Help with i5-750/Asus P7P55D Deluxe OC.

[mingsoup]

My OC to 4.0 Ghz at a base clock of 190 seemed to work just fine however Starcraft 2 would crash, restarting the computer after playing single player for a bit.

I tried reducing the overclock to 166 base clock for ~ 3.5Ghz. However, I still can't get Prime95 nor Starcraft 2 to not stop crashing. Prime95 fails after about .5 hours.

I've read two glowing articles on the Asus P7P55D Deluxe and its ability to handle voltages. I was trying to do the same, but I can't get a stable overclock. Legion Hardware reported success on 4 successive i5-750's. If you want the exact setting I'm using, they are here: http://www.legionhardware.com/articles_pages/intel_core_i5_750_overclocking_guide,4.html

Anyway, I'm also running 2x2Gb PC3 1600 , 2 ATI HD5770's, 3 120mm fans, a Sata 7200 hard drive, and one DVD writer.

This is powered by a Coolmax 550W power supply. I'm running latest bios revision 1702. (I can't revert either!!!, but thats my other thread.)

Any ideas? I bought this Mobo and Proc to specifically hit 4.0Ghz. I'm also running a Scythe Mugen and temps aren't an issue yet.

 

[mingsoup]

Thinking it might be a power issue or not enough power, I removed one of the ATI HD5770's and retested at 166 base clock. Prime 95 lasted 9 minutes.
I also don't think its a power issue as running both prime95 AND Unigen Heaven is possible for about 30 minutes until the Prime95 errors on one or more cores. I'm guessing it means its not a power issue, but I'm not totally sure.

I don't get what I'm doing wrong. Did I just get an absolute lemon. Uhg.
I'm going to try turn off all power saving options and turbo/speedstep, and just go for a strict OC w/ auto voltage. Maybe that will work.

 

[Termie]

One thing that seems a little fishy in the LegionHardware settings is their enabling of TurboMode in the bios. I have a hard time believing they were able to reach 4.2Ghz with Turbo turned on. If you're trying to overclock in the ~4Ghz range, turbo just won't work. Make sure that's turned off. Neither SC2 (nor Prime95, I believe) will load all four cores, so when you run those, you'd trigger turbo mode and the clocks would go sky high at 190 BCLK (or 166 for that matter), far too high to be stable.

 

[mingsoup]

I'll try turn off turbo and Prime95 again.

 

[Edrick]

I would not turn on Turbo with anything over 160 BLK if you are having issues with stability.

 

[Sp12]

Use linx, it's much more stressful than prime and will help you diagnose stability faster.

 

[mingsoup]

I can't figure it out. I turned off both Speedstep and Turbo. I locked in my 8-8-8-24 @ 1.65V. I ran at 1.25V VCORE. I've turned off all the power saving options I can find.
Something is wrong with the picture. This board was supposed to be able to auto everything at 200BCLK 100% stable.

3rd core failed in 3 minutes of prime95. I'll try bump up the VCORE and try linx? I'm using the latest BIOS, i'm stuck as its impossible to downgrade.
Any other suggestions would be massively appreciated.

Does the starcraft 2 crash and restart point to anything for anybody? Desktop has always been fine. Prime95 goes in and out of erroring time.

Asus autos my VCORE to 1.25V as reported by CPU-Z, with Load Line enabled which I'm thinking is correcting for VDroop. 20 passes of Linx went just fine.....

I'll try play Starcraft 2 some more.

 

[NumericalMethods]

Turning on Turbo on the Asus boards allows you to use the 21x multiplier. This gives you a shot at 21 *191 to hit 4GHz. If 21 is manually selected then the other turbo modes will not activate as long as you have disabled the c-state setting in the BIOS.

OP if you want to trouble shoot stability issues I would suggest going back to scratch and testing the stability of ONLY your bclk, leaving the CPU and memory below their specced speed. Use a memory intensive test such as OCCT large FFTs or Prime95 large to test for stability of the uncore.

Once you get that locked down move onto CPU core and test using CPU intesive tests - OCCT small FFTs, Prime95 small and of course the daddy, Linx/ Intel burn test.

