View Full Version : Help OCing my new E6420 for WinSMP
belorsch
04-24-07, 08:15 AM
I have a new E6420 running on GA-965P-S3 with 2x512 Adata DDR2 400.
My limitations: I'm not an uberclocker:) I tend to be conservative with it. Frankly I'm a little confused by the bios but I would think it would similar to the DS3's which seems to be popular. This setup is in an old SX800 case with the 2 rear fans removed to allow the Big Typhoon cooling the cpu. I will try to post more tonight when I'm at the machine.
PS In another thread pscout mentioned copying a wu for testing oc'ing. Does this mean I should have the pc disconnected from the internet so it can't report any bad results? TIA
RE the test wu ... since the test wu's i keep are already turned in successfuly, stanford will ignore any other results turned in for them. But since stanfords network has been overloaded, the nice thing to do is run with your lan diabled ... i usually do it through network properties in windows so i don't have a change of unplugging the wrong cable.
Also, with the win smp cleint, it doesn't like to have the lan disabled after it has started, so i just leave mine disabled most of the time while oc'ing.
A few OC'ing tools to install if you don't already have them:
- cpu-z to confirm your bios settings for cpu speed and memory. Also displays mobo bios level, cpu stepping info. It also diplace vc setting but i have not found it to be reliable across mobo's/cpus so i ignore it.
http://www.cpuid.org/cpuz.php
- orthos - runs p95 against 2 cores to stress cpu, cache and memory
Will generate higher temps than folding so if OC passes orthos testing, it should not eue when folding due to temps.
http://www.overclock.net/downloads/138142-orthos-v20060420.html
- P95 is an alternate to orthos but needs more work to setup since you need one instance setup per core with affintiy set. For dual cores, just use orthos. For quads, I run 4 instances of it since 2x orthos was harder to use.
- CoreTemp for measuring cpu temps on c2d's. http://www.thecoolest.zerobrains.com/CoreTemp/
- There is also an intel tool called TAT I have used but coretemp is light weight. and seems to report consistent temps.
- optional benching tools are not absoletuly necessary but i often run them as quick tests that also give you some performance #'s. Bottom line performance for me is folding tpf's, but is slower to run so i use these for quick initial tests. They prove little for stability but are very quick to run so if they fail after an adjustment you don't ned to bother running a potentially longer folding test.
-Sandra - I normally only use it for memory bandwidth measurments, but it has lots of function even in the free version so you may find it interesting/educational to play with its other features.
All you need is the free lite version http://www.sisoftware.net/index.html?dir=&location=downandbuy&langx=en&a=
- superPI modded - i usaully just run a 1M test with this ... takes 20-30 seconds. The modded version reports 3 decimals of accurancy vs just seconds for the standard version. Only runs 1 thread ie on 1 core
- Wprime similar to superpi but runs 1 thread on each core.
... i will update this some more later with links etc.
and then we can get you started with the oc settings.
I am pretty sure your mobo's bios is similar if not the same as my ds3.
Also, I beleive it oc's as well as the ds3 from my readings so you have a good base.
Are you running with the stock intel cooler? If so that may be the limiting factor in how high an OC you can get.
The other limiting factor may be your memory ... I am not familiar with yours. To date i have only run memory at 1:1 fsb ratios on 965 mobos with c2d's but bios improvements may have changed things although i don't recall the ds3 bios having a memory divider, just multiplers ... so how much your memory oc's may be the limiter.
I will add some more later
Well I just downloaded the manual for your mobo and it is the same manual as the one for the DS3, so i'd assume these boards are from the same pcb. To overclock you will ave to go into "MB inteligent Tweaker" and you will find all the overclocking options in there.
"CPU Host Clock Control" enable to allow you to change the fsb
"CPU Host Frequency (Mhz)" this is what you use to overclock, this is you FSB
"System Voltage Control" enable to allow change of voltage, voltage options are ligically named so I don't think I have to explain them to you.
"CIA" overclocks for you, it might be a safer way for you to go.
belorsch
04-24-07, 12:00 PM
Well I just downloaded the manual for your mobo and it is the same manual as the one for the DS3, so i'd assume these boards are from the same pcb. To overclock you will ave to go into "MB inteligent Tweaker" and you will find all the overclocking options in there.
"CPU Host Clock Control" enable to allow you to change the fsb
"CPU Host Frequency (Mhz)" this is what you use to overclock, this is you FSB
"System Voltage Control" enable to allow change of voltage, voltage options are ligically named so I don't think I have to explain them to you.
