Lian, it may be a good idea to explore the power supply issue. Your has 2 12V rails, 25A each and may not be giving you all the needed power.
I think one with a single 12V rail is superior, such as a Corsair HX650 or AX650 and provide 52A and 54A on the +12V.
The OCZ ModXStream has been known to fail after a short time as well.
Agree with Britton30, those voltage graphs don't look good.
Check your warranty, hopefully you can return it and get a better PSU.
PSU REVIEW DATABASE
Seasonic are very good PSUs and are the supplier for most of the Corsair HX and AX series of PSUs.
As you can see in our system specs we both have the Corsair HX650W, they are very good and Corsair has a great warrany service.
I'm not sure I did it right. The test failed a few times and I changed the settings until it passed for a first time, but the graphs look the same as before with voltages varying greatly during idle times. CPU Load Line calibration is set to level 5, load xmp settings changed from auto to X.M.P 1.2, and additional turbo V changed from auto to +0.035V
Max vcore was around 1.17v and max temps were (69,74,74,72).
edit: I did another test to see if it would go through all the way (same settings as first) and it failed after 45 secs. I uploaded that too. Not sure how high to put the additional turbo v but I'll try higher now.
edit2: Still failing at +0.203v, graphs look same as before. Vcore tops out around 1.15v and sticks around 1.1v-1.11v
edit3: failing at +0.301v, graphs look the same. I'm going to wait for an update because I don't really understand why it's continuing to fail.
Could it be a faulty psu? I thought this thing had enough power for what I'm doing, I know others have used this same processor with it without issue.
Last edited by lian9; 26 Jun 2012 at 21:17.
Sorry for the wall of text
The max actual vcore reported by OCCT or CPU-Z you'd want to see is 1.37-1.38v. Realistically you will hit a thermal wall long before that.
But you do have plenty of voltage 'wiggle room'.
Always a possibility, some of your previous idle fluctuations are interesting.Could it be a faulty psu? I thought this thing had enough power for what I'm doing, I know others have used this same processor with it without issue.
However, results from software based apps shouldn't be taken as gospel. At best they should be considered a guide, since the only way to get an accurate PSU volt reading is using a tester.
And despite not being able to check load fluctuations - the next best reporting is what's reported in the BIOS.
What is the BIOS reporting your volts as?
Some fluctuation is normal - but if they are fluctuating wildly, which one? (ie 12v, 3v etc)
***
Since the 24, 4/8 mobo pin are on the 12v1 Rail, your card should be on the 12v2 rail, which are the Red connectors.
Spec wise, the PSU is fine and is adequate for your card/setup. The 7850 amp recommendation is 24 and peripherals (HDD's etc) use bugger all.
But if it's running 'out of spec' (broken) - that is indeed one possible cause for the sporadic issues you've been having, since a spike or dip can occur at any moment causing a crash.
But it's best to to cover all angles first before having you buy a new PSU 'on the off chance' only to discover a new PSU makes no difference and the problem is something else.
For example it could be a faulty board, the CPU isn't seated 'quite right' (not making full and proper pin/pad contact) or damage like bent pins has occurred to the socket area.
(Of course you are welcome to buy a new PSU if you wish and are willing to take the 'off chance'. I'd personally opt for the Corsair AX650W - but there other good alternatives)
Just some basic double check questions:
Are you using the 8 pin CPU power plug, or the 4 pin plug? And is it firmly secure? Ideally the 8 pin plug should be used.
The same with the 24 pin mobo connector. All locked into place?
How have you got your 7850 connected? - Is it connected using the 6 or 6+2 PCI-E cable connected to one of the the Red plugs?
What you can try next:
It might be time to change tack and actually 'overclock' the system. Eliminate Turbo all together.
But it will be a 'very mild overclock'. 3.6ghz-3.8ghz across all cores, instead of 3.8ghz for just one. It will also be achievable with stock cooling.
It looks like the 1.7 BIOS has been pulled and is no longer available. The latest is now 1.8. - The first step would be to update to that straight away. Normally when a BIOS is pulled and no longer available, a fault has been found.
ASRock > Products > Motherboard > Z77 Extreme4 > Download
(The manual for your board isn't the most informative, so I'm working from mostly 'general' settings. But most of these settings should be found in the OC Tweaker section of your BIOS
►Put the Turbo voltage back to auto.
