Help me with i7 4770k!

[NirHahs]

what is the common stable voltage for 4770k at 4.5ghz 24/7?
does it need to run 4c/8t?
may i know best batch for this CPU? everybody knows that the stability of the CPU comes from its batch.
I need to know as im planning to get 4770k.

 

[nwo]

For 4.5 the "common" stable coltage is 1.25V for the CPU. 45 multi, 42 cache multi, 1.2v cache, system agent +0.250, I/O Analog +0.1950, I/O Digital +0.1950

Taken from:Overclocker's club

 

[BonzaiDuck]

For 4.5 the "common" stable coltage is 1.25V for the CPU. 45 multi, 42 cache multi, 1.2v cache, system agent +0.250, I/O Analog +0.1950, I/O Digital +0.1950

Taken from:Overclocker's club

Hmm. Interesting. [It should be common knowledge now that I don't own an i7-4770. At least . . . in this thread.]

Sandy was 32nm. Ivy was 22nm. And Haswell . . . is 22nm.

Is it a researchable consensus that 1.25 is proportionally "low enough" as 1.35 seemed to be for Sandy Bridge?

What would the core Tj temperatures be -- on average -- with top-end air-cooling, if it is actually possible? Anyone know firsthand? Without delidding [a range?] or With delidding?

I'm looking at building a Haswell-E when it's released. Probably good that I look more closely at these "K" processors, just for a better foundation . . .

 

[Tweakin]

For 4.5 the "common" stable coltage is 1.25V for the CPU. 45 multi, 42 cache multi, 1.2v cache, system agent +0.250, I/O Analog +0.1950, I/O Digital +0.1950

Taken from:Overclocker's club

That same reference also suggest that most chips have a low chance of anything north of 4.2Ghz. If you look at our own poll, you will see the majority of users on Haswell's fall under 4.5Ghz for stable 24/7 operation.

Speaking for myself, I can't even load windows at 4.5Ghz! 4.4 is barely alive with 1.30v, and that is with the cache at or below 4.0 and no XMP memory profile loaded, which requires additional voltage.

I'm just trying to be realistic here...

 

[nwo]

That same reference also suggest that most chips have a low chance of anything north of 4.2Ghz. If you look at our own poll, you will see the majority of users on Haswell's fall under 4.5Ghz for stable 24/7 operation.

Speaking for myself, I can't even load windows at 4.5Ghz! 4.4 is barely alive with 1.30v, and that is with the cache at or below 4.0 and no XMP memory profile loaded, which requires additional voltage.

I'm just trying to be realistic here...

I used to have constant BSOD with WHEA_UNCORRECTABLE_ERROR on my z87-a with 4770k. I updated the BIOS and I am now up and running at 4.2GHz manual OC without any issues. My temps are actually lower than what they were with a 4.1GHz auto OC.

I will aim for 4.4-4.5 one of these days, but stability is more important than a little bit extra clock speed since I am CPU and GPU mining on this PC.

Hmm. Interesting. [It should be common knowledge now that I don't own an i7-4770. At least . . . in this thread.]

Sandy was 32nm. Ivy was 22nm. And Haswell . . . is 22nm.

Is it a researchable consensus that 1.25 is proportionally "low enough" as 1.35 seemed to be for Sandy Bridge?

What would the core Tj temperatures be -- on average -- with top-end air-cooling, if it is actually possible? Anyone know firsthand? Without delidding [a range?] or With delidding?

I'm looking at building a Haswell-E when it's released. Probably good that I look more closely at these "K" processors, just for a better foundation . . .

According to the guide I linked:

If you do not have a high-end air or water cooler, it is possible to achieve a solid overclock, but you will need to be more cautious when rising the voltage past 1.1v. When you get up towards 1.2 to 1.25v, you get close to the limits of what the core can actually shed through the integrated heat spreader and what the cooling solution can shed.

With that said, I would conclude that 1.25v would actually be the upper limit for Haswells without lapping/de-lidding.

I am currently running my i7 4770k @ 1.2v @ 4.2GHz with a push/pull CM Hyper 212+ at full speed. Temps are solid at 64-65*C according to my Asus AISuite which seems to be off, even with the latest BIOS update.... Looking at CPUID HW monitor my temps are in the low-mid 70s... with ambient case temps in the upper 30s*C.

