4670k & 4770k purchased anwhere operating performance thread

Very nice results. My 4670K is prime95 stable for hours on end at 1.226v/4.5ghz and I wonder if that makes it a good candidate for delidding?

Personally I would take the plunge, but then again SonDa5 and me have an addiction to de-lidding chips heheh...I have done it to every IB build I've done, because it makes the cpu's be able to hang a decent 4-4.4Ghz o/c with just a cheap $25 HSF and not have to have a million CFM of air moving around to keep 'em cool..

I would imagine that you would be very happy with your results..If you can get your chip to 4.7Ghz @ 1.35V or less, I would pop that top!:cool:
 
How is tweaking going for everyone?


The more i play with the voltage the more I am discovering how complex Haswell is to over clock but I am having some success and I can tell you all the this chip requires alot of time in the BIOS(for me it has) to learn all the voltages and how they effect each other.

Coming from a 5GHZ 3770k it was a hard jump to haswell but I finally feel like it was a good move.
 
Well... I have a second 4770k coming. My current with some tweaking isnt going over 4.5ghz at 1.3v my motherboard is a MSI gaming Z87-GD65.

If the next CPU does well then Ill consider delidding it. Im not too sure if I want to yet. A little nervous with it. I have a razer blade to do it... but dontknow yet. I have a raystorm block that is light.
 
Well... I have a second 4770k coming. My current with some tweaking isnt going over 4.5ghz at 1.3v my motherboard is a MSI gaming Z87-GD65.

If the next CPU does well then Ill consider delidding it. Im not too sure if I want to yet. A little nervous with it. I have a razer blade to do it... but dontknow yet. I have a raystorm block that is light.

Did you play with all of your voltages in your BIOS? What Intel Microcode are you running in your BIOS?
 
To be honest I am new to overclocking haswell. What voltage should I play with? I can post screen shots I think. I've only played with vcore and multiplier and locked ram at 1600.
 
To be honest I am new to overclocking haswell. What voltage should I play with? I can post screen shots I think. I've only played with vcore and multiplier and locked ram at 1600.

Depending where you end up with the Vcore, you may also need to increase the VRIN (CPU Input Voltage). You normally want to keep that 0.4 to 0.5 above the Vcore. You'll also probably want to increase Vring (Cache Voltage) to make sure you aren't getting cache or ring bus issues.
 
What is a microcode?


Patched in BIOS that controls behavior of cpu. INTEL has put out a few of them and I think there are some custom ones floating around. Don't a search here in forums and there is info about how to edit bios with different micro codes.
 
Depending where you end up with the Vcore, you may also need to increase the VRIN (CPU Input Voltage). You normally want to keep that 0.4 to 0.5 above the Vcore. You'll also probably want to increase Vring (Cache Voltage) to make sure you aren't getting cache or ring bus issues.

No luck. I'm stuck a 4.5ghz. This sucks a lot!
 
i5-4670k purchased at Microcenter
L311B463
Made in Malayasia
Gigabyte Z87X-UD4H w/F7 bios

Seems relatively stable at 45x100 @1.2v
46x100 boots to windows but crashes on stress testing
Temps hitting 90C with a seidon 120XL (38mm rad CLC) are worrying me.
Think it's worth delidding?
Would it be possible to direct-die cool it with the seidon CLC?

 
i5-4670k purchased at Microcenter
L311B463
Made in Malayasia
Gigabyte Z87X-UD4H w/F7 bios

Seems relatively stable at 45x100 @1.2v
46x100 boots to windows but crashes on stress testing
Temps hitting 90C with a seidon 120XL (38mm rad CLC) are worrying me.
Think it's worth delidding?
Would it be possible to direct-die cool it with the seidon CLC?





I recommend Direct Die only with high quality light weigth water blocks with light mount pressure.

Nice voltage for the clocks. Delidding will more than likely help drop temps.
 
I recommend Direct Die only with high quality light weigth water blocks with light mount pressure.

Nice voltage for the clocks. Delidding will more than likely help drop temps.

How delicate are the Haswell packages? I applied more than a few heat sinks to K7's back in the day, it took more than just a little pressure to crack those dies. In fact you could apply a fair bit of pressure as long as you kept it evenly distributed.