In the mean time, save your current overclock in the BIOS and return all your settings to default and test Starcraft. If you are still having crashes then you need to troubleshoot elsewhere.

 

[mingsoup]

I can use the 21x multiplier with Turbo off and c-state off. I'm seeing it right now with CPU-Z. I think this is what you already pointed out.

I was able to stress with LINX for over an hour fifteen and I was stable. However, Large FFT's in OCCT, failed in ~ 20minutes as it does in Prime95. Does that give any hints?
I'm gonna go back to the drawing board. Starting at 166BCLK, lowering everything else below and testing Large FFTS. See where I get to.

Didn't even get 4 minutes in before errors when just testing BCLCK with lowered memory multiplier,cpu multiplier, and locked timings. Guess I'll fiddle w/ IMC now.

 

[NumericalMethods]

Yeah, failure in large FFTs most likely indicates an instability in your memory subsystem. That is - uncore and RAM on a P55 setup. You need to try raising your VTT voltage to see if it is the uncore that requires more voltage (max safe 1.4V) if not, you may need to increase the PLL voltage in addition.

If neither of those work try relaxing your RAM timings or use the x6 RAM multi to seriously underclock your RAM to eliminate it from the equation. You can concentrate on sorting your RAM timings once everything else is stable.

For reference my i5 take 1.2V VTT and 1.85V PLL to do a 190 bclk rock solid. Your settings will vary but could be similar. HTH

 

[mingsoup]

It seems I'm getting closer. I went >5 minutes at VTT (IMC on Asus Bios) of 1.175, I went 1:44 with a VTT of 1.18125V. I'll try one notch higher with large FFTs. As for PLL, I don't really know what is standard, but I didn't know what VTT was standard either I just kinda threw a dart. I'm sure the Asus bios will tell me what default PLL is.

I think I have ram out of the picture. I've got it locked at the voltage and timings that OCZ states for the modules and its on a multiplier that puts it well below 1600 Mhz which its rated at. However, letting the Asus bios Auto my timings and voltage for the memory doesn't produce instability, so I might be able to fall back on that if need be? (Memory is not an issue, its the uncore and VTT/PLL.

Its funny how I paid 50$ premium for this Asus board for its Auto Voltage OC, set your BCLK and leave. However, it seems I was misled. Won't do that again.

Thanks for the help Numerical. Every new platform oc's a little differently.

 

[mingsoup]

Well it seems its not the OC. Starcraft 2 is crashing on Default Settings in the Bios.

So now I need to figure out what it is. Thoughts on the PSU? The memory is not the issue, i've memtested it for over 12 hours, no problems.

Crashes are specifically reported as Kernel-Power Event ID 41. Searching, resulted in finding conflicting realtek audio drivers to be the issue. I'm using onboard audio as well...maybe I'll give that a try.

 

[NumericalMethods]

Is it this power supply?:

http://www.devhardware.com/c/a/Power-Supply-Units/CoolMAX-550W-Aluminum-PSU-Review/

Sorry I didn't check earlier. For what you are trying to do that thing is totally unsuitable. Only 264W from a single 22A 12V rail! Your system will need closer to 400W fully loaded! If it was a quality 550W PSU then you would be sorted but over half the rated power of that thing comes from the 3 + 5V rails. That is not very useful for modern computers as the big components ie mobo, CPU and GPUs take their power from the 12V rail. Anything that could supply over 450W 12V would work for your setup, 400W at a pinch but leaves you little wiggle room for case fans and other components.

I would recommend you buy a quality PSU if you intend to run an overclocked crossfire system. If I have your PSU wrong apologies.

 

[Termie]

Is it this power supply?:

http://www.devhardware.com/c/a/Power-Supply-Units/CoolMAX-550W-Aluminum-PSU-Review/

Sorry I didn't check earlier. For what you are trying to do that thing is totally unsuitable. Only 264W from a single 22A 12V rail! Your system will need closer to 400W fully loaded! If it was a quality 550W PSU then you would be sorted but over half the rated power of that thing comes from the 3 + 5V rails. That is not very useful for modern computers as the big components ie mobo, CPU and GPUs take their power from the 12V rail. Anything that could supply over 450W 12V would work for your setup, 400W at a pinch but leaves you little wiggle room for case fans and other components.

