Overclocking 4770k Results

I was using a pretty decent I7 950 x58 setup and decided to upgrade. My main reason was resell! I was able to sell my i7 950 for $100, Asus premium board for $100, and memory kit for $100. With a few hundred more, I was able to get an ASUS Maximus Hero, 4470k, and I went ahead and bought me 32g of Corsair Vengeance Pro RAM for some RAMdisk usage. I think if I waited a year, the old stuff would not be wanted and wouldn't sell.

I spent a good fun part of last weekend overclocking the CPU. How far can you push your 4770k? Have you ran all of the tests to show it is stable? I can only get 4.6GHZ stable at 1.375! Heat becomes a problem anything past 1.35 using a Corsair H110. I suggest that you start with 1.35 and see if you can get 4.8 stable and work your way down to 4.5. Anything higher than 1.35 is pointless under a synthetic test because you'll have heat issues. In reality, your CPU isn't going that take that type of stress though, so if you can get it stable at a high voltage, you may be okay. Honestly, unless you got some awesome cooling better than Corsair H110 (best sealed unit out) I don't recommend anything past 1.35 because the temps spike even if you're not 100% CPU... means your fans kick in all the time for a few secs, which is annoying. 34 idle and then 88 for a second is no fun... it bounces significantly less at stock and you computer will sporadically let of steam.

After a few tests and research, I think the lottery is like this (whether the chip can be stable under 1.35).

4.8 = 10%
4.7 = 20%
4.6 = 40%
4.5 = 70%

Chances are, you will either get an average chip that does 4.5 with ease and maybe 4.6... or an awesome chip that does 4.7 with ease and maybe 4.8. All in all, I am starting to think anything past 4.2 is kind of pointless with these chips. They indeed seem to be designed more for advanced mobile/efficiency features rather than brute speed. The adaptive voltage feature and turbo is actually really nice. The CPU can literally go down to near zero power usage usage...C state and etc are neat. At the same time, the CPU can eat 150+w continuously and on demand! Its a sleek power demon that turns into a fairy on demand. I can foresee how this move to "mobile efficiency" will eventually lead to cold tiny fast CPUs with great overclocking potential in desktops because less power = less heat/more performance.

I question the benefit you get going past 4.2ghz. Mind you, this is overclocking as the stock is at 3.5ghz... you still get a nice boost, but no stability and heat issues. I did benchmarks comparing 4.2 to 4.6 and honestly don't think there is anything there that is tangible for gaming or mainstream computer usage... its nice to say and know you got your chip sitting at 4.8ghz and your benchmarks are sky high, but unless you are doing crazy CPU stuff for work, you're just adding heat and making your computer louder... gaming is is all GPU nowadays. I understand why Intel doesn't have any motivation to make faster mainstream CPUs...they're already many times faster than anyone needs for anything. My 3-4 year old i7 950 was still ticking just fine and the 4770k is only 20% faster.

I am not saying this is a bad CPU. In fact, it is the best out at the moment aside from the server Xeons or the outrageously expensive X series x79 chips. I can see the way the new 4700k is a lot better refined and the 950 was old technology for sure. Haswell follows expectations of 10-15% faster than the previous generation and it can be had for less than $300! Sure there is going to be an X series chip that will come out and cost $1000 that is 10% faster than a Haswell chip...if you're shelling 1 grand for that, you are a lucky man. I'm not sure why a lot are complaining when this is a nice CPU and most likely don't even do anything CPU intensive. Last, the z87 platform is a lot better.

I expect the next generation to be a similar gradual upgrade jump and I think gone are the days of 40%+ CPU performance every generation years ago. I get tired of hearing "Moore's Law" from snotty college kids with zero real world knowledge.


-Nick

I've used the corsair and have had better results with the Swifttech h220 for a few builds for friends.

There are still a lot of games out there that are CPU intensive (two I play are Day-Z and DCS-A10), especially when going for 120hz @ 1080p, I have found real world benefit from 4.2 to 4.7.

Nice write up Nick. I also have the 4770K with the Hero MB.

