Intel Devil's Canyon: The Good, the Bad, the Ugly @ [H]

Here's another to add to the list:

Legitreviews.com - 4.9GHz w/1.4VCore. Kept getting BSOD so they dropped back to 4.7GHz @ 1.351 to get a stable OC. Temps around 76C. And that's all with Corsair H105 Extreme water cooler kit.

Legit Reviews 4790K CPU Review

Why didn't they crank the voltage and carry on @ 4.9 if they had the temperature headroom to do it? People seem to be arbitrarily scared of going above 1.4V on 22nm, but if you can handle the temperature these things will take 1.5+V all day long. Even pushed an Ivy up to 1.7V trying to get 5.2 out of it and *shock* *gasp* it didn't spontaneously explode (it also wasn't stable, but oh well I tried).
 
Why didn't they crank the voltage and carry on @ 4.9 if they had the temperature headroom to do it? People seem to be arbitrarily scared of going above 1.4V on 22nm, but if you can handle the temperature these things will take 1.5+V all day long. Even pushed an Ivy up to 1.7V trying to get 5.2 out of it and *shock* *gasp* it didn't spontaneously explode (it also wasn't stable, but oh well I tried).

They aren't afraid of hurting the free chip I think they are afraid of hurting their test kits possibly? Idk. I'm afraid to go past 1.3 and I am sorta just... following traffic, you know?
 
Thanks for the review Kyle.

Judging by what we are seeing, it is looking like these chips are not much better overclocking than many 4770Ks.



We just aren't seeing the performance gains that we once did. That and Intel really messed up their marketing message this time.

DC renamed 4770k IMO.
Haswell-E is what Intel want you to buy for high end.
 
Thanks for the review Kyle.

Judging by what we are seeing, it is looking like these chips are not much better overclocking than many 4770Ks.

I wonder what's causing it?

These DC dies are the same dies used for 4770Ks.

Why didn't they crank the voltage and carry on @ 4.9 if they had the temperature headroom to do it? People seem to be arbitrarily scared of going above 1.4V on 22nm, but if you can handle the temperature these things will take 1.5+V all day long. Even pushed an Ivy up to 1.7V trying to get 5.2 out of it and *shock* *gasp* it didn't spontaneously explode (it also wasn't stable, but oh well I tried).

Sure you can crank the voltage up on these, but you get to the upper 90Cs and these still throttle on ambient cooling.

It seems from the other reviews (to include yours) that the chip is already "OC'd" out of the box with little head room and that's not due to physical limitations. I'm not too worried about the hype machine that Intel tried to spin but the chip seems really good out of the box already. Its too sad that it isn't much for enthusiasts.

Basically, yes.

DC renamed 4770k IMO.
Haswell-E is what Intel want you to buy for high end.

Bingo. And I am not hearing good reports about Haswell-E either at this point.
 
Phoronix should give us another overclocking data point in a few days, for what it's worth. One thing that I thought was interesting was his measurement of power consumption during CPU benchmarks - the peak system consumption for a 4790k was 100W higher than a 4770k with both at stock clocks on the same motherboard/hardware.

http://www.phoronix.com/scan.php?page=article&item=intel_corei7_4790k&num=6

It could be erroneous software readings, but would that have anything to do with overclocking difficulties that people are seeing?
 
Anyone know what mobo and BIOS they got the supposed 5.5 OC on? Maybe it's a BIOS thing that only a certain mobo vendor was made privy to? Just grasping at air here.....
 
X99 platform is an incognito but the Xeons E7 v2 have very good base clocks on the 8 core parts. i7 5960X may not be a 4GHz SKU with 5GHz OC headroom but then again there is no need to reach higher frequency than X79 i7s and even a lower OC with marginally better IPC is more than compensate for the 2 extra cores.

Most X79 enthusiasts would gladly move to X99 [email protected] than stay at X79 [email protected].:p

On the other hand i am glad to hear that Haswell-E will be a good overclocker because it justifies my decision to build a dual X5675s SR-2 for 12 cores@ 4.2Ghz.;)

ummmm...your quote from Kyle is kind of saying the opposite isn't it?
 
