Can't OC is it CPU or MB? X79

[screen name]

Hello,
I have a MSI X79A-GD65 (8D) and a i7-3930k. When I got it I started OCing it to see how high i could get. I was able to achieve 4.8GHz @ 1.425V and finally stopped when I got to 5.0GHz @ 1.5V.

I decided to make my 24/7 OC to be 3.8GHz at 1.250V (running prime95 it would teter out at 57C). All was fine for a couple days but now things have deteriorated. If I try to overclock at all the system hangs at the win7 logo screen during boot. The stock 6-core turbo boost is 3.5GHz. Even if I just change that toe 3.6GHz the system freezes during boot. It does not matter how high I set the voltage. About 1 out of 5 boots its a success. When it does boot, everything runs perfectly. No stability problems. I'm clueless.

Even though it is the norm with SB to simply increase the Turbo Boost max frequency, one should still be able to simply turn off speed step and turbo boost and set a constant frequency. If I set a constant frequency above 3.5GHz though, I run into a new problem. It boot, but the cpu speed is set to 1.2GHz and the cpu temp readings are constant at 91C, even though there is no way it was ever that hot. If i open CoreTemp it reads "91C?".

Any help would be great. Thanks

Rest of my system:
Tuniq Ripper 1000W
G. Skill Ripjaws Z series DDR3 1600 4x4GB
MSI GTX 580 Lightning
Corsair H80
2x Corsair Force GT 120 ssd (RAID 0)

 

[screen name]

More Info:

I tried moving the ram around into different slots. Originally I was of course running quad channel with each stick in the primary dimm for each channel. I tried a bunch of other configurations including just using one stick and trying each channel. In any set up besides the original any OCing gives the result where the system boots at 1.2GHz and says the temperature is 91C. Lowering the speed of the RAM has no effect.

 

[Crashman]

Hello! You're having the same issue as I've had with both of my Engineering Sample C1-stepping CPU's. The problem appears to be related to C1 and C2 reporting, where disabling these causes the CPU to report 91C!!! This in turn forces thermal protection mode with minimum frequency and voltage.

I'm afraid to tell you that the only solution I've found is Asus. As described to me by an ASRock engineer, this is a flaw in the CPU that affects thermal reporting and, because it only occurs when overclocked settings are made, is not covered by the CPU warranty.

Asus has a workaround built into its UEFI. Oh yes, I have most motherboard brands on-hand.

Anyway, since you're stuck with an MSI board the next-best-workaround is to make sure C1, C2 and Turbo modes are enabled, then increase the base clock. I'm pushing 119 MHz BCLK at 1.35V, but you should be able to go higher with even more core voltage.

BTW, I'm writing about this RIGHT NOW and am looking to have a conversation HERE with ANYONE having a similar issue. Your biggest problem is that the issue your facing is so rare that the "experts" don't believe it exists, and many of them have assumed that your issue is a PEBKAC error. So, posts like yours get buried, which hurts me since I'm actually trying to document the whole fiasco.

PS: Please tell me you're running a RETAIL CPU

 

[Crashman]

Any progress?

 

[jaquith]

Might be worth looking at this PDF, explains the production delays as well -> http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=2&ved=0CEcQFjAB&url=http%3A%2F%2Fqdms.intel.com%2Fdm%2Fd.aspx%2FF051E059-17FA-4B21-92E7-025550FBF529%2FPCN111178-00.pdf&ei=VeYIT9W-NZO2twe3rMjGAw&usg=AFQjCNE3cJQNvjrxiinGPO0fBXhvaVJPMQ&sig2=eOlRjI-HSsKaflFbTNC-Fw

It's a crazy link address and seems cached.

Intel® Core™ i7-3960X Processor Extreme Edition & i7-3930K Processors,
PCN 111178-00, Product Design, C1 to C2 stepping Conversion for Boxed and Tray

Date of First Availability of Post-Conversion Material: January 20, 2012

changes apply:
• New S-spec and MM numbers for the converting products
• VT-d erratum removed for C-2 step. Refer to Processor Sighting# 3876388 in Sandy Bridge-EN/EP/EX Processor Sightings Report – NDA (#452856).

BTW - I wish Intel wouldn't screw around with disabling the 7th & 8th flipping cores and just enable; dumb games. Also, I cannot help some bozo posted this and Google cached it.

