Need Help With OC

[Zac w]

Hey forum, I am quite confused about my over clock I have read numerous guides and countless forums, I successfully found myself an over clock for my Intel 4670k starting with 4.5 at 1.22v 1:1 cache @ 1.25v through to 4.8 Ghz at 1.435v @ 45 cache with 1.25v. I have successfully validated my over clocks and am quite happy with them. 

 

Since then I have wanted to optimise my over clock further so I have found that adjusting my cores individually should technically give me a better final OC. 

 

I have ended up with this:

 

core 1: 49

core2: 49

core3: 49

core4: 47

 

@ 1. 43v with cache ratio at 45 with 1.25v 

 

I am extremely happy with this over clock but I feel my dreams have suddenly come to a halt. I noticed that in any CPU intensive task I have tried Aida64, Intel extreme tuning utility, passmark, fire strike, cinebench, prime 95, asus benchmarking utility and intel burn test all run my cpu at the lowest multiplier available which would be my 4.7Ghz instead of three of my cores running at 4.9 Ghz. I don't know if there is a bios setting I have missed, or if there is a setting in windows which I can adjust to achieve my my desired over clock.

 

I have an Intel 4670k and Asus v1 formula running windows 8.1 

 

but yeah if anybody could shed some light on this scenario I would appreciate it.

[nuclear465]

I don't know what is going on on your system but i wish my 4670K could reach those frequencies. I had to push to 1.22 V just to get to 4.2Ghz, so i guess you were lucky on the processor lottery.

 

Accordingly to information that i have been reading around the web, i believe the top end of the 4670K voltage threshold is 1.3 V (cannot say with certainty), but 1.435 V seems to be dangerously over.

[Faceman]

Wow, quite a chip you've got there.

 

The reason you are seeing 47 as the multiplier when running those benchmarks, is because all 4 cores are enabled during those benchmarks.

 

The 49 multiplier is when 1-3 cores are in use, but when all 4 cores are in use, which is what happens during a stress test, that is when a multiplier of 47 is used.

 

1 Core in use = 49

2 Cores in use = 49

3 Cores in use = 49

4 Cores in use = 47

 

What kind of cooler are you using?  What are your temperatures like?

[Zac w]

I don't know what is going on on your system but i wish my 4670K could reach those frequencies. I had to push to 1.22 V just to get to 4.2Ghz, so i guess you were lucky on the processor lottery.

 

Accordingly to information that i have been reading around the web, i believe the top end of the 4670K voltage threshold is 1.3 V (cannot say with certainty), but 1.435 V seems to be dangerously over.

So there is no way to bypass this and reverse it so it uses the highest multipllier?

 

I have my own custom loop, I have the XSPC Raystorm, with a 270mm D5 Photon Reservoir, and a XSPC RX 240mm radiator. I have it set up with push/pull with 2x cougar and 2x noctua 120mm. And yeah my case being the CM Storm Stryker, I have modded it a bit to have numerous fans so yeah I have excellent airflow.

 

Ah rather, then me typing it all out, I documented my results of my overclocking that I have experienced, Ill attach it to this post. Its my personal experience overclocking and the settings I used. Mind you it's not complete and I haven't input this current overclocks results.

 

Also after sifting through allot of forums I feel like the general consensus is that no manufacturing website will recommend anything over 1.30v just like on a Panadol packet, or energy drink they don't recommend more then two in any one day. I know going past the 1.3v threshold does incur a potential risk, but I am willing to go there. I believe 1.35v I would be my absolute maximum on a good air cooler like my old Noctua NH-D14 but 1.4-1.44 is the most I feel comfortable with on water. And at 4.7Ghz I only need 1.33v 1.35v was for the additional stability of some intense applications like prime 95 for a couple hours.

 

And yeah I am extremely happy with all of my components I believe I got extremely lucky.

 

I have a reference 780 TI that I can overclock with a 350 mhz core clock and 550 memory clock and it runs smooth as butter and at 55 degrees.

 

And I have my G.Skill Trident X 2400 overclocked nicely.. but I am not entirely sure if it is actually a good memory clock or not.

 

It runs at 2400 Mhz 1.5v instead of the stock 1.65v and the timings are as follows:

 

first timings: 9, 11, 11, 21, 1T

 

Secondary: 4, 150, 5500, 14, 4, 16, auto, auto, 6

 

But yeah extremely happy with all of my equipment, got extremely lucky with it all. Just want to figure out this overclock if not Ill just be reverting to my 4.7Ghz

[Zac w]

Ah, I have my file in an excel document, it says I am not permitted to upload this document. I am assuming this is a site restriction?, what file type is accepted?

