Official MSI P7N Diamond (780i) Discussion/Review/Overclock/Guide/BIOS Thread

[theYipster]

X-Built,

Congrats on achieving stability at 3.6ghz. That's great news

After going through a few more unsuccessful rounds of the screen freezing in OCCT, I dialed in your settings and passed the standard 30 minutes test!



In my experience, passing this usually means clear sailing to Prime95 stability, as I've never found OCCT to pass when Prime fails within four to eight hours. This gives me great confidence, as even if Prime does fail, all that should be needed now is some slight vCore tweaking. It's the home stretch!

Here are my settings for 3.6ghz, OCCT Stable:

CPU: 3600mhz (400x9)
FSB: 1600mhz (400x4)
RAM: 800mhz (400x2)
RAM timings: 5-5-5-15-2T

PCI-E freq: 100mhz (stock)
Auto-Disable: Disable

vCore: .3000 (the next step is to find the lower bounds of stability.)
vDIMM: 2.1
FSB: 1.45
SB: Auto
GTL Ref: 65
FSB Terminator: 25
All other voltages: Auto
Speed Spectrums: Disabled

Interestingly enough, the last unstable settings were similar: FSB was at 1.4v while FSB Terminator was at 50. These previous settings caused the same freezing issue within 5 minutes of the OCCT Test. This certainly adds weight to my suspicion that the freezing issues were caused by incorrectly "tuned" NB voltages. Bumping NB voltage just .25 and reducing FSB Term to 25 did the trick.

My next step is to see how low I can go on vCore while maintaining all other voltages. I should be able to cut down a few notches, based upon my experience with this chip. With that finished, I'll do a 12 hour Prime run and call it a day for 3.6ghz. Then I'll likely want to get my Vista environment running on this rig before I try for 3.8-4.0 stable.

This is very good news we're having today. With the earlier 4.05ghz benchmark and our success at 3.6ghz stable, I'm far more inclined to recommend this board as an able overclocker. That's very exciting, as this will be the first non-reference Intel nForce board (680i or 780i) that works well out of the box and provides for ample overclocking (even if OCing isn't as easy yet as on the reference boards.) Given that the solid caps and the advanced PWM system should provide for greatly increased reliability over the long term when compared to the reference boards, I'd say the P7N is indeed a winner!

[x-built-Stan]

theYipster,

I'm glad that you were able to reproduce my results. After all that's what this is about. I was able to get a post and into windows at 4GHz by pushing the Vcore to .38 (max). However, I could not get it stable enough to pass a 3DMark06 run . I'm tearing down now to lap the chip.

FYI: I have an E8400 (Q774A813) in the wings for testing later.

[BKA]

Yipster, which BIOS are you using for your successful overclock? I will probably get back into trying to get a stable 400 FSB with my E8400 later this evening. I have both of the beta's from the MSI forums, P05 & 1.14. Still using the 1.0 shipping BIOS.

I finally got my temperatures under control due to my 8800GT running in the 90's when gaming. At first I thought I could simply reapply some AS5 and hope the temps would go down but no go. My cards are factory overclocked and the stock fan was junk. After putting on a pair of Thermaltake DuOrb it dropped the temps to a max of 60c when playing Crysis and HGL for a couple of hours. So now I feel better about proceeding with overclocking the CPU now.

EDIT: Never mind, I just noticed in your post you mention 1.14.

[Overspeed]

theYipster,

I'm glad that you were able to reproduce my results. After all that's what this is about. I was able to get a post and into windows at 4GHz by pushing the Vcore to .38 (max). However, I could not get it stable enough to pass a 3DMark06 run . I'm tearing down now to lap the chip.

FYI: I have an E8400 (Q774A813) in the wings for testing later.

Awesome, I look forward to your results on the E8400.

[Exavier]

excellent work guys, I'm looking forward to seeing this proceed
I'm also interested in this board, looking for an SLI equivalent of the P5K Black Pearl, so far looking at the P7N as it doesn't look quite as bad as the reference 780i boards and their RMA turnaround..

[theYipster]

Hi all,

I've updated the front post with new information collected over the past two weeks. Please PM me if you see any errors or if you'd like anything added.

Thanks!

[loonym]

Thanks so much Yipster And thanks to everyone who has shown interest in this board and participated in the discussion. I see good progress being made and I think as more and more successes are reported interest will grow.