I was able to hit 4.5 but it needed over 1.30 volts to do so. Heat was under control using the CoolMaster Seidon 240.

I decided it is not worth going over 4.4g at 1.275 volts. At that speed I can run OCCT or prime all day and night and never go over 70C. It is 100% stable so I am going to keep it there.

I will play with the new memory I bought I picked up some great GSkill 1866 at 1.5V that was on the Hero QVL.

I am pretty sure I can raise it to 1.6v and hit at least 2000 or 2133.

I forgot to mention. The RAM is working at 2400 with 6.5v flawlessly regardless of the overclock I do. I read that supposedly some chips will not overclock well if you have 1600+ and it is recommended to start with that. I haven't had any issues with 32g of Corsair Vengeance Pro (red) at 2400....good because that darn stuff cost nearly $400!

What settings are you guys using on your hero?

I'm stable at 4.4ghz (42 cache) with the following:
CPU: 1.28v
Cache: 1.20v
CPU Input: 1.9v
Memory: 1200 (10-12-12-31-2) G.Skill Trident

Intel Burn Test temps: 78/81/80/75 (delid/swiftech 220)
AIDA64 is stable for 3 hours.

Make sure you are using AIDA64 for stability. Linx and Intel burn test passed at 4.6 @ 1.35volts but I crashed in BF3 / AIDA64.

wow that is expensive ram.

Neckemoff said:

The adaptive voltage feature and turbo is actually really nice.

Click to expand...

I haven't found a use for Adaptive voltage yet. I end up with blue screens until I bring Offset voltage up. And at that point I don't need any extra voltage boost for turbo speeds, so adaptive becomes pointless. To be clear, my Asus board presents "Adaptive Voltage" with two parameters: Offset voltage, and Additional Turbo Voltage.

Have you had success with Adaptive Voltage? What settings are you using?

Womper, Adapative voltage is required to allow you to use the power saving functionality built into Haswell according to the overclocking articles I've read.

IE - my 4.5Ghz Haswell overclock uses a max of about 2 watts at idle with adaptive turned on. if I enable speedstep it's in the 1 watt range.

Neckemoff what kind of test are you using to stress your cpu?

With Intel Burn Test it looks like most non delidded 4770ks that are decent will reach close to max temps of 100C with Intel Burn Test or Prime95 when over clocked to 4.6GHZ with any voltage over 1.25v. This is with good ambient temp water cooling or air cooling.


Have you delidded your 4770k?

The RAMdisk has some insane performance. It’s literally 5000k+ random 4k reads.

I’ve been able to use adaptive voltage by adding the appropriate offset; otherwise, it doesn’t seem to work otherwise with anything about 4.2GHZ. Yes, this adaptive voltage feature is part of what makes the CPU great for mobile.

I have used AIDA64, Prime95, and Intel Extreme Tuning Utility to benchmark. I find that the Intel tool will pass the CPU or it can run longer before crash. Prime95 is the easiest way to make it crash. AIDA64 is slightly less effective than Prime95 (literally a CPU killing program!) in getting a BSOD, but it seem to detect “hardware failure” and stop the stress test before then. I suggest running the Intel tool first and if that passes, run AIDA64. Prime95 seems to synthetically overheat the CPU and I don’t think it’s a realistic indicator of real world performance.

I have not delidded my CPU and don't think I will. Coming from a technical background, the only reason replacing the thermal paste may help is if it was applied incorrectly (doubtful...but you'll notice and should RMA it) or the wrong stuff was used (also doubtful, but perhaps a cheaper material is possible). Now I may be wrong, but replacing the factory paste will have unforeseeable consequences. You might get a few degrees cooler CPU at the cost of reliability and perhaps requiring you to reapply the paste periodically. I have no doubt you can put a compound on the die that act as a better heat transfer, but that compound may has issues like drying up. And delidding means your warranty is gone. Kudos to whoever manages to do this successfully though! I have heard others say the factory paste needs time to break-in before it functions better (i.e., temps will drop later). Myth? Probably.