Haswell was the first appearance of the FIVR, right? It's also present on Haswell-E, right?

Might this be the underlying issue?
 
This is the 2nd 4790k review that I have read that uses an engineering sample. The other review I read also didn't have very good over clocking results.

That 4770k that you guys did your review on was a very good over clocker compared to most other 4770ks.

I still think 4.7GHZ with 1.36v is decent at this point in time. So early in DC life to be making absolute opinions on it based on an engineering sample.

Looking forward to a review of your retail 4790k.
 
Let me quote a key phrase from a quick chat I just had with someone at ASUS:

We found one sample in 14 that can do 5G at 1.45V on our side.
To sum it all, the tweet from the Intel engineer that started all the hype was a pretty big mistake from his part.
 
Last edited:
Let me quote a key phrase from a quick chat I just had with someone at ASUS:

To sum it all, the tweet from the Intel engineer that started all the hype was a pretty big mistake from his part.

Interestingly, Francois has since deleted his original tweet. What I also read in his feed was when he was pressed on the quote by Ian Cutress of Anandtech, he said something the affect of "it's overclocking....it's a lottery." That tweet seems to be have been deleted as well now. Clearly the guy is backtracking now.

He has posted two new tweets today however....

In the next few days,I will twit about what I call "#ScientificOverclocking" explaining all the inner mechanisms that need to be understood.
Those OC recipes are my own, and shared with few pros of OC, I do not represent my employer on Twitter. The targets I give here are my own.

Can't wait to see what his "Scientific Overclocking" can provide that the reviewers have been unable to attain.
 
Phoronix should give us another overclocking data point in a few days, for what it's worth. One thing that I thought was interesting was his measurement of power consumption during CPU benchmarks - the peak system consumption for a 4790k was 100W higher than a 4770k with both at stock clocks on the same motherboard/hardware.

http://www.phoronix.com/scan.php?page=article&item=intel_corei7_4790k&num=6

It could be erroneous software readings, but would that have anything to do with overclocking difficulties that people are seeing?

The default voltage of the 4790k is 1.25. Isnt 4770k 1.18? That plus the mhz different is pretty substantial.
 
Was debating on whether to use this BNIB 4930k or to sell it off and build a Z97/DC to upgrade from my 4.9ghz 2600k. Looks like Ill just keep the 4930k.

Appreciate the hard work as always [H].
 
It looks like, that if you do nothing more serious on PC than gaming, and have 2500K or higher, you'd do better with buying 2nd GPU.

My Z68 has USB 3.0, Sata 6GB and even pair of 290s won't saturate the PCIe 2.0 bus. Guess, instead of new mobo and cpu, I'll be getting the LG UM65 21:9 display

As broadwell offers little to none improvement over Haswell, I'll run my CPU till Skylake (unless the mobo/cpu breaks) or even further on.
 
After recently building a system for a friend based around a 4770K and ASUS Z97 motherboard with a Samsung 250GB SATA III SSD and 16GB Kingston 2400Mhz Ram, I can happily report that my 4 year old core i7 970/X58 system with 12GB running at 4.2Ghz and 1600Mhz RAM and a SATA II SSD is doing just fine.

Sure, the newer one boots up faster and shuts down faster. I'll give you that. Otherwise I noticed very little difference in user experience when doing most stuff.
 
Even pushed an Ivy up to 1.7V trying to get 5.2 out of it and *shock* *gasp* it didn't spontaneously explode (it also wasn't stable, but oh well I tried).

afaik, SB dies at 1.55v under continuous load after a week or two.
 
How much of a upgrade would this be over a i5 4670k @4.1GHz?
 
How much of a upgrade would this be over a i5 4670k @4.1GHz?

Imagine the difference in clock speed and then add 5% ish and the difference HT would make.
Note this is only for CPU bound situations.
Games that are GPU bound will see no practical difference.
 
hmm wondering if I should pay for DC or fly one of you guru's over to my apartment to overclock my horrible 4770k (2/2).
 
afaik, SB dies at 1.55v under continuous load after a week or two.

Really?

Who told you that? Ive had my 2600k on a Noctua D14 its whole life running 1.488v no HT or can get stable w/ HT at 1.51v (temps get to mid-upper 80s though, high 60's without HT, Ive always ran it without HT due to temps and no need for it really).

