First time OC'ing, Core i7, Feedback?

[Ejjman1]

I am still trying to figure this overclock thing out. Here is what I have so far.

I set the voltage of the cpu around 1.316v. I set the Memory at 1.65v. I didn't change anything else, it was all at auto.

So everything seems dandy, but when I run prime95 for over 2 minutes, all of the little workers quit running. I can't find the reason, and there is no stress.txt file.

The bottom picture is really hard to see. But it says "FATAL ERROR: Rounding was 0.5, expected less than 0.4 -- Hardware failure detected, consult stress.txt file"

So I'm not sure what to do now.

 

[dcf-joe]

I believe it is because your overclock may not be stable.

Have you ever heard of IntelBurnTest? I believe this is the new, standard stress testing program, because it will detect instabilities in seconds instead of minutes.

 

[carrera997]

When I had this error it was because my memory timings were too slow (11,11,11 or something like that). Try tightening them up.

 

[PaulieG]

When I had this error it was because my memory timings were too slow (11,11,11 or something like that). Try tightening them up.

No. That isn't the problem. His clocks are just too ambitious for his voltages. He needs to find out what kind of voltages he needs for a particular speed. Many of the newer batches can't do 4.0 on that kind of vcore, let alone 4.2.


Back down that overclock. You need to slow down, and do this in steps. First drop your BCLK to about 180. Download LinX so that you can do some short stability testing. Run LinX at 180 BCLK for 20 cycles. If it passes, the increase it by 2. Keep doing this until you get errors. Then we'll start tweaking voltages.

Also, read this before you continue:

http://forums.techpowerup.com/showthread.php?t=108526

LinX

 

[Ejjman1]

No. That isn't the problem. His clocks are just too ambitious for his voltages. He needs to find out what kind of voltages he he needs for a particular speed. Many of the newer batches can't do 4.0 on that kind of vcore, let alone 4.2.


Back down that overclock. You need to slow down, and do this in steps. First drop your BCLK to about 180. Download LinX so that you can do some short stability testing. Run LinX at 180 BCLK for 20 cycles. If it passes, the increase it by 2. Keep doing this until you get errors. Then we'll start tweaking voltages.

Also, read this before you continue:

http://forums.techpowerup.com/showthread.php?t=108526

LinX

Ok, I'll do that. But when I was trying out these clocks, I was changing my memory ratios from 2:12 to like 2:10, so they never went above 1600 Mhz. What should I do about that?

 

[PaulieG]

Ok, I'll do that. But when I was trying out these clocks, I was changing my memory ratios from 2:12 to like 2:10, so they never went above 1600 Mhz. What should I do about that?

While you are trying to find a stable overclock, you really want to "rule out" as many things as you can. So, while you are doing this, drop the divider as low as possible, then bring it back up after you are stable. Don't try to overclock various components at the same time, nor try to change more than one clock/voltage setting at a time. That just makes things more convoluted.

 

[Deleted member 74752]

That memory divider would be 2:8 -

Your timings should be 9-9-9-24 1T

Off topic, but where in TN Paulieg?

 

[mastrdrver]

You need to find the limits of each system first. In other words, drop all the multis to their minimum and push up one clock at a time and make sure it is stable at that speed.

Read that link Paulieg posted and it should help you.

 

[PaulieG]

That memory divider would be 2:8 -

Your timings should be 9-9-9-24 1T

Off topic, but where in TN Paulieg?

Greeneville, East TN.

 

[Ejjman1]

That memory divider would be 2:8 -

Your timings should be 9-9-9-24 1T

Off topic, but where in TN Paulieg?

The guide did not explain much about timings. I have been using what is on the memory

7-7-7-24 @ 1.65v

And I haven't changed that around, just lowered the multiplier.

The test finished, and there were no errors. I will keep doing what you say, raising by 2, until I get errors.

 

[PaulieG]

The guide did not explain much about timings. I have been using what is on the memory

7-7-7-24 @ 1.65v

And I haven't changed that around, just lowered the multiplier.

The test finished, and there were no errors. I will keep doing what you say, raising by 2, until I get errors.

Don't think about memory at all. Just drop it to a low divider, and focus on cpu only for now. Just keep in mind that Core i7 chips are all over the place. Some need only 1.24v for 4.0, others struggle with 1.375v or even higher. Earlier batches seem a bit more consistent with low voltages.

 

[Ejjman1]

Ok, made it through those aswell!