I found that the temps don’t go above high 80’s under the highest load using a Corsair H110 and 1.35v. Anything higher and it gets to 88 and throttles. Seems adding voltage exponentially increases temps. Using 1.375 (only way to keep mine 100% stable at 4.6GHZ) and with an extreme Prime95 workload, it steadily goes to throttle point and the temps hit failsafe levels after 6-8 hours causing the computer to cut off. I think if I had better cooling, it would be doable, but the spikes of 90C from idle when just clicking around Windows 8 with 1.375 has me thinking it’s not worth it. It would be okay if it didn’t spike so high since the synthetics tests will never actually occur in real-world performance, but hearing your computer groan from CPU fans kicking in isn't appealing. I doubt even the highest end custom water loop can overcome these spikes…seems to be the nature of the beast. Perhaps liquid nitrogen will keep these spikes at bay, but probably not water!

Using 1.25v leads to very comfortable temps and performance. 4.5Ghz stability is no problem and I honestly think this is the highest and most comfortable overclock on my 4770k. Temps stay well under 80c under the highest Prima95 load. They don’t spike to the point of groaning either. I haven’t ran a long term stability on this though. My conclusion so far is mine can do 4.6Ghz with 1.375, but heat becomes the problem.

Last, when considering everything, I think 4.2 is a good overclock for these chips. It says cool, all the power saving features work, and it is quiet. It may seem like “man is CPU sucks…it can’t overclock,” but though overclocking is fun and all, I don’t think you reap any real world benefits from going 4.2 > 4.6 especially how new this CPU is and already overkill for everything. Your best bet is to buy the latest and greatest CPU after 1-3 years or when you believe you need an overclock to keep up. Buying a $1+ custom water cooling kit might be fun, but maybe that money is best used for other things...it really depends on your needs and what you are doing.

I'm surprised it hasn't caught on fire yet.

Just for the heck of it I ordered another 4770k from Amazon and wow what a disappointment! This one will not even boot at 4.5ghz with 1.4v! It says made in Costa Rica and the original in Malaysia...the Costa Rica one was from Amazon and the other was from Buy.com.

Neckemoff said:

Just for the heck of it I ordered another 4770k from Amazon and wow what a disappointment! This one will not even boot at 4.5ghz with 1.4v! It says made in Costa Rica and the original in Malaysia...the Costa Rica one was from Amazon and the other was from Buy.com.

Click to expand...

I have been reading about people waiting for Costa Rica 4770k because it is supposed to be higher quality. Doesn't look it so far.


I think they would be the same granted the same quality of materials and fabrication process is used.

4770k here, Malay batch# L312B323. I am getting a stable 4.5GHz @ 1.25v core (adaptive), 1.18v ring (adaptive), 45x multi on core, 42x multi on ring.

Cannot seem to get her stable at 4.6 and unwilling to go past 1.30v as AIDA64 temps spike to 100c and CPU begins to throttle.

I was a little disappointed at first but the more reading I do it seems I actually have a decent chip. And yes, at 1080p 120Hz I certainly AM noticing a difference in gaming framerates with dual 7970's compared to my old 4.13GHz i7 920.

Great chip and platform, I am glad I made the upgrade!

For the record the Intel Extreme Tuning Utility Stress Test results do NOT jive with the prime95 stress test results.

The longest I tried the IETU stress testing was for 30 minutes. I passed at 4.5Ghz at 1.85 volts on my Haswell 4770K. I saw in one of these threads to verify with prime95. In prime 95 it BSOD'd nearly immediately. I upped the voltage to 2.0 volts - it BSOD'd in nearly 10 seconds. I uppped it to 2.5volts. It BSOD'd in about 30 seconds.

Prime95 is either WAY more taxing - even though both show all cores at 100% - or there is a bug in my prime95 that causes BSOD. FWIW - at 1.85volts that Intel Extreme Tuning Utility vouched for I've never had a problem in games or regular PC use.

You are seriously running 1.85-2.0v? Jesus Christ you are going to kill that thing. Are you sure you have that right?

Archaea said:

For the record the Intel Extreme Tuning Utility Stress Test results do NOT jive with the prime95 stress test results.