Does anyone happen to know the exact thickness of the raised die over the substrate? Shouldn't be too hard to make a shim as long if I know how thick it needs to be.
 
Im afraid to use a razor blade to delid the cpu. I could go to a friends to use his vice and hammer...I'd rather return the chip to store.
 
How is tweaking going for everyone?


The more i play with the voltage the more I am discovering how complex Haswell is to over clock but I am having some success and I can tell you all the this chip requires alot of time in the BIOS(for me it has) to learn all the voltages and how they effect each other.

Coming from a 5GHZ 3770k it was a hard jump to haswell but I finally feel like it was a good move.

Personally I'm firming up two options for core speed/voltage:
4.7GHz with adaptive voltage. 1.3v has been fine, 1.275v might have an issue (unsure if the bsod was caused by the cache), and 1.25v is probably too low (but also may have been due to the cache). In AVX workloads that 1.3v increases to ~1.4v, so I don't plan on running AVX stuff in this configuration.

4.8GHz with manual 1.34v. I have not tested this on AVX workloads, but it hasn't blue screened yet in any of my games or non-AVX stress tests.

Cache speed is at 4.5GHz with a 1.250v - 1.3v adaptive in the above configurations. It's possible the cache @ 1.275v (or less) is causing rare bluescreens after several hours of testing. DDR is at 2133.

With a Swiftech H220, I peak in the mid-80C during the Intel Extreme Tuner stress test for that 1.34v config, otherwise under 80C. Games don't break 70C in any of these configs.
 
Personally I'm firming up two options for core speed/voltage:
4.7GHz with adaptive voltage. 1.3v has been fine, 1.275v might have an issue (unsure if the bsod was caused by the cache), and 1.25v is probably too low (but also may have been due to the cache). In AVX workloads that 1.3v increases to ~1.4v, so I don't plan on running AVX stuff in this configuration.

4.8GHz with manual 1.34v. I have not tested this on AVX workloads, but it hasn't blue screened yet in any of my games or non-AVX stress tests.

Cache speed is at 4.5GHz with a 1.250v - 1.3v adaptive in the above configurations. It's possible the cache @ 1.275v (or less) is causing rare bluescreens after several hours of testing. DDR is at 2133.

With a Swiftech H220, I peak in the mid-80C during the Intel Extreme Tuner stress test for that 1.34v config, otherwise under 80C. Games don't break 70C in any of these configs.

Wish I had your CPU. Mine maxes at 4.5 Im going to purchase another one and sell this one locally. Man, overclockers problem lol
 
Seems relatively stable at 45x100 @1.2v
46x100 boots to windows but crashes on stress testing
Temps hitting 90C with a seidon 120XL (38mm rad CLC) are worrying me.

Unless you plan on running Linpack constantly, stop worrying about your temps. If you want a worst-case scenario stress test, the latest version of Linpack is your ticket- you will throttle at voltages even less than 1.2v (but get 160 GFlops in the process).

Bottom line- Haswell overclocks should be tuned specifically for what you actually use your computer for, otherwise you're leaving performance on the table. Just keep Adaptive mode and Offset mode voltages under 1.3v- if you inadvertently encounter an AVX workload, that 1.3v becomes 1.4v.

Wish I had your CPU. Mine maxes at 4.5 Im going to purchase another one and sell this one locally. Man, overclockers problem lol

Maybe I'm lucky this time around. I still don't know what par is for haswell overclocking. Keep in mind that my 4.7 and 4.8GHz overclocks cannot be used with Prime95, Linpack, and other AVX based apps- 4.8GHz will crash due to low voltage and 4.7GHz will bump to 1.375v-1.4v and throttle instantly. But hey, Planetside 2 loves that 4.8GHz and stays under 70C, so we're good.

Regardless, Haswell is by far the most fun I have ever had with overclocking. So many things to tweak.
 
Unless you plan on running Linpack constantly, stop worrying about your temps. If you want a worst-case scenario stress test, the latest version of Linpack is your ticket- you will throttle at voltages even less than 1.2v (but get 160 GFlops in the process).

