Jump to content
Corsair Community

Gigabyte Z87X-UD4H and Corsair CMT16GX3M4X2133C9 - intermittent BSOD/app crashes


Cimexus

Recommended Posts

Hi all,

 

I'm running a Gigabyte Z87X-UD4H motherboard and the Corsair CMT16GX3M4X2133C9 kit (4x4 GB modules).

 

This memory is advertised and rated as running at 2133 Mhz with 9-11-10-27 timings. The only setting in the BIOS I adjusted was setting the XMP to 'Profile 1'. Haven't touched anything else whatsoever. The XMP profile appears to work correctly, as CPU-Z reports 2133 Mhz 9-11-10-27 as expected.

 

I am running the latest version of the motherboard BIOS (F7).

 

My CPU is a Haswell 4770K running completely at stock.

 

Since assembling this machine I've been having intermittent blue screens (roughly 1-2 per day). There's no pattern to them - they can happen at any time, at load or idle, in a game or just sitting at the desktop doing nothing. The BSOD reports show a different driver or file at fault every time ... I've seen everything from DirectX drivers to tcpip.sys to random kernel faults.

 

I've been trying to isolate the problem component for a while (these kind of issues are usually PSU or memory) and I'm fairly certain it's the memory. If I load optimized defaults in the BIOS, which downclocks the RAM to 1333, it seems stable (runs several days without a BSOD). Then I set it back to the XMP profile, and the crashes returned (had a couple later that same day). They are infrequent enough to make it hard to diagnose, but frequent enough to be very annoying.

 

Anyway, the first thing I turned my attention to was DRAM voltage. This is supposed to be 1.5V for this kit, and sure enough, the XMP profile (and all three JEDEC profiles) state 1.5V as the voltage (in CPU-Z). One thing I noticed however, was that, both when running at default settings (1333 Mhz) and when using the XMP profile, was that the actual DRAM voltage (as reported by the BIOS) is slightly lower: 1.488 V.

 

Interestingly, if I manually set the DRAM voltage in BIOS (previously set to use the XMP-specified voltage) to 1.5, it's still reported as 1.488V. This got me wondering whether it needed a bit more.

 

Increasing the manual voltage to 1.510 produced the following results:

 

- BIOS reports actual DRAM voltage as being exactly 1.5 V

- Bluescreens seem a bit less frequent. One every couple of days instead of several per day.

 

My question is: am I on the right track here with increasing voltage? Would I be safe in upping the voltage further above the 1.5V specification (maybe to 1.52, which would result in an actual voltage of ~1.51 given that it always seems to report around a tenth of a volt less than what you set). I'm not an overclocker and a bit of a noob when it comes to this stuff ... I just want to get this memory running stably at the rated/advertised speed without BSODS :)

 

PS. The memory itself isn't faulty, I don't think. Even clocked up at 2133, it runs through several Memtest86+ passes without a failure. It's just in Windows that it seems to intermittently BSOD.

Link to comment
Share on other sites

Hi all,

 

I'm running a Gigabyte Z87X-UD4H motherboard and the Corsair CMT16GX3M4X2133C9 kit (4x4 GB modules).

 

This memory is advertised and rated as running at 2133 Mhz with 9-11-10-27 timings. The only setting in the BIOS I adjusted was setting the XMP to 'Profile 1'. Haven't touched anything else whatsoever. The XMP profile appears to work correctly, as CPU-Z reports 2133 Mhz 9-11-10-27 as expected.

 

I am running the latest version of the motherboard BIOS (F7).

 

My CPU is a Haswell 4770K running completely at stock.

 

Since assembling this machine I've been having intermittent blue screens (roughly 1-2 per day). There's no pattern to them - they can happen at any time, at load or idle, in a game or just sitting at the desktop doing nothing. The BSOD reports show a different driver or file at fault every time ... I've seen everything from DirectX drivers to tcpip.sys to random kernel faults.

 

I've been trying to isolate the problem component for a while (these kind of issues are usually PSU or memory) and I'm fairly certain it's the memory. If I load optimized defaults in the BIOS, which downclocks the RAM to 1333, it seems stable (runs several days without a BSOD). Then I set it back to the XMP profile, and the crashes returned (had a couple later that same day). They are infrequent enough to make it hard to diagnose, but frequent enough to be very annoying.

 

Anyway, the first thing I turned my attention to was DRAM voltage. This is supposed to be 1.5V for this kit, and sure enough, the XMP profile (and all three JEDEC profiles) state 1.5V as the voltage (in CPU-Z). One thing I noticed however, was that, both when running at default settings (1333 Mhz) and when using the XMP profile, was that the actual DRAM voltage (as reported by the BIOS) is slightly lower: 1.488 V.

 

Interestingly, if I manually set the DRAM voltage in BIOS (previously set to use the XMP-specified voltage) to 1.5, it's still reported as 1.488V. This got me wondering whether it needed a bit more.