Purchased January of 2011 and still going strong. Granted its not 1.55v, but 1.488v is pretty damn close and way above what anyone thought they could hold for a length of time.
 
So early in DC life to be making absolute opinions on it based on an engineering sample.

First and foremost let me say that we buy retail processors for testing here so we do in fact get the retail experience when we can.

I am not making any "absolute opinions" but please be clear on this. I am basing my opinions on our ES experiences and on about 120 pieces of ES reports from ODMs. Maybe there is a pot of gold at the end of the rainbow, but I do not think we will see it live up to the bar that INTEL set.

Guru3D has a review up as well... 4.8 GHz at 1.4V but that is only stable enough to pass some benchmarks.

When we show you "benchmarks" these are shown at what most people would call "stable." If not, we sure let you guys know we could only sneak in a benchmark run before failure.

anybody confirmed if these have returned to a soldered IHS?

Covered in the first article.

http://hardocp.com/article/2014/06/03/intel_devils_canyon_core_processor_presentation/
 
I would have preferred that they put something up that was stable in their review and left it at that. I didn't want to leave the big * out of my post about stability as that part of their review felt like something you'd see on an enthusiast run.
 
hmm wondering if I should pay for DC or fly one of you guru's over to my apartment to overclock my horrible 4770k (2/2).

Only "tricks" I ever found to it were bumping System agent, Digital, and analog input a little bit. I use .05 .03 .03 offsets for these respectively. That got me completely stable at 2400Mhz RAM but as far as the chips that will boot post and run at above 4.2Ghz the "trick" seemed to me to keep buying them :(
 
Really?

Who told you that? Ive had my 2600k on a Noctua D14 its whole life running 1.488v no HT or can get stable w/ HT at 1.51v (temps get to mid-upper 80s though, high 60's without HT, Ive always ran it without HT due to temps and no need for it really).

Purchased January of 2011 and still going strong. Granted its not 1.55v, but 1.488v is pretty damn close and way above what anyone thought they could hold for a length of time.

1.48v to 1.55v is "not close" when it comes to those kind of voltages. You also never mentioned turning off HT.
 
Guru3D and LegitReviews has already been mention but here is some additional stuff about the 4790K I have found poking around the Internet..

Legitreviews.com - ASUS Z97-A got 4.7GHz at 1.351v with a Corsair H105 (were able to run a few benchmarks at 4.9GHz with 1.35V, but couldn’t get it fully stable)
http://www.legitreviews.com/intel-core-i7-4790k-devils-canyon-processor-review_143880/14

Digital Storm - ASUS (model not specificed but could have been Z97-A) got 4.8GHz at 1.37v using a 360mm radiator
http://www.digitalstormonline.com/unlocked/intel-4790k-overclocking-review-and-benchmarks-idnum295/

Guru3d.com - ASUS Z97 Deluxe got 4.8 at 1.399v with a H110 LCS Corsair cooler ("the 4800 MHz is at this stage not 100% stable, Prime95 will not run for over 15 minutes without crashing, however at the current settings it's stable enough to run some benchmarks")
http://www.guru3d.com/articles_pages/core_i7_4790k_processor_review,21.html

MultiTechnopark (youtube channel) - 4.7 ghz on air (Noctua NH-D15) (no MB or voltage info given)
http://www.youtube.com/watch?v=tBmttOQy8ls


From.. http://www.overclock.net/t/1493510/various-core-i7-4790k-devils-canyon-reviews/470
Last time I looked at the stats in that thread, only 6.3% of all the overclocks of 4770Ks hit 4.8GHz or above. If the percentage of 4970Ks that can hit 4.8GHz triples, then still only 19% of them will be able to reach those clocks or higher.
http://www.overclock.net/t/1411077/...guide-with-statistics/11600_100#post_22067510

The percentage of 4770Ks hitting 5GHz+ in the Haswell thread with statistics (thanks @Darkwizzie) is 1.4%, so if 5% of 4970Ks can hit 5GHz+, then that's a big improvement.