Bottom line- Haswell overclocks should be tuned specifically for what you actually use your computer for, otherwise you're leaving performance on the table. Just keep Adaptive mode and Offset mode voltages under 1.3v- if you inadvertently encounter an AVX workload, that 1.3v becomes 1.4v.



Maybe I'm lucky this time around. I still don't know what par is for haswell overclocking. Keep in mind that my 4.7 and 4.8GHz overclocks cannot be used with Prime95, Linpack, and other AVX based apps- 4.8GHz will crash due to low voltage and 4.7GHz will bump to 1.375v-1.4v and throttle instantly. But hey, Planetside 2 loves that 4.8GHz and stays under 70C, so we're good.

Regardless, Haswell is by far the most fun I have ever had with overclocking. So many things to tweak.

I guess its hard for me to not have it prime stable lol. something inside of me refuses. But I just got rid of my 4770k locally and used some gift cards to order obe from amazon that is coming thursday. I hope I get somewhat lucky.
 
I have an I7 4770k cpu. I ran 4.4GHZ @ 1.28v. I went to prime95. when I went few later on prime95, core 0 is 100, then 1-4 is double digit. it is like this:

100-95-96-85

I am wondering why core 0 is high than core 1-3.

how can I fix it?

my cooler is h80 w/ SP120 (non-pwm). I am wondering which one is better for i7 4770k? sp120 (P/P) or F12 ( one fan push)
 
is that normal? do I need to reseat the H80?

It's normal as far as I know. I see a 10-15C difference between Core 0 and Core 3 (#3 being the coolest all the time). This is true on both a hsf and the all-in-one water cooler I now have.
 
I change the fan few times, second time I change to stock fans and it works well. now I am using F-12 push. it works okay.
 
I have an I7 4770k cpu. I ran 4.4GHZ @ 1.28v. I went to prime95. when I went few later on prime95, core 0 is 100, then 1-4 is double digit. it is like this:

100-95-96-85

I am wondering why core 0 is high than core 1-3.

how can I fix it?

my cooler is h80 w/ SP120 (non-pwm). I am wondering which one is better for i7 4770k? sp120 (P/P) or F12 ( one fan push)

on my set up, Core 1 has consistently been the hottest while Core 3 has consistently been the coolest

temp difference between the 2 tend to be 7-8C for me

100-95-96-85

pretty sure you should not be getting temps this high with a H80 and running Prime 95. can possibly make sense if running some stress test with AVX but not with Prime 95.

might want to consider reseating your H80
 
on my set up, Core 1 has consistently been the hottest while Core 3 has consistently been the coolest

temp difference between the 2 tend to be 7-8C for me



pretty sure you should not be getting temps this high with a H80 and running Prime 95. can possibly make sense if running some stress test with AVX but not with Prime 95.

might want to consider reseating your H80

Prime95 uses AVX1, which will still force Haswell CPUs to apply more Vcore..He could try a reseat, but the H80 really isn't a super AIO cooler compared to the newer larger offerings on the market..
 
This is my second 4770K but both don't compare to some of the chips people are posting about. My definition of stability is several individual 24 hour runs of Prime95 (AVX), h264 encoding, the latest AVX2 Linpack, etc. with FurMark or AIDA64 GPU Stress test running simultaneously. If all I wanted to do was a validation or AIDA64's entire suite of "torture" tests then my voltages would be much lower.

Batch L312B515 (Others with this batch report 5Ghz results but not me.)

Here are my stable results (unless otherwise noted then everything is at AUTO, uncore is locked at 36x, memory is DDR3-1333Mhz. I am using a Swiftech H220 cooler.):

4.2 VCORE 1.20v VRING 1.10v @ 36x
4.3 VCORE 1.21v VRING 1.15v @ 36x
4.4 VCORE 1.26v VRING 1.15v @ 36x VCCSA +0.10v VRIN 1.81v
4.5 VCORE 1.35v VRING 1.15v @ 36x VCCSA +0.10v VRIN 1.83v

At 4.5Ghz, the CPU will thermal throttle during IBT, AVX2 Linpack, etc. At 4.4Ghz and lower, there is no thermal throttling. I haven't bothered trying 4.6Ghz because the voltage bump is probably going to be more than I'm comfortable with.