 

Increasing the manual voltage to 1.510 produced the following results:

 

- BIOS reports actual DRAM voltage as being exactly 1.5 V

- Bluescreens seem a bit less frequent. One every couple of days instead of several per day.

 

My question is: am I on the right track here with increasing voltage? Would I be safe in upping the voltage further above the 1.5V specification (maybe to 1.52, which would result in an actual voltage of ~1.51 given that it always seems to report around a tenth of a volt less than what you set). I'm not an overclocker and a bit of a noob when it comes to this stuff ... I just want to get this memory running stably at the rated/advertised speed without BSODS :)

 

PS. The memory itself isn't faulty, I don't think. Even clocked up at 2133, it runs through several Memtest86+ passes without a failure. It's just in Windows that it seems to intermittently BSOD.

 

Okay, you might think this is wierd, but this came from Gigabyte. I had the exact same issue as you do, but with a GA-Z87X-UD3h board. Go back into the Bios and press F2 to go into the old style Bios screen. In the memory tab, set your XMP profile to 1 again, ensure everything else is at Auto. It's wierd ,but the HD panel view or UEFI bios screen is not recognizing the RAM speed correctly, even though it says it is. Then save and exoit in the old style BIOS. I also am using BIOS F7. My system ran 14 days without a BSOD. i had only one after that. Look in your evet viewer and see the cause. Mine was always Kernal Power Error with a bugcheck code of 292.

Link to comment
Share on other sites

Oh, in the back of your motherboard manual is a support contact number, you can call them and they can walk you thorugh it. I had to do this, because I was like you, getting annoyed when I knew everything was working as it should. The bugcheck code 292 or in hex terms 124 is usually a Vcore (CPU) or a RAM voltage or timing issue. At least this is what I can gather from research on technical forums. In the windows event viewer, double click on the event. The actual event will open up in a seperate window. There is an option to do a internet search based on the event. Go to Control Panel-Adminastrative Tools-Event Viewer. After opening wait for it to gather the data. Click the drop down on Critical and double click the event. A window will open with a link to research online. It will also show the bugcheck codes and event ID.
Link to comment
Share on other sites

Thanks for the replies abn505. But I think my issue is different.

 

Firstly, I've always been using the 'old style' BIOS rather than the new mouse-driven one. I just prefer using it. So I definitely already have my XMP set via that method and everything else on Auto.

 

Secondly, my BSODs are always different - don't think I've seen the same one twice. USB drivers, network drivers, graphics drivers, miscellaneous kernel faults ... everything under the sun, a couple of times a day at unpredictable times. The error codes are different every time.

 

I've been progressively raising my DRAM voltage today but so far, no luck. Still getting BSODs even up at ~1.6v.

 

I think I need to try something else. Two things that sprang to mind were:

 

- Load line calibration. I'm not entirely sure what this does but someone suggested I should set this to 'extreme' to see if it helped.

 

- System Agent offset. This affects the voltage supplied to the IMC as I understand it. It's currently at the default +0.0V but perhaps I could bump this a bit (not sure by how much though ... 0.1V? What's an acceptable range of values for this on this mobo/CPU?)

 

Appreciate any advice people could give.

 

Beyond that I may try not using XMP at all and manually setting the DRAM frequency and timings. Don't see how that could change things since the XMP profile DOES appear to be working (enabling it does set 2133 Mhz, 9-11-10-27-2T and 1.5V as expected), but you never know...

Link to comment
Share on other sites

Secondly, my BSODs are always different - don't think I've seen the same one twice. USB drivers, network drivers, graphics drivers, miscellaneous kernel faults ... everything under the sun, a couple of times a day at unpredictable times. The error codes are different every time.

 

I've been progressively raising my DRAM voltage today but so far, no luck. Still getting BSODs even up at ~1.6v.

I would say it's another bad board. I would at the very least try a fresh windows instal first though.

 

Beyond that I may try not using XMP at all and manually setting the DRAM frequency and timings. Don't see how that could change things since the XMP profile DOES appear to be working (enabling it does set 2133 Mhz, 9-11-10-27-2T and 1.5V as expected), but you never know..
.

It's also very possible your CPU doesn't want to overclock to 2133mhz. Have you tried something more conservative like 1866mhz? Just because the XMP profile is active and sets the memory to 2133mhz, doesn't mean it's going to be stable. If your system is stable even at defaults(no XMP ) then I would give another vote towards a bad board. Given all the random BSOD errors i would say your memory is not the problem .

Link to comment
Share on other sites

Hi Cimexus,

 

Dang, hope that would have helped. I know a lot of people are having issues with Gigabyte Z87 boards and the bios. We have the same CPU as well. Hopefully someone with more knlowledge can help, did you try calling Gigabyte? The only thing I can think of is trying to play around with the CPU vcore voltage. I would try going down first one step at a time.

