E6750 & Gigabyte P35C-DS3R Overclocking Thread

[mrfatboy]

Originally posted by: stavesacre03
Hi guys, I posted a problem I'm having over in the video forum, but not getting much input. Hoping you all have some ideas. I have the P35C-DS3R board with a Q6600 G0 running at 3.2. System is Orthos stable for 13+ hours, memtest stable, temps at 68C under load (per Everest), etc. All latest drivers, Vista updates, and BIOS. See sig for full specs.

I had no problems until playing Bioshock, which works smooth, high framrates, etc, but will randomly crash the PC completely (no BSOD--I can hear sound looping sometimes, and then the system completely powers off). I tried various things such as reducing game settings, reverting to stock speeds, etc, all to no avail. At this point, I've given up on Bioshock due to frustration with these crashes--which sometimes happens after 5 minutes and sometimes happens after 45 minuts. At the time, I wrote this off as being a game problem and I hoped for a patch.

But then I tried TF2, which also crashed in the exact same manner under recommended settings (all high with AA/AF enabled). At this point, I'm thinking maybe my power supply is at fault, but here's the kicker: When I run TF2 in windowed mode, with AA off and AF set to trilinear, I can play for hours on end with no crashing. I can't imagine that the power supply would be the problem if it works with game settings slightly changed...

So, what is causing the crashing with full settings? Is the video card to blame??? Motherboard??? Power Supply??? Thanks in advance for any feedback. Let me know if I can offer more information.

I would like to know what your stressed coretemps are. Use Coretemp 0.95.4 and the latest Speedfan and run Orthos blend. Remember to add 15C to speedfan core temps. Speed fan should also tell us your video card temp.

68C for the CPU temp seems on the hotter side and the coretemps would be hotter than that. However, use said your system was orthos stable so that is a little perplexing. Orthos stressed the CPU more than video games do.

Also, run the 3dmark06 test for awhile and see if that crashes your computer.

 

[stavesacre03]

OK, I'll do some more testing tonight after work.
EDIT: before I left for work I tried to install Coretemp 0.95.4, but it is giving me errors on vista 64 and would not install...

 

[jcossin]

Originally posted by: stavesacre03
OK, I'll do some more testing tonight after work.
EDIT: before I left for work I tried to install Coretemp 0.95.4, but it is giving me errors on vista 64 and would not install...

I'm suspecting video card drivers. Have you updated to the latest, maybe even some beta ones?

Personally I'd move over to a GeForce 8800 series card if I was you, but if you love ATI, then try drivers.

Also, have you loaded on the latest Intel P35 chipset drivers?

Oh, and have you run memtest and made sure your RAM is totally solid?

 

[stavesacre03]

Originally posted by: jcossin
I'm suspecting video card drivers. Have you updated to the latest, maybe even some beta ones?

Personally I'd move over to a GeForce 8800 series card if I was you, but if you love ATI, then try drivers.

Also, have you loaded on the latest Intel P35 chipset drivers?

Oh, and have you run memtest and made sure your RAM is totally solid?

The ATI card is a stop-gap until the new GeForce cards arrive, so I'm with you on that. In the mean time, yes, I'm running the latest official drivers (haven't tried beta at this point), and memtest shows the RAM to be rock solid. I'm not 100% certain I have the latest Intel chipset drivers, but I thought I had. I'll check on that as soon as I can. Thanks for your input.

 

[Serj]

Hi Stavesacre,
1st of all nice mobo xD

Your problem to me seems like graphics card or power, because your system looks solid and moderately overclocked. i dont think you need all that overvoltage and heat.
recommendation: get a 8800 from a B&M store just for testing and return it once you're done testing. (or if you have someone who can lend you his for a day thats fine too)
I have heard people having problems with non-vista certified cards (rare and random) so i would think thats the problem.
You can also try to dual boot xp(if its possible) and see if bioshock functions well.
Good Luck!

PS: i just played bioshock demo and i think i will jump the $29.99 deal from Fry's

 

[stavesacre03]

Serj--thanks for the feedback.

