New BIOS, stability issues P8Z77-WS & i7-3770

[Goros]

Hey all.

So I loaded the 3302 bios today and lost stability, and have been messing with settings to get stable again but think I have a memory instability. I could be wrong.

My rig is a P8Z77-WS running 3302, with a i7-3770 (non-k) at 4.38GHz, using 32GB of Corsair Vengeance 1600MHz 10-10-10-27 2t 1.5v memory (QVL), and 2 - SLI'd MSI N580GTX Lightning Xtreme's.

Settings are as follows

107MHz FSB, x41 Mult, 100:100, 1712Mhz RAM
CPU - 1.23v (1.224 @ 100% load, sometimes 1.232)
RAM - 1.550
VCCSA - .92500
VCCIO - 1.05000
CPU PLL - 1.80000
PCH - 1.05

This was stable prior to this release of the bios.

So, I ran my usual Prime95 x64 (27.9) with roundoff and sum(inputs) checking on torture using 85% of ram (24GB) and it rebooted the system with 20 minutes each time.

Running it on standard with both roundoff and sum checking on torture with default settings, I got this error:

See the end of this message for details on invoking
just-in-time (JIT) debugging instead of this dialog box.

************** Exception Text **************
System.AccessViolationException: Attempted to read or write protected memory. This is often an indication that other memory is corrupt.
at System.Drawing.SafeNativeMethods.Gdip.GdipDrawString(HandleRef graphics, String textString, Int32 length, HandleRef font, GPRECTF& layoutRect, HandleRef stringFormat, HandleRef brush)
at System.Drawing.Graphics.DrawString(String s, Font font, Brush brush, RectangleF layoutRectangle, StringFormat format)
at System.Drawing.Graphics.DrawString(String s, Font font, Brush brush, PointF point)
at .RenderFooter(Graphics g)
at .RefreshBuffer()
at .FullRefresh(Boolean forceInit)
at .AddValue(Single value)
at .(Int32 ,  , Single &#57344
at .()
at System.Windows.Forms.Timer.OnTick(EventArgs e)
at System.Windows.Forms.Timer.TimerNativeWindow.WndProc(Message& m)
at System.Windows.Forms.NativeWindow.Callback(IntPtr hWnd, Int32 msg, IntPtr wparam, IntPtr lparam)


************** Loaded Assemblies **************
mscorlib
Assembly Version: 2.0.0.0
Win32 Version: 2.0.50727.5466 (Win7SP1GDR.050727-5400)
CodeBase: file:///C:/Windows/Microsoft.NET/Framework64/v2.0.50727/mscorlib.dll
----------------------------------------
nvidiaInspector
Assembly Version: 1.9.6.9
Win32 Version: 1.9.6.9
CodeBase: file:///E:/CleanInstallFiles/nvidiaInspector/nvidiaInspector.exe
----------------------------------------
System
Assembly Version: 2.0.0.0
Win32 Version: 2.0.50727.5467 (Win7SP1GDR.050727-5400)
CodeBase: file:///C:/Windows/assembly/GAC_MSIL/System/2.0.0.0__b77a5c561934e089/System.dll
----------------------------------------
System.Management
Assembly Version: 2.0.0.0
Win32 Version: 2.0.50727.5420 (Win7SP1.050727-5400)
CodeBase: file:///C:/Windows/assembly/GAC_MSIL/System.Management/2.0.0.0__b03f5f7f11d50a3a/System.Management.dll
----------------------------------------
System.Windows.Forms
Assembly Version: 2.0.0.0
Win32 Version: 2.0.50727.5468 (Win7SP1GDR.050727-5400)
CodeBase: file:///C:/Windows/assembly/GAC_MSIL/System.Windows.Forms/2.0.0.0__b77a5c561934e089/System.Windows.Forms.dll
----------------------------------------
System.Drawing
Assembly Version: 2.0.0.0
Win32 Version: 2.0.50727.5467 (Win7SP1GDR.050727-5400)
CodeBase: file:///C:/Windows/assembly/GAC_MSIL/System.Drawing/2.0.0.0__b03f5f7f11d50a3a/System.Drawing.dll
----------------------------------------
System.Configuration
Assembly Version: 2.0.0.0
Win32 Version: 2.0.50727.5420 (Win7SP1.050727-5400)
CodeBase: file:///C:/Windows/assembly/GAC_MSIL/System.Configuration/2.0.0.0__b03f5f7f11d50a3a/System.Configuration.dll
----------------------------------------
System.Xml
Assembly Version: 2.0.0.0
Win32 Version: 2.0.50727.5420 (Win7SP1.050727-5400)
CodeBase: file:///C:/Windows/assembly/GAC_MSIL/System.Xml/2.0.0.0__b77a5c561934e089/System.Xml.dll
----------------------------------------
ekaecsni
Assembly Version: 1.9.6.9
Win32 Version: 2.0.50727.5467 (Win7SP1GDR.050727-5400)
CodeBase: file:///C:/Windows/assembly/GAC_MSIL/System/2.0.0.0__b77a5c561934e089/System.dll
----------------------------------------

