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

dr/owned said:
Except a 4.5 ghz 4770k would benchmark faster than your 5.3 ghz 2500k, which is also no where near the "average" of what a SB will OC to.
Click to expand...
No. Guess you missed the benchmark I posted...... data shows your wrong-
http://www.ocaholic.ch/modules/smartsection/item.php?itemid=1158&page=13
Power Usage-
load (wPrime 1024M, peak during first minute) Watt Index
Intel Core i5-4670K 4.5G 156 130.00 %
Intel Core i5-3570K 4.5G 133 110.83 %
Intel Core i5-2500K 4.5G 130 108.33 %
Intel Core i5-4670K Stock 106 88.33 %
Intel Core i5-3570K Stock 107 89.17 %
Intel Core i5-2500K Stock 120 100.00 %
Peak Performance Index-
Intel Core i5-4670K 4.5G 103.08 %
Intel Core i5-3570K 4.5G 103.07 %
Intel Core i5-2500K 4.5G 102.17 %
Intel Core i5-4670K Stock 102.01 %
Intel Core i5-3570K Stock 103.10 %
Intel Core i5-2500K Stock 100.00 %
The 4670k shows less than a 1% increase over the 2500k @4.5ghz......first off the 4670/4770k do not run 4.5ghz ....stable while gaming period. Secondly, when the 2500k is elevated to 5.3ghz, and the 4670k/4770k can't run any higher than 4.4ghz....the performance index flips with the 2500k well into the 10-12% range faster/ and over 25% more power efficent than any Haswell cpu, this is a well known fact, that's why no one with a 2500k/2600k has switched, they know from the thousands of benchmarks out there, that Haswell is much slower than SB when overclocked. Sucks to be a Haswell owner!!! I glad I bought my 2500k when I did!!!
[email protected] 24-7/365 days a year-http://www.3dmark.com/3dm/1947312
 
Last edited:
ocaholic? That site is a joke. They test games on low settings which is stupid as some settings impact the cpu. Then they turn around and crank all settings and AA turning demanding games into 100% gpu limited scenario which sometimes are only in the 20s and 30s. :rolleyes:
 
Where is this tweet? Deleted?

1402338976c58SqSYffL_1_2.png
 
He did delete a lot of tweets, but I'm not sure if that's one of them. I think you could possibly get a 4.6ghz on a fanless heatsink. You'd need a lot of air flow, which after looking at his following tweets he's a huge fan of exhaust air flow. You would also need an open cooler, where the fin's aren't very dense.

Exhaust air flow is smart but you need to help it by providing SOME sort of intake. At least one intake fan will supply a medium sized case with enough air to forcibly flow across all internals using 2+ exhaust fans stationed above the CPU heat sink. I do this in my HTPC case using an i7 950, which is fanless right now.
 
Trimlock said:
He did delete a lot of tweets, but I'm not sure if that's one of them. I think you could possibly get a 4.6ghz on a fanless heatsink. You'd need a lot of air flow, which after looking at his following tweets he's a huge fan of exhaust air flow. You would also need an open cooler, where the fin's aren't very dense.

Exhaust air flow is smart but you need to help it by providing SOME sort of intake. At least one intake fan will supply a medium sized case with enough air to forcibly flow across all internals using 2+ exhaust fans stationed above the CPU heat sink. I do this in my HTPC case using an i7 950, which is fanless right now.
Click to expand...

I'm pretty interested in airflow (as it relates to most things but in particular racing). I almost always use positive pressure in my builds because it tends to get you smaller dust particles. Exhaust heavy does seem like it would bring in cold air more forcefully, but I have 3 cats and Intake heavy configs do work well in dust heavy environments because you can filter every possible intake and the "holes" in your case will just pour air out. I use much lower RPMs on the exhaust fans and crank the crap out of the intakes. So all the dust I end up with is particulate. If I used negative pressure... my case has tons of holes all over it, there would be all kinds of giant dust particles flowing in.
 
As Kyle pointed out, Francois Piednoel wasn't the only one making 5GHz on air claims..

It can be overclocked to 5GHz in air-cooled systems, said Renee James, president of Intel, during a keynote speech at the Computex trade show in Taipei.
Click to expand...


More 4790K overclocking results..