I would recommend you buy a quality PSU if you intend to run an overclocked crossfire system. If I have your PSU wrong apologies.

I doubt that's what he has, since that's a 2003 model. More likely he has this model: http://www.newegg.com/Product/Produc...82E16817159026. With 35 amps on the 12-volt rail, it's probably still not adequate, though. Running a crossfire system and a highly-overclocked, overvolted i7 on a 35 amp 12-volt rail is asking for trouble. My guess is that the reason Starcraft fails even where CPU stress-testing is working is because you're stressing both the CPU and dual GPUs...and poof, crash goes the 12-volt rail.

 

[mingsoup]

The PSU is inadequate and/or cutting is pretty close. The PSU 12V is rated at 420W, 18A on 12V1/12V2 (not sure what that means, the split 12V) However, I now believe that it might not be the immediate problem.
Every time Starcraft 2 crashed, I received a Kernel Event Power ID 41. Googled this, and it came back with a conflicting onboard audio driver issue. I removed the VIA audio driver I had installed and let Win7 alone take care of it. I haven't had another crash and Starcraft 2 performance must have increased by 50%. It seems there indeed was a conflict even to the point of hindering performance and system stability.

I just completed a 2hr Large FFT, followed by an 8hr Small FFT without any errors on the following settings. Whats interesting is that the max temperature achieved with these manual settings resulted in 10 degrees C less than the auto vcore that the Asus bios achieved, AND it was more stable. :S
Vcore-1.14375
VTT- 1.18750
Dram -1.65
PLL -1.80(stock)
PCH- 1.05 (stock)
BCLK -166 for 3.5Ghz

Should I still worry about the power supply? Thoughts now? I still might search for a new PSU. Would power be a performance detractor? I've run into inadequate power supplies before and more often than not, the system just won't boot OR it will shutoff very soon into windows. I'm sure that if a Vantage score matches up with a similar 5770 crossfire system, than I don't have problems yet. My OCD tells me that one day I may, as those caps go further. The other thing that makes me think my power is at least ~okay~, is that I ran BOTH prime95 Large FFTs and Unigen Heaven at the SAME time for over an hour and my system remained stable and on.

 

[betasub]

Sadly, the PSU is always going to be a nagging worry now that you are aware of its potential to cause issues for your shiney new Lynnfield-Crossfire system.

Futhermore, running two stress tests at the same time is not a good idea, because one simply detracts from the other (in this case Prime is likely to take a back seat). This is not better, or even the same, as running one alone for the same period of time.

 

[mingsoup]

There's alot of things I could let bother me. I'll stick with 430W on 12, for now. Its working and I'm not experiencing any problems now. I even knocked off 10C doing my own OC. However, I much appreciate all the help on this one, especially the exact process of establishing stability by going BCLK(VTT)>LargeFFT>CPU(VCORE)>SmallFFT.

 

[NumericalMethods]

That is great that you got a stable system now. Just to round things off, without turning this into a discussion for the PSU forum. Max power draw you are likely to see is 150W - CPU (overclocked); 2*100 W - 5770; 50W - Rest of system = 400W. Of course you will never see everything loaded to its maximum draw simultaneously anyway so real usage figures will be lower. At least you know that you are close to the rated maximum of the PSU but unlikely to be exceeding it currently. Good luck with the overclock!

 

[mingsoup]

Still getting Kernal Power error in Starcraft 2. Yet, I have never got it stressing. PSU? Maybe that is the reason for this motherboard not auto OC'ing as well. Its gotta be.
How much wattage should I shoot for. I was hoping 4.0~4.2Ghz on the Proc. 600W? 700W?Amps on the rail? Room for upgrades in the future?

 

[ehume]

BTW - what's your Vcore? Most of these guys run i7 9xx's. A Lynnfield cpu will need a higher Vcore for the same clocks.

