E6750 & Gigabyte P35C-DS3R Overclocking Thread

[fausto412]

Originally posted by: mrfatboy
Running orhos now. 5 minutes into it. My temps are

NB 45C
HD0 25C
Core 0 55C
Core 1 55C
CPU 38C
Case 48C
8800gts 50C


I believe the case sensor is by the NB. The cpu sensor is in the middle of the chip above the cores from what I read. Don't quote me.

There are aftermarket NB coolers if you want to use them. My Antec 900 is blowing all sorts of air on the NB now. If you want something cooler you need to buy an after market. I'm not sure it's worth it.

I'm still at 45C for NB at 10 minutes

that cpu temp seems odd to me, shouldn't it be closer to the core? does that 38C move up or down as the cores move up or down?

i have pretty good airflow in mine. thinking about getting a Scythe Kama bay to get cooler air on top of case for the cpu cooler.

 

[mrfatboy]

i think if you have temps anywhere near mine you are good to go. Not to say I'm the best but you can throw money at it for maybe a degree or two better.

I just try to give a baseline for people so that can see if their system is running right or really out of whack.

You (fausto & conjugal) have abnormally low voltage chips for OC. You should try to OC more and let us know your results. See what that baby can do at 1.45v.

From what I have read the OCZ memory is problematic and might be holding you back. I'm running 4-4-4-12 timings. Yours should be doing the same. If you are into super tweaking for cheap price get the Ballistix or Corsair.

update.... my NB is down to 41C with orthos 26 minutes running

 

[fausto412]

Originally posted by: mrfatboy
i think if you have temps anywhere near mine you are good to go. Not to say I'm the best but you can throw money at it for maybe a degree or two better.

I just try to give a baseline for people so that can see if their system is running right or really out of whack.

You (fausto & conjugal) have abnormally low voltage chips for OC. You should try to OC more and let us know your results. See what that baby can do at 1.45v.

From what I have read the OCZ memory is problematic and might be holding you back. I'm running 4-4-4-12 timings. Yours should be doing the same. If you are into super tweaking for cheap price get the Ballistix or Corsair.

update.... my NB is down to 41C with orthos 26 minutes running

I know my chip is a freakazoid. i had to let the board adjust the timings(much looser than rated 4-4-4-15 timings) to get it to run 1 to 1 at 900 MHZ on the ram, 450 FSB. This chip did 3.4 at 1.30625 setting, 1.28 actual volts in bios and stock at 1.26 and that 1.26 took me to 3.2 LOL. My OC is held back by the ram alright. i am planning to get some crucial ballistic pc8500 ram to get past that. i'll be ok with 3.6 ghz for now...don't know how much i'll get out of it by adding more voltage. and if i reach 3.8 ghz would it make much difference? not really. right now i want an nvidia 8800gt and some pc8500 ram from crucial. i want the video card first...then the ram which by then will be dirt cheap. i like having temp ranges within spec while i overclock that'sn why i bought a Ninja and added a fan that can move some air.

yeah i know my OCZ mem sucks....at first i thought "great mem" but then i tried overclock and it's useless. at rated timings and i could only get 850 MHZ at 5-5-5-15 timings. i don't know how much more i can get out of it but i'll try for more overclock when i get my crucial ram next year.

 

[fausto412]

btw...i discovered my chip had potential when i read on tom's hardware that some of them run at less than spec volts....so i tried undervolting and checking stability and found that all it needed was 1.26 actual volts in system health, 1.2865 setting to run. because i was doing 3.4 ghz when i was still on stock volts and i found that very odd.

your 1.45 volts, is it what is set at or what speedfan and cpu-z show? what speedfan and cpu-z show is pretty much actual volts(each program uses different rounding. ) i use f8 bios and that displays actual volts instead of pass or fail in bios.

 

[fausto412]

Originally posted by: mrfatboy
I eeked out 20 more points in 3dmark06 but adjusting my memory timings. 12828 3dmarks


and yes, i was bored

you know, there is something i don't understand about 3dmark. do you score higher when you run all the tests in the registered version because there is people with my same cpu and video card scoring 3000+ points higher than me and that makes no freaking sense to me. i run the free version

 

[mrfatboy]

i run the free version of 3dmark06

i set my vcore to 1.45 in bios.