[nuclear465]

So there is no way to bypass this and reverse it so it uses the highest multipllier?

 

I have my own custom loop, I have the XSPC Raystorm, with a 270mm D5 Photon Reservoir, and a XSPC RX 240mm radiator. I have it set up with push/pull with 2x cougar and 2x noctua 120mm. And yeah my case being the CM Storm Stryker, I have modded it a bit to have numerous fans so yeah I have excellent airflow.

 

Ah rather, then me typing it all out, I documented my results of my overclocking that I have experienced, Ill attach it to this post. Its my personal experience overclocking and the settings I used. Mind you it's not complete and I haven't input this current overclocks results.

 

Also after sifting through allot of forums I feel like the general consensus is that no manufacturing website will recommend anything over 1.30v just like on a Panadol packet, or energy drink they don't recommend more then two in any one day. I know going past the 1.3v threshold does incur a potential risk, but I am willing to go there. I believe 1.35v I would be my absolute maximum on a good air cooler like my old Noctua NH-D14 but 1.4-1.44 is the most I feel comfortable with on water. And at 4.7Ghz I only need 1.33v 1.35v was for the additional stability of some intense applications like prime 95 for a couple hours.

 

And yeah I am extremely happy with all of my components I believe I got extremely lucky.

 

I have a reference 780 TI that I can overclock with a 350 mhz core clock and 550 memory clock and it runs smooth as butter and at 55 degrees.

 

And I have my G.Skill Trident X 2400 overclocked nicely.. but I am not entirely sure if it is actually a good memory clock or not.

 

It runs at 2400 Mhz 1.5v instead of the stock 1.65v and the timings are as follows:

 

first timings: 9, 11, 11, 21, 1T

 

Secondary: 4, 150, 5500, 14, 4, 16, auto, auto, 6

 

But yeah extremely happy with all of my equipment, got extremely lucky with it all. Just want to figure out this overclock if not Ill just be reverting to my 4.7Ghz

I wish i could help you, but your knowledge surpasses mine by a long margin, regarding Overclocking. I just started OCing a couple of weeks ago using tutorials and trial and error methodology, and i'm really happy with what i have achieved.

 

Maybe there is someone out there able to help you. Sorry !

[Zac w]

I wish i could help you, but your knowledge surpasses mine by a long margin, regarding Overclocking. I just started OCing a couple of weeks ago using tutorials and trial and error methodology, and i'm really happy with what i have achieved.

 

Maybe there is someone out there able to help you. Sorry !

That's alright I appreciate your time and effort anyway and good job with your overclock.

[Wats]

Zac, very good overclock. raising only the first few cores is like cheating, it lowers your load temp and helps prevent degrading at full whack with very high speeds. Now you have to work on your adaptive voltage and get it as good as it will get.

 

after all of that is stable, memory! you have a cool kit, id try for 2666 at least. if not 1600cl7 would be nice to see

[Zac w]

Zac, very good overclock. raising only the first few cores is like cheating, it lowers your load temp and helps prevent degrading at full whack with very high speeds. Now you have to work on your adaptive voltage and get it as good as it will get.

 

after all of that is stable, memory! you have a cool kit, id try for 2666 at least. if not 1600cl7 would be nice to see

Thanks for your appreciation, I have adaptive voltage enabled, it is quite temperamental for example my stable overclock for 4.7 Ghz is actually at 1.325 manual voltage and 1.35 adaptive. And in regards to my memory, I read tons and tons and tons of information regarding raw speed vs tight timings, there appears to be diminishing returns on mhz speed from around 1866 speed ram and then jumps up again  slightly at 2400 and then again at 3000. but moving to a 2666 or even 3000 which my ram is capable of posting wont yield as much a benefit as my tighter timings. I used memtweak it for dram efficiency score, I know its a rough score but it also seems to support this. I wish I had the maths in front of me to explain, but based off my calculations low timings raw at 1866 is better then 2400 ram with loose timings, just like 2400 tight timings is better then both 1866 and 3000. In my mem tweak it im supposedly receiving a score of 56000 and my stock X.M.P profile yields me 3600. In the scope of things, unless your using programs which directly draw their efficiency from DRAM your not going to really notice the differences except for in benchmarks and potentially 1-3 fps in particular games.