Link to comment
Share on other sites

I would say it's another bad board. I would at the very least try a fresh windows instal first though.

 

Yeah I suppose I should. (Sigh, I hate reinstalling Windows and it's not an old installation - I was having the BSODs even in the first couple of days of the build, but then, I had the RAM clocked at the XMP profile speed and timings since before I even installed Windows, so that's not surprising if it is indeed the problem)

 

It's also very possible your CPU doesn't want to overclock to 2133mhz. Have you tried something more conservative like 1866mhz?

 

True, but that seems hard to believe with a 4770k running at stock, doesn't it? I've heard Haswell can be tricky when you're overclocking both CPU and memory at once, but in this case the CPU is at stock so the CPU/IMC really shouldn't have any issues driving some 2133 RAM. But I'm a noob, so I'll take your suggestion on board...

 

I've only tried XMP on (2133 Mhz) and everything at default (which is a paltry 1333 Mhz). Note that it seems stable at 1333. I notice that, because my CPU voltages are all set to Auto, the motherboard automatically bumps the CPU Vcore up from 1.138 to 1.215 V when I enable XMP on the memory.

 

Only reason I haven't tried an intermediate speeds like 1833 is that I don't really know what I'd set the timings to. It's rated at 9-11-10-27 @2133, and when I leave it at default it ends up being 9-9-9-24 @1300. Suppose I could leave the timings at the looser end, reduce frequency, test for stability and then tighten timings a bit (though the difference between 9-11-10-27 and 9-9-9-24 is pretty minor, either way).

 

Just because the XMP profile is active and sets the memory to 2133mhz, doesn't mean it's going to be stable. If your system is stable even at defaults(no XMP ) then I would give another vote towards a bad board. Given all the random BSOD errors i would say your memory is not the problem.

 

Did you mean, "even if your system is UNstable even at defaults"? I'm pretty sure it is stable at defaults (no XMP), so I think the board is OK. At least I hope it is - the thought of having to disassemble the whole machine again to replace the board makes me cringe (I'm pretty time-poor!)

 

I'll try 2000/1833/1600 if I can't fix this with some voltage tweaks. But I'm still hoping I can get it stable at 2133. If not with more DRAM voltage, then with system agent/IMC voltage or by manually specifying the DRAM settings rather than using XMP. I've just seen so many machines in the past plagued by "random BSODs" that have been fixed with the right voltage tweak ... the tricky part is just finding which voltage to tweak! (e.g. I had a machine years ago that would randomly crash every 15 minutes in 3D games ... an extra 0.1 or 0.15V to the AGP port completely fixed it!)

Link to comment
Share on other sites

The only thing I can think of is trying to play around with the CPU vcore voltage. I would try going down first one step at a time.

 

CPU stuff is all on default/auto at the moment and until I get the memory stable I don't want to play around with it (introduces another variable into my troubleshooting). If it is somehow related to the CPU not quite coping with the RAM clock, if anything, I'd think system agent voltage offset would help more than Vcore.

 

I note that enabling XMP bumps my Vcore automatically up by about +0.1 V. Given that motherboard 'auto' voltage settings almost always ~overcompensate~ and provide too much voltage rather than too little, Vcore isn't high on my list of suspects at this point :) But stay tuned and we'll see.

 

I just wish that the BSODs were more frequent! Might sound weird to say that, but it's so hard to test whether changes I make to the settings helped or not, since they are so infrequent. I can run for hours and hours and start thinking "hey, maybe that setting change I made actually fixed it ... and then an hour later again, bam BSOD". :(

Link to comment
Share on other sites

CPU stuff is all on default/auto at the moment and until I get the memory stable I don't want to play around with it (introduces another variable into my troubleshooting). If it is somehow related to the CPU not quite coping with the RAM clock, if anything, I'd think system agent voltage offset would help more than Vcore.

 

I note that enabling XMP bumps my Vcore automatically up by about +0.1 V. Given that motherboard 'auto' voltage settings almost always ~overcompensate~ and provide too much voltage rather than too little, Vcore isn't high on my list of suspects at this point :) But stay tuned and we'll see.

 

I just wish that the BSODs were more frequent! Might sound weird to say that, but it's so hard to test whether changes I make to the settings helped or not, since they are so infrequent. I can run for hours and hours and start thinking "hey, maybe that setting change I made actually fixed it ... and then an hour later again, bam BSOD". :(

 

Hey Cimexus,

 

I feel your pain. I just had one yesterday, so that is 2 in 21 days since Gigabyte had me go into the bios (Old version) and reset my XMP1 profile even though it was already there. I was getting at least two a day. For me, it never happens when I am actually doing anything. I might play a game or two, leave the office for a bit, come back to see my computer restarted due to a critical error, which is always the same. Kernal Power Error Bugcheck code 292. So, now I disabled Intel Turbo Boost and running my CPU at stock 3.5 ghz. I will see how that goes. It does get frustrating, I built this in November and have been looking for a fix since. If it is due to Turbo boost, that will be a pisser because I have a I7-4770k and it should run fine at 3.9ghz with Turbo Boost according to Intel, but I also heard that is not always the case.