I updated my temps in the OP, stressed and non stressed.

 

[fausto412]

Originally posted by: mrfatboy
i run the free version of 3dmark06

i set my vcore to 1.45 in bios.

I updated my temps in the OP, stressed and non stressed.

so what is your actual vcore?
if i set 1.45 it would probably be 1.41 actual. if you have cpu-z or speedfan Vcore1 shows actual voltage.


are people with similar systems outscoring you by 3000 points?

 

[mrfatboy]

Originally posted by: fausto412

Originally posted by: mrfatboy
i run the free version of 3dmark06

i set my vcore to 1.45 in bios.

I updated my temps in the OP, stressed and non stressed.

so what is your actual vcore?
if i set 1.45 it would probably be 1.41 actual. if you have cpu-z or speedfan Vcore1 shows actual voltage.


are people with similar systems outscoring you by 3000 points?

cpuz report 1.408v at idle. I don't think anybody has out scored me significantly in 3dmark06 with the same rig setup. What are you referring to?

 

[fausto412]

Originally posted by: mrfatboy

Originally posted by: fausto412

Originally posted by: mrfatboy
i run the free version of 3dmark06

i set my vcore to 1.45 in bios.

I updated my temps in the OP, stressed and non stressed.

so what is your actual vcore?
if i set 1.45 it would probably be 1.41 actual. if you have cpu-z or speedfan Vcore1 shows actual voltage.


are people with similar systems outscoring you by 3000 points?

cpuz report 1.408v at idle. I don't think anybody has out scored me significantly in 3dmark06 with the same rig setup. What are you referring to?

you have about same vdrop i have. i wonder if there is someone out there that gets close to perfect voltage regulation....well...people with an e6750 and x1950pro's outscoring me by 3000+ points is weird to me....and i have a pretty good overclock LOL

 

[MaDMvD]

Any experience with this cooler, anyone?

http://www.newegg.com/Product/...x?Item=N82E16835129048

 

[mrfatboy]

@ fausto
Maybe they are using a x1950 in SLI. There should not be a 3000 point difference in 3dmark06 with the same rig. Or someone is BS'ing

@MAdmvd
Just get the tuniq and be done with it You will have to take your board out to install it but it's worth it. Make sure it fits in your case.

 

[Conjugal Visit]

Originally posted by: MaDMvD
Any experience with this cooler, anyone?

http://www.newegg.com/Product/...x?Item=N82E16835129048

If it's a down-firing heatsink you're looking for, then it's the Big Typhoon for you...

http://www.newegg.com/Product/...x?Item=N82E16835106061

Same price, and some of the best reviews across the board you're going to find. Quiet too. Simply one of the ugliest cooling solutions in existence, though. If your case doesn't have a window, who cares anyway.

Regards,

 

[biltong]

Originally posted by: fausto412
Talked to gigabyte about the voltage issues we see with these boards on ram. they said that the board only defaults to 1.8 volts on Vdimm when the ram is generic ddr2 800. so since my ram is OCZ platinum rev 2, it requires 1.9 to 2.1 volts, the board defaults to 1.9 and that is what throws off the voltages, can anyone verify this?

i know we all are performance freaks so we have performance ram. does anyone have some generic cheap a$$ ram to see if the voltage is 1.8 like gigabyte says and is upping voltage works correctly...ie +0.4 gives about 2.2 volts.

One of my Ballistix 1GB 8500 modules was failing memtest86+ test 5 occasionally, so I RMA'd both sticks (it was a kit, so thought I'd send both back).

Anyway, since they wouldn't cross-ship, I got hold of a cheap generic ddr2 800 512MB module to be going on with. Rated at 1.8V only. Stuck it in and surprise surprise, with the offest set to 0.0, I'm still getting 2.0V. Thats with the F4 BIOS (Anything higher than F4 and the CPU fan runs much slower when using voltage/PWM control, so I'm sticking with F4 for now).

I'll give the F8 BIOS a go tonight, but I'm betting I'll get about 1.9V with a 0.0 offset.