 

While I have this post open I was wondering if anybody could grant me clear clarification on Bclk frequency.

 

I know allot of people instantaneously say leave it alone, but I understand the mechanics of it or atleast the general mechanics of it and I understand I could probably raise my 4.7 Ghz to around 4.75-9 by increasing my black by 1-4 my question however is does it or does it not really hurt other components?, the great debate seems to illustrate the concept that increase the base clock frequency also extends to PCI express slots, sata and yeah basically everything running on the bus and so is not good for your computer. If that's the definitive case then I wont touch it. Im just looking for the sweet spot with my overclock.

 

4.7 Ghz minimum voltage 1.32 45 cache ratio @ 1.25v

 

4.8 Ghz 1.44v 45 cache @ 1.25v 1.95 input voltage.

 

per core overclock 49,49,49,47 @ 1.43v 45 cache, 1.9 input voltage.

 

Ideally I would love to get an overclock around 1.4 volts.

[Zac w]

Honestly how can I upload my file to this post? I have an excellent excel spreadsheet I designed and In the future I would like to contribute a post containing my results and also consolidating my knowledge I researched on overclocking haswell. I have read every guide and watched every youtube video and I have honestly got some great knowledge I would love to share with community. My first overclock attempt took me about 3 days of non stop trial and error and allot of cigarettes haha.

 

Now I have refined a process that is safe and I could overclock my Haswell from stock to 4.1 > to maximum overclock in a couple hours, and would love to share this process along with lots more information to help those new or getting into the overclocking scene.

 

but yeah for now how can I add an excel spreadsheet or should I turn it into a a PDF ? what is accepted at Linus tech tips?

[ProKoN]

Honestly how can I upload my file to this post? I have an excellent excel spreadsheet I designed and In the future I would like to contribute a post containing my results and also consolidating my knowledge I researched on overclocking haswell. I have read every guide and watched every youtube video and I have honestly got some great knowledge I would love to share with community. My first overclock attempt took me about 3 days of non stop trial and error and allot of cigarettes haha.

 

Now I have refined a process that is safe and I could overclock my Haswell from stock to 4.1 > to maximum overclock in a couple hours, and would love to share this process along with lots more information to help those new or getting into the overclocking scene.

 

but yeah for now how can I add an excel spreadsheet or should I turn it into a a PDF ? what is accepted at Linus tech tips?

 

just take a screen shot of the excel file and upload to http://tinypic.com/

 

post the link in your message.

[Wats]

Thanks for your appreciation, I have adaptive voltage enabled, it is quite temperamental for example my stable overclock for 4.7 Ghz is actually at 1.325 manual voltage and 1.35 adaptive. And in regards to my memory, I read tons and tons and tons of information regarding raw speed vs tight timings, there appears to be diminishing returns on mhz speed from around 1866 speed ram and then jumps up again  slightly at 2400 and then again at 3000. but moving to a 2666 or even 3000 which my ram is capable of posting wont yield as much a benefit as my tighter timings. I used memtweak it for dram efficiency score, I know its a rough score but it also seems to support this. I wish I had the maths in front of me to explain, but based off my calculations low timings raw at 1866 is better then 2400 ram with loose timings, just like 2400 tight timings is better then both 1866 and 3000. In my mem tweak it im supposedly receiving a score of 56000 and my stock X.M.P profile yields me 3600. In the scope of things, unless your using programs which directly draw their efficiency from DRAM your not going to really notice the differences except for in benchmarks and potentially 1-3 fps in particular games.

 

While I have this post open I was wondering if anybody could grant me clear clarification on Bclk frequency.

 

I know allot of people instantaneously say leave it alone, but I understand the mechanics of it or atleast the general mechanics of it and I understand I could probably raise my 4.7 Ghz to around 4.75-9 by increasing my black by 1-4 my question however is does it or does it not really hurt other components?, the great debate seems to illustrate the concept that increase the base clock frequency also extends to PCI express slots, sata and yeah basically everything running on the bus and so is not good for your computer. If that's the definitive case then I wont touch it. Im just looking for the sweet spot with my overclock.

 

4.7 Ghz minimum voltage 1.32 45 cache ratio @ 1.25v

 

4.8 Ghz 1.44v 45 cache @ 1.25v 1.95 input voltage.

 

per core overclock 49,49,49,47 @ 1.43v 45 cache, 1.9 input voltage.

 

Ideally I would love to get an overclock around 1.4 volts.