[Overspeed]

theYipster, one change I think you can make to your top post is the value you assign to GTL Ref. This board has two settings that refer to GTL ... NB GTL and CPU GTL. It would helpful know which of these two your are modifying. Great work.

[memory]

Quick test on my MSI P7N Diamond on Bios v1.14(Beta)

Testbed:
CPU: E8400(Q746A476)
Memory: gskill GBPK PC8500 2gb x 2kit
Mobo: MSI P7N Diamond SLI (Bios V1.14 Beta)
GPU: XFX 7600GT
HDD: SATA II Hitachi 160gb 7200rpm
PSU: CoolerMaster iGreen 600watt
OS: XP Pro SP2

nb=1.4volt
FSB Termination Voltage = 63
vcore = 1.28volt
vdimm = 2.1volt
Rest = AUTO

500mhz x 8 = 4.0ghz


Completed 06 run

[x-built-Stan]

All,

Great news. I installed the E8400 (Q774A813, VID=1.1125v) chip today.
After some trial and error, I was able to boot into Windows at 4.275GHz .
I haven't had any time to run the stability test series yet. But hey, I just had to tell someone.

UPDATE: Ran 3DMark06 and updated the picture - 19185

Here are the basic settings:

BIOS: v1.0 (yes - believe it)
FSB: 1900MHz (475 x 9)
DIMM: 950MHz
Vcore: 1.448 (0.20000 no load)
Vdimm: 2.1
NB V: 1.4
CPU GTL: 65
FSB Term: 63
All other settings to AUTO

Major tuning and testing needed - but this should provide you with a starting point.

FYI - I did try to push to 4.5 but just got BSOD.
These results have been posted to the MSi Forum @ http://forum.msi.com.tw/index.php?to...6366#msg866366

[Overspeed]

With great sadness, I must report I am unable to produce the same results as x-built-stan and memory. Using both posts, I tried each settings configuration and countless others ... literally. After two hours of trying every combination possible (yes I kept a chart) my setup will not boot windows stable beyond 3.54Ghz. I did all the right things:

33Mhz FSB incremental jumps
matched ram timings
correct bios settings
even temperate management

I tried the 9x and 8x mulitpier configurations without success. What's funny is I am able to boot Vista stable at 3.54Ghz with only 0.125v on the vcore and nb at 1.35, all others auto. One might speculate that simply increasing these values while increasing FSB accordingly would work but nope. GTL and TERM did not seem to have any notable effect.

A couple times I was able to post successfully and vista actually loaded ... for about 10 seconds before I started getting various BSOD's including:

Failure in ecache.sys
Failure in CI.dll
page_fault_in_nonpaged_area

I went even so far as to remove 2 or my 4 dimms to reduce stress on NB. I also increased all volt measurements to max and still no boot. A couple times got really scary because after it was clear that a post failed I was unable to power down the machine by holding power button. In some cases not even the clear CMOS button would work and I would have to unplug all power and reset.

I have a feeling that increasing volts higher and higher can also lead to instability even with the best cooling solutions. My 3.54Ghz stable configurations fails if I add more volts to vcore or NB, so I have learned that more is not always better, and that sometimes you need to find the happy medium. Officially I tried all of the below configurations:

Shooting for 3.888Ghz -->

For all tests, FSB set to 1728, ram set to 864, multi at 9x and dimms set to 2.10v:

NB Tests >
1.45 + all possible vcore voltages
1.4 + all possible vcore voltages
1.35 + all possible vcore voltages
1.3 + all possible vcore voltages

NB Tests + GTL at 65 and TERM at 63 >
1.45 + all possible vcore voltages
1.4 + all possible vcore voltages
1.35 + all possible vcore voltages
1.3 + all possible vcore voltages

NB Tests + GTL at 65 and TERM at 63 + SB at 1.6 >
1.45 + all possible vcore voltages
1.4 + all possible vcore voltages
1.35 + all possible vcore voltages
1.3 + all possible vcore voltages

No possitive results. Total tests 144. I also tried a few tests with DIMM's at 2.20v for the hell of it. To avoid potential FSB holes, I did about 20 tests using different FSB and multipier settings. No go. I think it's safe to say for whatever reason (BIOS, RAM type, temps) my board will not ever exceed 3.54. I do not believe this will be corrected by BIOS updates. x-built and memory have clearly demonstrated that no BIOS update is needed to push the E8400.

In closing I would like to thank all who have assisted me in my quest. I will continue to monitor this thread just in case I have missed something or a magical fix is discovered. Many thanx.

