Finding the sweet spot for my EP45-UD3R setup...

[Studabaker]

Check it out check it out, just a bit of tweaking and trying to figure it out. See, my RAM is just crap, it will not run at much higher speed, it really doesn't like getting out of spec. So I played in the BIOS long enough to get the multiplier + CPU bus to make 3.8GHz that the UD3R can also divide properly for my RAM. Boom, 800MHz RAM FSB with these settings, and lookit how stable it is:

Ignore the CPU speed on Orthos, I don't know what kind of crack it's been smoking.

CPU-Z Validation: http://valid.canardpc.com/show_oc.php?id=560751

The CPU cooler in use here is the super duper cheap Masscool 8WA741, and it super duper cools too!

 

[sneekypeet]

Well if you are looking for a better sweet spot, see if you can run the 10X or 10.5X multi's with not too much voltage. That way the ram OC and bus stay the same. I think 4.0 is attainable but be prepared to run the volts in the 1.4-1.41V range, oh and this will bring on the heat.

As far as orthos is concerned it bases the math off of the 10.5X multi that the proc uses and multiplies by your rated FSB. It isnt on crack, it just cant read well.

Other than that it looks like if you ram is limiting your OC there isnt much you can do at this point aside from the multi to sweeten your OC.

 

[Studabaker]

Hey, thanks for the suggestions. Right now I think I gotta go back to looking at overall case temperature. I was playing COD4 for a good few minutes yesterday and then it didn't crash or anything, it simply froze up and disappeared off my screen. This seems to me like maybe CPU voltage isn't strong enough to go the distance? I'm starting to think it could be the GPU because it can't reach the clocks it could a couple months ago and is experiencing greater heat. Not entirely sure which way to go here.

 

[sneekypeet]

try adding a bit more VTT and see if it corrects the issue, I found running mine a tad higher stabilized my crash to desktop / freeze issues. Its alot of trial and error, maybe just my specific chip wanted that more than Vcore.

 

[Mussels]

higher VTT and NB voltages can stabilise a lot of things.

I was spending hours trying to get my machine more stable, only to say F it and raise the NB voltage... now i can run everything but NB and CPU at stock, at 475 FSB

 

[Studabaker]

Awesome, awesome, thanks to both. I will give both the CPU Termination and NB voltages a little bump and run it like that for a while.

 

[Hayder_Master]

like my friends say here , add more Vcore and try rise up NB voltage to 1.4

 

[Studabaker]

like my friends say here , add more Vcore and try rise up NB voltage to 1.4

Right, the vcore itself doesn't seem to be an issue, but I think the VTT (CPU Termination on a Gigabyte P45 board) might have something to do with it, as raising it just a notch helped decrease vdroop and now instead of 1.36 it runs at 1.38 even though I still have a 1.4 vcore set in BIOS. So I am definitely looking more at what VTT/Termination does. I raised the NB voltage from 1.18 to 1.20 also, just to be on the safe side.

Once again it is Orthos CPU-Burn stable, been running for 8 minutes with no problem. Now just have to game-game-game to see how stable it really is.

 

[Mussels]

Right, the vcore itself doesn't seem to be an issue, but I think the VTT (CPU Termination on a Gigabyte P45 board) might have something to do with it, as raising it just a notch helped decrease vdroop and now instead of 1.36 it runs at 1.38 even though I still have a 1.4 vcore set in BIOS. So I am definitely looking more at what VTT/Termination does. I raised the NB voltage from 1.18 to 1.20 also, just to be on the safe side.

Once again it is Orthos CPU-Burn stable, been running for 8 minutes with no problem. Now just have to game-game-game to see how stable it really is.

8 minutes is nothing. try 4 hours or more.

 

[Studabaker]

8 minutes is nothing. try 4 hours or more.

I know, but it's still a pretty good preliminary test because most crazy OC's I try the first time end up hard-locking the system within the first couple of minutes. I find most tests that go for 5 minutes can go for 5 hours. I know it's always good to run the long test just to be sure, but I'm just saying, for now it's 'stable'.

 

[Studabaker]

Could one of you guys tell me which app/test will generate the most CPU+GPU stress+heat combined?

 

[silkstone]

Could one of you guys tell me which app/test will generate the most CPU+GPU stress+heat combined?

3D mark is pretty intensive oth on CPU and GPU.
You could also try running bionic manager and folding at home at the same time.

higher VTT and NB voltages can stabilise a lot of things.