"CIA" overclocks for you, it might be a safer way for you to go.
I didn't know how reliable the CIA would be. I did have the cpu host clock control enabled and host frequecy at 275 for roughly day and everything seemed ok. I then bumped it to 285 and booted up fine. I started up WinSMP and blue screened. So I took the oc back down and luckily hadn't corrupted to wu. I will try pscouts steps above and will put the oneunit flag on current wu when I get home. pscout if I back up the current wu before it finishes can I use it for stability testing along with your other listed software. Thanks.
Thanks Avg ... the only think i recalled being different with the ds3 is it uses solid state caps, and threads i have read on it say they oc equally well.
While the built in 'easy' oc'ing stuff may work fine, i have always preferred the manual route since it doesn't introduce the need to understand how each vendors and specific bios actually behaves.
One key thing to remember is with the gigabyte bios, you need to use ctrl-F1 to enable all the overclocking options in bios including things like memory timings. Use the ctrl-F1 on the first bios screen before you go into the MIB and other subpanels.
I've been using memtest86+ v. 1.70 to test memory stability and to quickly see the effects on ram bandwidth of changes in timings and FSB. It is very handy to be able to make your adjustments, boot into memtest86+ and within seconds know the unbuffered ram bandwidth. (memtest86 does this as well but is not as up to date on chipsets as memtest86+)
I didn't know how reliable the CIA would be. I did have the cpu host clock control enabled and host frequecy at 275 for roughly day and everything seemed ok. I then bumped it to 285 and booted up fine. I started up WinSMP and blue screened. So I took the oc back down and luckily hadn't corrupted to wu. I will try pscouts steps above and will put the oneunit flag on current wu when I get home. pscout if I back up the current wu before it finishes can I use it for stability testing along with your other listed software. Thanks.
Yes, it is not essential to use a 2610, the other wu's should work fine for now.
You can always catch one later.
If you are careful (ie lan disabled and wu backed up) you can use a live wu to test with so you don't really need to put the oneunit on ... and maybe the next wu you get will be a 2610.
I will describe it some more tonight.
belorsch
04-24-07, 08:04 PM
Well with cpuz you all have helped me find at least part of my problem. I checked it and noticed cpu frequency at 275 but the memory frequency was at 412. I went into the bios and changed the memory clock from auto to 2.00 then saved and rebooted. When checking cpuz again cpu freq and mem freq are the same now. I am currently running 8x300 and testing on my current wu which is a 2610, and I have the nic disabled.
Well with cpuz you all have helped me find at least part of my problem. I checked it and noticed cpu frequency at 275 but the memory frequency was at 412. I went into the bios and changed the memory clock from auto to 2.00 then saved and rebooted. When checking cpuz again cpu freq and mem freq are the same now. I am currently running 8x300 and testing on my current wu which is a 2610, and I have the nic disabled.
Sounds good :)
What rev is your mobo 1.0 or 3.3? Also what level of bios are you running?
If not running the latest, it might be worth moving to but may not be necessary unless you have a really old bios. The early bios level on 965 mobo's were quite buggy, especailly around memory.
cpuz will show you bios level.
On my ds3 r3.3 it does NOT show the mobo rev level but it shows the chipset rev level c2 ... and the early 965's were c1. Also, on my rev 3.3 manual is is clearly labeled on the cover.
These details may not matter at all to your oc, but it is useful to know to make sure you are getting the right bios before you try flashing it. The flash utilities should check anyway that you have the right one.
Here are some of the things i do before oc'ing:
note: use ctrl-f1 before you select the sections so that you see all availble settings.
- in bios under advanced settings, turn off the cpu options like eist, C1E and TM so they don't interfere with oc'ing and voltage settings. I assume it is going to run 724 folding, so none of those features will be useful to you anyway.
- integrated peripherals ... turn off all features you don't use. I turn off virtually everything on my dedicatied folding rigs, usb extra io , serial, parallel etc. I leave sound enabled since some tools can use it. I never have speakers plugged in so i disable what i can. If this is your main rig, you need sound anyway.
-MBIT set pci-e to 100 or 101.
- set voltages to manual but leave them at their default settings until you need to raise them. If you leave system voltage on auto, bios will increase them as you raise the fsb. Your objective is to get the highest oc with the lowest voltage so you will want to control all the voltages manually working up from the minimum.
Always run the temp monitor program. coretemp only reads the cpu.