Computershopper.com - MSI Z97 MPower Max AC got 4.2GHz using Thermaltake’s Water 2.0 self-contained liquid cooler on the CPU, and mounted two 120mm fans on the radiator in a push-pull setup. From the article.. "Bumping up the core voltage (eventually, as high as 1.36 volts), we were able to get temporary stability with the base clock at nearly 4.5GHz.. when we tried to nudge the clock up the tiniest bit more to hit a true 4.5GHz, the system would reset, and we were never able to get close to that speed again with any stability at all."
http://www.computershopper.com/components/reviews/intel-core-i7-4790k/(page)/4#review-body


Overclockers. com - ASUS Maximus VII Gene got 4.9GHz at 1.5125v (memory dropped to 1600) using EK-Supreme LTX Water Block with a 360 mm Radiator and MCP35X Pump for complete wPrime 32M and SuperPi 1m runs
http://www.overclockers.com/intel-i7-4790k-devils-canyon-cpu-review


Hothardware.com - Gigabyte Z97X SOC Force got 4.8GHz at 1.375v using "large Cooler Master air-cooler "
http://hothardware.com/Reviews/Devils-Canyon-Intel-Core-i74970K-Review/?page=7


Hitechlegion.com - MSI Z97 Gaming 9 got 4.7GHz at 1.28v using Noctua NT-HI paste & Corsair H100i.. "the only changes made in the BIOS was a simple change of the multiplier to 47 and a voltage bump to only 1.28V on the 4790K to achieve a solid and stable 'performance tune' of 4.7GHz on all 4 cores"
http://www.hitechlegion.com/reviews/processors/40369-on-the-trail-with-intel-s-devil-canyon


Overclock3d.net - ASUS Z97A got 4.8GHz at 1.35v using Corsair H100i (but ran benchmarks at 4.6GHz at 1.26v)
http://www.overclock3d.net/reviews/cpu_mainboard/intel_i7_4790k_devils_canyon_cpu_review/3
 
Last edited:
Intel can do one thing to correct this giant fuck-up: immediately switch back to Indium Solder and offer a no cost RMA to every owner that already bought a paste crippled DC processor.
 
Intel can do one thing to correct this giant fuck-up: immediately switch back to Indium Solder and offer a no cost RMA to every owner that already bought a paste crippled DC processor.
Click to expand...


I don't think adding solder will make much difference, it seems the Haswell architecture is maxed. TIM and temperature are not the limiting factors here.
 
misterbobby said:
ocaholic? That site is a joke. They test games on low settings which is stupid as some settings impact the cpu. Then they turn around and crank all settings and AA turning demanding games into 100% gpu limited scenario which sometimes are only in the 20s and 30s. :rolleyes:
Click to expand...

They're also addicted to OCahol. Sad...
 
Sampsa said:
Where is this tweet? Deleted?

1402338976c58SqSYffL_1_2.png
Click to expand...

Yes, his crass tweets have mysteriously vanished~

DejaWiz said:
Intel can do one thing to correct this giant fuck-up: immediately switch back to Indium Solder and offer a no cost RMA to every owner that already bought a paste crippled DC processor.
Click to expand...

Why would they do that when they can push you onto LGA2011 for more money?
 
DarkStryke said:
Why would they do that when they can push you onto LGA2011 for more money?
Click to expand...

I do not believe lga2011 will be any better. I still say the problem is with the 22nm tri-gate silicon. And it is not fixed yet and I do not believe 14nm has it fixed either hence the reason for a 1 year delay.
 
Chongmagic said:
I don't think adding solder will make much difference, it seems the Haswell architecture is maxed. TIM and temperature are not the limiting factors here.
Click to expand...

Kyle himself has stated numerous times now that thermal throttling is inhibiting the ability to get to the overclocks that key intel personnel had essentially promised us all. Solder is what, about 10 fold better at conducting heat than this shit TIM paste they are sticking under the IHS? They need to quit being cheap-asses and just do the right thing for their customers...I don't think anyone would complain if DC was $10-20 more for a soldered part in return.
 
So I've looked over a few leaked slides of Intel's roadmap. I want to upgrade to a cpu/mobo that will have 6 or 8 cores, thunderbolt available, and plenty of pcie lanes.

DC doesn't fit the bill on cores. Maybe some forthcoming mobo's will hit the other marks.

But from the roadmaps it's looking like skylake (tock, or enthusiast version), 100 series chipset, (so 2016 probably)? not 100% sure because the one I found was in a foreign language. Anyone have a feel for the specs on skylake or maybe broadwell?

I'm still rocking an i7-920 @3.6, just over 5 years old now. I want to upgrade, but DC isn't giving me goosebumps.
 
GoodBoy said:
So I've looked over a few leaked slides of Intel's roadmap. I want to upgrade to a cpu/mobo that will have 6 or 8 cores, thunderbolt available, and plenty of pcie lanes.