The longest I tried the IETU stress testing was for 30 minutes. I passed at 4.5Ghz at 1.85 volts on my Haswell 4770K. I saw in one of these threads to verify with prime95. In prime 95 it BSOD'd nearly immediately. I upped the voltage to 2.0 volts - it BSOD'd in nearly 10 seconds. I uppped it to 2.5volts. It BSOD'd in about 30 seconds.

Prime95 is either WAY more taxing - even though both show all cores at 100% - or there is a bug in my prime95 that causes BSOD. FWIW - at 1.85volts that Intel Extreme Tuning Utility vouched for I've never had a problem in games or regular PC use.

Click to expand...

I hope you mistyped with those volt numbers because those are way too high. Are you talking about your core CPU volts? 1.8v will kill a cpu on anything other than LN2.

Archaea said:

For the record the Intel Extreme Tuning Utility Stress Test results do NOT jive with the prime95 stress test results.

The longest I tried the IETU stress testing was for 30 minutes. I passed at 4.5Ghz at 1.185 volts on my Haswell 4770K. I saw in one of these threads to verify with prime95. In prime 95 it BSOD'd nearly immediately. I upped the voltage to 1.20 volts - it BSOD'd in nearly 10 seconds. I uppped it to 1.25 volts. It BSOD'd in about 30 seconds.

Prime95 is either WAY more taxing - even though both show all cores at 100% - or there is a bug in my prime95 that causes BSOD. FWIW - at 1.185volts that Intel Extreme Tuning Utility vouched for I've never had a problem in games or regular PC use.

Click to expand...

Sorry I don't know what happened there - I really messed up typing the voltages -
yes I've never taken my CPU above 1.25 volts.... I've fixed the above in red.

Neckemoff said:

I have not delidded my CPU and don't think I will. Coming from a technical background, the only reason replacing the thermal paste may help is if it was applied incorrectly (doubtful...but you'll notice and should RMA it) or the wrong stuff was used (also doubtful, but perhaps a cheaper material is possible). Now I may be wrong, but replacing the factory paste will have unforeseeable consequences. You might get a few degrees cooler CPU at the cost of reliability and perhaps requiring you to reapply the paste periodically. I have no doubt you can put a compound on the die that act as a better heat transfer, but that compound may has issues like drying up. And delidding means your warranty is gone. Kudos to whoever manages to do this successfully though! I have heard others say the factory paste needs time to break-in before it functions better (i.e., temps will drop later). Myth? Probably.

Click to expand...

People on XS have shown results of delidding, temp drops of over 30c running prime95 when using the right thermal compounds. I guess you can call that nothing but it seems quite significant to me.

Justintoxicated said:

People on XS have shown results of delidding, temp drops of over 30c running prime95 when using the right thermal compounds. I guess you can call that nothing but it seems quite significant to me.

Click to expand...

I agree.

Neckemoff said:

I expect the next generation to be a similar gradual upgrade jump and I think gone are the days of 40%+ CPU performance every generation years ago. I get tired of hearing "Moore's Law" from snotty college kids with zero real world knowledge.

Click to expand...

I agree with this and is in part to no AMD competition pushing Intel. Competition is in the end good for the consumer.

wow 30c difference, that is a real kick in the balls to Intel. What a poor job they did with the hardware.

Update: Tested 4 more 4770k CPU's. Please don't hate me Amazon! Results have me feeling good about the original CPU in the original post...all of the others were worse. If anyone is curious, I got a completely different batches and manufacture locations by othering multiple chips from Amazon at once...seems the stocks are mixed batches on Amazon.

2 were Costa Rica and 2 were Malaysia. Seems like the Costa Rica chips were strangely a good margin worse...they required a lot of voltage to do even 4.2ghz or 4.4ghz.

I noticed setting the dram freq matters. It would generally be more stable with a lower dram overclock. If you're trying to get 4.7 or 4.8, try decreasing dram to 1600.