 

[stavesacre03]

well, i think we can close the books on my issue. I swapped in a spare power supply i had laying around, and just played 2 straight hours of Bioshock with no crashes. Crazy, the free after rebate Ultra X-finity works, but the $80 Antec doesn't...anyway, thanks for your help everyone.
EDIT: nope, still same problems despite the power supply change. I'm still thinking this may be related to the video card. I did load Coretemp, which gives me basically Speedfan temps plus 15 degrees, which confirms my Orthos load temps are mid 60's. My memory also remains stable. No crashes on 3dmark06, but the free version doesn't use AA/AF, which seems to be what is crashing my PC. If anyone else has any ideas, please let me know.

 

[mrfatboy]

Gigabyte Trouble Ticket Update


Here is the latest from Gigabyte. I have spelled out specificaly what the memory voltage problem is. They keep giving me BS answers like this. Of course, I countered.

I can't believe there is that big of a languange gap.

I am utterly amazed!!!!!!!



Software may not be reporting accurate at all times, we have tested this board in our facility running benchmark for 24 hours. We were unable to experience any stability issue with the system crashing. Can you provide us with more information to replicating similiar symphons wen you are having?

 

[mrfatboy]

Gigabyte Trouble Ticket Update


Well guys, it looks like Gigabyte has had it with me Not only have they refused to answer any of my DIRECT questions, they have now stopped/deleted my trouble ticket.

My last question was directed soley to the memory voltage issure. Poof! no more trouble ticket

The guilty always run

Just a note to everybody. I do belive the MB is over volting the ram. Another poster just PM'd me that he was having BSOD's until he set his voltage according to my chart. Now everything works fine for him

So don't fry your ram! The manual is wrong

 

[Ballashoes]

What would be the performance of this combo with water cooling? What could I expect to overclock it to?

 

[biltong]

First off, a big thanks for all the really useful info in this thread ! This has been really helpful e.g. If it wasn't for mrfatboy, my mem would still be running severely overvolted !

Secondly, a question (and don't take this the wrong way)...how many of the overclocks we see are really stable ? The reason I ask is that I'm having a bit off difficulty getting a stable overclock.

First off, a bit of background:

I picked the E6750 because of its overclocking ability and reasonable price (especially when compared to the E6700). The P35C DS3R I chose again because of its overclocking ability and price. The memory, DDR8500 because I'd read that the minimum FSB:Mem ratio was 1:1 with the P35 chipset. I figured that to reach the 3.6Ghz that seemed reasonable, I'd need DDR1066 if I didn't want to overclock the memory whilst overclocking the CPU (450MHz x 2 = DDR900), and at only a little more pricewise than DDR800, seemed a reasonable choice.

Anyway, I messed around with various overclocks before settling on 3.552GHz with a mem multiplier of 2.4 (1:1.2) and timings of 5-5-5-15. This would max out the memory without overclocking it (444 x 2.4 = 1066MHz), whilst still giving a reasonable overclock of the CPU. So I started messing with the settings to see if I could achieve this. Eventually, I got a 12 hour+ stable Orthos Small FTT run at stock voltage settings with the exception of VCore 1.4875V and VDimm +0.4V (this was before I'd seen mrfatboy's posts on the memory voltage settings being way high). All good I thought until I tried the Orthos Blend test......

No matter what I tried (upping the MCH, NSB etc), it just wasn't stable. By this time I realised my memory voltage was probably a bit high (its rated at 2.2V), but again, no matter what setting I chose, it wasn't stable. So I went back to running Orthos small FTT and guess what, this wouldn't pass either, at least not for 12 hours. It ran for 8 hours, but that was the longest I could get it to run. Again, I tried adjusting the MCH, NSB, Vdimm etc, but no joy.

So it looks like I'm going to have to up the VCore again. Which is odd because everyone else seems to be able to run 3.6GHz @ 1.45V or 1.46V. I know a lot of people have better cooling than the AC Pro7, but its pretty cool over here at the moment. Typically, my temps are 56/57 for the cores with the fan on Auto, which doesn't seem excessive ?

Anyway, it seems that a single long(ish) Orthos run doesn't prove a system stable, at least not in my case ! I think I'll need maybe 3 x 12 hour runs on Small FTT and the same again with the Blend test before I'm happy.

Which brings me back to my question. How many people have assumed a stable overclock after a single, long Orthos run ?

Anyway, back to upping my Vcore !!

 

[biltong]

Actually, I just noticed a mistake in my last post.....My Vcore is currently at 1.46875, not 1.4875, so maybe not too excessive.....yet......

