Please help me getting OC step by step

[IIIdefconIII]

 

Hi all,

 

Can someone please help me step by step how to archieve the best OC result.

i can get 4.6 stable @ 1.42 V, 30 min stresstesting in AIDA64. when i just 1.41 V aida stress test failed 3 times within the 20 minutes.

Strange thing is when I didnt use my XMP profile before, i was able to get an OC of x47 @ 1.4V stable. (this was before i traded my dead mobo for a new one).

I see lots of people having much lower voltage on 1.42, most using 1.38 V.

 

Please help me out, what to do and what to try.

I got ripjaws V5 DDR4-3000 CL 15 enabled the XMP.

Multiplier x47

cpu V 1.42

 

stable @ a max 85 degrees. (need to repaste soon with a good thermal pasta, any recommendations? i head the pk3 is perfect).

 

Am i stress testing correctly?

Why does cpuid from aida shows its using 1.440 V while stress testing, whil i put 1.42 in BIOS.

 

Most people say 1.42 is top Voltage for 24/7 daily use. So i dont want to exceed it, but i surely had x 47 before stable @ 1.4V.

 

Thank you!

 

 

 

[RS2007GOD]

you may have lucked out on the silicon lottery. I friend is getting 4.8 stable with 1.4v unsure how he done it though.

[YubinTheBunny]

I personally like the Gelid GC extreme or the Noctua NT H1 as these 2 pastes seem to be some of the best performing ones (at least on benchmarks). The gelid paste is better for higher temp applications though as I found out it produced a more consistent temp on gpus. But for cpus they are both fine. A lot of people recommend the Artic MX4 but I personally never used it before, but most pastes are fine and perform 1 to 2C from each other unless you decided to buy some weird off brand.

[Imakuni]

1 hour ago, IIIdefconIII said:

Strange thing is when I didnt use my XMP profile before, i was able to get an OC of x47 @ 1.4V stable. (this was before i traded my dead mobo for a new one).

A RAM OC can cause a CPU OC to be unstable, and vice versa.

1 hour ago, IIIdefconIII said:

Can someone please help me step by step how to archieve the best OC result.

Let me teach you how to OC:

1. Put everything at stock (leave XMP on if you intend to use it, or leave if off if you'll manually OC your RAM later) and CPU voltage at a set 1.2v.
2. Get Prime95 28.9 (latest version).
3. Fire the small test for around 15min and:
   1. If temps are below 81C, increase the voltage by a notch.
   2. If temps are around 81~83, leave voltage as is.
   3. If temps are at 84 or above, decrease voltage a notch.
4. Repeat 3 as necessary until you find your max safe voltage.
5. Now that you've found your voltage, put your desired multiplier.
6. Stress test for around 30min. If it fails / crashes, bump it down by one. Repeat until "stable".

And remember, if it ever crashes / BSOD, decrease the multi again.

1 hour ago, IIIdefconIII said:

Why does cpuid from aida shows its using 1.440 V while stress testing, whil i put 1.42 in BIOS.

I don't really trust Aida's sensor... from your HWMonitor screenshot, seems like it's getting to 1.42v under load, just as you wanted.

[IIIdefconIII]

Currently i have x45 @ 1.33 V @ max 78 degrees, it is stable with xmp @ 3000 Mhz. So x46 should be easy with 1.4 you would say right

[IIIdefconIII]

 

[IIIdefconIII]

What about this i got it very stabel at 1.34 V in BIOS @ x45, but different Voltage reports.. Im just unsure cause cpuid goes up to 1.4V

 

[IIIdefconIII]

its all fucking failing

[RS2007GOD]

have you consider a better cooler? An AIO 280mm or 240mm would probably help alot with temps esp for higher overclocks.. 

[RS2007GOD]

Wait you have an rog board right?

[RS2007GOD]

Disable "epu power saving"  and "cpu svid support" then test

 

[IIIdefconIII]

On 7/29/2016 at 1:22 PM, RS2007GOD said:

Disable "epu power saving"  and "cpu svid support" then test

 

 

thanks, will try this tonight. Strange this stays, realier it worked on x47, could i or asus have fried the cpu somehow so the oc is worse? aisuite reported alot of voltage in the beginning with auto oc

[IIIdefconIII]

still not getting a stable oc, i maybe will go for the intel 4.2  cry, altought the pc runs fine, playing games etc, only halted on aida54 system stress

[RxL]

If you are using xmp make sure you set the ram voltage to 1.35v instead of auto and maybe try add some system agent voltage like maybe try 1.02v for system agent voltage

 

Hard to say why you of fails if you don't give full Info on all settings you changed in the bios

[IIIdefconIII]

just did an bios updated recommended on overclockers.com testing now ans seems ver more stable

[Djxinator]

I don't know why someone told you to get the latest version of prime95 but unless you want to kill your CPU use 26.6 instead.

[Imakuni]

6 minutes ago, Djxinator said:

I don't know why someone told you to get the latest version of prime95 but unless you want to kill your CPU use 26.6 instead.