Phoronix.com and Hexus.net also have 4790K reviews but no overclocking results (phoronix says they will later)

FYI: Intel 4790K OC Report (Intel 4790K ES C0 Cascade and LN2 limit testing) at http://www.xtremesystems.org/forums/showthread.php?289119-Intel-4790K-OC-Report&p=5233330

Have to remember.. IT'S A REFRESH!! This is how refreshes work. They up the base clocks a bit. I don't understand how anyone thought it would do anything else. But I'll gladly take this 'free' performance boost over choosing the i7-4770K.
 
Its worth remembering that the 4.4GHz boost clock is only with 1 core.
It doesnt guarantee a 4 core 4.4GHz overclock at stock voltage or with higher voltage.
It looks likely that higher voltage will achieve it with decent cooling.
 
Have to remember.. IT'S A REFRESH!! This is how refreshes work. They up the base clocks a bit. I don't understand how anyone thought it would do anything else. But I'll gladly take this 'free' performance boost over choosing the i7-4770K.

Well Intel themselves claimed that if you're not getting 5GHz on air, you're doing something wrong. That is why there is all the fuss over these numbers that would have otherwise fallen about where we would have expected.
 
still rocking a 920 @3.8ghz and I'm not even sure why I would need to upgrade it still.
 
Well Intel themselves claimed that if you're not getting 5GHz on air, you're doing something wrong. That is why there is all the fuss over these numbers that would have otherwise fallen about where we would have expected.


Exactly, although the INTEL engineer's tweet seems to have mysteriously disappeared now.

After all the disappointed ES chips right now, I'm really debating whether to keep my preorder or still with my poorly overclocking Ivy chip.
 
Exactly, although the INTEL engineer's tweet seems to have mysteriously disappeared now.

After all the disappointed ES chips right now, I'm really debating whether to keep my preorder or still with my poorly overclocking Ivy chip.

I cancelled mine. Legit 4.6 to 4.7 isn't a good move no matter how filthy rich you are, imo. And I'm not.

But 4.2Ghz to 4.7 might be worth it.
 
ummmm...your quote from Kyle is kind of saying the opposite isn't it?

which means that i partially disagree with Kyle. That kind of thing happens here about once every 4 years, and is one of the primary reasons i am still an active member.

Just to clarify. my point is that a 5960X does need to be an amazing OCer to beat 4960X. all it need to do is to beat the 3.4Ghz 8c 2687w ,

there are rumors that it will be a X99 version of the E5-1680 v2, a fabled single socket Xeon supposedly unlocked.

When i say X99 may be an incognito i mean that only a new chipset with DDR4 can hamper Haswell-E that much to make an 8-core with that architecture suck, because if X99 manages to be the desktop equivalent l of current unlocked Xeons platforms, it will be a desktop landslide


unless you, like me, is the happy owner of one of the surviving SR-2 mobos and dual unlocked 6c xeons.:cool:
 
I cancelled mine. Legit 4.6 to 4.7 isn't a good move no matter how filthy rich you are, imo. And I'm not.

But 4.2Ghz to 4.7 might be worth it.

My Ivy (3470k) stops at about 4.4ghz (1.35v) , so either way I figure it will be an upgrade, just not sure the difference would be worth $700 Canadian (cpu and motherboard)

I'm further concerned by the fact that a lot of 4770k ES seemed to clock better then there retail equivalent, so this 4.6-4.8ghz range were seeing on the current 4790k may be higher then the retail results.

Gonna wait for some more reviews I guess, but I am kinda wishing the retailer hadn't shipped my Maximus hero vii already.
 
Francois started tweeting his #ScientificOverclocking

https://twitter.com/FPiednoel


Francois:

"If you Quit temp is lower than 50C, you need to get some help :eek:n my solution, an additional heatsink in direct contact with heat speader"

I would love to know what me means by this. A lot of what he is saying is known but its cool hes covering this now as this might help fuel future enthusiasts who find OC'ing too daunting.

My Q9450 is a bitch to figure out, my i950 was quite easy. I was going to upgrade to DC and maybe get the 3258 for shits and giggles but I'm waiting for Broadwell now. All I want is a 4c/4t processor for now.
 
Back
Top