Strangely, 2 of the CPUs got Clock Watchdog Timeout BSODs, which couldn't be fixed by increasing voltage. They got this at 4.4 or 4.2. Both were Costa Rica.

The best chips seem to be batch L311 (Malaysia). Oddly, I got 2 of these in a similar batch numbers (end digits were different) and each preformed similarly, but one of them (the original) was better. Reading feedback, I think it is agreed that this is a better batch, but not the best. I suppose above average is fair to say.

Out of all of them, the chip in the original post was the only one able to do 4.8 ghz at a high voltage. The rest simply couldn't boot at this overclock at all despite any setting.

Honestly, I am a bit shocked in the variance in these chips. From my standpoint, if you're getting at least 4.5ghz at 1.35v, you've got an above average chip. If your chip can do 4.7 or 4.8 at any voltage, I think you probably are getting a rare good chip and heat will be the bottleneck rather than overclockablity. Seems some chips simply can't go above a certain overclock...but this lottery style gamble has been in previous generations, though the variance might be higher this time around. If you can do 4.5 at 1.25v or 4.6 at 1.375, hold on to that SOB...you got a nice chip.

It might not be the best chip, but my testing makes me glad i didn't get some of these others... Not sure if they are defective, but they can't overclock much.

Can people stop spreading bullshit about the "bad" compound that intel uses on the die? Delidding helps temps NOT because of a TIM swap. It helps because of the gap reduction between IHS and die! By delidding, you get rid of that black gunk and slather of TIM that is creating an offset between the IHS and die, allowing to close the gap and creating better contact. It has been tested by "idontcare" on anandtech forums in a very scientific and controlled manner.

"Notice that the Intel stock CPU TIM outperforms the NT-H1 replacement TIM once the CPU-to-IHS gap is identical" (idontcare)

Sorry, I'm just angry that people don't understand WHY delidding is helping their temps and just assume it is the TIM. IT IS NOT!

http://forums.anandtech.com/showthread.php?t=2261855
http://forums.anandtech.com/showpost.php?p=34053183&postcount=570

Get reading and inform yourselves and stop spreading BS please.

Neckemoff said:

Update: Tested 4 more 4770k CPU's. Please don't hate me Amazon! Results have me feeling good about the original CPU in the original post...all of the others were worse. If anyone is curious, I got a completely different batches and manufacture locations by othering multiple chips from Amazon at once...seems the stocks are mixed batches on Amazon.

2 were Costa Rica and 2 were Malaysia. Seems like the Costa Rica chips were strangely a good margin worse...they required a lot of voltage to do even 4.2ghz or 4.4ghz.

I noticed setting the dram freq matters. It would generally be more stable with a lower dram overclock. If you're trying to get 4.7 or 4.8, try decreasing dram to 1600.

Strangely, 2 of the CPUs got Clock Watchdog Timeout BSODs, which couldn't be fixed by increasing voltage. They got this at 4.4 or 4.2. Both were Costa Rica.

The best chips seem to be batch L311 (Malaysia). Oddly, I got 2 of these in a similar batch numbers (end digits were different) and each preformed similarly, but one of them (the original) was better. Reading feedback, I think it is agreed that this is a better batch, but not the best. I suppose above average is fair to say.

Out of all of them, the chip in the original post was the only one able to do 4.8 ghz at a high voltage. The rest simply couldn't boot at this overclock at all despite any setting.

Honestly, I am a bit shocked in the variance in these chips. From my standpoint, if you're getting at least 4.5ghz at 1.35v, you've got an above average chip. If your chip can do 4.7 or 4.8 at any voltage, I think you probably are getting a rare good chip and heat will be the bottleneck rather than overclockablity. Seems some chips simply can't go above a certain overclock...but this lottery style gamble has been in previous generations, though the variance might be higher this time around. If you can do 4.5 at 1.25v or 4.6 at 1.375, hold on to that SOB...you got a nice chip.

It might not be the best chip, but my testing makes me glad i didn't get some of these others... Not sure if they are defective, but they can't overclock much.

Click to expand...

Thanks for the info and testing. Makes me feel a little better about my chip. Felt it was a did. I'm at 1.35v.