 

[Crashman]

Might be worth looking at this PDF, explains the production delays as well ->

Talked to Gigabyte yesterday and was told 2% of retail C1-stepping cores are impacted by this issue. Mine still overclocks "as expected" only when using an Asus motherboard.

 

[jaquith]

I thought 'I' discovered something but, Tecmo34 sent me this link -> http://www.tomshardware.com/news/intel-cpu-sandy-bridge-e-c2-virtualization,14252.html and after I posted here I found other news and posts as well. Frankly, after your or somebody's here lack luster OC'ing article it doused my interest with SB-E. I've built to date four R4E + i7-3960X and settled in with4.8GHz with no problems. But I only played around with them for a day once completed.

I won that i7-3930K + Sabertooth but I'm on hold waiting for the lithography 'changes' and rerelease. I'll play around with it a little longer before it's donated, but it's on an ASUS. However, with the CPU changes it'll be interesting how the BIOS handles the updates.

 

[Crashman]

I thought 'I' discovered something but, Tecmo34 sent me this link -> http://www.tomshardware.com/news/intel-cpu-sandy-bridge-e-c2-virtualization,14252.html and after I posted here I found other news and posts as well. Frankly, after your or somebody's here lack luster OC'ing article it doused my interest with SB-E. I've built to date four R4E + i7-3960X and settled in with4.8GHz with no problems. But I only played around with them for a day once completed.

I won that i7-3930K + Sabertooth but I'm on hold waiting for the lithography 'changes' and rerelease. I'll play around with it a little longer before it's donated, but it's on an ASUS. However, with the CPU changes it'll be interesting how the BIOS handles the updates.

Intel sent me a 3960X that goes 4.4 GHz at 1.35V. A motherboard manufacturer sent me another CPU from around the same production date that went 4.7 GHz at 1.35V. I hear that 4.6 GHz is about average for 3960X's at 1.35V, and it's interesting to note that the sub-par CPU came directly from Intel

However, BOTH of the C1 CPUs I received have the same overclocking issue. MSI motherboards do less/worse with these affected CPU's, so they're the easiest to find complaints about on the 'net. Asus firmware has a built-in workaround, but I'm not even sure Asus knows that. After all, sometimes things get fixed by accident.

 

[jaquith]

I couldn't find any suitable and 'stable' settings for 5.0GHz but 4.7GHz~4.8GHz required 1.42v~1.45v even using Koolance CPU-370. I have not as of yet tried the i7-3930K, but I'll report back; it'll be limited by the H80. I never use MSI, only ASUS, EVGA and Gigabyte but so far only the R4E. Damn I wanted to win the 3960X + R4E

 

[Crashman]

I couldn't find any suitable and 'stable' settings for 5.0GHz but 4.7GHz~4.8GHz required 1.42v~1.45v even using Koolance CPU-370. I have not as of yet tried the i7-3930K, but I'll report back; it'll be limited by the H80. I never use MSI, only ASUS, EVGA and Gigabyte but so far only the R4E. Damn I wanted to win the 3960X + R4E

Ah yes, somewhere between 1.35V and 1.40V most of these 32nm processors turn into heat monsters. Diminishing returns. Which is why "max OC" at something below 1.40V will continue to be my preferred method.

I certainly wouldn't jump from 1.35V to 1.42V just to gain 100 MHz.

And in other news:
http://www.digitimes.com/news/a20120108PD202.html

 

[jaquith]

First, I 'get' you're looking for the 90C 'bug'. So bottom-line, I didn't have the issue on any ASUS R4E + 3960X; 0/4 issues. My 'bitch' is the lack luster OC'ing and poor BCLK implementation which I hoped/expected to be in leagues with X58 -- big disappointment. So hopefully this will all get better, but my expectations are low. The 3960X does handle what you throw at it benches aside, seems more like a Xeon than what I expected for SB-E to be.

Other than that, it depends on how close you are to TJmax which IMO is the most damaging, in all my instances temps weren't the issue, the issue is my reluctance in providing folks with >1.45v; a 4.4GHz -> 4.8GHz is use of design with the cooling at hand.

Who knows, we'll / you'll know if the lithography improves vCore soon enough. Honestly, if it does I know a few folks who'll be wanting an exchange! The Article needs to not 1+1=3. Bouncing in my head is the affects of disabling the two cores, both thermally and electrically.