Link to comment
Share on other sites

Oh, what kind of video card do you have? I have a gigabyte 760GTX-OC-4GD. It has two PCI power connectors, I had a issue with the 8 Pin PCI connector not being seated properly and caused some BSOD issues for me. The eight pin has a 6 pin and seperate 2 pin that has to be flush together before inserting. There is a little notch to line the two together before inserting. I didn't have the 2 pin side correct, it was sticking in further than the 6 pin portion. I hope that made sense.
Link to comment
Share on other sites

Update: I put turbo Boost to "Enable" and set all my cores to multiplier of 38. I left the uncore at 35. I stressed out my computer by running Future Mark 11, PC Mark7 and 3DMark back to back. No BSOD's and then I left at idle for an hour with no BSOD's. I also played BF4 for an hour and Assassins Creed 4 for an hour and no issues. I will keep you posted if this helped.
Link to comment
Share on other sites

Oh, what kind of video card do you have?

 

I'm running an ASUS DirectCU II GTX 770. It also has the same 6 pin + 2 pin power arrangement.

 

Given that most (but not all) of my crashes occur in games (Diablo 3 is a particularly bad offender ... BSODs within an hour almost every time, where as strangely I can run Minecraft all day and it's fine), I might crack the case open and check the power cables.

 

I'm actually still not 100% convinced it's a hardware issue to be honest. I set everything back to 'normal' (stock CPU, XMP on and RAM at 2133). These are the settings when I was definitely getting BSODs. Ran Memtest for 7 hours and Prime95 for 4 hours and you know what - zero errors whatsoever. That's enough to make me think my crashes might be a weird driver issue or related to graphics card (since they mostly occur when gaming). The problem with that hypothesis is that, even though most BSODs occur in game, I have had a few just sitting at the desktop, or browsing the web. Plus the fact that the BSOD errors themselves rarely seem to have anything to do with graphics drivers (they're usually kernel related ... tcpip.sys is another one I see quite regularly ... I'm using onboard NIC so maybe the board IS faulty, who knows).

 

It'll be a while before I get to the bottom of this, as testing is a really slow process (since the BSODs are so infrequent). Summary of results so far:

 

1. BIOS defaults (stock CPU, RAM running at only 1333). Seems stable.

 

2. BIOS defaults except XMP ON (RAM at 2133). I get BSODs. I noticed CPU Vcore gets automatically set to ~1.21 when XMP is enabled which seems quite high considering the CPU is still at stock. That leads to some insanely high temps when running Prime95 in this configuration (closing in on 100C, which is throttling territory for the 4770k). This is despite having a high end Noctua cooler and idle temps around 28 C. But the heat issue is a separate issue to my BSOD issue, as the BSODs happen even when the CPU is quite cool. The heat is purely it's cause the BIOS auto voltage is setting such a high Vcore (1.21, but seems to adaptively jump up to 1.34 (!!) under load). If I manually reduce Vcore (see config (5) below), temps are much better and it still passes all the stress tests I can throw at it, so I think the mobo's auto voltage is just stupidly over-aggressive.

 

3. AS above with DRAM voltage manually set to 1.525, 1.5, 1.575 and 1.6 V: no improvements, BSODs persist, but Memtest/Prime95 still seemingly run without error (for at least 4+ hours).

 

4. As above with DRAM voltage manually set to 1.61 V. So far no BSODs, despite a good 4 hours in Diablo 3 (which usually attracts a BSOD within 1 hour). Fingers crossed! Seems weird to need 1.61V on a memory kit rated at 1.5V, but that's still well within the kit's tolerances. If it's as simple as this I'll be happy, since this fix doesn't require me to boost IMC/sys agent/uncore/Vcore voltages (which would hurt my temps and reduce my ability to OC CPU later on).

 

5. As in (2) but with Vcore manually set to 1.14V (just a fraction above stock, which is 1.13 on my chip). Runs Prime95 with no errors for 5+ hours, and the RAM is still at 2133 in this config, so this convinces me that my cooling is fine and Gigabyte's auto-voltage is just WAY TOO aggressive when XMP is enabled. Temps barely hit 65-70 C (amazing that a mere 0.09V extra Vcore gives me 100C temps!). Haven't had time to test for BSODs in games yet in this config, but since it's the same config as (2) just with a lower Vcore, I doubt this will be any different. This test was just to confirm that the auto-Vcore sucks and I can get away with lowering it manually without losing stability.