[wolf1974]

X-Built,

Congrats on achieving stability at 3.6ghz. That's great news

After going through a few more unsuccessful rounds of the screen freezing in OCCT, I dialed in your settings and passed the standard 30 minutes test!



In my experience, passing this usually means clear sailing to Prime95 stability, as I've never found OCCT to pass when Prime fails within four to eight hours. This gives me great confidence, as even if Prime does fail, all that should be needed now is some slight vCore tweaking. It's the home stretch!

Here are my settings for 3.6ghz, OCCT Stable:

CPU: 3600mhz (400x9)
FSB: 1600mhz (400x4)
RAM: 800mhz (400x2)
RAM timings: 5-5-5-15-2T

PCI-E freq: 100mhz (stock)
Auto-Disable: Disable

vCore: .3000 (the next step is to find the lower bounds of stability.)
vDIMM: 2.1
FSB: 1.45
SB: Auto
GTL Ref: 65
FSB Terminator: 25
All other voltages: Auto
Speed Spectrums: Disabled

Interestingly enough, the last unstable settings were similar: FSB was at 1.4v while FSB Terminator was at 50. These previous settings caused the same freezing issue within 5 minutes of the OCCT Test. This certainly adds weight to my suspicion that the freezing issues were caused by incorrectly "tuned" NB voltages. Bumping NB voltage just .25 and reducing FSB Term to 25 did the trick.

My next step is to see how low I can go on vCore while maintaining all other voltages. I should be able to cut down a few notches, based upon my experience with this chip. With that finished, I'll do a 12 hour Prime run and call it a day for 3.6ghz. Then I'll likely want to get my Vista environment running on this rig before I try for 3.8-4.0 stable.

This is very good news we're having today. With the earlier 4.05ghz benchmark and our success at 3.6ghz stable, I'm far more inclined to recommend this board as an able overclocker. That's very exciting, as this will be the first non-reference Intel nForce board (680i or 780i) that works well out of the box and provides for ample overclocking (even if OCing isn't as easy yet as on the reference boards.) Given that the solid caps and the advanced PWM system should provide for greatly increased reliability over the long term when compared to the reference boards, I'd say the P7N is indeed a winner!

nice yip looking good, i am at work and my rig is running quad prime right now it has been good for 8 hours so far, i am at 4 ghz and some change!!! hey i need no know how to get my ram to run at 1 to 1 could you post you cell settings and mem settings so i can see what i am messin up but i am getting happy and i added a spot cooler for nb,,temp is 37,!! 1.5 v wolf,,,

[theYipster]

Wolf,

For your 4ghz setup, what is your FSB speed? To run 1:1, set your RAM speed to half of that. If you need to increase the speed of your RAM to get to a 1:1 divider, loosen your timings to compensate so that you remain stable. Once you've established stability at 1:1, you can then work to tighten timings, testing for stability as needed.

For my 4.05ghz run with my OCZ PC-6400 SLI 2x2 gig set, my FSB was set to 1800, RAM to 900mhz, and timings to 5-5-5-15-2T (other timings on auto.) vDIMM was set to 2.2v. Remember that proper settings for RAM at any frequency depends greatly on what RAM you have. If you have PC-8500 RAM, for example, you'll likely be able to run 900mhz with tighter timings. Also be sure not to exceed the maximum warranted voltage for your RAM, as DIMMS will fry due to overvolting much more easily than any other component in your rig.

With a QX9650, you have a great deal of flexibility in setting your FSB. Ultimately, you'll want to set your RAM so that your RAM's bandwidth is as high as possible with latencies as low as possible, while your proc remains at 4ghz. At times, lowering your RAM and FSB speed, while increasing your processor's multiplier (so that 4ghz is maintained) will result in more bandwidth and lower latency than with a higher set FSB. This is due to the ability to set tighter timings (and a 1T command rate) at lower speeds, and due to NB strap differences. Passing a certain FSB threshold may cause the NB to switch to a different strap which can have a drastic effect on bandwidth, despite your RAM and FSB running at higher speeds. If you can achieve 4ghz with your multi set to x9 or x10, test both and see which combination of multi, FSB, and RAM will result in highest bandwidth. Unfortunately I haven't tested the strap tendencies of this board, and it may very well differ between boards and BIOSes. Ultimately, the best way to find the best settings is by testing different combinations... overclocking is always a game of trial and error.

[theYipster]

X-Built, Memory, and Overspeed

Can you please post the following information about your E8400s, if you have it.

