I give up [i7 920]

the way i learnt was set vtt to 1.4 then work down from there when you reach your dEsired clock

i'd up your vcore a notch on the fault
 
Huh ... this time got a system service exception BSOD after two days uptime (after nothcing vcore up). That's new. Rebooted, system froze in Firefox :\. Vcore this time too? Just bumped it some more, up to 1.1875
 
Last edited:
yha, you have a long while before too much vcore. Bump it a bit more. Watch your temps.

Are you rjunning any vdroop control? LLC on in bios?
 
LLC is OFF. Also, as I increase vcore I notice my temperatures under load DROPPING. Is this normal? 0.o I was loading 66-67C, now I'm loading 62.

EDIT: I'm starting to get strange new BSODs and actual errors in benchmarking programs in place of BSODs. Combined with less BSOD-less hangs, I think I might be getting closer.
 
Last edited:
Well your paste could be settling in.

Bump vcore a notch, or the ioh core might need a bit too.... I think its the ioh core.
 
Patonb said:
Well your paste could be settling in.

Bump vcore a notch, or the ioh core might need a bit too.... I think its the ioh core.
Click to expand...

What's a good setting? My IOH is currently 1.14. Also, sometimes while gaming my onboard NIC disconnects and I have to disable / re-enable the connection :\. It's been more common lately, but has happened since day one. Related maybe?
 
Get everything stable on stock clocks first and then try overclocking. Doing otherwise complicates the troubleshooting process significantly.

Stock speeds/volts would be a good place to start. Set the RAM divider to run the ram at 1066 with timings 9-9-9-27 and figure out if your cpu can handle stock speeds at stock volts. Run it this way for several days to figure out how often, if at all, it crashes. Change ONE thing at a time and run it for a few days taking note of the # of crashes. Eventually you'll have results, just takes patience.
 
I have a single GPU.

Here are my current voltages after having some linpack errors w/o BSOD.

154jdrd.jpg


errors, second run. First produced errors in 8 minutes.

2sbmydg.png


earlier that day I was able to do this though 0_o. Flops are lower due to web browsing during test.

fl97wl.png
 
Hum ... system hung, looped sound forever, and turned off the keyboard / mouse during GTA IV. Rebooted, started linpack, and it immediatelyrestarted with no BSOD within 20 seconds. Gonna bump IOH to 1.2 now ...
 
Maybe up yhour qpi volts... you have 6 sticks which is hard on the controller.
 
I've upped QPi to 1.325 and vcore to 1.225, as well as DRAM to 1.66 from 1.64. I seriously can't get past the first LinX test without an error anymore, so these settings are still no good.

EDIT: Vcore chart from OCCT

sqggg8.png


EDIT2: More :\

16bhb29.png


EDIT: Prime95 blend also found errors pretty quickly (worker crashed due to rounding error). Two hours of small FFT did nothing. Ran another blend and a different worker (different core too) failed with the exact same error (expected less than 0.4, got 0.5.). More vtt? 0.o. I'm already at 1.325 ...
 
Last edited:
I'd say more Vcore.

Like I said, set 1.4 vtt, then work down after getting a stable overclock.
 
Alright. Now ... this is disturbing. When I run some tests .. a NOISE comes from my case. I can't seem to tell exactly where, but it seems to be coming from the cpu itself and not the HDD, DIMMS, or PSU. If I run Linpack with 16MB memory it goes all rapid fire. If I run Prime large FFT / Blend I hear it but not if I run small FFT. If I run 7GB set in linpack I don't hear anything, but when stopping linpack hear kind of a squeal ... I just now noticed this because normally my case window keeps almost all noise inside. It definately seems to be coming from the actual CPU, at least when I get close to the case.
 
Sounds kinda like this.

Prime large fft -- crickets.

Prime blend - constant whirr.

Linpack 16MB -- crickets, but not as frequent as large fft.