 

So in summary, a big DRAM voltage increase (1.5 --> 1.61V) seems to be the only thing helping so far. But I'm still not confident that that's solved it. It'll take several BSODless days before I'm convinced it's 'fixed'. So I'll report back when I have new info :) If OTOH no amount of DRAM voltage seems to help, then my remaining options will be:

 

- Increase system agent voltage (i.e. IMC voltage, to cope with the faster memory)

 

- Run memory at DDR 2000 or 1866 instead. Real world difference to 2133 is pretty minimal after all.

 

- Conclude that it's software-related (drivers?).

 

- Conclude that the board itself is bad. Hard to really say this for sure though given that it's stable when XMP is off (and even when XMP is on, Memtest and Prime95 have never produced a single error ... it's just at random times when I'm not stress testing that I get the blue screens).

 

Apologies for wall of text in this post, but hopefully I'll find a solution and record it here for anyone else in the future who has a similar problem.

Link to comment
Share on other sites

Yes, I noticed the same thing with voltages on my Gigabyte board. Possibly it doesn't regulate properly in Auto with Turbo Boost, also the core speeds. I also noticed with my Turbo boost in auto, Cores 1&2 ran at 39 core 3 at 38 and core 4 at 37. When I disabled Turbo boost and ran at stock 3.5 it seemed stable. Then, now with Turbo Boost set to "Enable and maunually setting each core at 38 it seems fine. Oh, with all the stress test I put on it my CPU it didn't go over 42c. I am using a Zalman 9900max cooler. I have never used prime95. I also disable the onboard video in the Bios.
Link to comment
Share on other sites

Sorry for the late reply guys, lets see if i can help with some of this, and whew!, Do I have some catching up to do! :)

Quote:

It's also very possible your CPU doesn't want to overclock to 2133mhz. Have you tried something more conservative like 1866mhz?

True, but that seems hard to believe with a 4770k running at stock, doesn't it? I've heard Haswell can be tricky when you're overclocking both CPU and memory at once, but in this case the CPU is at stock so the CPU/IMC really shouldn't have any issues driving some 2133 RAM. But I'm a noob, so I'll take your suggestion on board...

Your absolutely correct, and most of them should have no problems running at 2133,mhz. I've read many of articles though about some of them not binning real high . I'm just saying there is an off chance , no matter how minute.

 

 

I've only tried XMP on (2133 Mhz) and everything at default (which is a paltry 1333 Mhz). Note that it seems stable at 1333. I notice that, because my CPU voltages are all set to Auto, the motherboard automatically bumps the CPU Vcore up from 1.138 to 1.215 V when I enable XMP on the memory.

XMP will boost vcore to compensate for the adde3d load on the memory controller. However I have another suggestion , but i'll get to that at the end of this.

 

Only reason I haven't tried an intermediate speeds like 1833 is that I don't really know what I'd set the timings to. It's rated at 9-11-10-27 @2133, and when I leave it at default it ends up being 9-9-9-24 @1300. Suppose I could leave the timings at the looser end, reduce frequency, test for stability and then tighten timings a bit (though the difference between 9-11-10-27 and 9-9-9-24 is pretty minor, either way).

Okay, your going to have two timing specs.One is SPD or defaultsI.E.1333mhz. Then you have XMP settings or your overclocked settings.

I.E.2133mhz XMNP will relax the timings to achieve the higher speed. Just try 1866 you can use the same timings as XMP , which will be a little loose and a bit easier on the IMC. Nothing says you can't go back and tighten them up at a later date.

Did you mean, "even if your system is UNstable even at defaults"? I'm pretty sure it is stable at defaults (no XMP), so I think the board is OK. At least I hope it is - the thought of having to disassemble the whole machine again to replace the board makes me cringe (I'm pretty time-poor!)

Damn, yes I did mean that Sorry!

 

I'll try 2000/1833/1600 if I can't fix this with some voltage tweaks. But I'm still hoping I can get it stable at 2133. If not with more DRAM voltage, then with system agent/IMC voltage or by manually specifying the DRAM settings rather than using XMP. I've just seen so many machines in the past plagued by "random BSODs" that have been fixed with the right voltage tweak ... the tricky part is just finding which voltage to tweak! (e.g. I had a machine years ago that would randomly crash every 15 minutes in 3D games ... an extra 0.1 or 0.15V to the AGP port completely fixed it!)

Reply With Quote Multi-Quote This Message

Just stay away from 2000mhz. That is not a supported speed with your chip[set. It's almost a guarantee it wont be stable.