************** JIT Debugging **************
To enable just-in-time (JIT) debugging, the .config file for this
application or computer (machine.config) must have the
jitDebugging value set in the system.windows.forms section.
The application must also be compiled with debugging
enabled.

For example:

<configuration>
<system.windows.forms jitDebugging="true" />
</configuration>

When JIT debugging is enabled, any unhandled exception
will be sent to the JIT debugger registered on the computer
rather than be handled by this dialog box.

This is telling me that there is a memory error somewhere. My temps on all 4 cores at max load cap at 83c with the average being around 75c.

This error continuously pops up when I close the box. None of the 8 workers have stopped yet. When I close the offending program, it's my copy of Nvidia Inspector v1.9.6.9 which I was using to monitor GPU activity just in case.

Any advice? I'll run small fft's soon to verify it's a RAM issue and not a cpu settings issue. I'm leaving Nvidia Inspector closed and using GPU-Z to monitor just in case it was isolated.

Thanks,

 

[Goros]

Ok, well 30 seconds after posting this, the system rebooted after completing 672k self test.

Going to start small fft's to see if I can isolate this.

grrr...

 

[Goros]

Just got BSOD Page Fault In Non Paged Area - memory of some kind (weird that it was on small fft's).

Moved the memory down to 1.495 since it was overvolting slightly and is 1.5v memory only slightly overclocked. See if that helps.

 

[Goros]

30 minutes in got a driver_irql_not_less_or_equal.

Gotta be a memory issue, because the cards are both powered down and only running @ rest rates.

 

[Redoitall]

Can't you flash back to the earlier version and re-test.

 

[x2ezx]

Try memtest, to check your memory.

 

[Goros]

Got memtest 86+ cooking now. See what happens.

So fricken weird.

 

[coffeejunkee]

Set everything to stock first and test again. If stable, change everything one step at a time. HCI memtest is good for ram too.

You're running 32GB of overclocked ram without increasing either ram voltage or vccio + you overclock bclk, I'm actually surprised it was stable at all.

 

[tweakboy]

Ok, well 30 seconds after posting this, the system rebooted after completing 672k self test.

Going to start small fft's to see if I can isolate this.

grrr...

Your going to have to up the voltage, that is what a restart means. Do it in tiny increments until your stable. The BIOS update could have messed up the voltage situation for CPU. gl

 

[Ketchup]

Set everything to stock first and test again. If stable, change everything one step at a time. HCI memtest is good for ram too.

You're running 32GB of overclocked ram without increasing either ram voltage or vccio + you overclock bclk, I'm actually surprised it was stable at all.

Yes, when upgrading the BIOS and finding some instability, testing everything as stock should be the first thing to do.

 

[tweakboy]

Well go to your BIOS and make sure all settings are good how you had it before. gl

 

[Goros]

YUP, I shot everything back to stock and it seems to be stable SO FAR.

I was 8 hours OCCT Linpack / 24 hours prime95 x64 / 8 hours OCCT GPU stable on the previous bios with memory voltage at stock (1.5v) 1712 effective MHz (on 1600mhz ram @ 100:100 divider) with pll 1.6v, CPU at 1.22v (4.38ghz) all other voltages auto. GPU's were stable at 940/1880/2400 with some minor tweaks to voltages on the cards.