DC doesn't fit the bill on cores. Maybe some forthcoming mobo's will hit the other marks.

But from the roadmaps it's looking like skylake (tock, or enthusiast version), 100 series chipset, (so 2016 probably)? not 100% sure because the one I found was in a foreign language. Anyone have a feel for the specs on skylake or maybe broadwell?

I'm still rocking an i7-920 @3.6, just over 5 years old now. I want to upgrade, but DC isn't giving me goosebumps.
Click to expand...

Last I saw on broadwell it was mainly just a die shrink, so I would expect same 4 core limit, and quite possibly same PCI-E lane limit. If your looking for 6-8 cores and more PCI-E lanes i think your going to want to look into Haswell-E on the X99 chipset (6-8 cores, Socket 2011-3, DDR4).
 
The processor has not even been released and already people make conclusions based on a couple reviews of engineering samples..... Until we actually have a good amount of users that actually have these in hand with overclocking results to go with it, not sure how anyone can make a conclusion at this point.
 
Slightly disappointed, but not surprised. Was also looking to upgrade, but i guess one more year the way it is isn't a bad thing.

Kyle, thanks for the review.
 
raidflex said:
The processor has not even been released and already people make conclusions based on a couple reviews of engineering samples..... Until we actually have a good amount of users that actually have these in hand with overclocking results to go with it, not sure how anyone can make a conclusion at this point.
Click to expand...
A couple of reviews? Lol you have to be freaking joking? And if anything retail is usually worse than ES. Go look at the 4770k reviews and you will see basically every ES sample did 4.7 where that is not even close to normal for a retail 4770k.

Intel talked up the 5ghz for 4790k and made it sound like anyone can get that which has turned out to be nonsense. :rolleyes:
 
raidflex said:
The processor has not even been released and already people make conclusions based on a couple reviews of engineering samples..... Until we actually have a good amount of users that actually have these in hand with overclocking results to go with it, not sure how anyone can make a conclusion at this point.
Click to expand...

It would help you to read what happened.
Intel engineers stated what overclocking performance we should expect and that we would need training if we couldnt achieve 5GHz on air.
They showcased an air cooled processor at 5.5GHz, but the air was cooled with liquid nitrogen.
Subsequently there has only been one report of a 5GHz overclock on air.

We rightly had a high expectation because thats what they insisted we think.

Note they have now recalled the tweet saying the above.
Why would they do that unless there was something to hide.
 
Nenu said:
It would help you to read what happened.
Intel engineers stated what overclocking performance we should expect and that we would need training if we couldnt achieve 5GHz on air.
They showcased an air cooled processor at 5.5GHz, but the air was cooled with liquid nitrogen.
Subsequently there has only been one report of a 5GHz overclock on air.

We rightly had a high expectation because thats what they insisted we think.

Note they have now recalled the tweet saying the above.
Why would they do that unless there was something to hide.
Click to expand...

I agree what Intel did was wrong and it may be that these early reviews is what we will see in the long run. I am just saying we are talking what maybe 100-200 processors tested between all the review sites out of the 1000's of retail samples that will be produced.
 
raidflex said:
I agree what Intel did was wrong and it may be that these early reviews is what we will see in the long run. I am just saying we are talking what maybe 100-200 processors tested between all the review sites out of the 1000's of retail samples that will be produced.
Click to expand...

Nope. Same shit with lower than expected OC's and high temps has been experienced since IB. And IB just happened to be the introduction of non-soldered TIM. Instead of LISTENING TO THE REVIEWERS AND CUSTOMERS about how shitty the paste is, Intel keeps playing fuck-around with each new generation they've released since.
 
raidflex said:
I agree what Intel did was wrong and it may be that these early reviews is what we will see in the long run. I am just saying we are talking what maybe 100-200 processors tested between all the review sites out of the 1000's of retail samples that will be produced.
Click to expand...

If you agree it was wrong, then you agree that the performance wasnt up to par.
As they showcased the CPUs performance and stated that training would be needed if we couldnt hit 5GHz on air, what would your expectation have been?

They have not issued a statement along the lines of what you suggest, considering the effort they put into making us believe they are high performers, why stop now?
There is no information to allow us to believe that the release CPUs will perform better.
 
DejaWiz said:
Nope. Same shit with lower than expected OC's and high temps has been experienced since IB. And IB just happened to be the introduction of non-soldered TIM. Instead of LISTENING TO THE REVIEWERS AND CUSTOMERS about how shitty the paste is, Intel keeps playing fuck-around with each new generation they've released since.
Click to expand...
I think its more to do with their 22nm process as the Ivy Bridge E cpus don't oc all that good either and they are soldered.
 