As we both know no one is going to be running at 100% 7/24.

 

[Crashman]

First, I 'get' you're looking for the 90C 'bug'. So bottom-line, I didn't have the issue on any ASUS R4E + 3960X; 0/4 issues. My 'bitch' is the lack luster OC'ing and poor BCLK implementation which I hoped/expected to be in leagues with X58 -- big disappointment. So hopefully this will all get better, but my expectations are low. The 3960X does handle what you throw at it benches aside, seems more like a Xeon than what I expected for SB-E to be.

Other than that, it depends on how close you are to TJmax which IMO is the most damaging, in all my instances temps weren't the issue, the issue is my reluctance in providing folks with >1.45v; a 4.4GHz -> 4.8GHz is use of design with the cooling at hand.

Who knows, we'll / you'll know if the lithography improves vCore soon enough. Honestly, if it does I know a few folks who'll be wanting an exchange! The Article needs to not 1+1=3. Bouncing in my head is the affects of disabling the two cores, both thermally and electrically.

As we both know no one is going to be running at 100% 7/24.

Yeh, well I'm sorry that the average of 4.6 GHz at less than 1.40V is unimpressive.

As for me, I'm just trying to gather user info, like you said on a bug. But you'll never even know if your CPU has the bug, because you have an Asus board. And Asus is immune to the bug. That's nice for you...

 

[jaquith]

If my only concern was Prime95 then the vCore's could have been -0.02v~-0.03v less, but I encountered problems looping Futuremark benches in particular 3DMark11. See -> http://www.legitreviews.com/article/1803/14/ ; these folks had to exceed the '1.45' vCore. Further, my idea of 'stable' is not an hour or two of running Prime95.

The choices for the builds were very limited: ASUS Rampage IV Extreme, EVGA X79 Classified, or Gigabyte GA-X79-UD7. Folks who pay for 4-WAY EVGA 03G-P3-1591-AR systems want '5.0 GHz' which couldn't be delivered. Using an i7-2600K it is possible to achieve 5.0GHz+, so @ x3 the price your expectations get clear.

So if there's a 'bug' encountered or not it's a matter of perspective. I felt there is 'something' going on there and the lithography changes will be a tell all.

Again, I can almost guarantee if the i7-3960X 'fix' improves performance all hell is going to break loose. Furthermore, while Intel could take the stance it performs as 'Rated' the underlying issue is that Intel knowingly released a CPU with a defect and later corrected an error in production.

 

[screen name]

Thanks for the info. Yes I am using a RETAIL CPU. Sorry I did not reply earlier. The post had been sitting around for a while with no response so I stopped checking it.

Guess I have to suck it up and buy an ASUS.

If I touch the BCLK at all the board completely fails. It powers on for about two seconds then powers down and back up for another two seconds (and loop). Same goes for adjusting the BCLK ratio.

 

[screen name]

Best answer selected by screen name.

 

[screen name]

Talked to Gigabyte yesterday and was told 2% of retail C1-stepping cores are impacted by this issue. Mine still overclocks "as expected" only when using an Asus motherboard.

Could you clarify what the Asus workaround is? Do you still simply change the CPU ratio like one normally would and the ASUS UEFI simply ignores the bug?

 

[Crashman]

Could you clarify what the Asus workaround is? Do you still simply change the CPU ratio like one normally would and the ASUS UEFI simply ignores the bug?

Magic. Isn't that how we explain an observed behavior that we can't otherwise explain?

According to everything I've found, Asus must be doing "something different" with C-states. I can't go farther than that because I simply don't know.

 

[screen name]

I disagree with Intel when they say its not covered under warranty because the chip "performs as described." I don't even need to look at the description, the dam name of the product has the K at the end which means the CPU ratio is unlocked. Hmm, why would that be there if I was not supposed to be able to change the ratio? I would say that that means part of the product's proper performance is to be able to change the CPU ratio. But I can't do this... so wouldn't that mean it is not performing as described? (i would like to emphasize my sarcasm)

 

[jaquith]

Magic. Isn't that how we explain an observed behavior that we can't otherwise explain?

According to everything I've found, Asus must be doing "something different" with C-states. I can't go farther than that because I simply don't know.

Now the big question is How will ASUS and others handle the updated C state?!

As I mentioned to Chris to know fully a BIOS update would be needed to avoid gibberish results and findings.