 

[mrfatboy]

I ran orthos blend for 14 hours with the settings in my first post. I stopped it because I thought it was enough. That was over a month ago. I have not checked it since.

 

[biltong]

mrfatboy, If you get a chance, I'd be interested to know if you pass another run. Maybe I was just lucky (or unlucky depending how you look at it......I did waste a lot of time trying to get the Blend test to pass at the same Vcore, suspecting my problem must be to do with the memory)......

 

[mrfatboy]

try using my same settings. See my sig. I need to up the volts to (g)mch +.1v to get it stable. I will try to do another run tonight after my bf2 playing

 

[biltong]

I tried upping the (g)mch but it didn't make any difference for me. Also tried upping the FSB voltage, but that didn't make any difference either. I notice you've got your DDR2 voltage set to +0.3, so I'm giving that a go now.

After that, its the Vcore. I reckon I might be OK with one more step on the Vcore. The other difference is I have Refresh to ACT Delay set to 42 rather than 0. It seems to give me slightly better performance (tested with Folding SMP client). I also have different Vcore, and memory timings.

Maybe I'll also try your exact settings from your first post !

 

[mrfatboy]

Originally posted by: biltong
I tried upping the (g)mch but it didn't make any difference for me. Also tried upping the FSB voltage, but that didn't make any difference either. I notice you've got your DDR2 voltage set to +0.3, so I'm giving that a go now.

After that, its the Vcore. I reckon I might be OK with one more step on the Vcore. The other difference is I have Refresh to ACT Delay set to 42 rather than 0. It seems to give me slightly better performance (tested with Folding SMP client). I also have different Vcore, and memory timings.

Maybe I'll also try your exact settings from your first post !

yes, try the exact settings from my OP. It was the only way it worked for me. Also, look for 5t3vo's posts' in this thread. He had to up his vcore 1 notch but he got to lower his (g)mch to normal to get his 3.6.

 

[biltong]

mrfatboy, tried your exact settings, but Orthos small FTT wouldn't run for more than 10 seconds ! I even upped the VCore back to my original 1.46875, but Orthos would fail after a couple of minutes. I think I'd need much more VCore to get stable at 3.6GHz, so you obviously got lucky with your CPU (or I got unlucky with mine). I'm going back to my original settings (4.552GHz) and I'll see if upping the VCore to 1.475V does the trick. Orthos small FTT is running for 2 or 3 hours with my settings, so I'm now hoping one more increment of the VCore will do the trick. I'll post my results when I get stable

 

[darkenedsoul]

I've got the motherboard + E6850 and am wondering what tool will show me the actual voltage on the memory. In the Easytune it is showing it at 2.1. I don't want to go tweaking in BIOS unless I know what it actually is if it's off from Easytune.

 

[biltong]

Originally posted by: darkenedsoul
I've got the motherboard + E6850 and am wondering what tool will show me the actual voltage on the memory. In the Easytune it is showing it at 2.1. I don't want to go tweaking in BIOS unless I know what it actually is if it's off from Easytune.

Speedfan seems to show the memory voltage as Vcore2. As per MrFatBoy's earlier posts, you should see it at approx 2.0V with DDR2 set at +0.0 or +0.1 and then +0.2 should give approx 2.13V and +0.3 results in about 2.26V

 

[mrfatboy]

@biltong, I'm currently at 11 hours Orthos blend stable at my OP settings. It looks like my chip is still holding the OC nicely My core temps are only 48C but it's earily in the morning and cool.

@darkenedsoul, I don't believe Easytune is acually reading the memory voltage from the board. I think it's just added you offset (+.1, +.2, +.3, etc) to 1.8v and then reporting it. Crappy programming if you ask me. Like I said before, don't use Easytune for anything.

 

[biltong]

mrfatboy, sounds like your definitely stable with your settings. Your CPU is obviously from a better batch than mine.

I've just done a couple of orthos runs with the DDR2 offset set to +0.3 and I had a 12 hour run (before I stopped it) and then an 8 hour run (before I got an error).