Neckemoff said:

I expect the next generation to be a similar gradual upgrade jump and I think gone are the days of 40%+ CPU performance every generation years ago. I get tired of hearing "Moore's Law" from snotty college kids with zero real world knowledge.

-Nick

Click to expand...

Moore's Law are fine but thanks to certain company who can't produce any resonably fast cpu the additional transistors are wasted on shitty iGPUs race.

Looking at transistor count we should be getting 6 core desktop Haswells if whole die area was cpu

Dizon248 said:

Can people stop spreading bullshit about the "bad" compound that intel uses on the die? Delidding helps temps NOT because of a TIM swap. It helps because of the gap reduction between IHS and die! By delidding, you get rid of that black gunk and slather of TIM that is creating an offset between the IHS and die, allowing to close the gap and creating better contact. It has been tested by "idontcare" on anandtech forums in a very scientific and controlled manner.

"Notice that the Intel stock CPU TIM outperforms the NT-H1 replacement TIM once the CPU-to-IHS gap is identical" (idontcare)

Sorry, I'm just angry that people don't understand WHY delidding is helping their temps and just assume it is the TIM. IT IS NOT!

http://forums.anandtech.com/showthread.php?t=2261855
http://forums.anandtech.com/showpost.php?p=34053183&postcount=570

Get reading and inform yourselves and stop spreading BS please.

Click to expand...

So what you're saying is the intel heat sink gets closer to the die is what causes temps to drop (rather than using a different thermal paste)? I am not so sure how that can happen when it still needs to fit in the same CPU slot, which I can't imagine will tension the lid lower... maybe possible because I did notice a tight fit. I looked at my CPUs and the black rubber stuff is really thin... perhaps reducing this tiny gap means less thermal compound and closer contact to the intel heat sink, but despite the reports of "-30c" after delid, I have a hard time believing this. Maybe a few degrees is more realistic, which I doubt really matters in grant scale of things... its not going to let you go from 1.35v to 1.5v or anything like that to achieve a 4.8 overclock.

To me it seems the CPU temp spike is the problem... no matter what cooling you have, the cores are going to spike to 90c and that heat can't be dissipated fast enough.... with the intel heat sink touching a water cooled copper plate anyway. I suppose closer contact of your cooling system to the die may help.

I got an email from Amazon not happy over me getting a lot of CPUs and returning them, so please be cautious doing what I did!

Neckemoff said:

I got an email from Amazon not happy over me getting a lot of CPUs and returning them, so please be cautious doing what I did!

Click to expand...

What they said?

And have you paid some costs for returning used stuff?

Raghar said:

What they said?

And have you paid some costs for returning used stuff?

Click to expand...

Nah I'll have full refund - it was more of a inquire as to whether there is something wrong with the product and what they can do to help. I sent them my explanation that the chips didn't overclock well.

I buy a lot of stuff and never had an issue with Amazon - always great service!

^^ It's a very prickish thing to do, return all those cpus. You man up and sell them for a loss. I don't have an entitled child speech for you, because I don't normally do it, but you my friend would be very worthy of someone who does.

Dizon248 said:

Can people stop spreading bullshit.....

"Notice that the Intel stock CPU TIM outperforms the NT-H1 replacement TIM once the CPU-to-IHS gap is identical" (idontcare)

Sorry, I'm just angry that people don't understand WHY delidding is helping their temps and just assume it is the TIM. IT IS NOT!

Get reading and inform yourselves and stop spreading BS please.

Click to expand...

Fact of the matter is that most people that delid try a number of different TIMs between IHS and Die and block combinations and most of the people that I have seen that have delidded see a drop in temps when they use Coolaboratory Liquid Pro or Liquid Ultra TIM(CL LP, CL LU).

Right now amongst the delidders that are in the know have experienced the lowest drops in temperatures using CL LP or CL LU.



Have you delidded yet? What has been your experience with delidding and using different TIMs?