 

[Tweakin]

With that said, I would conclude that 1.25v would actually be the upper limit for Haswells without lapping/de-lidding.

Intel recommends a limit of 10% above default core voltage with 22nm processing. With most chips coming in around 1.1xx core, upper limit is roughly 1.2 ~ 1.22 if one is wanting to be safe.

I am currently running my i7 4770k @ 1.2v @ 4.2GHz with a push/pull CM Hyper 212+ at full speed. Temps are solid at 64-65*C according to my Asus AISuite which seems to be off, even with the latest BIOS update.... Looking at CPUID HW monitor my temps are in the low-mid 70s... with ambient case temps in the upper 30s*C.

That's a sound OC, but not what the OP was asking regarding 4.5Ghz. As you can see in my sig, I run 4.3 with my cache at 4.0 pushing 1.215. Thermals during IBT are Mid 60's. Running Seti@home overnight will bring me in around 62c. I'm running under water with 660mm of rad space and fans spinning 700rpm or less, and the pump is probably only pushing 1.3gph as I have pwm control on it so it is only spinning around 25% duty cycle.

 

[BonzaiDuck]

I used to have constant BSOD with WHEA_UNCORRECTABLE_ERROR on my z87-a with 4770k. I updated the BIOS and I am now up and running at 4.2GHz manual OC without any issues. My temps are actually lower than what they were with a 4.1GHz auto OC.

. . .


According to the guide I linked:


With that said, I would conclude that 1.25v would actually be the upper limit for Haswells without lapping/de-lidding.

I am currently running my i7 4770k @ 1.2v @ 4.2GHz with a push/pull CM Hyper 212+ at full speed. Temps are solid at 64-65*C according to my Asus AISuite which seems to be off, even with the latest BIOS update.... Looking at CPUID HW monitor my temps are in the low-mid 70s... with ambient case temps in the upper 30s*C.

OK . . . things could have changed between generations of these ASUS boards, but tweaking guides I've read per my SB-K either discourages using ASUS or any other utility using a single thermal value. What I discovered begins with what I knew years ago: The sensors can be expected to have an error of +/- 5+%. While other causes like TIM application or cooler-mounting might affect individual cores, the biggest error would be that which comes from the sensor itself. IN other cases, some software versions had actually misunderstood or otherwise failed to receive accurate Tj throttling spec from Intel, so the software miscalculated the actual reported temperatures by some 10C degrees for more than a year before BIOS corrections were issued.

Here, we are speaking of thermal measurement issues for previous CPU generations, and it is not entirely obvious that the expected error of sensors would have changed much. Only i7-4770K owners would know, but if you're confident about your cooler installation, it seems likely the sensor error would be obvious even from the range or max less min for either the maximum stress-test values for all cores, or a prevailing set of averages for each core.

Not speaking for other makes or mfgr'r, the ASUS AI Suite on my system seems to report 1C to 2C below a CORE AVERAGE at any point in time, or:

SUM(T, 1, 2, . . . Tn) / n - 2C

for n Cores. Another measure from well-calibrated software: the maximum values of these temperatures over the entirety of the stress test.

People need to know about these possibilities if they read only the highest of the four core values as indicative of the prevailing thermal condition under the IHS. And this might be even more obvious with successfully delidded applications. Someone would have to prove, disprove or qualify, since I cannot do so firsthand.

 

[nwo]

What's weird is that the Asus AISuite temps were accurate before I did the BIOS update. But with the original BIOS, I had a BSOD way too often when I would put the CPU at full load for prolonged periods of time, which seems to be gone now (2 days without any random restarts/BSOD).