I manually set the memory as auto was tightening the timings to 9-9-9-24 instead of stock 10-10-10-27.

This board/chip combo is really stable and was very easy to OC with minimal voltage changes. I actually ran a pll of 1.6 until this happened, I'll probably under volt it again.

Since my guess is this is the last bios update, I'll probably try to get my RAM clocked up around 2200/2400, even if I loosen my timings one or two notches on Ivy it seems that MHz>Timing in benchmarks, which is NOT what was seen traditionally.

I hate starting from scratch, but I guess it's a good excuse to push the envelope again. This bios seems to have corrected some of the vdroop and voltage inconsistency to the CPU. Might be able to drop voltage safely after I get the memory stable I'll work on the CPU overclock again.

 

[coffeejunkee]

Well, it could be the new bios just messed something up, that's not unheard of. Or maybe you weren't as stable as you thought and somehow this new bios exposes it, just speculating.

Must say I don't really understand why you didn't get K version, but why don't you leave the bclk as is and just max out the turbo multipliers? IIRC you can go up to 43x, 100MHz extra not worth it messing with the bclk imho. Also I wouldn't bother with the ram oc, it's a waste of time for rl return. Vdroop is perfectly normal too and according to Intel spec, you don't really need llc for such mild oc.

 

[dqniel]

107 bclk is very high.

 

[Goros]

107 bclk is very high.

I know, but by the same token I used 107blck since building the system last year and it was stable.

Windows Experience Index: 7.8
Processor: 7.8
Memory: 7.9
Graphics: 7.9
Gaming Graphics: 7.9
Primary Hard Disk: 7.9
CPU-Z Validation: Validator
PCMark Vantage x64: 25112
PCMark 7: 6166
3DMark Vantage: P43096; X30500
3DMark 11: P13259; X4949
Aida 64:
Memory Read: 19251 MB/s
Memory Write: 21776 MB/s
Memory Copy: 22734 MB/s
Memory Latency: 42.2 ns
CPU Queen: 55381
CPU Photoworxx: 67201
CPU ZLib: 356.7 MB/s
CPU AES: 460460
CPU Hash: 3549 MB/s
FPU VP8: 4371
FPU Julia: 23134
FPU Mandel: 12263
FPU SinJulia: 5908
SiSoft Sandra 2012: 16.350 kPT (#796 Higher than 99.13% ranked results 16.350kPT Excellent Performance )
Dhrystone: 169.21GIPS
Whetstone: 108.62 GFLOPS
Cinebench R10 CPU 29967
Cinebench R11.5 CPU: 8.84
Passmark: 4663.2
CPU Mark: 12544.1
3D Graphics Mark: 4589.8
Memory Mark: 9262.6
Disk Mark: 4673.2

I ran memtest 86+ overnight and on the third pass it showed 16 errors. I upped the voltage on the RAM from the 1.525 it was running at and started over this morning. Once the memory is stable again I'll work on the processor.

Just time consuming.

 

[Goros]

Well, it could be the new bios just messed something up, that's not unheard of. Or maybe you weren't as stable as you thought and somehow this new bios exposes it, just speculating.

24 hours prime95 x64 @85% memory used, 24 hours memtest86+ 4.20, and 8 hours OCCT Linpack stable, but somehow still unstable? Uh...

Must say I don't really understand why you didn't get K version,

K versions were sold out when I built this computer and I missed them 4 different times when they came in stock and sold out same day. Got sick of waiting

but why don't you leave the bclk as is and just max out the turbo multipliers? IIRC you can go up to 43x, 100MHz extra not worth it messing with the bclk imho. Also I wouldn't bother with the ram oc, it's a waste of time for rl return. Vdroop is perfectly normal too and according to Intel spec, you don't really need llc for such mild oc.

Can only go to 41, 42 and 43 can be entered but never goes above x41 in practice.

 

[dqniel]

24 hours prime95 x64 @85% memory used, 24 hours memtest86+ 4.20, and 8 hours OCCT Linpack stable, but somehow still unstable? Uh...

K versions were sold out when I built this computer and I missed them 4 different times when they came in stock and sold out same day. Got sick of waiting



Can only go to 41, 42 and 43 can be entered but never goes above x41 in practice.

