I5 4690k overclocking settings

[sethmac82]

Just switched over to Intel and have my I5 4690k overclocked to 4.6 at 1.26v. Was wondering if I need to be messing with any other settings besides multiplier. Have been reading about people using bclk strap and other settings. Do I need to be setting this too or am I good with just multiplier? I also notice my frequencies jumping in core temp too and didn't know if it's because of the adaptive setting or is it something I did wrong? Any help would be great as this is my first time ocing Intel.

 

[TechyInAZ]

You will be fine with just multi/ratio overclocking. Bclk is if you want to get a very specific overclock and/or want to overclock other things too.

The only think i'd recommend to make your CPU even more speed is to up the cache to around 43 ratio (you overclock the cache just like the cpu).

I have my cache at 42x with 1.2v running through it (however I put in extra vcore until I get time to test and lower it).

Since you and I have nearly the exact same parts in our systems, i'm curious. What temps are you getting when stress testing?

I got around 70C with cpu at 4.6ghz at 1.275v with the H100i (had to return it for h105 because it was defective).

 

[JackNaylorPE]

4.6 Ghz at 1.26 is an incredible voltage (I required 1.38 for stability) . While many argue that cache does nothing (and it doesn't for OC validation) it does have an effect in certain apps like image editing... Asus recommends cache be kept with 4 of CPU multiplier.

The question I have is .... is it stable ? Is it manual or adaptive ? what have you used to test stability ?

I would pass many stress tests fine but Asus RoG Real bench woud always be what determined any instabilities.

 

[sethmac82]

You will be fine with just multi/ratio overclocking. Bclk is if you want to get a very specific overclock and/or want to overclock other things too.

The only think i'd recommend to make your CPU even more speed is to up the cache to around 43 ratio (you overclock the cache just like the cpu).

I have my cache at 42x with 1.2v running through it (however I put in extra vcore until I get time to test and lower it).

Since you and I have nearly the exact same parts in our systems, i'm curious. What temps are you getting when stress testing?

I got around 70C with cpu at 4.6ghz at 1.275v with the H100i (had to return it for h105 because it was defective).

Not quite sure on how to do the cache. Ill have to read up on it. Bios has min and max setting and not sure where to start. Rerunning aida64 now to double check but getting temps in high 40s to mid 50s at 4.6 with 1.2603v. Thats with cpu, fpu, cache and memory stressed. Left gpu and local disks unchecked. Do they need to be tested too?

 

[sethmac82]

4.6 Ghz at 1.26 is an incredible voltage (I required 1.38 for stability) . While many argue that cache does nothing (and it doesn't for OC validation) it does have an effect in certain apps like image editing... Asus recommends cache be kept with 4 of CPU multiplier.

The question I have is .... is it stable ? Is it manual or adaptive ? what have you used to test stability ?

I would pass many stress tests fine but Asus RoG Real bench woud always be what determined any instabilities.

Ya it seems to be stable. No freeze ups or bsods. Only thing ive noticed is sometimes i cant get a internet connection on startup so i restart and its good to go after that. Played bf4 for about 2-3 hours today no prob . Read to use manual for ocing then switch back to adaptive once overclock is set so thats what i did. Ive only used aida64 so far but only ran it for 30 min.

Just realised cpu core voltage is still on manual.

 

[TechyInAZ]

You will be fine with just multi/ratio overclocking. Bclk is if you want to get a very specific overclock and/or want to overclock other things too.

The only think i'd recommend to make your CPU even more speed is to up the cache to around 43 ratio (you overclock the cache just like the cpu).

I have my cache at 42x with 1.2v running through it (however I put in extra vcore until I get time to test and lower it).

Since you and I have nearly the exact same parts in our systems, i'm curious. What temps are you getting when stress testing?

I got around 70C with cpu at 4.6ghz at 1.275v with the H100i (had to return it for h105 because it was defective).

Not quite sure on how to do the cache. Ill have to read up on it. Bios has min and max setting and not sure where to start. Rerunning aida64 now to double check but getting temps in high 40s to mid 50s at 4.6 with 1.2603v. Thats with cpu, fpu, cache and memory stressed. Left gpu and local disks unchecked. Do they need to be tested too?

wow nice. My overclock of 4.5ghz at 1.275 brought temps to nearly 70C at the maximum with H100i (however it was defective so that could of been the issue).