REALLY? Geez, then I guess I'm super lucky for mine not blowing up afer 29 days running P95 nonstop.... wow, thx for the reminder. /s

 

Joke aside, that's BS. Prime95 doesn't kill processors. DUMB PEOPLE (that have no clue on what they are doing) kill CPUs, using Prime95. So long as you do things the proper way and manage your expectations, there's no risk in using it.

[Djxinator]

11 minutes ago, Imakuni said:

REALLY? Geez, then I guess I'm super lucky for mine not blowing up afer 29 days running P95 nonstop.... wow, thx for the reminder. /s

 

Joke aside, that's BS. Prime95 doesn't kill processors. DUMB PEOPLE (that have no clue on what they are doing) kill CPUs, using Prime95. So long as you do things the proper way and manage your expectations, there's no risk in using it.

The later versions use AVX instructions and push way too much heat and wattage into your cpu. 

 

If you're using a 3rd to 6th gen Intel CPU any of the newer versions are not recommended and are in no way a good test of "stability".

[Imakuni]

Just now, Djxinator said:

The later versions use AVX instructions and push way too much heat and wattage into your cpu. 

 

If you're using a 3rd to 6th gen Intel CPU any of the newer versions are not recommended.

You forgot to mention about AVX2 (specifically FMA3). Also, 2nd gen supports AVX as well, so by your argument, it shouldn't be recommended either.

 

Are you sure you really know what you are talking about, rather than just repeating what you've heard from other people?

2 minutes ago, Djxinator said:

If you're using a 3rd to 6th gen Intel CPU any of the newer versions are not recommended and are in no way a good test of "stability".

Sorry to disagree, but latest version P95 can find instability that other tests (and sub 28.7) would be unable to otherwise, both on the CPU and RAM side of things. For one's system to be *rock solid*, it MUST be able to pass every single test possible, which includes P95. Not to say that it is a perfect test (one can pass P95 but fail somewhere else), but it's a damn good one!

 

Of course, if someone doesn't want such level of stability and would rather go with a "just game" quick OC, they would most likely skip P95. And it's certainly not a sin. But if one goes by the "stable as possible" definition, which is what I use (and seemingly OP as well), then it has to pass P95.

 

Whether you agree with us or not, it's your choice. But be sure to know WHY before you give any sort of advice / recommendation to someone else.

[Djxinator]

24 minutes ago, Imakuni said:

You forgot to mention about AVX2 (specifically FMA3). Also, 2nd gen supports AVX as well, so by your argument, it shouldn't be recommended either.

 

Are you sure you really know what you are talking about, rather than just repeating what you've heard from other people?

Sorry to disagree, but latest version P95 can find instability that other tests (and sub 28.7) would be unable to otherwise, both on the CPU and RAM side of things. For one's system to be *rock solid*, it MUST be able to pass every single test possible, which includes P95. Not to say that it is a perfect test (one can pass P95 but fail somewhere else), but it's a damn good one!

 

Of course, if someone doesn't want such level of stability and would rather go with a "just game" quick OC, they would most likely skip P95. And it's certainly not a sin. But if one goes by the "stable as possible" definition, which is what I use (and seemingly OP as well), then it has to pass P95.

 

Whether you agree with us or not, it's your choice. But be sure to know WHY before you give any sort of advice / recommendation to someone else.

I'm not arguing about using Prime95 as a whole. It's an excellent tool for finding instability. I'm a bit startled you would be so petty as to build your argument on a missed number 2, you know full well what I meant.

 

Earlier versions of Prime95 are perfectly capable of finding instability in a system. Newer versions create instability by forcing way too much power into the CPU. 

 

Almost every overclocking guide for Haswell onwards has mentioned that the newer versions of P95 are not suitable for testing unless you are intending to use AVX/2 instructions during normal workloads.

 

I am using Prime95 myself. I use an older version because the newer versions, specifically on an X99 CPU can suck upwards of 350w through your CPU, which is not healthy by any standard.

 

Some people do like to use the newer versions for whatever reason. I'm suggesting that in this instance the OPs cooling solution is not up to the task for P95 AVX/2, which is in no way a marker for instability because the workload is so unrealistic it's ridiculous.

 

If he uses an earlier version of Prime95 or a test that uses AVX/2 properly and much more realistically (AIDA64) then he should find that his overclock is much easier to get stable.

 

[EDIT] - I know that Sandy Bridge supports AVX but the issue lies with the way the IHS has been applied to later CPUs, using Thermal Compound instead of Solder (the latter having significantly better cooling performance) thus rendering them unable to deal with the heat generated by AVX workloads without Delidding.

[Imakuni]

5 minutes ago, Djxinator said:

I'm a bit startled you would be so petty as to build your argument on a missed number 2, you know full well what I meant.

*2 missed 2s. One for AVX2.0 and the other for 2nd gen CPUs.