On my Z77 Deluxe and i5 3570k, there is a much larger discrepancy between AISuite values and CPUID HW monitor values. Almost 20*C difference. But I have not tried updating the BIOS, which I definitely will have to look into ASAP and also manually OCing the 3570k. Speaking of which, any good guides out there :awe: I really like the on OCers club because it was noob friendly. I don't consider myself an OCing noob, but I have been out of the desktop OCing for a good decade. Last desktop CPU I OCed was a socket 939 A64 x2 :$

 

[BonzaiDuck]

What's weird is that the Asus AISuite temps were accurate before I did the BIOS update. But with the original BIOS, I had a BSOD way too often when I would put the CPU at full load for prolonged periods of time, which seems to be gone now (2 days without any random restarts/BSOD).

On my Z77 Deluxe and i5 3570k, there is a much larger discrepancy between AISuite values and CPUID HW monitor values. Almost 20*C difference. But I have not tried updating the BIOS, which I definitely will have to look into ASAP and also manually OCing the 3570k. Speaking of which, any good guides out there :awe: I really like the on OCers club because it was noob friendly. I don't consider myself an OCing noob, but I have been out of the desktop OCing for a good decade. Last desktop CPU I OCed was a socket 939 A64 x2 :$

Well, what you may be describing is how I read the Guide: It offers important insights, but it is short, to the point, and reasonably well-written. But it would of course be more troublesome for someone who just wandered into the practice we follow here.

And you're right about the BIOS update effecting better accuracy of readings -- this was evident with my initial and second BIOS revisions, but was resolved quickly with the third that followed. Sorry I forgot about that. It pays to remember these things, and no reason the initial errors in BIOS revisions isn't repeated from one chip generation to another.

 

[nwo]

Is it safe to say I can pretty much follow the Haswell guide to OCing my 3570k? Or are there some major differences I would have to pay special attention to?

 

[rgallant]

What's weird is that the Asus AISuite temps were accurate before I did the BIOS update. But with the original BIOS, I had a BSOD way too often when I would put the CPU at full load for prolonged periods of time, which seems to be gone now (2 days without any random restarts/BSOD).

On my Z77 Deluxe and i5 3570k, there is a much larger discrepancy between AISuite values and CPUID HW monitor values. Almost 20*C difference. But I have not tried updating the BIOS, which I definitely will have to look into ASAP and also manually OCing the 3570k. Speaking of which, any good guides out there :awe: I really like the on OCers club because it was noob friendly. I don't consider myself an OCing noob, but I have been out of the desktop OCing for a good decade. Last desktop CPU I OCed was a socket 939 A64 x2 :$

most people on the rog forums strongly advise not to use that and some even offer custom removal guides for it.
-only use one monitoring tool per hardware like don't set up msi-ab for the gpu then add the cpu while running core temps.

 

[OCGuy]

Intel recommends a limit of 10% above default core voltage with 22nm processing. With most chips coming in around 1.1xx core, upper limit is roughly 1.2 ~ 1.22 if one is wanting to be safe.

That is a rather sweeping generalization that is basically not true. Intel doesn't recommend specific over-voltages, and what is "safe" depends on your cooling solution. On air in a good case, 1.3v isn't going to kill your processor. On a side note, VRIN is just one of the settings that people having problems probably are not adjusting.

I believe 1.5v is where definite immediate degradation can occur.

Edit:

http://www.overclock.net/t/1401976/the-gigabyte-z87-haswell-overclocking-oc-guide

 

[nwo]

Tried following the OCer's club guide for 4.4-no dice... My i7 4770k could not get past windows loading screen without locking up. Also tried running it at 4.2GHz @1.1 and 1.15v but it was unstable so I went back to 1.2 @ 4.2GHz. Haswell OCing sucks, although I'm pretty happy with my temps at 4.2!

 

[Tweakin]

Tried following the OCer's club guide for 4.4-no dice... My i7 4770k could not get past windows loading screen without locking up. Also tried running it at 4.2GHz @1.1 and 1.15v but it was unstable so I went back to 1.2 @ 4.2GHz. Haswell OCing sucks, although I'm pretty happy with my temps at 4.2!

That's about where I am also.

 

[The|Hunter]

Nah dont use those extreme values

For example VCCSA up to 0.950v is enough for 3ghz ram, default 0.808v.

Digital & Analog IO affect ram and only if you manually OC from idk 2133mhz to 2400mhz+.

Im doing this now @ 2400mhz CL10 and use 0.015v+ offset, by both and Vccsa 0.030v+ offset.