I could delid, add another radiator, and spend lots of money but it is too much work and the return is questionable at best. Honestly, I think 4.3Ghz is the sweet spot for this particular CPU. I'll probably bump up the uncore and run the memory at 2133Mhz (the rated speed) and call it a day. Thoughts?
 
Last edited:
Prime95 uses AVX1, which will still force Haswell CPUs to apply more Vcore..He could try a reseat, but the H80 really isn't a super AIO cooler compared to the newer larger offerings on the market..



what does this mean: newer larger offerings on the market? what cooler u prefer to?
 
I'd de lid it man.

I thought about it but I'm already uncomfortable with a 1.35v vcore for 24/7 use. I suspect 4.6Ghz will take well north of 1.4v so I don't think I'd push it there.

Delidding would, at best, buy me a cooler (non-throttled) 4.5Ghz overclock. At 4.4Ghz, it's running between 77-81C max on the most demanding stress tests. 100Mhz isn't worth voiding the normal warranty and "overclocking warranty" on this particular CPU.

Dropping it down to 4.3Ghz buys me even more thermal headroom so I might tear out the H220 and drop in a Noctua NH-U12S air cooler. I think I am done having my "fun" with two 4770K CPUs. I suppose I can consider delidding next year if/when Broadwell is coming out on the desktop but I can pick up Intel CPUs for 50% off MSRP whenever I want so I think I'll head to Ivy Bridge-E next.

what does this mean: newer larger offerings on the market? what cooler u prefer to?

He's referring to the Corsair H100i, Swiftech H220, etc. that have 240mm (2x140) or larger radiators. I played with many mainstream cooling solution out there and I didn't find the H80i to be that much better than a Noctua NH-U12S. I'm currently using the Swiftech H220 and it is better than the H100i.
 
what does this mean: newer larger offerings on the market? what cooler u prefer to?

I have always used custom WC loops since I got out of using HSF units for air cooling..but the H100(i),NZXT Kraken, Swiftech H220 etc (as hrana mentioned) are the cream of the crop when it comes to AIO WC units, and can be picked up for great prices if you watch the sales..You could even sell your H80 for $30-35 and recoup some of the costs...

I thought about it but I'm already uncomfortable with a 1.35v vcore for 24/7 use. I suspect 4.6Ghz will take well north of 1.4v so I don't think I'd push it there.

I wouldn't have any concerns running @ 1.35V for 24/7 use as long as you can keep the temperatures down to earth (instead of sky high!)..If you don't use AVX 1/2 enabled applications, then I am sure you will be fine, even if you choose not to de-lid that puppy. Personally I would, since the results are wonderful, even if you choose not to increase the speed due to voltage concerns..
 
I thought about it but I'm already uncomfortable with a 1.35v vcore for 24/7 use. I suspect 4.6Ghz will take well north of 1.4v so I don't think I'd push it there.

Delidding would, at best, buy me a cooler (non-throttled) 4.5Ghz overclock. At 4.4Ghz, it's running between 77-81C max on the most demanding stress tests. 100Mhz isn't worth voiding the normal warranty and "overclocking warranty" on this particular CPU.

Dropping it down to 4.3Ghz buys me even more thermal headroom so I might tear out the H220 and drop in a Noctua NH-U12S air cooler. I think I am done having my "fun" with two 4770K CPUs. I suppose I can consider delidding next year if/when Broadwell is coming out on the desktop but I can pick up Intel CPUs for 50% off MSRP whenever I want so I think I'll head to Ivy Bridge-E next.



He's referring to the Corsair H100i, Swiftech H220, etc. that have 240mm (2x140) or larger radiators. I played with many mainstream cooling solution out there and I didn't find the H80i to be that much better than a Noctua NH-U12S. I'm currently using the Swiftech H220 and it is better than the H100i.


Then why say, if you already know best? :rolleyes:

I could delid, add another radiator, and spend lots of money but it is too much work and the return is questionable at best. Honestly, I think 4.3Ghz is the sweet spot for this particular CPU. I'll probably bump up the uncore and run the memory at 2133Mhz (the rated speed) and call it a day. Thoughts?
 