It won't go to 42 or 43 even with the new BIOS and when BCLK is set to 100?

 

[Goros]

It won't go to 42 or 43 even with the new BIOS and when BCLK is set to 100?

Correct. Always been that way, it's working as intended. Everyone with a non-k 3770 sees the same exact result.

X43 Multiplier is for one (physical) core only, and the rest would scale down to x39.

X41 is the fastest for all of them.

 

[Goros]

I think I have this thing stable now.

I think.

I changed the divider to 100:133 and was able to get it stable at looser timings (11 11 11 33) at 1996MHz using 1.55v.

Now I' messing with the processor overclock and am (hopefully) on my last prime95 stability test.

Here is a quick overview of my settings:
100:133
X41
107 BLCK

4.386GHz processor
1996MHz RAM

VCC = 1.23 (1.272v under 100% Load, VID recommends 1.271v under 100% Load)
DRAM = 1.5500
VCCSA = .968
VCCIO = 1.07
PLL = 1.6

LLC = All Extreme
CPU Fixed Frequency = 350
DRAM Fixed Frequency = 500

Prime95 x64 started at 11:30AM 2/18/13, 27,756MB Ram (94%)
High temp as of 12:15 - 90c (2 cores)
Average temp as of 12:15 - 76c
Idle temp - 39c

I'll run this for 24 hours and report any failures here. Then I'll run 24 hours of memtest86+ 4.20 to be extra-sure, followed by 8 hours of OCCT linpack.

Then I'll start on the GPU over clocks. Won't get them all done this week, heading out of state on vaca.

Once it's complete I'll post benchmarks here.

 

[coffeejunkee]

24 hours prime95 x64 @85% memory used, 24 hours memtest86+ 4.20, and 8 hours OCCT Linpack stable, but somehow still unstable? Uh...

Uh...yeah. These cpu's are advanced thingies with lots of instruction sets. Most stresstesters use only a fraction of those, meaning under other scenarios it can still be unstable. Did you check Windows event viewer for WHEA errors too?

K versions were sold out when I built this computer and I missed them 4 different times when they came in stock and sold out same day. Got sick of waiting

Figured something like that, but with the money you already invested it wouldn't be too much to sell it and get a K version instead.

Can only go to 41, 42 and 43 can be entered but never goes above x41 in practice.

I know, it's just regular turbo boost on steroids. But still, most games use just 2 threads so you get x43 and if they do use more threads x41 is still pretty good.

 

[Goros]

Uh...yeah. These cpu's are advanced thingies with lots of instruction sets. Most stresstesters use only a fraction of those, meaning under other scenarios it can still be unstable. Did you check Windows event viewer for WHEA errors too?

The last time I overclocked, yes I did. Today I opened it up (thanks for reminding me) and saw 201 Kernel errors - dated 5/14/12 thru 5/25/12 - (20 total when I was setting this this up the first time) and the next ones are on 2/16/13 and 2/17/13 For the other 181, right after loading the new bios.

Figured something like that, but with the money you already invested it wouldn't be too much to sell it and get a K version instead.

yeah I know, but I've been overclocking since you had to solder processors (P2/P3 days) so BLCK overclocking is old hat now, and it does show some decent returns on the benchmark side of things.

I know, it's just regular turbo boost on steroids. But still, most games use just 2 threads so you get x43 and if they do use more threads x41 is still pretty good.

I'll take what I can get. I don't want to swap processors again until "Tock" is ready for release.

Currently prime95 is looking good. A little over 6 hours in and no stopped workers, all cores still within one - two tests of each other (after 6 hours that's better than the last OC) and my max temp hit 92.

 

[Goros]

Just an update, lost stability 2 more times after the last post. 3-6 hours in for both...adjusted the vccsa up to 1.0 and put the vccio to 1.08xxxx

Adjusted some voltages and brought the processor frequency down to 300, and shed 3-4 degrees at max load.

Ran memtest 86+ for 12 hours (4-1/2 passes) with zero errors so I think I'm good there. Once I get prime95 x64 at 95% memory to pass for 24 hours I'll go back and run a 24 hour memtest just to double check that it's 100%.

Then comes linpack for 8 hours, and then OC on the 580's.

Will keep updating till I leave for vaca, and finish when I get home.