I don't like the Easy Tune utility but i have run it to try to validate readings from nicer tools like speedfan which doesn't always work properly (eg with quads).
As you raise the fsb, both cpu and motherboard temps will go up so you should monitor them. The Northbridge will get hot, and you will likely need to raise the v to it (MCH setting in MBIT) once you get to 350+ fsb.
I find it useful to keep track of all setting and changes and measurments.
ie. a sheet with columns for fsb, VC, MCH, Vfsb, VDDR, mem timings, superpi/wprime, bandwidth, idle temps, orthos load temps, folding temps, TPF
It is best to make 1 change at a time, record the measurments so you can figure out what combinations work best. So on the sheet, record all changed setting and measurments on a separate line for each test.
Keep this also in case you need to reset cmos at some point and it will be easy to reenter the values you were last at. Or to check settings again in a month.
I have sheet like this in a binder ... 1 (or more) for every rig i have, or had.
Initially, you should increase FSB in increments of 5-10. With experience you will find you can probably try 320-350 right off the bat with known good components. Since i don't know how your memory oc's, i would suggest bumps of 10-20 at the lower ranges <350, and 5-10 above 350 fsb. On stock coling you might not be able to get above 350 folding stable. Case temps are also a big factor and some of the pwm and chipsets may need extra fans in a case esp at high clocks.
Once you raise fsb too high for current settings, you will either fail to post or fail to boot into windows. Once you get these, it is usually time to start adjusting memory timings and/or raising voltages.
As a strategy, it is best to run your memory as slow as possbile with loose timings so that you can find your cpu oc limit first. After that, you try to tighten memory timings. On c2d's with 965 chipsets, you will prolly never run anything but 1:1 cpu:ram speed. On your bios this means leave your memory muliplier at 2 which is the lowest setting. Memory timings seem a bit less critical with lots of cache on the c2d's. So don't worry about tight timings for now.
Before we start raising voltages, cooling is very important.
Are you riuning the stock intel HSF?
I assume it is in a case? What video card is in there (ie how much heat does it produce) , and what kind of fans do you have?
Tests to run after a change
1 - run orthos first ... use the default blend test on let it go through at least 10-20 minutes. keep track of highest temps. C2D's will aften pass orthos testing with temps over 70C, but will eue folding in the mid-high 60's based on coretemp readings.
Before starting a folding test with a live wu, make a new backup incase it eue's at new settings - i just add -x to the directory name where x is the % complete. This way after an eue, you can restore to the last known good backup for your next test.
If you are useing a completed wu, you can just restore the original backup since you don't care about saving progress.
The orthos test is good to do first since it will often fail immediately, saving you the bother of restoring wu's from backups. Let it run at least a few test iterations before bothering with a folding test.
I don't bother running benchmark tests every time ... depending on how much i am changing fsb between tests just to save time.
On a 2610 if it can complete 5-10% without error, there is a decent chance it is stable. So then it is time to stop the client, make a backup and start another test at higher speed, more volts if it failed, etc
belorsch
04-26-07, 07:42 AM
Update: I was able to get a little testing in lastnight. I noticed a 7-8c difference between ETC's core temps and that other utility. I went as high as 325x8 running orthos and seeing core temps as high as 46c. I also disabled those 3 items in the advanced settings in the bios.
To answer an early question from pscout My cooling consists of 3x80mm fans in front of the case(antec sx800) and TT Big Typhoon cooling the cpu. The rear 2x80mm fans have been removed to make room for the typhoon. I have a TT Pure power 480w psu powering the rig. Thanks to pscout chasr and AVG for all the help so far. I hope to spend more time on the oc this weekend.
Your temps sound ok for now.
as fsb gets raised, your northbridge should get pretty hot so it might need a small fan on it.
I am sure you will get to at least 350 on the weekend.
belorsch
04-26-07, 08:13 AM
I forgot to add that bios had the normal vcore at 1.35. I'm not sure of a lower range for the cpu so I went down to 1.3 and have been working from that level. I updated the bios with the onboard utility and I believe it is at ver 8 now but I will check that and board revision tonight and edit the post.
According to cpuz mb is c1 rev and bios is F6. I got the oc up to 365x8 tonight and had to bump vcore up to 1.3125. Temps reached 50c while running orthos. Testing a little every night and then backing it down a little to fold till the next night :)
That Jon feller is gonna pass me but I'm trying to delay it as long as possible. :p
vBulletin® v3.8.7, Copyright ©2000-2012, vBulletin Solutions, Inc.