For cache, i'd start at 1.2 at 43 and if it's stable, just keep going down on the voltage until you get a bsod.

Here is my thread on cache: http://www.tomshardware.com/answers/id-2475409/overclocking-cpu-cache.html#xtor=EPR-8809

 

[sethmac82]

Do I need to leave vcore on manual or should I set it to adaptive?

 

[TechyInAZ]

Do I need to leave vcore on manual or should I set it to adaptive?

Cache or CPU? Assuming cpu since you said vcore.

I've noticed that manual keeps the cpu and voltages at a constant speed, no power savings on. Whereas when I use adaptive it keeps all my C states and speedstep on. I'd use adaptive. No issues with adaptive on my end.

 

[sethmac82]

You will be fine with just multi/ratio overclocking. Bclk is if you want to get a very specific overclock and/or want to overclock other things too.

The only think i'd recommend to make your CPU even more speed is to up the cache to around 43 ratio (you overclock the cache just like the cpu).

I have my cache at 42x with 1.2v running through it (however I put in extra vcore until I get time to test and lower it).

Since you and I have nearly the exact same parts in our systems, i'm curious. What temps are you getting when stress testing?

I got around 70C with cpu at 4.6ghz at 1.275v with the H100i (had to return it for h105 because it was defective).

Not quite sure on how to do the cache. Ill have to read up on it. Bios has min and max setting and not sure where to start. Rerunning aida64 now to double check but getting temps in high 40s to mid 50s at 4.6 with 1.2603v. Thats with cpu, fpu, cache and memory stressed. Left gpu and local disks unchecked. Do they need to be tested too?

wow nice. My overclock of 4.5ghz at 1.275 brought temps to nearly 70C at the maximum with H100i (however it was defective so that could of been the issue).

For cache, i'd start at 1.2 at 43 and if it's stable, just keep going down on the voltage until you get a bsod.

Here is my thread on cache: http://www.tomshardware.com/answers/id-2475409/overclocking-cpu-cache.html#xtor=EPR-8809

Used Intel XTU and set to 4.2 at 1.2v on adaptive. 20 min in and everything seems to be fine. Still testing tho. Do I need to mess with offsets too or am I good on those?

 

[TechyInAZ]

You will be fine with just multi/ratio overclocking. Bclk is if you want to get a very specific overclock and/or want to overclock other things too.

The only think i'd recommend to make your CPU even more speed is to up the cache to around 43 ratio (you overclock the cache just like the cpu).

I have my cache at 42x with 1.2v running through it (however I put in extra vcore until I get time to test and lower it).

Since you and I have nearly the exact same parts in our systems, i'm curious. What temps are you getting when stress testing?

I got around 70C with cpu at 4.6ghz at 1.275v with the H100i (had to return it for h105 because it was defective).

Not quite sure on how to do the cache. Ill have to read up on it. Bios has min and max setting and not sure where to start. Rerunning aida64 now to double check but getting temps in high 40s to mid 50s at 4.6 with 1.2603v. Thats with cpu, fpu, cache and memory stressed. Left gpu and local disks unchecked. Do they need to be tested too?

wow nice. My overclock of 4.5ghz at 1.275 brought temps to nearly 70C at the maximum with H100i (however it was defective so that could of been the issue).

For cache, i'd start at 1.2 at 43 and if it's stable, just keep going down on the voltage until you get a bsod.

Here is my thread on cache: http://www.tomshardware.com/answers/id-2475409/overclocking-cpu-cache.html#xtor=EPR-8809

Used Intel XTU and set to 4.2 at 1.2v on adaptive. 20 min in and everything seems to be fine. Still testing tho. Do I need to mess with offsets too or am I good on those?

I've never used offsets in my life, and everything's running fine for me.

 

[JackNaylorPE]

Ya it seems to be stable. No freeze ups or bsods. Only thing ive noticed is sometimes i cant get a internet connection on startup so i restart and its good to go after that. Played bf4 for about 2-3 hours today no prob . Read to use manual for ocing then switch back to adaptive once overclock is set so thats what i did. Ive only used aida64 so far but only ran it for 30 min.

Just realised cpu core voltage is still on manual.