VID number (read from Core Temp)
Batch number (located on product box label, under FPO/Batch #)
Stepping & Revision Information (read from CPUZ)

Overspeed,

Two things come to mind before giving up:

It is not certain that there exists a linear relationship between GTL, FSB Term, and NB voltage. If X-Built's theory is correct, in that the setting for FSB Term is a percentage added or subtracted to NB voltage, you may very well need to decrease FSB Term as you increase vNB. Also, on other boards, GTL settings are best left to AUTO unless working with Quad Cores. Try setting GTL Ref to auto and see what that does.

Last, but certainly not least, it could very well be that the issue lies with your memory. Before deciding that the board is at fault, you should test each DIMM to ensure that your memory isn't faulty. ecache.sys and cl.dll issues can easily be memory related, and a page-fault is certainly a memory problem. To conduct a proper memory diagnostic, you will want to download memtest86 (there are two alternatives, memtest86 and memtest86+ -- some prefer one over the other. I don't see a difference between the two.) This is a memory diagnostic that you burn to a CD and run in its own bootable environment. Running the diagnostic will be quite time consuming, but it is important that you do it properly.

You will want to run memtest86 on one DIMM at a time. Remove all but one DIMM in your system and place the DIMM in the slot furthest away from the cpu socket. Then boot up your PC with the memtest CD in the drive, and let the test run for 5 to 8 hours. (Some suggest longer, but reducing the amount of memory being tested often will make errors pop up sooner.) If memtest display an error, you've found your culprit. If not, simply swap out for the next DIMM and repeat. Again, this is time consuming, but it is the best way to ensure that the issue does not lie with your memory.

[wolf1974]

Wolf,

For your 4ghz setup, what is your FSB speed? To run 1:1, set your RAM speed to half of that. If you need to increase the speed of your RAM to get to a 1:1 divider, loosen your timings to compensate so that you remain stable. Once you've established stability at 1:1, you can then work to tighten timings, testing for stability as needed.

For my 4.05ghz run with my OCZ PC-6400 SLI 2x2 gig set, my FSB was set to 1800, RAM to 900mhz, and timings to 5-5-5-15-2T (other timings on auto.) vDIMM was set to 2.2v. Remember that proper settings for RAM at any frequency depends greatly on what RAM you have. If you have PC-8500 RAM, for example, you'll likely be able to run 900mhz with tighter timings. Also be sure not to exceed the maximum warranted voltage for your RAM, as DIMMS will fry due to overvolting much more easily than any other component in your rig.

With a QX9650, you have a great deal of flexibility in setting your FSB. Ultimately, you'll want to set your RAM so that your RAM's bandwidth is as high as possible with latencies as low as possible, while your proc remains at 4ghz. At times, lowering your RAM and FSB speed, while increasing your processor's multiplier (so that 4ghz is maintained) will result in more bandwidth and lower latency than with a higher set FSB. This is due to the ability to set tighter timings (and a 1T command rate) at lower speeds, and due to NB strap differences. Passing a certain FSB threshold may cause the NB to switch to a different strap which can have a drastic effect on bandwidth, despite your RAM and FSB running at higher speeds. If you can achieve 4ghz with your multi set to x9 or x10, test both and see which combination of multi, FSB, and RAM will result in highest bandwidth. Unfortunately I haven't tested the strap tendencies of this board, and it may very well differ between boards and BIOSes. Ultimately, the best way to find the best settings is by testing different combinations... overclocking is always a game of trial and error.

i think my fsb is 1371 with muliti at 11.5,,, idid have the ram set manual at 1066, so i should ste it to 685, and my ram is 1066 corsair 2 gig strips 2.1 v 5-5-5-15, this is good info thanks,me setting it to 1066 manualy has been keeping me from getting a good fsb overclock!!! now i will put multi to 9 and try like 1600 fsb and set ram to 800, this if it boots up, should look like 533 with 1-1 divider in cpuz,,correct??,, then i can up the multi and be smokin!!! hey what did ya think of my 3dmark score i sent ya?? it should only get better,once i get the ram running good,,,thanks for the help,,,wolf

[theYipster]

Try dropping the multi to 10, set FSB to 1600 and RAM to 800. This will be 1:1 with CPU-Z reporting that your RAM is running at 400mhz. (400 x 2 = 800 DDR2.)