►BCLK should be left at AUTO or 100.00
►Adjust the multiplier to 38 - this 'should' grey out/disable Turbo. If not, set Intel Turbo Boost Technology to Disabled.
►Leave the LLC at 5, or max settings for now.
► Leave all Power saving settings like Speedstep, EIST etc Enabled.
►CPU Voltage - use a +offset.
ie +0.040 is a good starting place. Check with CPU-Z what the reported voltage is. Depending on temps/crashes etc, you can adjust this up or down. Most likely it may have to go down.
(Just for reference, with my 3770K I use a +0.85 offset with a multi of 47. This gets me a 4.7ghz OC with an actual vcore usage of ~1.28v. Since you are aiming for a much lower OC - you will not need a vcore this high. Tbh, you shouldn't even need to go above to 1.2v. So if you see the voltage getting this high, lower the offset.
*Offset works by taking your CPU's VID (voltage identification) that is factory set to 'ensure' that individual chip runs at the specified stock clock speed. ie 3.4ghz. Using a positive offset + increasing the multiplier, adds extra voltage to the VID to make the CPU run at the new clock speed ie 3.8ghz. Same idea behind increasing the Turbo voltage.)
All other settings like VTT, PCH, CPU PLL, VCCSSA, Spread Spectrum etc should be left on AUTO.
The fun bit
► Start Testing again. Under load, all your cores should now be 3.8ghz. Keep an eye on your temps. For personal comfort sake, aim for low 90c-mid 90c as your 'max temp limit'.
You may have to fiddle with the offset a bit.
If you start passing OCCT without problems, consider a longer stress test with Prime 95 27.7, or test 'real world usage' like games.
If you still experince OCCT core failures or game crashes - it might be time to look at either re-seating your CPU (checking for bent pins when you do) or replacing your PSU.
Hopefully this gives you something to work with.
Your voltage fluctuations are concerning. I agree with smarteyeball you need to test your CPU to be sure that a setting isn't causing the issues. Check all the power connections, CPU for bent pins, etc. these types of issues can be hard to nail down. If you know someone with a known working PSU you can test with it will help narrow down the possibilities.
AsRock has removed the 1.7 BIOS version, update yours to the 1.8 version, as smarteyeball mentioned it usually means there is an issue with it.
Since your build is only about a month old, check with your supplier about the return policies, this will be good to know in case you need it.
Thanks!
Updated bios to 1.8. My graphics card uses both PCIe cables (Newegg.com - XFX Double D FX-785A-CDFC Radeon HD 7850 2GB 256-bit GDDR5 PCI Express 3.0 x16 HDCP Ready CrossFireX Support Video Card), I am using the 8 and 24 mobo pins and they are securely attached. I took out the CPU and the pins were very small but I didn't notice any that were bent (hopefully didn't miss any). I reinserted it and did another OCCT test which failed. I had followed your overclocking instructions. The Vcore was getting to about 1.2v with +0.030 and 1.9 with +0.020. One of the tests failed due to a core going over the max of 86C, but it also failed due to core errors as before.
In the BIOS the voltages are:
Vcore - This was constantly fluctuating (several times each second), 1.048 seemed to be most common and the midpoint, it would go down to 1.040 and often up to 1.056, the highest I saw it go up to was 1.072.
+12V - 12.355V no fluctuations
+5V - 5.064V no fluctuations
+3.3V - 3.376V no fluctuations
I got the AX650 that you recommended but before I put it in I want to make sure I add more details:
I haven't tested games a lot recently, but I played diablo again for about ~30 mins a few times yesterday, before the overclock, and did not get any blue screens (but I can't say I won't anymore). However, at one point, maybe after 20mins, the game/computer simply froze and I had to restart the computer.
In addition, the game will occasionally freeze for ~5seconds, I minimize it, and then a tab shows up saying "AMD driver has stopped responding and has successfully recovered" **This has actually been happening sporadically ever since I got the game about two weeks ago, sorry for forgetting to mention it. It happens maybe 3 times an hour, but that's an estimate and it's not consistent. There have been times when it hasn't happened at all I'm not sure if this is an issue with the graphics card, voltage/psu issues as mentioned before, diablo 3 itself, etc.
Also, the cinematics run choppy.
Let me know if this post changes anything.
Last edited by lian9; 29 Jun 2012 at 00:14.
Good. It never hurts to double check.