Previously, I'd tried +0.4 (before you pointed out how high that really was) and +0.2, but you've got yours set to +0.3 and you're also running crucial ballistix, so figured I'd give that a go (keeping all my other settings the same). Seems to be the best results so far. I'm just going to up the VCore a notch and I reckon that'll be stable. A shame I have to go so high for less GHz Still below Intel's official spec I guess - 0.85V to 1.5V:

http://processorfinder.intel.c...tails.aspx?sSpec=SLA9V

 

[Conjugal Visit]

Some questions on benching, stress tests, and voltage monitoring tools:

1. Why are you all using Orthos blend to establish whether your CPU is stable? I've read that the Small FFT test is the appropriate one, and that the blend test constantly pages info to the ram modules, obscuring results. So I use Orthos Small FFT to stress the CPU, and Memtest to stress the ram. Am I wrong?

2. Speedfan ver4.33 - VCore2 does seem to report VDIMM voltages. Mine are set to +0.2v and Speedfan reports 2.14V. However, Speedfan seems to require the +15C offset to show accurate temps at the VCore. Am I wrong?

General comment - I remain convinced that a proper O/C is a step-by-step, and step one is ensuring your ram is stable at a given O/C using Memtest at boot. Only then can a person determine what other changes are enhancing or killing an O/C. Going into BIOS and altering two, three, four parameters at a time would make it pretty difficult to determine what's working and what isn't.

Regards,

 

[mrfatboy]

Originally posted by: Conjugal Visit

Some questions on benching, stress tests, and voltage monitoring tools:

1. Why are you all using Orthos blend to establish whether your CPU is stable? I've read that the Small FFT test is the appropriate one, and that the blend test constantly pages info to the ram modules, obscuring results. So I use Orthos Small FFT to stress the CPU, and Memtest to stress the ram. Am I wrong?

2. Speedfan ver4.33 - VCore2 does seem to report VDIMM voltages. Mine are set to +0.2v and Speedfan reports 2.14V. However, Speedfan seems to require the +15C offset to show accurate temps at the VCore. Am I wrong?

General comment - I remain convinced that a proper O/C is a step-by-step, and step one is ensuring your ram is stable at a given O/C using Memtest at boot. Only then can a person determine what other changes are enhancing or killing an O/C. Going into BIOS and altering two, three, four parameters at a time would make it pretty difficult to determine what's working and what isn't.

Regards,

1) From what I have read, Orthos Blend is the more stressful test. If your computer passes that then you should be good to go for anything.

2) You are absolutely right on all counts.

3) You are right again. I have been seeing a trend that people are becoming more impatient about overclocking. They just want the setttings and be done with it. Some don't even bother to read this thread even when all the information they need is in the top 20 posts or so. All chips are different and it is a step by step process But at least this thread gives you a roadmap on how to do it.

But hell, if people didn't ask questions this thread would be dead

 

[biltong]

Originally posted by: Conjugal Visit

Some questions on benching, stress tests, and voltage monitoring tools:

1. Why are you all using Orthos blend to establish whether your CPU is stable? I've read that the Small FFT test is the appropriate one, and that the blend test constantly pages info to the ram modules, obscuring results. So I use Orthos Small FFT to stress the CPU, and Memtest to stress the ram. Am I wrong?

2. Speedfan ver4.33 - VCore2 does seem to report VDIMM voltages. Mine are set to +0.2v and Speedfan reports 2.14V. However, Speedfan seems to require the +15C offset to show accurate temps at the VCore. Am I wrong?

General comment - I remain convinced that a proper O/C is a step-by-step, and step one is ensuring your ram is stable at a given O/C using Memtest at boot. Only then can a person determine what other changes are enhancing or killing an O/C. Going into BIOS and altering two, three, four parameters at a time would make it pretty difficult to determine what's working and what isn't.

Regards,

There's a new Beta version of Speedfan that reports the correct temperatures.

And I agree with your comments. I decided to see if I could get stable with FSB and mem multiplier that would max out my memory without overclocking it (444, 2.4). I've been upping the VCore one step at a time and if I'm not stable, Adjusting other voltage settings (e.g. MCH, FSB, DDR2) one at a time to see if it makes a difference. As you say, it wouldn't make sense to change 3 or 4 parameters simultaneously. So for example, I'd set MCH to +0.1 and see how stable it is. If that doesn't make a difference, I'd set it back to normal and then maybe set the FSB to +0.1 etc. I'm using Orthos small FTT, Orthos Blend and memTest86+ as stress tools.