 

You will have to use a tiny bit more Vcore with an Offset or Adaptive voltage, but that's alright because you end up with very low idle voltage.

 

As for memory, I use ADIA64's Memory benchmark along with HyperPi 1M to display my progress in readable numbers. Saves you the work of doing maths -does most of it for you.

 

BCLK, I like to modify the BCLK a bit. Honestly, yes, it is ok to mess with it. But you have to be careful, as you said it does speed up stuff like your PCI-e lanes and your various controllers to the point where the board will not post. Enable BCLK Recovery, if the option is in your bios, to prevent that. Yes, play with BCLK, but make sure you can "punch in any overclock frequency willy nilly" first. As in, you know exactly voltage what it takes for every clock. That way, if you're running 100x48 with 1.44V, you would instantly know if running 102.1x47 was better or not. Of course you do have to re-test everything, even if it is better...

..Then there is the strap, but I dont think that works very well for cpu overclocking. memory overclocking though..

 

 

Hmm.. Yes, I think you should try different core clocks. ex: 49, 49, 48, 47. or even 49, 48, 47, 47. Haswell isn't effected by Vdroop as bad as previous gens. If it were effected, it'd make hitting 4.7@1.4v under 100% load easy while still giving you the 1.45v+ needed for 4.9 when at low loads.

 

 

 

 

By the way, how are you testing your single threads? I use Cinebench11.5 Single CPU along with HyperPi 32M and IntelBT with 1 Thread. Most other tests you have to go into task manager and set the affinity to 1 thread to get it set up.

[Socratic]

Have you tried increasing your vrin a bit more? I could never be bothered with tinkering with 4.8. Can get 4.7 easily, but stuck at 4.6, quite a drop in voltage, and damn that 4.8 took so much voltage and still couldn't get it stable. i tried different uncore settings, can't remember what I done though, but it never worked. You could push 4.8 for a test by increasing your bclk with x47 to 102.2mhz. and sync all cores. 

 

Oh wait you're wanting 4.9, damn i'm tired. 1stly, check if your ram can handle more, if not, take it down a notch and tighten timings. and 104.2mhz bclk, and best also take down pcie frequency if you're worried, I dunno I'm too tired to think.

[Sonefiler]

Honestly how can I upload my file to this post? I have an excellent excel spreadsheet I designed and In the future I would like to contribute a post containing my results and also consolidating my knowledge I researched on overclocking haswell. I have read every guide and watched every youtube video and I have honestly got some great knowledge I would love to share with community. My first overclock attempt took me about 3 days of non stop trial and error and allot of cigarettes haha.

 

Now I have refined a process that is safe and I could overclock my Haswell from stock to 4.1 > to maximum overclock in a couple hours, and would love to share this process along with lots more information to help those new or getting into the overclocking scene.

 

but yeah for now how can I add an excel spreadsheet or should I turn it into a a PDF ? what is accepted at Linus tech tips?

google drive m8

[Zac w]

Thanks all for the quick responses, I have been in the process of moving houses and as consequence have had no internet. Also my power supply decided to die so it's being rma at the moment. Once my computer is reassembled I'll post my graphs I think I could help new overclockers a bit.

Wats: thanks for your detailed response once my computer is set up I'll muck around with the blck but in regards to single thread testing I never though about trying it out. But I do feel like my 4th core is a bit funny seems to bottleneck me on single core oc . And it also runs 20 degrees cooler then other cores regardless of how my waterblock is mounted or my thermal compound it's quite different as all other cores are within 3 degrees of each other. Idle my 4th core is 20 degrees warmer then the others but under load it's 20 degrees under the other cores. I really don't get it and I have rempunted my cpu a good 10x +

[Wats]

Thanks all for the quick responses, I have been in the process of moving houses and as consequence have had no internet. Also my power supply decided to die so it's being rma at the moment. Once my computer is reassembled I'll post my graphs I think I could help new overclockers a bit.

Wats: thanks for your detailed response once my computer is set up I'll muck around with the blck but in regards to single thread testing I never though about trying it out. But I do feel like my 4th core is a bit funny seems to bottleneck me on single core oc . And it also runs 20 degrees cooler then other cores regardless of how my waterblock is mounted or my thermal compound it's quite different as all other cores are within 3 degrees of each other. Idle my 4th core is 20 degrees warmer then the others but under load it's 20 degrees under the other cores. I really don't get it and I have rempunted my cpu a good 10x +

luckily, i stumbled across this. Use @Zac w to mention someone =p

 

No one really knows why it happens. Some would say that problem is suppose to go away after a delidding. Just ignore it pretty much, focus on the highest core temps.