2. BIOS defaults except XMP ON (RAM at 2133). I get BSODs. I noticed CPU Vcore gets automatically set to ~1.21 when XMP is enabled which seems quite high considering the CPU is still at stock. That leads to some insanely high temps when running Prime95 in this configuration (closing in on 100C, which is throttling territory for the 4770k). This is despite having a high end Noctua cooler and idle temps around 28 C. But the heat issue is a separate issue to my BSOD issue, as the BSODs happen even when the CPU is quite cool. The heat is purely it's cause the BIOS auto voltage is setting such a high Vcore (1.21, but seems to adaptively jump up to 1.34 (!!) under load). If I manually reduce Vcore (see config (5) below), temps are much better and it still passes all the stress tests I can throw at it, so I think the mobo's auto voltage is just stupidly over-aggressive.

You can certainly try reducing vcore a bit with XMP enabled. But again, yeah, it's going to generate some extra heat. And the voltage increase's your seeing are all completely normal. HOWEVER, given the jump in heat your seeing, I would remount your cooler with some fresh thermal paste. that cooler should have no problems with that CPU even at XMP settings. 2133mhz is really a mild overclock to begin with.

 

 

Now to fast forward a bit and goo back to my suggestion from earlier...This is what I would do, and I know it's a pain, but reset your BIOS one more time. And enable the XMP profile for 2133mhz...Then,DISABLE Intel Turbo Boost. I have a feeling your board is having an issue with the voltage swings from idles settings to load settings. Disabling this will set everything to run at a constant..

 

But i would definatly look into your cooler and thermal paste as well.

Link to comment
Share on other sites

Cimexus,

 

I downloaded Prime95 to see if I could replicate my random BSOD issues. Below is what I found.

 

Scenario 1:

 

Bios Settings= Turbo Boost=Enable

XMP Profile Auto

All 4 cores @ 38 multiplier

All voltages @ auto

Onboard Video Off

Audio= Enabled

All other settings Auto

 

Software running Prime95 FFT small test 100% CPU small memory usage.

Gigabyte App Center/Easy Tune to monitor temps

Screen saver on

 

Pass 1=Passed

Pass 2=BSOD mid way through, Kernal Power Bugcheck 292 hex 124 (Same as Always)

 

Scenario 2:

 

Bios Settings= Turbo Boost=Auto

XMP Profile

All 4 Cores set to Auto

CPU Voltages Auto

Dimm volts to 1.55 from 1.5 (Peanutz Recomendation)

Onboard video disabled

Onboard Audio Enabled

All other settings Auto

 

Software running Prime95 FFT small test 100% CPU small memory usage.

Gigabyte App Center/Easy Tune to monitor temps

Screen saver on

 

Pass 1=Passed

Pass 2=Passed

Pass 3=BSOD mid way through, Kernal Power Bugcheck 292 hex 124 (Same as Always)

 

Scenario 3:

 

Bios Settings= Reset back to default

Turbo Boost=Auto

All 4 cores set to Auto

All voltages to Auto

XMP Profile Auto

Onboard Video Disabled

Onboard Audio Auto

All other settings Auto

 

 

Software running Prime95 FFT small test 100% CPU small memory usage.

Intel Extreme Tuner to monitor temps

Screen saver off

 

All 8 passes=Passed

 

I am wondering if this is more software related? I could point to Gigabyte App center, or the screen saver or both working at the same time. My Gigabyte App center has several apps, On/Off charge, Easy Tune, Live Update and @Bios. I have read some people having issues with On/Off Charge and Easy Tune. Any thoughts?

 

I am going to run the Prime95 balanced test next with scenario 3. The one that uses memory as well.

Link to comment
Share on other sites

You can certainly try reducing vcore a bit with XMP enabled. But again, yeah, it's going to generate some extra heat. And the voltage increase's your seeing are all completely normal. HOWEVER, given the jump in heat your seeing, I would remount your cooler with some fresh thermal paste. that cooler should have no problems with that CPU even at XMP settings. 2133mhz is really a mild overclock to begin with.

...

 

But i would definatly look into your cooler and thermal paste as well.

 

Remember: that crazy heat only occurs when under very heavy load with a Vcore is at ~1.34V, which as I understand it is very high for the 4770k. Overclocking guides I've read tell me things like "1.25V is the highest you should be using with air cooling". That seems to match my findings, as if I manually set Vcore so that it remains at or under the 1.25V level, my temps stay in the low 60s C when stress testing, which is fine. In fact, I can set the Vcore way down to 1.15V (which isn't much above stock), turn XMP ON and it will happily Prime95 all day (well, I tested 6 hours without errors, Vcore 1.15, XMP on).

 

What I was trying to say was that I think the auto voltage is 'overreacting' to XMP being on. With a freshly reset BIOS, the default/auto Vcore is at 1.13V. Under heavy load/turbo boost, it might peak up around 1.2V at most. If I then turn XMP on, leaving everything else at stock, that auto voltage gets set to 1.22 or something ridiculous (which means under load it skyrockets to 1.34ish).