EDIT: The noise just shifted in Prime blend, followed immediately by Firefox crashing.
 
What fan is on the heatsink?

I'd shut down and check for anything around the cpu thats odd looking.
 
H50. No noise coming from it other than the normal pump swish. I looked with a flashlight around the CPU and don't see any charred MOSFETs or popped caps. Still disturbing. I dialed VTT down to 1.30 again and upped the vcore in its place, so far so good in prime blend, much longer than the previous settings, which is really weird since that's the most vtt sensitive test in prime95 and I didn't have issues with the pure CPU test (small fft). Perhaps the noise is because I set my IOH to 1.20 instead of the 1.10 default?

EDIT: My temps are weird now too. Typically my CPU temp (tC) is 5-10C lower than the lowest core temp, right now it is 2C higher than the lowest core temp.

EDIT 2: If this was the vcore, it is currently only a single increment above the setting that produced errors quickly, it will likely need to be bumped one or two more times.

EDIT 3: Still no errors after almost an hour, previously thery happened immediately, but HWmon is a little weird atm.

6tjuhg.png


That Auxfan0 is the H50 pump. Those RPMs are not normal :D. It's registered 0 since day one from time to time, but I always assumed that to be a simple monitoring bug. I think I've seen the 10k rpm before too, but idunno. The CPUTIN is the real concern. That's tcase running at 73C. The highest I have ever seen this is 68, and that was in linpack, not Prime. Linpack makes my chip significantly hotter than prime does (5C easily). It is also typically 6-10C below any core temp. Here we can see it is above cores 1 and 3. I have never seen it within 5C of ANY core temp since I bought the chip. The Amps and Watts used by the CPU are the same as those shown when Tcase was below 70 in linpack. After stopping prime, it is currently at about 50. Normally this idles in the mid 30s, and is in the mid 40s by now.
 
Last edited:
Man this thread is full of fail.
1. You have 6 sticks of memory populating your board. That's double the stress on your IMC. With that density of memory, you're looking at 1.40v+ vQPI to get memory stability. MINIMUM.
2. For 6 sticks, you're also looking at additional IOH voltage - in the range of 1.2-1.25v.
3. Stock voltage on your i7 920 is probably between 1.2 and 1.25v. 4GHz will require 1.3 on even the best cpus with wimpy single radiator cooling like the H50.
4. That noise is called coil whine. Many motherboards, and power supplies will make these noises when they're stressed and this is normal. If your components didn't make this sound previously, it's a sign that your power circuitry for that component is failing(dried caps, polarized coils).
5. 10 Linx loops is not enough. You need to do at least 100 loops. Even after 100 loops I've seen my PC fail.
6. You should have LLC enabled for i7 series processors.

Edit: That tech reaction guide is pretty bad.
Try this one instead:
http://www.techpowerup.com/forums/showthread.php?t=108526

This template isn't for your board, but is for a similar configuration(930 with 3x4gb sticks)

Code: 
Target CPU Frequency                  4200MHz
Target DRAM Frequency                 1683MHz

LN2 Mode                              [Disabled]***
QPI Loadline Calibration              [Enabled]***

Sync Mode                             [Enabled]
AI Overclock Tuner                    [Manual]
OC From CPU level Up                  [Auto]
CPU Ratio Setting                     [20]
CPU Turbo Power Limit                 [Disabled]

>CPU Configuration----------------------------
CPU Ratio Setting                     [20]
C1E Support                           [Enabled]
Hardware Prefetcher                   [Enabled]
Adjacent Cache Line Prefetcher        [Enabled]
MPS and ACPI MADT ordering            [Modern ordering]
Intel(R) Virtualiyation Tech          [Enabled]
CPU TM Function                       [Enabled]
Execute Disable Bit                   [Enabled]
Intel(R) HT Technology                [Enabled]
Active Processor Cores                [All]
A20M                                  [Disabled]
Intel(R) SpeedStep(TM) Tech           [Enabled]
Intel(R) Turbo Mode Tech              [Disabled]
Intel(R) C-STATE Tech                 [Enabled]
C State package limit setting         [Enabled]
----------------------------------------------