Also, I'll try the other offsets, but I'm guessing there's still the big leap from 2.13V to 2.4V that was introduced with the F7 BIOS i.e. +0.3=2.13V, +0.4=2.4V

Prior to F7, you used to be able to get 2.26V with +0.3V

I've just got my replacement ballistix, so I'll have a play with the value RAM stick in place with the F8 BIOS and try the various offsets.....since its only rated at 1.8V, it probably won't like 2.4V that much :Q

 

[fausto412]

Originally posted by: biltong

Originally posted by: fausto412
Talked to gigabyte about the voltage issues we see with these boards on ram. they said that the board only defaults to 1.8 volts on Vdimm when the ram is generic ddr2 800. so since my ram is OCZ platinum rev 2, it requires 1.9 to 2.1 volts, the board defaults to 1.9 and that is what throws off the voltages, can anyone verify this?

i know we all are performance freaks so we have performance ram. does anyone have some generic cheap a$$ ram to see if the voltage is 1.8 like gigabyte says and is upping voltage works correctly...ie +0.4 gives about 2.2 volts.

One of my Ballistix 1GB 8500 modules was failing memtest86+ test 5 occasionally, so I RMA'd both sticks (it was a kit, so thought I'd send both back).

Anyway, since they wouldn't cross-ship, I got hold of a cheap generic ddr2 800 512MB module to be going on with. Rated at 1.8V only. Stuck it in and surprise surprise, with the offest set to 0.0, I'm still getting 2.0V. Thats with the F4 BIOS (Anything higher than F4 and the CPU fan runs much slower when using voltage/PWM control, so I'm sticking with F4 for now).

I'll give the F8 BIOS a go tonight, but I'm betting I'll get about 1.9V with a 0.0 offset.

Also, I'll try the other offsets, but I'm guessing there's still the big leap from 2.13V to 2.4V that was introduced with the F7 BIOS i.e. +0.3=2.13V, +0.4=2.4V

Prior to F7, you used to be able to get 2.26V with +0.3V

I've just got my replacement ballistix, so I'll have a play with the value RAM stick in place with the F8 BIOS and try the various offsets.....since its only rated at 1.8V, it probably won't like 2.4V that much :Q

i look forward to your results. i have f8 bios.
any issues configuring the ballistix 8500 ram? i'll be upgrading to it next year. i agree that prior to F7 you used to get +0.3= 2.13 since i got the same.

 

[biltong]

@fausto412 - OK, I did my testing......seems that the VDimm voltage settings changed in F7, but in F8, they're back to what they were in F4, F5, F6.

Here's the testing I did:

Offset..........F4..............F7.............F8
+0.0.........2.02V.........1.89V.........2.02V
+0.1.........2.02V.........1.89V.........2.02V
+0.2.........2.14V.........2.14V.........2.14V
+0.3.........2.27V.........2.14V.........2.27V
+0.4.........2.40V.........2.40V.........2.40V

Thats with the single stick of generic DDR2 800 Ram. With the ballistix, I got similar results. The voltages were slightly lower e.g. 2.13 rather than 2.14V but thats probably because of the extra load of 2 sticks of ballistix rather than the single stick of the generic.

So looks like they tried something with F7, but gave up and in F8, they undid whatever they were trying to do in F7.

No problems configuring the ballistix 8500, although I initially had a faulty stick. Got the replacements today, so I'll be doing some memory testing to make sure this batch is OK. The big test seems to be test 5 of memtest86+.

I was running it at FSB 444 with a memory multiplier of 2.4 (giving 1066MHz) with timings of 5-5-5-15 and VDimm of 2.26V. Gave better results than running at 444 with a memory multiplier of 2 (888MHz) with timings of 4-4-4-12. I posted some results in an earlier post.

 

[fausto412]

Originally posted by: biltong
@fausto412 - OK, I did my testing......seems that the VDimm voltage settings changed in F7, but in F8, they're back to what they were in F4, F5, F6.

Here's the testing I did:

Offset..........F4..............F7.............F8
+0.0.........2.02V.........1.89V.........2.02V
+0.1.........2.02V.........1.89V.........2.02V
+0.2.........2.14V.........2.14V.........2.14V
+0.3.........2.27V.........2.14V.........2.27V
+0.4.........2.40V.........2.40V.........2.40V

Thats with the single stick of generic DDR2 800 Ram. With the ballistix, I got similar results. The voltages were slightly lower e.g. 2.13 rather than 2.14V but thats probably because of the extra load of 2 sticks of ballistix rather than the single stick of the generic.