This is worst case scenario core0 70-74C, Lost Planet2 - fixed Test B, 720p dx9 mode, jobthread = 8

4.7Ghz @ 1.284v with Corsair H90 custom push pull fans. It also the max OC for me anything higher its too much for this cooler.

4.6Ghz @ 1.230v, 4.5Ghz @ 1.176v


And check this good OC guide
http://www.xbitlabs.com/articles/cpu/display/haswell-overclocking.html

if on Asus board, leave digi+ stuff at auto only cpu current at 120 - 130% and in main ai tuner menu below digi+, cpu power set long cpu power duration to 200 - 250w.

 

[Tweakin]

Tried following the OCer's club guide for 4.4-no dice... My i7 4770k could not get past windows loading screen without locking up. Also tried running it at 4.2GHz @1.1 and 1.15v but it was unstable so I went back to 1.2 @ 4.2GHz. Haswell OCing sucks, although I'm pretty happy with my temps at 4.2!

Are you running manual voltage or adaptive?

I've been tweaking my "auto" mode so that I can enjoy all the lower c states for quiet operation by keeping my thermal load as low as possible. I have been able (two weeks + of work) to get mine to throttle properly while only using 1.2 for 4.2Ghz and keeping the cache at 4Ghz.

I have thoroughly thrashed this OS and really need to do a clean install with all the driver and utilities I've thrown at Haswell...

 

[Tweakin]

...And check this good OC guide
http://www.xbitlabs.com/articles/cpu/display/haswell-overclocking.html

That was a good read, thanks. Nothing really new but it was nice to have reinforcement on the approach I've been using.

OP, have you seen this yet?

 

[nwo]

That's about where I am also.

:hmm: I think it needs more voltage :twisted:

Are you running manual voltage or adaptive?

Manual

4.7Ghz @ 1.284v with Corsair H90 custom push pull fans. It also the max OC for me anything higher its too much for this cooler.

4.6Ghz @ 1.230v, 4.5Ghz @ 1.176v


And check this good OC guide
http://www.xbitlabs.com/articles/cpu...rclocking.html

if on Asus board, leave digi+ stuff at auto only cpu current at 120 - 130% and in main ai tuner menu below digi+, cpu power set long cpu power duration to 200 - 250w.

Wow, that's a great OC for the voltage you have it running at. What are your temps like? I'm scared to use AISuite to be honest. Half the features don't even work properly/accurately, so I would definitely not want to overclock using it.

Did you OC using the BIOS or AIsuite?

 

[Tweakin]

:hmm: I think it needs more voltage :twisted:

Voltage starts to ramp up for me above 4.2Ghz. I'm at 1.35 for 4.4 and not 24/7 stable. The extra 200Mhz is meaningless for me, I want quiet with low thermals.

I do agree with the op...you've got a really nice chip to get what you've got at those voltages.

 

[The|Hunter]

:hmm: I think it needs more voltage :twisted:


Manual



Wow, that's a great OC for the voltage you have it running at. What are your temps like? I'm scared to use AISuite to be honest. Half the features don't even work properly/accurately, so I would definitely not want to overclock using it.

Did you OC using the BIOS or AIsuite?

Im Oc'ing through uefi. Yeah that AI suite is kinda so so.. At first I tested fixed voltage to find stability, now at adaptive voltage.


Yeah looks like its a good batch. Wished it was even better something like 5ghz @ 1.25v


Anyway at 4.6ghz ~ 1.234v its up to 63-67C max max, prime95 etc will heat more, but i dont run that.. In games 35-45C, more demanding 45-55C (Bf3, Bf4, SeriusSam3).

The pic above with voltages and stuff is at 4.7ghz 1.284v, that's max ~ 70-75C, in games 40-55, more demanding 55-64C on avg.

Digi+ in uefi OC section was kinda bugged too, but now with latest 1707 all at auto gives the best perf./temp. balance. Only cpu current 110-130% and ram current 100-120% needs to be tweaked if Oc'ed beyond 1.25v or ram idk 2133 to 2400mhz+.

 

[tweakboy]

You should get 4Ghz without touching the voltage.