Overclock.net › Forums › Intel › Intel CPUs › [SOLVED] Q9550 @ 3.8GHz failing Prime95 Large FFTs
New Posts  All Forums:Forum Nav:

[SOLVED] Q9550 @ 3.8GHz failing Prime95 Large FFTs - Page 3

post #21 of 38
Thread Starter 
Quote:
Originally Posted by PanicProne View Post
I hardly think VTT is the cause of instabilty. But if you think so, then lower it back to 1.20v and play with the CPU GTLs.

You could also try to put RAM @ 2.1v just to rule it out.

But before that, run Prime 95 and see which core(s) actually fail, and then adjust CPU GTL accordingly.

If core 0 or 2 fails, adjust CPU GTL ref 0/2.

If core 1 or 3 fails, adjust CPU GTL ref 1/3.

If it doesn't work, you'll have to increase VTT again. And then, calibrate the GTLs once again.

It won't be easy.

In case you didn't know, a Core 2 Quad is two separate dual core dies in a single package, so these pairs of cores communicate externally through the FSB.
Ok I'll try that. Core #2 and #3 fail consistently in Prime large fft, all the time. They're the only ones that fail. Core #0 and #1 keep running after the other 2 fail.. hmmm.. I'll try messing with the GTL Refs, 0.63 isn't working
post #22 of 38
Also, you should raise all your voltages a little bit higher than they need to be, this way, you can rule out many parts of your system as the source of instability.

For example, set Vcore to 1.35v with LLC enabled. just because you have an E0 revision, doesn't mean they all clock the same.

You may very well need a higher Vcore to achieve stability.
Son of a Bridge
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7 2600k @ 4.2ghz | 1.24v Asus P8P67 EVO [B3 Revision] Sapphire HD6870 1GB GDDR5 @ 1ghz Core 4GB Kingston HyperX DDR3 1600mhz 
Hard DriveOptical DriveOSMonitor
Hitachi Deskstar P7K500 500gb 16MB Cache SATA II SATA LG DVD burner Windows 7 Ultimate x64 Acer LCD 17" 1280x1024 
PowerCaseMouseMouse Pad
Corsair TX650 Modded regular case X7 X-760H 2000 DPI N@NO 
  hide details  
Reply
Son of a Bridge
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7 2600k @ 4.2ghz | 1.24v Asus P8P67 EVO [B3 Revision] Sapphire HD6870 1GB GDDR5 @ 1ghz Core 4GB Kingston HyperX DDR3 1600mhz 
Hard DriveOptical DriveOSMonitor
Hitachi Deskstar P7K500 500gb 16MB Cache SATA II SATA LG DVD burner Windows 7 Ultimate x64 Acer LCD 17" 1280x1024 
PowerCaseMouseMouse Pad
Corsair TX650 Modded regular case X7 X-760H 2000 DPI N@NO 
  hide details  
Reply
post #23 of 38
Thread Starter 
Quote:
Originally Posted by PanicProne View Post
Also, you should raise all your voltages a little bit higher than they need to be, this way, you can rule out many parts of your system as the source of instability.

For example, set Vcore to 1.35v with LLC enabled. just because you have an E0 revision, doesn't mean they all clock the same.

You may very well need a higher Vcore to achieve stability.
I know, but based on my previous OCs, the vcore necessary stays below 1.30v. I tried running 1.30v up to 1.35v and it would BSOD when entering windows. Now I have it at 1.21875v LLC enabled and FSB term down to 1.20v. I have the following screenshots from before to give you an idea of how much vcore it needs:

http://img207.imageshack.us/i/prime9...206v44hrs.jpg/
(these turned out to be unstable)

Since my OC is failing Large FFT, it's got to be the FSB Term, NB Voltage that's the culprit.. possibly even the CPU GTL refs, but I know it isn't the vcore. Vcore would be tested on small fft.
post #24 of 38
I also hit the stability wall at 3.8GHz. Sounds fairly common with this chip. Cant recommend much, as mine was stable on AUTO voltages at that speed. Good luck though.
post #25 of 38
does the same .8v rule apply to the NB Gtl Ref? Im having the exact same problem. I went over an hour in small ffts with no issues, blend will fail immediately on worker 1 and 2 with a rounding error. Im sure its not ram.

I messed with the GTL ref for the NB, and whats strange is I went from 1.39v to 1.45v changing the GTL ref appropriately, yet no matter what the voltage is, it will not boot unlees its set to .63x.
post #26 of 38
Thread Starter 
Quote:
Originally Posted by Mattb2e View Post
does the same .8v rule apply to the NB Gtl Ref? Im having the exact same problem. I went over an hour in small ffts with no issues, blend will fail immediately on worker 1 and 2 with a rounding error. Im sure its not ram.

I messed with the GTL ref for the NB, and whats strange is I went from 1.39v to 1.45v changing the GTL ref appropriately, yet no matter what the voltage is, it will not boot unlees its set to .63x.
Same here. Worker 3 & 4 would always fail large fft within 10 mins. I found that tweaking the CPU GTL Refs and NB GTL Ref helped stabilize the OC. Small fft is fine, that means our vcore's set appropriately. It's either the FSB Term, NB voltage or the correct CPU GTL Refs. Tweaking the NB voltage does nothing for me too. It's all on the GTL Refs (methinks)

I now have the following settings and they seem fairly stable:
CPU Ratio Setting: 8.5
FSB Frequency: 450 MHz
FSB Strap to North Bridge: AUTO
DRAM Frequency: 900 MHz
DRAM Timings: 5-5-5-18-50

CPU Voltage: 1.21875v
CPU GTL Voltage Reference (0/2): 0.645
CPU GTL Voltage Reference (0/2): 0.685

CPU PLL Voltage: 1.54v
FSB Termination Voltage: 1.20v
DRAM Voltage: 2.0v
NB Voltage: 1.32v
NB GTL Reference: 0.645

Load-Line Calibration: Enabled

I've been replaying Crysis for 30 mins now and haven't encountered any problem. I'm sick and tired of failing large fft so I'm just going to stick with this and see if I BSOD during the week.
post #27 of 38
Yes, I actually forgot about NB GTL Ref.