I was spending hours trying to get my machine more stable, only to say F it and raise the NB voltage... now i can run everything but NB and CPU at stock, at 475 FSB

You got it to 4.5ghz? sweet man

 

[Studabaker]

hahaha, i just ran orthos CPU stress and then launched OCCTs GPU stress and after i exited OCCTs GPU test i came back to like 6 finished threads on Core0 and 2 finished threads on Core1

 

[Hayder_Master]

Right, the vcore itself doesn't seem to be an issue, but I think the VTT (CPU Termination on a Gigabyte P45 board) might have something to do with it, as raising it just a notch helped decrease vdroop and now instead of 1.36 it runs at 1.38 even though I still have a 1.4 vcore set in BIOS. So I am definitely looking more at what VTT/Termination does. I raised the NB voltage from 1.18 to 1.20 also, just to be on the safe side.

Once again it is Orthos CPU-Burn stable, been running for 8 minutes with no problem. Now just have to game-game-game to see how stable it really is.

my friend i have mobo like yours so i give you some tips
1- i do high overclock without change VTT and PLL with my q6600
2- are set the" load line calibration to enable "
3- you can see how i go far in VTT and PLL with my mobo in this thread with my cpu
http://forums.techpowerup.com/showthread.php?t=84106

so i hope you find here what you need

 

[JBravo]

I tested my E7200 last night @ 3400mhz ( 8.5 * 400 ) on 1.168V load and it was nice and stable. But only tested for 25min before going to bed. It looks like these need alot of extra volts to go higher from there? I was kinda hoping to get to 3800mhz with about 1.28-1.3V Load

Oh well

 

[Studabaker]

I tested my E7200 last night @ 3400mhz ( 8.5 * 400 ) on 1.168V load and it was nice and stable. But only tested for 25min before going to bed. It looks like these need alot of extra volts to go higher from there? I was kinda hoping to get to 3800mhz with about 1.28-1.3V Load

Oh well

Nah I cannot get a stable 3.8GHz OC without at least 1.38 volts showing in-system. Anything less hard locks and reboots or BSODs.

 

[Studabaker]

Dammit, Max mode Linpack CPU testing with OCCT fails out after about 10 minutes and it wasn't even during a high heat moment (this thing gets my cores to 70 and CPU to 75 easy).

 

[JBravo]

Dammit, Max mode Linpack CPU testing with OCCT fails out after about 10 minutes and it wasn't even during a high heat moment (this thing gets my cores to 70 and CPU to 75 easy).

Are you getting those temps even after you did the temp calibration?? Or are you using OCCT's temp readouts? I use realtemp, and you have to basically calibrate the sensor readouts. I find that other programs read about 10C higher than realtemp after calibration.

I'm getting like 63C max @ 1.168V load but thats stock intel cooling...

 

[Studabaker]

Are you getting those temps even after you did the temp calibration?? Or are you using OCCT's temp readouts? I use realtemp, and you have to basically calibrate the sensor readouts. I find that other programs read about 10C higher than realtemp after calibration.

I'm getting like 63C max @ 1.168V load but thats stock intel cooling...

How do I do a calibration? Every app I use has reported the same temps for CPU/cores and only been off by around 1C, with SpeedFan usually reporting the 1C higher.

I can't figure out this calibration thing in RealTemp, it seems as though it's asking me to know how much off it is and enter it in by hand, I see no automatic calibration here...

 

[Mussels]

How do I do a calibration? Every app I use has reported the same temps for CPU/cores and only been off by around 1C, with SpeedFan usually reporting the 1C higher.

I can't figure out this calibration thing in RealTemp, it seems as though it's asking me to know how much off it is and enter it in by hand, I see no automatic calibration here...

underclock your CPU as low as you can. let it run at idle. find out ambient temp, calibrate to all cores read as ambient temp.

 

[Studabaker]

underclock your CPU as low as you can. let it run at idle. find out ambient temp, calibrate to all cores read as ambient temp.

Idle temps have always been displayed as 37C (core temps) in all apps under all running speeds in 3 different motherboards.

 

[JBravo]

How do I do a calibration? Every app I use has reported the same temps for CPU/cores and only been off by around 1C, with SpeedFan usually reporting the 1C higher.

I can't figure out this calibration thing in RealTemp, it seems as though it's asking me to know how much off it is and enter it in by hand, I see no automatic calibration here...

What you can do is you set you cpu fsb back to 266mhz and drop your multi as low as it can go? mine is 6 so that gives me a cpu speed of 1600mhz. Also, drop your voltage as low as it will boot ( shouldn't be more that 1.1V bios I think) Then boot too windows and let it sit for like 10min or so too cool down. This in theory should be the smallest amount of heat its ever gonna produce. Check your ambient temps, go to the calibration page in realtemp and enter a number like -5.0 or so. See what that gives you. Keep dropping the number till you get within I would say 2C or so of your ambient showing in realtemp. You can add a little extra to be safe. I added 10C to my ambient seeing as I'm on stock Intel cooling.

I think you should see about 25-30C depending on your ambient...

Hope that helps