And no, I really didn't know if you missed it or just copy pasted what you've heard, hence why I pressured you. That's a common mistake, and one that I myself do often: I type something and people misunderstand it, not because I don't know, but because my thought was so clear in my head that I'd believe it would be clear to everyone else.

10 minutes ago, Djxinator said:

Earlier versions of Prime95 are perfectly capable of finding instability in a system. Newer versions create instability by forcing way too much power into the CPU.

You've got the "cause-effect" order swapped out. It's not that the high load causes instability. It's the other way around: instability is only found, provided it exists, when you try to squeeze every droplet of performance out of your CPU (and even P95 doesn't get close to doing so on a clock level, but you can ignore that).

19 minutes ago, Djxinator said:

If he uses an earlier version of Prime95 or a test that uses AVX2 properly and much more realistically (AIDA64) then he should find that his overclock is much easier to get stable.

Back to the -rock solid- definition. If one wants true stability, everything has to pass. And that includes AIDA64, obviously, but it also includes P95 latest version.

 

*Cue* for when intel finally lets us put our hands into AVX-512 and P95 releases an update taking advantage of that.

[IIIdefconIII]

i have updated the bios and the system runs way better at oc now, anyway aida64 will do the job fure sure. back on x47 mulitplier test, stable for 3 hours now, and my system fans werent all driving on full speed while i tested this. So cooling should be ok. i can keep it under 86 degrees on 1.4v higest spike was(86)

[done12many2]

9 minutes ago, Imakuni said:

*2 missed 2s. One for AVX2.0 and the other for 2nd gen CPUs.

And no, I really didn't know if you missed it or just copy pasted what you've heard, hence why I pressured you. That's a common mistake, and one that I myself do often: I type something and people misunderstand it, not because I don't know, but because my thought was so clear in my head that I'd believe it would be clear to everyone else.

You've got the "cause-effect" order swapped out. It's not that the high load causes instability. It's the other way around: instability is only found, provided it exists, when you try to squeeze every droplet of performance out of your CPU (and even P95 doesn't get close to doing so on a clock level, but you can ignore that).

Back to the -rock solid- definition. If one wants true stability, everything has to pass. And that includes AIDA64, obviously, but it also includes P95 latest version.

 

*Cue* for when intel finally lets us put our hands into AVX-512 and P95 releases an update taking advantage of that.

 

This is a thread started by a guy asking for a step by step on a basic overclock.

 

Just about any method of testing can be customized to make it "safe" for sustained use, but Prime95 does not fall into a "click and go" category.

 

I agree that the instability is not a product of the load, but the undue stress is indeed a product of the ridiculous amount of power before forced through that poor chip.  

 

Stability is subjective and what you believe is required for an overclock to be considered "rock solid" has become more of an archaic view.  I'm not saying this is bad, but more extreme.

[Djxinator]

52 minutes ago, done12many2 said:

 

This is a thread started by a guy asking for a step by step on a basic overclock.

 

Just about any method of testing can be customized to make it "safe" for sustained use, but Prime95 does not fall into a "click and go" category.

 

I agree that the instability is not a product of the load, but the undue stress is indeed a product of the ridiculous amount of power before forced through that poor chip.  

 

Stability is subjective and what you believe is required for an overclock to be considered "rock solid" has become more of an archaic view.  I'm not saying this is bad, but more extreme.

High temperatures create instability. An overclock that is 100% stable at 70c may fail immediately at 85c.

 

I use older Prime to test stability because at maximum load it produces temperature values around 10c over normal use (gaming etc). If I run the latest version my temperatures skyrocket to ridiculous levels, still under thermal control but enough to cause a failure where there isn't under already above-normal operational loads.

 

Right now I'm testing 26.6 to around 67c maximum. If I move to 28.9 or whatever, it will probably jump above 80c - not only causing instability but forcing my somewhat conservative overclock down 100-200mhz again, and I'm using a 280mm rad with two of the best fans on the market. To put it into perspective, my gaming loads top out around 57c on the hottest core.

 

 

[done12many2]

 

16 minutes ago, Djxinator said:

High temperatures create instability. An overclock that is 100% stable at 70c may fail immediately at 85c.

 

I use older Prime to test stability because at maximum load it produces temperature values around 10c over normal use (gaming etc). If I run the latest version my temperatures skyrocket to ridiculous levels, still under thermal control but enough to cause a failure where there isn't under already above-normal operational loads.

 

Right now I'm testing 26.6 to around 67c maximum. If I move to 28.9 or whatever, it will probably jump above 80c - not only causing instability but forcing my somewhat conservative overclock down 100-200mhz again, and I'm using a 280mm rad with two of the best fans on the market. To put it into perspective, my gaming loads top out around 57c on the hottest core.

 

Thanks for the explanation, but I'm not sure where I made you believe that I didn't already know that heat contributes, not causes, instability?

 

I absolutely agree with the necessity to maintain cooler CPU temps in order to promote better stability.  Same applies to GPUs.  That's why I went with 5 x 360mm radiators with 15 EK Vardar F4 ER fans strapped to them.   I like my water to remain at ambient temp regardless of load.