Not sure where your negative attitude towards TIM and delidding comes from but for you and anybody that really wants to take some time and do research on delidding the delidded club is the best place that I can think of to read up on delidding and learn more about it.


http://www.overclock.net/t/1313179/official-delidded-club

texuspete00 said:

^^ It's a very prickish thing to do, return all those cpus. You man up and sell them for a loss. I don't have an entitled child speech for you, because I don't normally do it, but you my friend would be very worthy of someone who does.

Click to expand...

To each his own, I guess. I have a couple of 4670k's and one 4770k at the moment to begin dong some "binning." I plan on selling the ones I will not use.

SonDa5 said:

Fact of the matter is that most people that delid try a number of different TIMs between IHS and Die and block combinations and most of the people that I have seen that have delidded see a drop in temps when they use Coolaboratory Liquid Pro or Liquid Ultra TIM(CL LP, CL LU).

Right now amongst the delidders that are in the know have experienced the lowest drops in temperatures using CL LP or CL LU.



Have you delidded yet? What has been your experience with delidding and using different TIMs?



Not sure where your negative attitude towards TIM and delidding comes from but for you and anybody that really wants to take some time and do research on delidding the delidded club is the best place that I can think of to read up on delidding and learn more about it.


http://www.overclock.net/t/1313179/official-delidded-club

Click to expand...

^ 100% correct.

Don't speak of delidding, unless you've actually delidded your chip, and used CL LP or LU. It makes a massive difference.

texuspete00 said:

^^ It's a very prickish thing to do, return all those cpus. You man up and sell them for a loss. I don't have an entitled child speech for you, because I don't normally do it, but you my friend would be very worthy of someone who does.

Click to expand...

I don't have a problem with what he did. He bought a product which didn't perform to his satisfaction and the vendor that he bought it from honored his negative experience and dissatisfaction with the product and honored the refund.


I am on my 2nd 4770k and I exchanged my first one for the current one that I have because the first one didn't perform to my satisfaction. The vendor that I bought from had a 15 day return or exchange policy for customer satisfaction and I took advantage of it.

Neckemoff said:

So what you're saying is the intel heat sink gets closer to the die is what causes temps to drop (rather than using a different thermal paste)? I am not so sure how that can happen when it still needs to fit in the same CPU slot, which I can't imagine will tension the lid lower... maybe possible because I did notice a tight fit. I looked at my CPUs and the black rubber stuff is really thin... perhaps reducing this tiny gap means less thermal compound and closer contact to the intel heat sink, but despite the reports of "-30c" after delid, I have a hard time believing this. Maybe a few degrees is more realistic, which I doubt really matters in grant scale of things... its not going to let you go from 1.35v to 1.5v or anything like that to achieve a 4.8 overclock.

To me it seems the CPU temp spike is the problem... no matter what cooling you have, the cores are going to spike to 90c and that heat can't be dissipated fast enough.... with the intel heat sink touching a water cooled copper plate anyway. I suppose closer contact of your cooling system to the die may help.

Click to expand...

I have honestly read mixed results so this could be true, unless using some particular compounds, and from my briff reading on the issue, the temp drop might only be 5-8 degrees. I spose it could be because you have reduced the gap as that would make sense. In either case though it seems an OBVIOUS area for improvement. I mean they keep trying to make these things more efficient and run cooler yet they neglect to fix one of the most obvious problems, except on their higher end CPUs. If I delid the IHS isn't going back on anyways so it really wouldn't matter.

The 30c temp drop was using Coollaboratory Liquid Pro which is basically liquid metal and becomes permanent.

Honestly, why even care. It may be interesting but whether or not its bc of better TIM or the gap, its a reliable 10-20c drop in temps from a delid. I happen to also think its the gap but there's no point in arguing one way or another.

Batch# L307B240
4.8 ghz/1.445V/2.0VCCIN/4.0Ring/1.2RingV. Undelided. Only ran Aida for 30 min on cpu test. Max core temps of 82C.


Can do 4.7 ghz 24/7 using EIST/enhanced turbo/adaptive voltage/cpu offset +0.15.
Most likely wont be able to do this even at 4.8 cause the cache voltage spikes at 1.49v and the vcore hits 1.45v with 4.7ghz during Aida testing.