BCLK Frequency                        [210]
PCIE Frequency                        [102]
DRAM Frequency                        [DDR3-1683MHz]
UCLK Frequency                        [3365MHz]
QPI Link Data Rate                    [7572MT/s]

Memory Configuration Protect          [Disabled]

>DRAM Timing Control--------------------------
CAS# Latency                          [7]
RAS# to CAS# Delay                    [8]
RAS# PRE Time                         [7]
RAS# ACT Time                         [20]
RAS# to RAS# Delay                    [Auto]
REF Cycle Time                        [88]
WRITE Recovery Time                   [Auto]
READ to PRE Time                      [Auto]
FOUR ACT WIN Time                     [Auto]
Back-To-BackCAS# Delay                [Auto]

Timing Mode                           [1N]
Round Trip Latency on CHA             [Auto]
Round Trip Latency on CHB             [Auto]
Round Trip Latency on CHC             [Auto]

WRITE To READ Delay(DD)               [Auto]
WRITE To READ Delay(DR)               [Auto]
WRITE To READ Delay(SR)               [Auto]
READ To WRITE Delay(DD)               [Auto]
READ To WRITE Delay(DR)               [Auto]
READ To WRITE Delay(SR)               [Auto]
READ To READ Delay(DD)                [Auto]
READ To READ Delay(DR)                [Auto]
READ To READ Delay(SR)                [Auto]
WRITE To WRITE Delay(DD)              [Auto]
WRITE To WRITE Delay(DR)              [Auto]
WRITE To WRITE Delay(SR)              [Auto]
----------------------------------------------

CPU Differential Amplitude            [Auto]
CPU Clock Skew                        [Auto]
IOH Clock Skew                        [Auto]
------------ Extreme Engine Digi+ ------------
Digi+ PWR Mode                        [T-Balanced]
PWR Volt. Control                     [Auto]
Load-Line Calibration                 [Full Calibration]
CPU Voltege OCP                       [Enabled]
CPU PWM Frequency                     [1000KHz]

Extreme OV                            [Disabled]
Extreme OC                            [Auto]

CPU Voltage                           [1.38125]
CPU PLL Voltage                       [Auto]
QPI/DRAM Core Voltage                 [1.48750]
DRAM Bus Voltage                      [1.65625]

>DRAM REF Voltages----------------------------
DRAM DATA REF Voltage on CHA          [Auto]
DRAM CTRL REF Voltage on CHA          [Auto]
DRAM DATA REF Voltage on CHB          [Auto]
DRAM CTRL REF Voltage on CHB          [Auto]
DRAM DATA REF Voltage on CHC          [Auto]
DRAM CTRL REF Voltage on CHC          [Auto]
----------------------------------------------

IOH Voltage                           [1.20525]
IOH PCIE Voltage                      [Auto]

ICH Voltage                           [Auto]
ICH PCIE Voltage                      [Auto]

---------- Spread Spectrum Control -----------
CPU Spread Spectrum                   [Disabled]
DRAM Spread Spectrum                  [Disabled]



***on-board jumper dependant
 
Last edited:
I figured the 6sticks were gonna be a source of issue. I've never dealt with that much ram.

I use F@H bigadv to dial in my oc, after 20-30min of LinX.
 
Alright, I'll keep moving QPI up then. I turned LLC on and dipped vcore back down to 1.21 for starters, and moved QPI to 1.34V. The regular CPU temp is back down now ... which is weird. Cores still push 80, but now the entire CPU die doesn't follow them.
 
Don't forget to use the all button on linx. it makes the loops a lot longer and stresses all of your memory.
 
Zoson said:
Don't forget to use the all button on linx. it makes the loops a lot longer and stresses all of your memory.
Click to expand...