So looks like they tried something with F7, but gave up and in F8, they undid whatever they were trying to do in F7.

No problems configuring the ballistix 8500, although I initially had a faulty stick. Got the replacements today, so I'll be doing some memory testing to make sure this batch is OK. The big test seems to be test 5 of memtest86+.

I was running it at FSB 444 with a memory multiplier of 2.4 (giving 1066MHz) with timings of 5-5-5-15 and VDimm of 2.26V. Gave better results than running at 444 with a memory multiplier of 2 (888MHz) with timings of 4-4-4-12. I posted some results in an earlier post.

thanks dude. that's exactly what i saw when i tried different bioses. i found it odd that in f7 +0.3 and +0.2 were 2.10 and 2.13 in my case. would be cool if someone came up with a bios editor for us to edit the giga bioses to settings that we want for the voltage. correct the variance you know?

 

[fausto412]

Originally posted by: mrfatboy
@ fausto
Maybe they are using a x1950 in SLI. There should not be a 3000 point difference in 3dmark06 with the same rig. Or someone is BS'ing

@MAdmvd
Just get the tuniq and be done with it You will have to take your board out to install it but it's worth it. Make sure it fits in your case.

maybe...i didn't think of that.

 

[biltong]

Well, something odd going on with my replacement ballistix.........

Stuck the 2 modules in the orange slots (2 x 1GB, ddr2 8500) and ran memTest86+ test 5. Got loads of errors. So swapped the modules round (still in the orange slots). No errors. Swapped them a couple of times (still in the orange slots) and same results. One way round, I get errors, the other way I don't.

Tried the yellow slots and the same results. Stick the modules in one way round and errors. Swap them over and no errors.

I can't believe I've got another duff set of ballistix. And why would it only fail with the memory modules in one postition ?

I'm not overclocking the memory. Running to spec - 444FSB x 2.4 (1066MHz) 5-5-5-15 and 2.26V

Strange

 

[fausto412]

Originally posted by: biltong
Well, something odd going on with my replacement ballistix.........

Stuck the 2 modules in the orange slots (2 x 1GB, ddr2 8500) and ran memTest86+ test 5. Got loads of errors. So swapped the modules round (still in the orange slots). No errors. Swapped them a couple of times (still in the orange slots) and same results. One way round, I get errors, the other way I don't.

Tried the yellow slots and the same results. Stick the modules in one way round and errors. Swap them over and no errors.

I can't believe I've got another duff set of ballistix. And why would it only fail with the memory modules in one postition ?

I'm not overclocking the memory. Running to spec - 444FSB x 2.4 (1066MHz) 5-5-5-15 and 2.26V

Strange

i ran into a problem with my OCZ memory. i can't do stock timings and speed at 1 to 1 with fsb. maybe that it your problem. setimings on auto and OC your ram a little bit and see if the behavior changes...or leave it clocked as sto but set timings to auto. get latest bios too. let us know if any of that helps.

one thing to keep in mind also, if it works error free one way just leave it that way

 

[biltong]

Originally posted by: fausto412

Originally posted by: biltong
Well, something odd going on with my replacement ballistix.........

Stuck the 2 modules in the orange slots (2 x 1GB, ddr2 8500) and ran memTest86+ test 5. Got loads of errors. So swapped the modules round (still in the orange slots). No errors. Swapped them a couple of times (still in the orange slots) and same results. One way round, I get errors, the other way I don't.

Tried the yellow slots and the same results. Stick the modules in one way round and errors. Swap them over and no errors.

I can't believe I've got another duff set of ballistix. And why would it only fail with the memory modules in one postition ?

I'm not overclocking the memory. Running to spec - 444FSB x 2.4 (1066MHz) 5-5-5-15 and 2.26V

Strange

i ran into a problem with my OCZ memory. i can't do stock timings and speed at 1 to 1 with fsb. maybe that it your problem. setimings on auto and OC your ram a little bit and see if the behavior changes...or leave it clocked as sto but set timings to auto. get latest bios too. let us know if any of that helps.

one thing to keep in mind also, if it works error free one way just leave it that way

yep, I think I'll leave it running overnight with the working positions. Its done 36 passes on test 5 so far. The other way round, it fails on the 1st pass.