I couldn't run 191x21 because i needed too much voltage and the cores got too hot - or hotter than I'd like, anyway. So I run 182x22, with Turbo on Auto and Speed Step disabled.

From the numbers I have seen, absolute max for Vtt is 1.21v. I hope you didn't go over that when you nudged it up from 1.8-something. I'm running mine at 1.9v.

Most of the OC guides tend to ignore the Lynnfield oc'ers, but we're out there.

Edit: running my RAM at 10x182=1820MHz, staying well within its 2000MHz stated envelope. I got my Vdram by running my rig on Auto voltage, and watching what it did on 4004MHz. I didn't push it too hard, of course. Had to keep the heat down. But I saw what the mb thought was the right voltage for my sticks. I haven't had any problems.

 

[Kenmitch]

How much wattage should I shoot for. I was hoping 4.0~4.2Ghz on the Proc. 600W? 700W?Amps on the rail? Room for upgrades in the future?

Just make sure you get a good quality power supply such as a Corsair for example. Having some xtra power for things like video card upgrades wouldn't hurt. Unless you plan on going dual GTX 480's something in the 750-850 watt range should be good.

 

[mingsoup]

Yet its only in SC2 where my OC will restart the computer. Not in stressing or Borderlands. but i'm just so close to the wall on the power supply, it can't be avoided. Is OCZ any good?

Vcore-1.14375
VTT- 1.18750
Dram -1.65
PLL -1.80(stock)
PCH- 1.05 (stock)
BCLK -166 for 3.5Ghz
The voltages above gave me a supposed stable OC at 3.5Ghz with some pretty heavy duty testing. 8hr Large, 2hr Small FFT. However, the auto memory timings that the asus board are interesting. They range from 77720 to 99924, depending on the BCLCK I choose. I thought my modules were 88824, so I set them there at 1.65V, but blue screened 30 minutes into DOTA. So I had to auto it again. I just need to get the module models of the stick as I think CPUZ is reporting wrong.

Another disappointment. I tried Bit Tech's 4.2Ghz oc settings and they do work I can boot and then run linpack. However, with their Noctua NH-U12P SE2 they had 85C temps. This is also the same exact result TechSpot received with this mobo and a 4.2Ghz OC., 85C However I immediately spike to 90, and am soon at 93C which is way way too hot. Is this just as far as a Scythe Mugen Rev B will get me, should I even expect to hit 4.2Ghz with it and see reasonable temps?

 

[yottabit]

I have the Asus P7P55D and my OCZ ram runs nowhere near its rating. I have the 4 gig obsidian kit which is rated 9-9-9 at 1600 mhz and 1.65 V. The highest stable rating I've got it to was 8-8-8 at 1056 Mhz or 9-9-9 at 1200something like that, even with 1.65 V. (Stable 24 hrs on Memtest)

That's in dual channel, with just one stick I was able to run much better timings and frequency. It doesn't really matter what the memory manufacturer rates it at if its not on ASUS's approved memory list for that speed. From what I understand P55 is tricky with memory compatibility so I wouldn't be surprised if that was what was holding you back. However I was getting page fault BSOD nothing to do with power.

It is interesting how ASUS chooses the memory timings for you in AUTO. I wish I knew the reasoning behind the way it chooses them. If I were you I would set the DRAM frequency really modest and let it reboot a couple times to find the best timings, then run a memory stress test. Or run a mem stress test now and see if you find any errors.

 

[mingsoup]

Yeah, my 8-8-8-24, auto to 9-9-9-24 at the rated 1600Mhz they are at. If I run them at the rated 888....they BSOD. Guess it's like you said, Asus and its approved memory.

SUPER EDIT::::: Moral of story.

Wasn't any of my voltages, the power supply.

It was the ram. When I typed in what the OCZ ram what rated at, memtest went crazy on me. However, I had eliminated memory earlier (or so I'd thought) at the bios default settings of 1066Mhz, but thats not what this ram is rated at. Thus typing in rated speeds later when I started oc'ing and becoming more wise, caused me great heartache. I need to RMA these sticks.