 

I say this is an overreaction, because if I manually set Vcore back to near-stock (1.14 or 1.15), and leave XMP on ... it's still just as stable (no Prime95 errors in 6 hours) and temps are 60 instead of 100 C. Merely by turning XMP on, it sets a Vcore that's equal to what they recommend for a 4.5 Ghz CPU overclock! (despite the CPU still being at 3.5 Ghz stock (3.9 w/turbo boost)).

 

From what I've read, 1.3+V is water cooling territory anyway, so the fact my Noctua can't cope with load at 1.34V I don't think is evidence that the cooler isn't seated/thermal pasted correctly. I thought I did a decent job with the thermal paste. Pea-sized amount in the middle, press down hard...ah well, practice makes perfect I guess? :)

 

This isn't to say I think you're wrong. Perhaps I do need to redo the cooler mounting. So I will check it out - but that's a job for after I've figured out the RAM stability issue. The crashes are definitely NOT heat related since they occur when the CPU is quite cool (and in fact, the crashes have never happened even once while the CPU is being stress tested ... I've yet to have a error in Prime95 at all, in any configuration).

 

I'm still 80% sure more DRAM voltage will fix this. I had a whole night, XMP on, without a crash giving the RAM 1.61 V, whereas on any lesser voltage I'd get BSOD within an hour or two. If I get a few crashless days in a row at this config, I think that's strong enough evidence that it really is just DRAM voltage that's lacking rather than something more complex. I'll keep you posted.

 

Appreciate all your help - learning more every day (in particular, thanks for telling me that DDR-2000 won't work on my chipset, as you can bet I would have tried that before 1866 ;)

Link to comment
Share on other sites

Update: I ran prime95 balanced test for 6.5 hours all passed no BSOD's from scenario 3 above. I turned my screensaver back on a left the computer run overnight and no issues. Below is the dump file findings. It's always the same.

 

 

 

Crash List

 

 

Created by using BlueScreenView

 

Dump File Crash Time Bug Check String Bug Check Code Parameter 1 Parameter 2 Parameter 3 Parameter 4 Caused By Driver Caused By Address File Description Product Name Company File Version Processor Crash Address Stack Address 1 Stack Address 2 Stack Address 3 Computer Name Full Path Processors Count Major Version Minor Version Dump File Size Dump File Time

012314-21387-01.dmp 1/23/2014 17:25 0x00000124 00000000`00000000 fffffa80`12acc028 00000000`bf800000 00000000`00200401 hal.dll hal.dll+12a3b x64 ntoskrnl.exe+75bc0 C:\Windows\Minidump\012314-21387-01.dmp 8 15 7601 292,952 1/23/2014 17:27

 

This is the same all the time I get a BSOD Hall.dll+12a3b is the culprit.

Link to comment
Share on other sites

Originally posted by Cimexus:

What I was trying to say was that I think the auto voltage is 'overreacting' to XMP being on. With a freshly reset BIOS, the default/auto Vcore is at 1.13V. Under heavy load/turbo boost, it might peak up around 1.2V at most. If I then turn XMP on, leaving everything else at stock, that auto voltage gets set to 1.22 or something ridiculous (which means under load it skyrockets to 1.34ish).

Again, this would be Intel's turbo boost setting voltages when needed. This is another reason i suggested to disable it and leave all voltages at a constant.

 

I say this is an overreaction, because if I manually set Vcore back to near-stock (1.14 or 1.15), and leave XMP on ... it's still just as stable (no Prime95 errors in 6 hours) and temps are 60 instead of 100 C. Merely by turning XMP on, it sets a Vcore that's equal to what they recommend for a 4.5 Ghz CPU overclock! (despite the CPU still being at 3.5 Ghz stock (3.9 w/turbo boost)
).

 

Yes, and no. It may be what is needed for a 4.5ghz overclock, but that same adjustment could also be needed for a high memory overclock too. I guess what I'm getting at is to not get too hung up on just what your Vcore is being set to. As you have found there is no problem lowering that as long as you are stable.

From what I've read, 1.3+V is water cooling territory anyway, so the fact my Noctua can't cope with load at 1.34V I don't think is evidence that the cooler isn't seated/thermal pasted correctly. I thought I did a decent job with the thermal paste. Pea-sized amount in the middle, press down hard...ah well, practice makes perfect I guess? :)

Yes and no again. Yes, 1.3+ is watercooling territory, but if you have a good air cooler that can dissipate the heat,. there isn't a problem there either. You may want to just for giggles reseat your cooler. As far as your method goes, make sure you check with the pastes manufacturer website,many of them have their own methods depending on CPU Not to mention cure times with some of them..

This isn't to say I think you're wrong. Perhaps I do need to redo the cooler mounting. So I will check it out - but that's a job for after I've figured out the RAM stability issue.

Please just keep in mind that the two of them can be connected. Improper tension on any of the mounting bolts can cause some really weird memory and CPU issues.