The all button never works for me, just spits out an error about running out and cancels. I use the preset closest to how much I have, works well enough. 9G free = 8GB problem set. I still get errors, and my core temps are starting to push to 80 in LinX :\. Bear in mind I'm not looking for 4 GHZ I'm happy gunning down 3.2 or 3.6. My case doesn't have the cooling for 4 as it is.
 
myounage said:
The all button never works for me, just spits out an error about running out and cancels. I use the preset closest to how much I have, works well enough. 9G free = 8GB problem set. I still get errors, and my core temps are starting to push to 80 in LinX :\. Bear in mind I'm not looking for 4 GHZ I'm happy gunning down 3.2 or 3.6. My case doesn't have the cooling for 4 as it is.
Click to expand...

If you have 12gb ram and can only run 9gb on the test, you're not doing a thorough test. You need to free up the extra memory. Otherwise the consumed parts won't be stressed. If the 'slowest' part of your memory happens to be in that last 3gb, it will never get tested and you'll get random crashes later on during normal use.

Remember, 'safe' vQPI is up to 1.45v according to intel's spec for 45nm cpus. You WILL need a LOT of QPI to run your memory correctly. More than 1.4875v for sure since you're 6x2gb sticks instead of 3x4gb.
 
Zoson said:
If you have 12gb ram and can only run 9gb on the test, you're not doing a thorough test. You need to free up the extra memory. Otherwise the consumed parts won't be stressed. If the 'slowest' part of your memory happens to be in that last 3gb, it will never get tested and you'll get random crashes later on during normal use.

Remember, 'safe' vQPI is up to 1.45v according to intel's spec for 45nm cpus. You WILL need a LOT of QPI to run your memory correctly. More than 1.4875v for sure since you're 6x2gb sticks instead of 3x4gb.
Click to expand...

Goddamn .. my temps are already getting too high too. 69C surface temp during linpack. I'm not sure 1.4xx is even an option, I'm at 1.35 right now.
 
1.40, still an error after 38 minutes. I've actually been able to dial vcore all the way down to 1.18 without any changes to my (in)stability.
 
I'm telling you... 1.4 is just not enough for that amount of ram.
 
If you have no reason to have 12GB then you are just fighting yourself having 6 sticks installed and trying to OC. If you can't get that point then you really have no business even asking for help.
 
myounage said:
Goddamn .. my temps are already getting too high too. 69C surface temp during linpack. I'm not sure 1.4xx is even an option, I'm at 1.35 right now.
Click to expand...
69C CPU temp sounds good to me with linpack (same as IBT?); RealTemp temperatures?

Sell your RAM and go with 3×4GB if you want 12GBs. I'm at 1.35V on QPI with 3×2GB.

You can do it, BRO.
 
djoye said:
69C CPU temp sounds good to me with linpack (same as IBT?); RealTemp temperatures?

Sell your RAM and go with 3×4GB if you want 12GBs. I'm at 1.35V on QPI with 3×2GB.

You can do it, BRO.
Click to expand...

He's talking 69C tCase, not 69C core temp. 69C is the limit for tCase, and is indicative of significantly higher core temps.

1.35v is totally normal for 3x2GB modules. It's upping the density of the memory available to the IMC that's the problem. 12GB 3x4GB and 12GB 6x2GB is very similar as far as voltage requirements.

From what I've seen, and what I've experienced myself, 45nm bloomfields need 1.4875v to run 3x 4GB sticks. I've owned two i7 930's and four i7 950's. The best chips I had ran my mem at 1.4875v, and the worst of them needed 1.56v!!!

With 6x2GB sticks I would expect to be in the 1.5-1.55v range to get proper stability.
 