This is my second 4770K but both don't compare to some of the chips people are posting about. My definition of stability is several individual 24 hour runs of Prime95 (AVX), h264 encoding, the latest AVX2 Linpack, etc. with FurMark or AIDA64 GPU Stress test running simultaneously. If all I wanted to do was a validation or AIDA64's entire suite of "torture" tests then my voltages would be much lower.

Batch L312B515 (Others with this batch report 5Ghz results but not me.)

Here are my stable results (unless otherwise noted then everything is at AUTO, uncore is locked at 36x, memory is DDR3-1333Mhz. I am using a Swiftech H220 cooler.):

4.2 VCORE 1.20v VRING 1.10v @ 36x
4.3 VCORE 1.21v VRING 1.15v @ 36x
4.4 VCORE 1.26v VRING 1.15v @ 36x VCCSA +0.10v VRIN 1.81v
4.5 VCORE 1.35v VRING 1.15v @ 36x VCCSA +0.10v VRIN 1.83v

At 4.5Ghz, the CPU will thermal throttle during IBT, AVX2 Linpack, etc. At 4.4Ghz and lower, there is no thermal throttling. I haven't bothered trying 4.6Ghz because the voltage bump is probably going to be more than I'm comfortable with.

I could delid, add another radiator, and spend lots of money but it is too much work and the return is questionable at best. Honestly, I think 4.3Ghz is the sweet spot for this particular CPU. I'll probably bump up the uncore and run the memory at 2133Mhz (the rated speed) and call it a day. Thoughts?

Yeah, 4.3GHz is your ticket. I'm seeing a 10C jump in core temps going from DDR 1600 to 2133, so that 1.26v may begin to see throttling on AVX2 linpack. I also think that some, including myself, don't put their "stable" configs through nearly as much torture testing as you, and/or avoid AVX stress testing since it's not something we plan on encountering. So your numbers are very conservative. It takes a lot of juice to run those AVX instructions, but in return you're getting more performance than you could by overclocking.
 
Well, been tweaking with the modded F7 Bios for my board, and I'm pretty happy with how things are starting to look. Was able to do some benching at 4.8 @1.375V, but was hitting thermal limits, didn't really see if I could lower the voltage much (started high). Sitting at 4.6 @ 1.22, cache at 4.2 @ stock voltage and gaming just fine where as before at 4.5, no cache OC I was running into instability with the NV driver in games. Guess this chip/board are late bloomers. :D
 
Well, been tweaking with the modded F7 Bios for my board, and I'm pretty happy with how things are starting to look. Was able to do some benching at 4.8 @1.375V, but was hitting thermal limits, didn't really see if I could lower the voltage much (started high). Sitting at 4.6 @ 1.22, cache at 4.2 @ stock voltage and gaming just fine where as before at 4.5, no cache OC I was running into instability with the NV driver in games. Guess this chip/board are late bloomers. :D

Did you change anything else (between your early 4.5 OC vs. your current 4.6 @ 1.22)?
I know Sin's guide mentions a pretty broad range for the stock vRing volts (1.0-1.1 for 3.5-4.5 Uncore). I personally wasn't really planning on touching the Uncore at all unless I recognized serious bottlenecking haha.
 
Nothing else changed, just the bios. Picked up the modded one from tweaktown with the updated ucode, and my system is much happier. If I raise the V Ring by .05 V I can get the cache happy at 4.5, but haven't tried much further. It's real slow going testing that one out. It does help out as I'm using the Intel XTU for stress/benching and raising cache has given me a score of 1020 from 1000 even.
 
Can't get past 4.2 with my 4670k. May exchange.

Swapped out my dud 4670k for a 4770k. Rock solid stable (Prime95 for over 4 hours and Intel Burn Test x 20) at 4.5 Ghz @ 1.3 Vcore. Memory at 1866 too.
 
4770k (batch L310B562) is OCCT stable at [email protected]. Having a heck of a time with my UD4H. Updated to the F7 bios and I can't seem to get the same stability I did with the older bios (not sure what version it was).
 
Back
Top