I'll do what you said, and keep working on it. I'll tell you when I hit a wall!

Thanks

 

[Hayder_Master]

i guess this voltage is low for this clocks, try 4.0 ghz maybe you will decrease the voltage too also it depend on cpu release version, and what about qpi voltage you set

 

[Thrackan]

I believe it is because your overclock may not be stable.

Have you ever heard of IntelBurnTest? I believe this is the new, standard stress testing program, because it will detect instabilities in seconds instead of minutes.

OCCT is the only testing program I trust nowadays

But yeah, this OC is lacking stability. Some more vCore most likely.

 

[Binge]

I agree with Thrack's post, but it's good he's seeing errors and not getting full crashes. It means he's very close to stability.

 

[Ejjman1]

i guess this voltage is low for this clocks, try 4.0 ghz maybe you will decrease the voltage too also it depend on cpu release version, and what about qpi voltage you set

QPI would be default. With the first setup, I changed nothing but VCore. (And 1.65v DIMM)

OCCT is the only testing program I trust nowadays

But yeah, this OC is lacking stability. Some more vCore most likely.

I am now doing with paulieg said, and am going to move up very slowly in steps. So we'll see if I can reach the first clock with ease, and see if it was the vCore that was causing the failure.

 

[Binge]

QPI would be default. With the first setup, I changed nothing but VCore. (And 1.65v DIMM)

You will need to change your QPI most likely. Voltages for 4.0+ range from stock to 1.35V QPI.

 

[bpgt64]

push your volts up to 1.34 for your CPU or so and it will stabilize at 4.0-4.2

 

[Binge]

push your volts up to 1.34 or so and it will stabilize at 4.0-4.2

Don't say that... you make it sound like he needs that kind of help. You aren't being specific as to what volts he should increase, and you have no idea if it would stabilize his overclock. Furthermore if you read any of these previous posts then you'd know he has an idea of what he needs to do. The OP is a smart guy, so give him more credit please by giving him more than a vague/unnecessary response.

 

[bpgt64]

Sorry, I didn't mean to make it sound like it was an exact solution, given though I have an i7 920, and thats what I did to make it stable at 4.0 with linpack and prime 95 with 8 threads, I figured it was relevant. You obviously want to find the minimum voltage that will stabilize your system...but heh.

 

[overclocking101]

yes I would notch up the vcore to 1.3ish and as binge said notch QPI up to 1.3/1.35V backing down the oc would be sort of counter productive. if it posted and didnt crash prime instantly he's getting close to stable. are you doing blend? small or large ftt?? your ram gets stressed with large and cpu with small, blend is both if you do one at a time it will help determine stability. try upping the volts a tad before moving the oc down. Temps dont seem to bad but I cant read highest temp due to the SS being so small.

 

[bpgt64]

There's a really nice sweet spot with a blck (atleast in my system) at 4.0 ghz with turbo boost 4.2ghz. It allows my DDR3-1066 to run at 1600 7-7-7-20 CR1 QPI 3.6ghz

 

[Ejjman1]

paulie,

I ran LinX at BCLK of 184, and memory multiplier was 2:8. When I came back, I had a Blue Screen.

I then ran LinX at BCLK of 184, and memory multiplier was 2:6. When I came back, it had finished.

I will keep the memory multiplier that low (2:6 is the lowest), and try out the test with adding 2 more BCLK to be 186!

 

[PaulieG]

paulie,

I ran LinX at BCLK of 184, and memory multiplier was 2:8. When I came back, I had a Blue Screen.

I then ran LinX at BCLK of 184, and memory multiplier was 2:6. When I came back, it had finished.

I will keep the memory multiplier that low (2:6 is the lowest), and try out the test with adding 2 more BCLK to be 186!

Yup, taking the memory out of the equation. You can go back later and tweak that. I should reiterate what Binge said. At some point you will need to increase your QPI voltage too. My suggestion is increase that to the same voltage as your vcore, and at some point it may need to go higher than your vcore.

You are starting to get the hang of it!

 

[Ejjman1]

Yup, taking the memory out of the equation. You can go back later and tweak that. I should reiterate what Binge said. At some point you will need to increase your QPI voltage too. My suggestion is increase that to the same voltage as your vcore, and at some point it may need to go higher than your vcore.

You are starting to get the hang of it!

Well, 186 failed pretty fast.

So now, what do I start doing?