I'll also give it a go in the failing positions with the timings set to auto. Good idea.

I'll also try the latest BIOS, but the CPU fan runs slower with later versions of BIOS (voltage control or PWM) so my CPU runs hotter under load.

Just seems odd that it fails with the modules in one position and passes when you swap them round. Suggests that there's something different about the slots........

 

[Conjugal Visit]

@biltong - I know I'm setting myself up for one major flaming from somebody, but what the heck. I use Easytune5 Pro to adjust the heatsink fan. I DO NOT use Easytune for anything else. For PWM control, it works beautifully. Extremely low overhead, as in, I can't spot it in processes running in task manager. Zero instability on this rig. Really excellent control of the fan - it uses a linear approach - you'll see it when and if you try it. I suggest you load the best Bios version for you and then use Easytune5 pro. You don't like it? It uninstalls very nicely.

Regards,

 

[biltong]

Originally posted by: Conjugal Visit
@biltong - I know I'm setting myself up for one major flaming from somebody, but what the heck. I use Easytune5 Pro to adjust the heatsink fan. I DO NOT use Easytune for anything else. For PWM control, it works beautifully. Extremely low overhead, as in, I can't spot it in processes running in task manager. Zero instability on this rig. Really excellent control of the fan - it uses a linear approach - you'll see it when and if you try it. I suggest you load the best Bios version for you and then use Easytune5 pro. You don't like it? It uninstalls very nicely.

Regards,

Thanks for the tip ! I'll give it a go. I'd given up on BIOS versions later than F4 because they've messed with the fan control. If easytune works well, then it might be worth upgrading the BIOS and using easytune to control the fan.

For my memory problems, after a bit of messing about with various settings, loosening the timings, latest BIOS etc, all with no success, I finally hit on the solution. I raised the G(MCH) by 0.1V and hey presto. No more errors with the RAM modules in any slots.

I need to do some more testing over a longer period, but with the RAM modules in the "failing" position, if (G)MCH is set to 0.0, it always fails on the 1st pass of test 5 of memTest86+. With (G)MCH set to +0.1 and the RAM in the "failing" position", I've run a few cycles of test 5 without any errors. Just need to leave it to soak to be sure, but looking good.

 

[Conjugal Visit]

Originally posted by: biltong
For my memory problems, after a bit of messing about with various settings, loosening the timings, latest BIOS etc, all with no success, I finally hit on the solution. I raised the G(MCH) by 0.1V and hey presto. No more errors with the RAM modules in any slots.

I need to do some more testing over a longer period, but with the RAM modules in the "failing" position, if (G)MCH is set to 0.0, it always fails on the 1st pass of test 5 of memTest86+. With (G)MCH set to +0.1 and the RAM in the "failing" position", I've run a few cycles of test 5 without any errors. Just need to leave it to soak to be sure, but looking good.

That's good information on the Northbridge voltage, and it makes sense, too, at a 3.552GHz O/C.

As to your Mem, perhaps we've gone down this road before, and apologies if we have, but you've tried setting your Ram at 1:1, tighter timings (4-4-4-12), voltage to 2.1-2.2v? To achieve your O/C, that's setting the Ram to 444? Was this a no-go? Because my bet is it will run there, stable. I really don't think you'll miss any performance at all by slightly reducing your Ram's "effective clock." Tighter timings should offset this.

Regards,

 

[biltong]

@Conjugal Visit - I could run the memory at a lower frequency with tighter timings, but previously I found I got better performance with a higher frequency and looser timings. I did a few tests a few pages back. You're probably right that the performance gain on everyday tasks is negligible, but having bought the higher spec memory, I figure it should run at its rated settings. In fact that's the reason I chose a FSB of 444. With a memory multiplier of 2.4, I get the memory's rated frequency of 1066MHz.

I think I just hate paying for something and then finding it won't do what's claimed

 

[Conjugal Visit]

@biltong - I hear you!