As a general rule, they are relatively easy to spot if bent. So for now, that's another thing we can scratch from the list.I took out the CPU and the pins were very small but I didn't notice any that were bent (hopefully didn't miss any).
1.9? I presume there was a number missing there?I reinserted it and did another OCCT test which failed. I had followed your overclocking instructions. The Vcore was getting to about 1.2v with +0.030 and 1.9 with +0.020. One of the tests failed due to a core going over the max of 86C, but it also failed due to core errors as before.
Is it the same core failing, or different cores each time?
Although it doesn't really matter as it would be highly unlikely for the CPU itself to have a problem. (It's pretty damn rare to have a faulty CPU. 99% of the time it's always another component)
All good, all normal. Vcore fluctuation is normal. The PSU volts are within spec.In the BIOS the voltages are:
Vcore - This was constantly fluctuating (several times each second), 1.048 seemed to be most common and the midpoint, it would go down to 1.040 and often up to 1.056, the highest I saw it go up to was 1.072.
+12V - 12.355V no fluctuations
+5V - 5.064V no fluctuations
+3.3V - 3.376V no fluctuations
(However as mentioned, it doesn't measure load usage, since there is no load so to speak.
Good choice sir :)I got the AX650 that you recommended but before I put it in I want to make sure I add more details:
A freeze is good as a BSOD - meaning that it's still highlighting a h/w issue.I haven't tested games a lot recently, but I played diablo again for about ~30 mins a few times yesterday, before the overclock, and did not get any blue screens (but I can't say I won't anymore). However, at one point, maybe after 20mins, the game/computer simply froze and I had to restart the computer.
That's okay - it gives us more to work with.In addition, the game will occasionally freeze for ~5seconds, I minimize it, and then a tab shows up saying "AMD driver has stopped responding and has successfully recovered" **This has actually been happening sporadically ever since I got the game about two weeks ago, sorry for forgetting to mention it. It happens maybe 3 times an hour, but that's an estimate and it's not consistent.
There are multiple reasons for the display driver to stop working, such as:
Bad driver / game incompatibility (software based essentially. Unlikely that what's what were are dealing with)
Too high an overclock on a video card
A h/w fault with the video card itself.
PSU not delivering enough power to sustain the card under load.
etc
It could be part of the underlying cause, or something completely separate. Focus on that later.Also, the cinematics run choppy.
It does. Extra info is always useful.Let me know if this post changes anything.
***
If this was just a bad video card, the CPU shouldn't be failing OCCT at AUTO settings, adjusted settings etc. It doesn't mean the card is 100% in the clear - but it does put it down the list.
(In the meantime - smell the card, especially around the PCI-E plugs. What you're 'smelling for' is a burnt electrical/burnt plastic smell. Sometimes a bad card can cause an damaging electrical feedback which can have a flow on affect on other components.)
Before you swap the PSU:
And to eliminate one more variable during the testing process
► Remove the 7850, adjust the BIOS to run off the iGPU (integrated GPU) and run your display from the iGPU. This will both reduce any excessive PSU load and eliminate the 7850 from the equation for the time being.
(This method turns it into a potential bad motherboard or PSU trouble
shoot. Don't forget to install the iGPU drivers so you can still play D3).
► Run the CPU at stock 3.4ghz with Turbo disabled. 34 multi, or simply disable the Turbo option. All that matters is that it's a flat stock 3.4ghz
For expediency, use a manual vcore. Use a relatively high vcore like 1.2v-1.25v. This will eliminate any questions regarding insufficient voltage causing the crashes. LLC at 5. All other options to AUTO
►Test again with OCCT.
If you pass - awesome. The CPU and motherboard are most likely fine. Without the 7850, start adding your 'daily' CPU settings. ie Turbo etc
If it fails at stock settings, replace the PSU
►Re-run the test again with the last failed/unchanged BIOS settings and new PSU. ie The stock 3.4 no Turbo Boost.
*If it passes - Add the 7850. Re-Run/game etc
*If it fails with the new PSU and no 7850 - RMA/return the board. It's a dud.
*If it passes with the new PSU but fails with the 7850, RMA the card.
*If it passes with both the new PSU and 7850 installed - time to start re-instating all your daily settings :)
Basically, start off stripped with as little variables as possible. Success or failure should determine the next step. If you need any further clarification - just ask
Quote