[Zac w]

luckily, i stumbled across this. Use @Zac w to mention someone =p

 

No one really knows why it happens. Some would say that problem is suppose to go away after a delidding. Just ignore it pretty much, focus on the highest core temps.

Once again sorry for the slow response, I am living out at my rural address for the next two months for ski season work and don't have internet. 

 

I have been contemplating a delid from what I have seen if done carefully with the vice and hammer method. It would appear to have extremely effective results, I'm just scared that being a clumsy individual ill stuff it up and lose my chip haha. 

[Wats]

Once again sorry for the slow response, I am living out at my rural address for the next two months for ski season work and don't have internet. 

 

I have been contemplating a delid from what I have seen if done carefully with the vice and hammer method. It would appear to have extremely effective results, I'm just scared that being a clumsy individual ill stuff it up and lose my chip haha. 

 

No need to apologize, I can't see a single neighbor here. I know how bad internet can be...

Rightfully nervous though, Haswell is more difficult than Ivy when it comes to delidding due to the on chip vrms. They're on the left side in a line along the die which is a vertical rectangle.

 

Vice method, I remember reading you should use a soft wood for smacking the hammer on, I think Ceder would be the choice.

Razor method, I do that any day over the vice. Steady hands.

 

 

Should knock off a solid 15C, if not 20-30c lower has been reported. imangine that.. going from 80c to 50c, in 10 minutes.

 

 

just be careful if you do go through with it... you know, have the money to replace it before you do it yada yada.

[Zac w]

No need to apologize, I can't see a single neighbor here. I know how bad internet can be...

Rightfully nervous though, Haswell is more difficult than Ivy when it comes to delidding due to the on chip vrms. They're on the left side in a line along the die which is a vertical rectangle.

 

Vice method, I remember reading you should use a soft wood for smacking the hammer on, I think Ceder would be the choice.

Razor method, I do that any day over the vice. Steady hands.

 

 

Should knock off a solid 15C, if not 20-30c lower has been reported. imangine that.. going from 80c to 50c, in 10 minutes.

 

 

just be careful if you do go through with it... you know, have the money to replace it before you do it yada yada.

haha yeah,  I am a pretty broke uni student however I do have quite a bit of financial reserves as I manage a ski hire in winter. But yeah buying a new chip if I did stuff it would be frustrating in the form of playing the chip lottery again .

I also would feel far more comfortable using a razor blade over the vice but nobody seems to recommend it, as I can see numerous individuals have killed their chips this way but I am very careful. Would you happen to know what kind of glue I could use that is near identical to Intel's black default goop? I know you don't need to reseal it up as it can just sit there clamped in, but I would like to seal it up again.

 

Also thanks for the tip in tagging individuals through the @name I am new to the forums I think Ill make myself a nice looking profile I quite like this site. Can't wait to show off my updated rig to the community.

[Wats]

haha yeah,  I am a pretty broke uni student however I do have quite a bit of financial reserves as I manage a ski hire in winter. But yeah buying a new chip if I did stuff it would be frustrating in the form of playing the chip lottery again .

I also would feel far more comfortable using a razor blade over the vice but nobody seems to recommend it, as I can see numerous individuals have killed their chips this way but I am very careful. Would you happen to know what kind of glue I could use that is near identical to Intel's black default goop? I know you don't need to reseal it up as it can just sit there clamped in, but I would like to seal it up again.

 

Also thanks for the tip in tagging individuals through the @name I am new to the forums I think Ill make myself a nice looking profile I quite like this site. Can't wait to show off my updated rig to the community.

I've forgotten what glue it is..prolly for the best, if you mess up your application of coollaboratory...well you know what youll have to do again to fix it.

 

if you ever do want to sell it or move it, just use a rubber band to wrap it together. Even that isn't needed, as long as you dont chip the die it'll survive a lot.

 

Some P4s are delidable. they're difficult though, prolly cuz they're so old and hard, good practice. Find them for free in old retired machines. http://www.ebay.com/itm/Intel-P4-2-8GHz-512k800FSB-socket-478-SL6WJ-HT-Northwood-core-CPU-/121040614626?pt=CPUs&hash=item1c2e9534e2 scroll down and you can see the glue on this one oozing out =p

 

oh and i know.. that one guy impressed everyone swinging a hammer and now everyone has shaky hands.