With 1066mzh RAM rated for 2.1v, you should be able to tighten timings at 800mhz to 4-4-4-12 while keeping a 2.1v voltage. Honestly, there is really no reason to have RAM rated higher than PC-6400 / 800mhz unless you are pairing with a lower end, lower-multiplier chip.

It's hard to comment on your 3d mark score w/o knowing more about your setup. With the specs in my sig and the proc at 3.6ghz, I hit 18,300 in 3DMark06.

[loonym]

Very nice numbers Stan. These are the results I knew would start coming once people exercise a little patience. Overclocking new hardware is very time consuming but extremely satisfying at the same time.

Memory (longsiew), That is a very impressive thread you have at hardwarezone forum... link, and also a very nice 8400 for benchies like that with 1.28v

[BKA]

Good job Stan, those settings allowed me to do a successful OC of 400 FSB for 3.6 on my E8400. I'm not fully sure what the CPU GTL and FSB Term do, but it allowed me to run a full 3DMARK06 and system appears stable for the moment. The only thing I'm concerned about is NB temperatures which I have no idea of measuring exactly. Tommorow I will stop by the local PC store and pick up some AS5 ceramique and pull the heat pipes off and reapply the thermal paste for reassurance. I will let Prime95 run overnight to see how it goes. Thanks again.

[x-built-Stan]

theYipster,

I added my VID to the my 4.275GHz post.

BKA,

Very nice indeed.
I'm going to start running my stability tests tomight. I'll start with a Prime95 small FFT.

If you want to monitor the NB temp you should read my previous post on the subject. I use a thermal sensor connected to my LCS monitoring panel. You can put together for less than $40.

Good luck on your tests

[BKA]

theYipster,

I added my VID to the my 4.275GHz post.

BKA,

Very nice indeed.
I'm going to start running my stability tests tomight. I'll start with a Prime95 small FFT.

If you want to monitor the NB temp you should read my previous post on the subject. I use a thermal sensor connected to my LCS monitoring panel. You can put together for less than $40.

Good luck on your tests

Thanks, I saw that post but I assumed it was for water cooling. I'm on air, using a Tuniq Tower 120. I had a front panel with temperature probes but I'm not sure what I did with it or if it even works any longer if I remember correctlty. I'll have to look into that.

[memory]

Very nice numbers Stan. These are the results I knew would start coming once people exercise a little patience. Overclocking new hardware is very time consuming but extremely satisfying at the same time.

Memory (longsiew), That is a very impressive thread you have at hardwarezone forum... link, and also a very nice 8400 for benchies like that with 1.28v

Thx.

Here are max FSB taht I could achieve last night with blend test for 1 hr. 514mhz x 6 nb at 1.6volt. Dual Core Centre report about 1.57xx volt. On Bios P05

[Overspeed]

theYipster, x-built-stan, memory,

I just can't figure out my rig. You three are the only others who have the E8400 and you have all been able to manage decent OC's using this mobo and common settings. I tried the P05 tonight and I couldn't even get windows vista to load at stock settings. BSOD every time. I was only able to restore PC by reflashing 1B4. I did as theYipster suggested and ran memtest on all my dimms ... no errors after several runs. My Ram must be ok, they use the D9 Micron chip and are always cold to the touch, nor are they overclocked. I am nervous now that maybe my CPU or mobo is bad, or something else is going on. I just don't know what to do and i'm very bothered.

[BKA]

So far so good. Prime ran all night with no errors. So I will increase the FSB to see how far I can get with the current voltage. I really don't want to go past 1.4 on the NB but if more CPU voltage is required I'll bump that up some. My 3.6 OC is with the default 1.0 BIOS also, I never flashed any of the betas.

[Psycho_eddie]

Im interested in this board, but will a True 120EX fit vertically alligned?

I.e conflict with NB rollercoaster cooler?

Any1 got 450x8 fsb on a quad stable (3600mhz) ??

[memory]

Does anybody of you here experiences some significant performance decrease by upping FSB more than 451mhz?? It is some kinda 'nb strap' kicked in(sorry I do not know the correct term to describe the performance decease)

I run some tesing for 450mhz x 9 on Pi4mil, eve and sandra to compared the memory bandwidth and the timings for pi 4million run



This is with 454mhz x 9. It is very clear that memory bandwidth on eve and sandra had decreased a lot. Same as Pi 4 million even the seeting is all the same as per 450mhz x 9.


Can anybody test this decrease on yr P7N Diamond or is this a common issues on nvidia chipset as intel 965 chipset??