I dont know, it just seems to me maybe Engineering really thought that heat transfer was a bigger cause of the speed block seen on haswell, and they were wrong, its more a product of its process and node tech. Would hardly be the first time Engineering was wrong in how something would scale (see Pentium 4). Given the short timeline for the product ( ~6 months ) maybe Marketing just worked off assumptions instead of real world practice.

Certainly doesn't forgive anything, just looking for an explanation.
 
Good 'ol fugger, watch out, you'll get banned~

That cite is a cesspool of shills and fanboys.
 
Don't care, coming from a Duo Core E8400 to the Core i7 4790K is going to be absolutely amazing, especially loaded with DDR3 and of course a SSD.
 
PhoneyVirus said:
Don't care, coming from a Duo Core E8400 to the Core i7 4790K is going to be absolutely amazing, especially loaded with DDR3 and of course a SSD.
Click to expand...
Wow you certainly waited long enough. My oced E8500 was not cutting it even 3 years ago and had to be upgraded.
 
Phoronix.com finally got around to overclocking their 4790K.. ASRock Z97 Extreme6 got 4.5GHz (no voltage or cooler info given)
http://www.phoronix.com/scan.php?page=news_item&px=MTcyMDA


Odds & Ends

List of Motherboards Compatible with Intel Core i7-4790K and Core i5-4690K

Intel Core i7-4790K overclocked to 4.50GHz with passive cooling

Intel Core i7 4790K 'Devil's Canyon' CPU already pushed to 7GHz - ASRock Z97 OC Formula got 7.0GHz at 1.792v using a 4790K Engineering Sample with two cores disabled


Intel Broadwell K, Broadwell-E and Skylake Release Date Revealed in Leaked Intel Roadmap - Broadwell for desktop and the Broadwell K series (unlock processors) towards the middle of 2015. By second half of 2015, Intel will reveal probably the 6th Generation Intel Skylake processors. Skylake will have a new type of socket. Broadwell-E, successor of Haswell-E, will support the same X99 Chipset and same LGA 2011-3 socket


Thought about Engineering Samples

Both Kyle and Ryan used 4770K ES CPUs during their original 4770K reviews (both dated June 1, 2013) but neither states this fact during that article, although Kyle mentions "a "golden" sample directly from Intel". Ryan's picture of the 4770K he used shows it to be a Intel Confidential ES chip. Ryan, using a Intel DZ87KLT-75K, got 4.6GHz with 1.326v using a Corsair H100 while Kyle, using a ASUS Z87-A, got 4.8GHz/2000MHz at 1.26v vCore using a Koolance Exos 775 system with 370 series water block.

Kyle mentions using an ES chip in the July 1, 2013 Gigabyte Z87X-UD5H LGA 1150 Motherboard Review "I was able to use the Easy Tune to further push the 4770K to a solid 4.8GHz. (Keep in mind this a engineering sample I am using and Dan is using a retail purchased processor.)" Dan only achieved "4.6GHz @ DDR3 2400MHz speeds" on the same motherboard and using a Koolance Exos 2.5 and CPU-370 water block.

Kyle seemed to have kept using his 4770K ES until the July 30, 2013 Gigabyte Z87X-UD3H LGA 1150 Motherboard Review where he wrote "the Intel Core i-7 4770K that we have been using for testing over the last couple of months. It seems that electromigration has damaged our CPU, at least in an overclocking sense when stressed for long periods of time. It seems that it has also been damaged for high GHz (4.8) for short periods of time. That all said, we had the same results with this Gigabyte motherboard that we have had with the two previous high end motherboards we have reviewed in that these would not hold a 48 hour 4.5GHz overclock. Since this review we have pulled our two month "old" testing 4770K and replaced it, and are now back on track with our stress testing."

Concerning the original 4770K ES, Kyle wrote in the September 5, 2013 GIGABYTE Z87X-UD4H LGA 1150 Motherboard Review "We no longer run any testing at default CPU speeds, and we have recently purchased a new Core i7-4770K to test with as our original engineering sample CPU from Intel seemed to be affected by some of the long term torture testing we do as it would not longer hold a 4.5GHz overclock for any considerable amount of time."

In Kyle's original June 1, 2013 article http://www.hardocp.com/article/2013/06/01/intel_haswell_i74770k_ipc_overclocking_review/6 he mentions..

ASUS has tested a couple hundred Haswell processors at this time and this is ASUS’ specific feedback from that overclock testing.