I don't know how important it is, but I don't think it needs to be the same value as VTT * CPU GTL. But it is related to VTT and NB voltage.

You can try and change it to see if affects stability in any way. Try .67x on the NB GTL Ref.

I still stand by my point that VTT and CPU GTLs and probably CPU PLL will be the most important settings to play with when trying to get your quad stable.
Edited by PanicProne - 1/3/11 at 8:10am
Son of a Bridge
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7 2600k @ 4.2ghz | 1.24v Asus P8P67 EVO [B3 Revision] Sapphire HD6870 1GB GDDR5 @ 1ghz Core 4GB Kingston HyperX DDR3 1600mhz 
Hard DriveOptical DriveOSMonitor
Hitachi Deskstar P7K500 500gb 16MB Cache SATA II SATA LG DVD burner Windows 7 Ultimate x64 Acer LCD 17" 1280x1024 
PowerCaseMouseMouse Pad
Corsair TX650 Modded regular case X7 X-760H 2000 DPI N@NO 
  hide details  
Reply
Son of a Bridge
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7 2600k @ 4.2ghz | 1.24v Asus P8P67 EVO [B3 Revision] Sapphire HD6870 1GB GDDR5 @ 1ghz Core 4GB Kingston HyperX DDR3 1600mhz 
Hard DriveOptical DriveOSMonitor
Hitachi Deskstar P7K500 500gb 16MB Cache SATA II SATA LG DVD burner Windows 7 Ultimate x64 Acer LCD 17" 1280x1024 
PowerCaseMouseMouse Pad
Corsair TX650 Modded regular case X7 X-760H 2000 DPI N@NO 
  hide details  
Reply
post #28 of 38
I would run that RAM unlinked until you get a stable OC.
    
CPUMotherboardGraphicsRAM
Q9550 3.6Ghz/1.27v EVGA 680i SLi 2x EVGA-GTX480 SLI 8Gb GSkill Pi series (CL4) 4-4-4-12 
Hard DriveHard DriveOptical DriveCooling
1Tb Caviar Black OCZ Agility ASUS DVD-R/W ASUS Silent Square EVO 
OSMonitorKeyboardPower
Win7 x64 Ultimate 24'' Samsung MERC Stealth 1000W Thermaltake 
MouseMouse PadOther
Logitech G9x ZERO Gaming pad USB 3.0 pcie x1 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
Q9550 3.6Ghz/1.27v EVGA 680i SLi 2x EVGA-GTX480 SLI 8Gb GSkill Pi series (CL4) 4-4-4-12 
Hard DriveHard DriveOptical DriveCooling
1Tb Caviar Black OCZ Agility ASUS DVD-R/W ASUS Silent Square EVO 
OSMonitorKeyboardPower
Win7 x64 Ultimate 24'' Samsung MERC Stealth 1000W Thermaltake 
MouseMouse PadOther
Logitech G9x ZERO Gaming pad USB 3.0 pcie x1 
  hide details  
Reply
post #29 of 38
Also:

LOWER NB Voltage to 1.20v. Yes, I mean NB voltage.

I used to run my NB at over 1.3v thinking it was helping with stability, when in fact it was oversensitive to voltage.

1.20v was much more stable than 1.30v on the NB on my P5Q Pro.

Try it
Son of a Bridge
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7 2600k @ 4.2ghz | 1.24v Asus P8P67 EVO [B3 Revision] Sapphire HD6870 1GB GDDR5 @ 1ghz Core 4GB Kingston HyperX DDR3 1600mhz 
Hard DriveOptical DriveOSMonitor
Hitachi Deskstar P7K500 500gb 16MB Cache SATA II SATA LG DVD burner Windows 7 Ultimate x64 Acer LCD 17" 1280x1024 
PowerCaseMouseMouse Pad
Corsair TX650 Modded regular case X7 X-760H 2000 DPI N@NO 
  hide details  
Reply
Son of a Bridge
(13 items)
 
  
CPUMotherboardGraphicsRAM
i7 2600k @ 4.2ghz | 1.24v Asus P8P67 EVO [B3 Revision] Sapphire HD6870 1GB GDDR5 @ 1ghz Core 4GB Kingston HyperX DDR3 1600mhz 
Hard DriveOptical DriveOSMonitor
Hitachi Deskstar P7K500 500gb 16MB Cache SATA II SATA LG DVD burner Windows 7 Ultimate x64 Acer LCD 17" 1280x1024 
PowerCaseMouseMouse Pad
Corsair TX650 Modded regular case X7 X-760H 2000 DPI N@NO 
  hide details  
Reply
post #30 of 38
Thread Starter 
Quote:
Originally Posted by PanicProne View Post
Also:

LOWER NB Voltage to 1.20v. Yes, I mean NB voltage.

I used to run my NB at over 1.3v thinking it was helping with stability, when in fact it was oversensitive to voltage.

1.20v was much more stable than 1.30v on the NB on my P5Q Pro.

Try it
I tried it before it failed But I'll try it now with my newly set GTL Refs
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Intel CPUs
Overclock.net › Forums › Intel › Intel CPUs › [SOLVED] Q9550 @ 3.8GHz failing Prime95 Large FFTs