use Crucial Ballistix.
See the Tom's article on DDR2 memory http://www.tomshardware.com/reviews/ddr2-ram-memory,2079.html.

Guides:
If you are an OC noob (don't worry, we all were at some time), you need to look at at least 3 guides. You need to find a general on overclocking C2D's and C2Q's. You need to find a guide on OC'ing your specific motherboard (or at least motherboard brand and chipset). And you need to find something about maximum thermal loads and voltages.

----------
Overclocking since 1978 - Z80 (TRS-80) from 1.77 MHz to 2.01 MHz

 

[vanillasnake21]

Thanks guys You answered my questions perfectly. I will definitely look into more guides. But as a last question, out of the people who overclock how many times do you guys see errors in your system (ie blue screen, freeze, other glitches). Any one can chip in beginners, veterans, and please say from which speeds you OCed it to, thanks.

 

[MaDMagik]

Hi,

I run XP pro on asus p5qpro with e5200 OC`d from 2.5 to 3.6ghz. Its been running like that for about 5 months now and it never froze or bsod,not once. The only errors were those found while stress-testing with prime. fixed by bumping vcore though

 

[jsc]

First overclock - 1978 - Z80 (TRS-80) from 1.77 MHz to 2.01 MHz. Did it in hardware by piggy-backing chip, cutting PCB traces and running jumpers.

Second overclock - P233MMX eventually to 333 MHz. Discarded the motherboard last year because it was really, really obsolete.

Third OC - the famous PII 300A Celeron to 450 MHz. Discarded the motherboard last year because it was really obsolete. The Intel 440BX chipset was great. The Intel engineers did everything right.

(N-2)th - E6600 in an nVidia 680i motherboard with an ArcticCooling Freezer 7 Pro to 3.3 GHz in a hand me down case. Then to 3.6 GHz with a ThermalRight Ultra 120 Extreme (TRUE). Tested stability with Orthos for 24 hours.

(N-1)th - Q6600 in a Gigabyte GA-EP35-DS3P with the TRUE in an Antec 900 case - 2.4 GHz. to 3.6 GHz. Tested stability with Prime 95 for 24 hours. When that passed, I did a 72 hour test run. Core temps hit 60 - 65 C.

Nth - Replaced the Q6600 with a Q9550 in the Gigabyte motherboard. Took me 5 minutes to reach 3.6 GHz. I expect that I can hit 4.0 GHz with some tinkering. Put the Q6600 and ACF7P on the 680i motherboard.

The whole point of stress testing is to find the point where the CPU doesn't work.Then you decrease the speed a little and test again.