70% of CPUs can clock to 4.5GHz
30% of CPUs can clock to 4.6GHz
20% of CPUs can clock to 4.7GHz
10% of CPUs can to 4.8GHz

Overall you will find most CPUs capable of reaching 44x to 45x with varying levels of voltage.

These ASUS results were obtained with sealed water cooling systems that are comparable to a Corsair H80/H100 configuration or extremely efficient air cooling with 120mm push/pull fans while applying a maximum core voltage of 1.275v under full thread load conditions.

Obviously it is very early in terms of Haswell retail processor production, and we are still not sure what we will see in terms of i7-4770K retail purchased Haswell overclocking headroom. We will be buying a couple of Core i7-4770K processors this weekend and see what comes out of that. Going on the above information, it would seem that getting a 4.8GHz-worthy Haswell sample will be about a one in ten shot with "normal" sealed water cooling systems.

Overclocking the Haswell i7-4770K processor that was sampled to HardOCP directly from Intel was extremely simple using two ASUS Z87 series motherboards and we will describe on that on the following pages. From the information above we can assume that we got a "golden" sample directly from Intel, or were extremely lucky when it came to the luck of the draw.
Click to expand...

I point this all out only because of all the attention being paid to the 4790K being overclocked by review sites are "Engineering Samples" while the originally tested 4770K by Kyle and Ryan were also ES models, and neither made any case for or against it being such a chip. This time, however, Kyle wrote at the top of his 4790K review "We have now had a single Core i7-4790K engineering sample in our hands for 72 hours and this is what we have found." And Ryan wrote "after talks with motherboard vendors that have used quite a few more engineering samples of the Core i7-4790K". Just curious on why using ES chips wasn't a big deal with the 4770K but is with the 4790K.

Kyle above talked about the possibility of getting some "retail" 4770Ks to test but I couldn't find any evidence he posted any tests on these and seemed to still using his original 4770K ES until his mentioning of the "electromigration" problem in the July 30, 2013 Gigabyte Z87X-UD3H review.
 
Asus said:
70% of CPUs can clock to 4.5GHz
30% of CPUs can clock to 4.6GHz
20% of CPUs can clock to 4.7GHz
10% of CPUs can to 4.8GHz
Click to expand...
What an absolute pile of horseshit.
 
spawn447 said:
what does it all mean ?
Click to expand...

Nothing really, we could be looking at the peek of the architecture or we could be seeing intel hit a wall with their current platform. As it stands the processor is fine, it runs cooler than the 4770k and runs faster at stock as well for only $20 more.
 
Filiprino said:
Well, these ones have TSX-NI and VT-d.
Click to expand...

TSX-NI may be a feature I can use - however, VT-d is actually useless to me; EPT, however, IS a feature I can leverage.

TSX-NI has to be supported by software; therefore, at this point, isn't it a niche feature? VT-d (VT with Directed I/O) is even more niche than EPT - while type 1 hypervisors can leverage it, none of them even need it (VT-d, that is - dedicated I/O still makes more sense than VT-d, even where it's available, and it's no harder to leverage).

EPT stands for Extended Processor Tables support - every Intel Core i-series (and their AMD counterparts, and all AMD APUs) supports it. HyperV Gen2 (Windows 8 and later clients) requires EPT support for graphical-performance reasons; however, Windows Server 2008 and later - which ALSO supports HyperV Gen2 - does not, though it can leverage it if present.

The biggest mainstream use for TSX (let alone TSX-NI) is for security - specifically, cryptography; however, how much in the way of even security software makes use of even base TSX, let alone TSX-NI? That is why I said it's a niche use (for now). EPT (despite the utter ubiquity) is quite capable of becoming mainstream, but has hardly moved since the ubiquity of even VT-x became obvious - how much has even virtualization moved into the mainstream? (In this example, I'm referring to usage, not availability.)
 
Trimlock said:
Nothing really, we could be looking at the peek of the architecture or we could be seeing intel hit a wall with their current platform. As it stands the processor is fine, it runs cooler than the 4770k and runs faster at stock as well for only $20 more.
Click to expand...

There are some who would say that Intel hit the wall with Sandy, and has been spinning their tires ever since. When a 920, a 6 going-on-seven year old CPU, is still competitive for most tasks, that's a sign that something has gone very wrong in the engineering department.

Increasing tick-tock delays (The roadmap still had 10nm in Q1 2015 as recently as 2011), and stagnant performance increases makes me wonder how long before Intel just stops trying like AMD, or finally dumps the architecture they've been miniaturizing for 20+ years and releases something new.
 
You must log in or register to reply here.
Back
Top