Synthetic benchmarks like AIDA / Prim, etc are not recommended for Haswell on adaptive voltage control. Voltage will pop up 0.10 to 0.13 on adaptive when AVX instructions are present. After switching to adaptive, run a 2 hour stress test with RoG Real Bench. AIDA is like throwing a ball chest high to a practicing fielder in baseball from 2nd base to home plate every 2 seconds .... Real Bench is like throwing them high, low, fly, ground, fast, slow from every direction. AIDA / Prime show that it can take a single load, RB shows it can take varying loads in a multitasking environment.

 

[TechyInAZ]

Ya it seems to be stable. No freeze ups or bsods. Only thing ive noticed is sometimes i cant get a internet connection on startup so i restart and its good to go after that. Played bf4 for about 2-3 hours today no prob . Read to use manual for ocing then switch back to adaptive once overclock is set so thats what i did. Ive only used aida64 so far but only ran it for 30 min.

Just realised cpu core voltage is still on manual.

Synthetic benchmarks like AIDA / Prim, etc are not recommended for Haswell on adaptive voltage control. Voltage will pop up 0.10 to 0.13 on adaptive when AVX instructions are present. After switching to adaptive, run a 2 hour stress test with RoG Real Bench. AIDA is like throwing a ball chest high to a practicing fielder in baseball from 2nd base to home plate every 2 seconds .... Real Bench is like throwing them high, low, fly, ground, fast, slow from every direction. AIDA / Prime show that it can take a single load, RB shows it can take varying loads in a multitasking environment.

I personally switched from prime95 to OCCT, is OCCT a good program?

 

[JackNaylorPE]

I like OCCT because it graphs the results but it's pretty much doing the same thing .... it takes standards synthetic tests and wraps them in a nice package. So in that sense it's mo better or worse than the tests that it packages.

I started using it less when nVidia started detecting it and throttling its cards in response.

One can use as many synthetic tests as they want ..... I like to use them for the purposes of establishing cooling parameters .... i.e how hot could my system possibly get ? This is great for that purpose as Furmark and the OCCT GPU test is for GPUs. But I didn't build a PC to run synthetics.... I built it to run programs. So if you want to test it's stability for the reasons you built it running programs / playing games .... you need to test the machine's ability to do those things.

After passing 8 hours on synthetics, I had "stable" OC settings that were unstable under RoG Real Bench. My GPU OC settings were stable on synthetics and all the GFX test benchmarks but in Metro 2033, loading would stall on the opening credits,.... it went away after some tweaking.

At my 24/7 settings I have been running for 14 months running every game (incl. Far Cry 3/4, Tomb Raider, the Metros, all the Crysis versions, etc) and every program I have thrown at it .... now but BF4 will still crash after anywhere from 20 minutes to 2 hours at that settings. I created a more conservative set of CPU and GPU settings for BF4 so my son can play that when he's here. But I lay that on BF4's lap rather than the puter.

 

[TechyInAZ]

I like OCCT because it graphs the results but it's pretty much doing the same thing .... it takes standards synthetic tests and wraps them in a nice package. So in that sense it's mo better or worse than the tests that it packages.

I started using it less when nVidia started detecting it and throttling its cards in response.

One can use as many synthetic tests as they want ..... I like to use them for the purposes of establishing cooling parameters .... i.e how hot could my system possibly get ? This is great for that purpose as Furmark and the OCCT GPU test is for GPUs. But I didn't build a PC to run synthetics.... I built it to run programs. So if you want to test it's stability for the reasons you built it running programs / playing games .... you need to test the machine's ability to do those things.

After passing 8 hours on synthetics, I had "stable" OC settings that were unstable under RoG Real Bench. My GPU OC settings were stable on synthetics and all the GFX test benchmarks but in Metro 2033, loading would stall on the opening credits,.... it went away after some tweaking.

At my 24/7 settings I have been running for 14 months running every game (incl. Far Cry 3/4, Tomb Raider, the Metros, all the Crysis versions, etc) and every program I have thrown at it .... now but BF4 will still crash after anywhere from 20 minutes to 2 hours at that settings. I created a more conservative set of CPU and GPU settings for BF4 so my son can play that when he's here. But I lay that on BF4's lap rather than the puter.

Interesting, have you used it for cpu overclocking?

 

[JackNaylorPE]

I use them all for overclocking.... for the purposes described in the above post

Synthetics for cooling
RoG Real Bench and application / game demos for stability

As above, synthetic stable OCs often crash under RoG Real bench