 

I'm may be totally wrong, it's just one more thing that could potentially add to your issue.

'm still 80% sure more DRAM voltage will fix this. I had a whole night, XMP on, without a crash giving the RAM 1.61 V, whereas on any lesser voltage I'd get BSOD within an hour or two. If I get a few crashless days in a row at this config, I think that's strong enough evidence that it really is just DRAM voltage that's lacking rather than something more complex. I'll keep you posted.

1.65v is the max you would want to go. In fact I would add more DIMM voltage before I would raise your Vcore. For testing purposes, i would drop your your Vcore and take your DIMM voltage to 1.65v and then maybe raise your IMC voltage .2v. But leave Vcore at stock settings.

 

Originally posted by abn505:

Update: I ran prime95 balanced test for 6.5 hours all passed no BSOD's from scenario 3 above. I turned my screensaver back on a left the computer run overnight and no issues. Below is the dump file findings. It's always the same.

 

 

 

Crash List

 

 

Created by using BlueScreenView

 

Dump File Crash Time Bug Check String Bug Check Code Parameter 1 Parameter 2 Parameter 3 Parameter 4 Caused By Driver Caused By Address File Description Product Name Company File Version Processor Crash Address Stack Address 1 Stack Address 2 Stack Address 3 Computer Name Full Path Processors Count Major Version Minor Version Dump File Size Dump File Time

012314-21387-01.dmp 1/23/2014 17:25 0x00000124 00000000`00000000 fffffa80`12acc028 00000000`bf800000 00000000`00200401 hal.dll hal.dll+12a3b x64 ntoskrnl.exe+75bc0 C:\Windows\Minidump\012314-21387-01.dmp 8 15 7601 292,952 1/23/2014 17:27

 

 

Looks to be driver related abn. i really don't think you have a hardware issue.

Thats part of the windows OS. I would seriously think about reinstalling windows at this point.

 

P.S. Sorry for all the grammar mistakes...it's early yet ...on a sunday to boot!:p:

Link to comment
Share on other sites

Hey Peanutz,

 

I don't think so either, I ran Memtest86+ for 21 hours 11 passes and 0 errors. I have been getting these random BSOD's 10 days after building my machine. The only thing I can see is in Windows updater the following was installed and that it the first day I had a BSOD ironically.

 

Microsoft .NET Framework 4.5.1 for Windows 7 x64-based Systems (KB2858725)

 

Installation date: ‎11/‎27/‎2013 7:56 PM

 

Installation status: Successful

 

Update type: Recommended

 

The Microsoft .NET Framework 4.5.1 is a highly compatible, in-place update to the .NET Framework 4 and 4.5. By using the .NET Framework 4.5.1 together with the C# or Visual Basic programming languages, you can write Windows Store apps. The .NET Framework 4.5.1 delivers better performance, reliability, and security than previous versions. After you install this update, you may have to restart your computer.

 

More information:

http://go.microsoft.com/fwlink/?LinkId=245423

 

Help and Support:

http://support.microsoft.com

Link to comment
Share on other sites

Hi people,

I also have the same kit DRAM 16 GB 4x4 - 2133-c9 that I use on a Rampage 4 And with a 3930K CPU and Ram @ 4690 @ 2177 MHz. 9-11 timings-10-27-128-1t with VDRAM to 1,575, and I've never had BSOD caused by memories,, so I should be happy ... but I'm not anymore.

A friend of mine with the same kit of RAM, the same MB, same CPU, the same VCCSA, the same VTT, the same timings primary, run the DRAM frequency @ 2400 Mhz, and I can not get over 2240Mhz and also raiseing VDRAM tweaking the timings that I know.

But why when it comes to the setup of the memory, nobody give the specific setup of the secondary and terthiarytimings that are of fundamental importance, and why Corsair no indication of them?

Greetings to all,

sandro c.

Link to comment
Share on other sites

A friend of mine with the same kit of RAM, the same MB, same CPU, the same VCCSA, the same VTT, the same timings primary, run the DRAM frequency @ 2400 Mhz, and I can not get over 2240Mhz and also raiseing VDRAM tweaking the timings that I know.

Even though the hardware is the same , they are still vastly different and each system will overclock differently. Your just may not do as well or need further BIOS adjustments.

But why when it comes to the setup of the memory, nobody give the specific setup of the secondary and terthiarytimings that are of fundamental importance, and why Corsair no indication of them?

All other memory settings are best left to auto for the board to set accordingly. This is why they are not published. Secondary timing set-ups are pretty much a thing of the past.

Link to comment
Share on other sites

Thank peanutz94,

I've also read your answers on the new thread that I opened it and I answered you there ... you were very clear and precise.

hello sandro c

Thanks, and no problem. You can always give me a high five by hitting the rep button on the left! :p:

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.

×
×
  • Create New...