I've come to the conclusion that unless I find some way to get those voltages down there's no point deliding. Heat's not going to kill it, the voltage will.

Rob94hawk said:

Batch# L307B240
4.8 ghz/1.445V/2.0VCCIN/4.0Ring/1.2RingV. Undelided. Only ran Aida for 30 min on cpu test. Max core temps of 82C.

Can do 4.7 ghz 24/7 using EIST/enhanced turbo/adaptive voltage/cpu offset +0.15.
Most likely wont be able to do this even at 4.8 cause the cache voltage spikes at 1.49v and the vcore hits 1.45v with 4.7ghz during Aida testing.

I've come to the conclusion that unless I find some way to get those voltages down there's no point deliding. Heat's not going to kill it, the voltage will.

Click to expand...

Have you tried the FPU test on Aida? I don't think you'll be able to keep running 1.445V without delidding once you start running tests that use AVX extensions.

And OP, Clock Watchdog Errors (which are x124 errors) don't seem to be Vcore related. I've done extensive testing because my chip gets a lot of them, and reducing the cache speed or increasing VCCIOD (and VCCIOA) stop them much better than Vcore. For example, at 4.5 I get 124 errors at any voltage up to 1.4V, but if I increase the VCCIOD to +0.05 I can get 4.5 stable at 1.37V.

Rob94hawk said:

Batch# L307B240
4.8 ghz/1.445V/2.0VCCIN/4.0Ring/1.2RingV. Undelided. Only ran Aida for 30 min on cpu test. Max core temps of 82C.


Can do 4.7 ghz 24/7 using EIST/enhanced turbo/adaptive voltage/cpu offset +0.15.
Most likely wont be able to do this even at 4.8 cause the cache voltage spikes at 1.49v and the vcore hits 1.45v with 4.7ghz during Aida testing.

I've come to the conclusion that unless I find some way to get those voltages down there's no point deliding. Heat's not going to kill it, the voltage will.

Click to expand...

Damn, I'm shying away from over 1.4v. You have it 1.45v...

SonDa5 said:

Fact of the matter is that most people that delid try a number of different TIMs between IHS and Die and block combinations and most of the people that I have seen that have delidded see a drop in temps when they use Coolaboratory Liquid Pro or Liquid Ultra TIM(CL LP, CL LU).

Right now amongst the delidders that are in the know have experienced the lowest drops in temperatures using CL LP or CL LU.



Have you delidded yet? What has been your experience with delidding and using different TIMs?



Not sure where your negative attitude towards TIM and delidding comes from but for you and anybody that really wants to take some time and do research on delidding the delidded club is the best place that I can think of to read up on delidding and learn more about it.



http://www.overclock.net/t/1313179/official-delidded-club

Click to expand...

Did those guys who try different TIMs control for gap distance like idontcare did? He tested a whole bunch of TIMs INCLUDING CLP/CLU, which does show an extra decrease in temps when normalizing all TIMs to the same gap distance. But the temps are NOT soley attributed to the TIM, it is also a factor of distance. Distance between die and IHS is actually a much larger factor than what type of TIM you use.

http://forums.anandtech.com/showpost.php?p=34053183&postcount=570

So after all that testing, it is found that intel's stock TIM is better than Noctua's NH-1 TIM when controlling for gap between die and IHS.

http://forums.anandtech.com/showpost.php?p=34297433&postcount=67

Using these 2 links, you can extrapolate the performance of CLU/CLP vs Intel's TIM by comparing CLU/CLP's performance to Noctua's NH-1 WHEN controlling for gap between die and IHS!

Conclusion? Reduction of gap matters more than TIM being used, a huge factor. However, using TIM worse than Intel's TIM will hurt the gains you SHOULD HAVE expected due to gap reduction.

As far as the "negative attitude", I don't have one towards delidding. I just don't like it when people tell others false information about why delidding helps temps. A bunch of people believe it is soley due to TIM, not true. It is due to gap/height reduction between IHS and die. That is all I wanted to prove.