I run 6x2GB at stock speeds (1600) with a QPI of 1.3v. Maybe you need more for overclocking the RAM, but I have a completely stable setup with only 1.3v. I think the OP needs to put his vcore up as people have been telling him since page 1 of this thread.
 
weuntouchable said:
I run 6x2GB at stock speeds (1600) with a QPI of 1.3v. Maybe you need more for overclocking the RAM, but I have a completely stable setup with only 1.3v. I think the OP needs to put his vcore up as people have been telling him since page 1 of this thread.
Click to expand...

Your experience is kind of atypical for bloomfields really... How many chips have you owned? Every single one of my 930 and 950 D0's needed a lot of QPI on my R3E to get the 12GB kit running.

My 32nm Gulftown needs 1.3375v QPI for the same 12GB kit. But that's because UCLK is 1.5x instead of 2x. If i boost my UCLK to 2x, I need 1.4875v again.
 
You need 1.30~36 to get a oc of 4.0ish... So if its 1.3 for 12Gigs and 1.35 for an OC, then thats gonna be a overly high setting for an OC and 12Gig.
 
Alright, I downed QPI all the way back to 1.35, and upped vcore to 1.23 in its place and it is more stable than it was when vcore was 1.20 and QPI was 1.40. I'm thinking it actually is vcore at this point, as I have not had any crashes and linpack / prime run quite a bit longer before picking out any errors (These setting changes have garnered me about an extra 30 mins in each before any errors happen).

EDIT: Also, wouldn't too low qpi/vtt produce errors in memtest86+? Bear in mind my OC is not 4GHZ or anything crazy like that. It's just 160 base clock which is 3.2GHZ at 20.0x core mult. I also increased my uncore mult by 1 past the minimum, which I heard can help with stability and it seems to have helped, though I can't be sure.
 
Last edited:
Patonb said:
I figured the 6sticks were gonna be a source of issue. I've never dealt with that much ram..
Click to expand...

Yup. Remember mem controller on die with Bloomfield. Try running 6GB, reducing stress on mem controller. One of my 920s didn't like 6 sticks, had to go with 3x4GB.
 
bitbum said:
Yup. Remember mem controller on die with Bloomfield. Try running 6GB, reducing stress on mem controller. One of my 920s didn't like 6 sticks, had to go with 3x4GB.
Click to expand...

this is exactly what i'm saying, with 6 sticks you need loads of QPI to get full stability. 1.35v is just not enough when you're populating all slots.... and like I said, with my 3x4gb kit I still needed 1.4875v qpi to run my kit.

Gulftown changed that nicely though.
 
Alright, might take it higher. But what about this shutting down and restarting nonsense? What's that indicative of? That only started just recently. Before if my QPI was low I would get those hangs or BCC 124 BSODs. Now I don't even get errors in Linpack anymore, but after a while the system will just reboot. Before I raised vcore I would get errors in around 20-30 minutes. Last night it ran for over an hour before the system just rebooted. Possible PSU or the UPS it's hooked up to? Motherboard? This has never happened before, and I'm starting to think it might not be a symptom of voltage but something else entirely.
 
Alright, the more I test the more I'm convinced the current settings are fine. The system hangs without BSOD have been pinpointed as QPI, and I have not had a single one at any QPI setting above 1.37. It is currently sitting at 1.40 after getting an 0x050 BSOD (page fault in non-paged area) at 1.38, but I still get the problem where the system restarts itself suddenly, at a rate of once every two days or so in normal use, and within an hour or an hour and a half of stress testing. I'm starting to become convinced that this is caused by actual hardware failure, and not settings, as it never happened before now, but the system halts without BSOD did. Now those are gone completely thanks to QPI tweaking, and I no longer get any errors in linpack or prime after increasing vcore, just sudden restarts. Perhaps the IMC is overheating and triggering thermal shutdown? Also, one of the lights on the front of my case has begun to flicker sporadically. A precursor for power supply death maybe? In addition, my UPS is now several years old and can no longer handle even a momentary lapse in power.
 
You must log in or register to reply here.
Back
Top