Intel "Devils Canyon" Launch - Reviews, OC Info, Availability

[Idontcare]

Translation from pclab.pl

https://translate.googleusercontent...5.html&usg=ALkJrhjcd6GNpvebgwrtRraOFFhvrUZpcg

Seems to be able to OC at 4.7GHz on AIR using the best AIR coolers. Im sure some SKUs will reach 5GHz (or close) with custom WC setups but im not sure they will be stable for 24/7 use.

Edit: I believe the sweet spot will be around 4.4-4.5GHz with the lowest voltage possible. Temperatures (due to higher Voltage) and power consumption are raising rapidly after 4.5GHz and 1.25v.

​

Well we should be honest with ourselves and acknowledge that if Intel could have clocked their version of the FX-9590 to 5GHz turbo as a stock config then they would have...which leaves us with really only one conclusion in terms of where our OC'ing expectations ought to be - the chip has maybe 5% OC headroom left in it after Intel is done binning and "optimizing the TIM".

All I'm interested in now is delidding and bare-die with WC results (to highlight why the extra caps are on the package). Everything else we already know about from delidded 4770Ks.

 

[BFG10K]

The pclab.pl review is the best so far. They have a suite of games - logical given this is a gaming CPU. With ~12% over the 4770K in some cases, the gap should be even wider over my 2500K, assuming CPU limitations of course.

Also the temperatures are very good with the first overclocked result (max 75C). At stock clock/voltage they'll be even better.

This chip is perfect for me - a high base clock with fixed thermals, and large clock increase over the 2500K so we don't have to rely on IPC or threading to get a gain.

 

[BFG10K]

Here come the mainstream reviews: http://www.tomshardware.com/reviews/core-i7-4790k-devils-canyon-overclock-performance,3845.html

 

[RussianSensation]

I think some people don't understand that most overclockers don't care about 5Ghz. Of course it would be nice, but many of us are just mild to moderate overclockers. With the 4790K it seems all but certain you can get 4.5 or 4.6GHz with almost no effort, whereas on the 4770K that was far from certain. To me that is a win.

It's not about some mythical 5Ghz, 6Ghz, 10Ghz figure but about the performance increase from SB/IVB and upgrade path. A lot of people in this thread were ready to dismiss the X99 socket with 5820. If 4790K can consistently hit 'only' 4.5-4.7ghz, but a 6 core Haswell-E can hit 4.2-4.3Ghz, many people will easily accept the trade-off of 9-12% lower clock speed to get 50% more cores if 5820 is really a sub-$400 CPU. If HW-E overclocks even better than 4.3Ghz, then the decision will be even easier. The average gap between E8400 and Q9550's top overclocking on air was more than 15% and E8400 still turned out to be an inferior choice overall when the dust settled. I think 5820 vs. 4790K may become a turning point that we saw last time with E8400 vs. Q9550.

In numbers, I would wager that lot less PC enthusiasts want to buy a CPU every 1-2 years now. Many members of our forum are still on i5 2500K/i5 2600K and those are 3.5 years old CPUs! The times have changed from the past when every 2 years CPUs went from single to double to quad core or frequency increased 40-50%.

Now, a PC gamer can buy an Intel CPU and keep it 4-5 years as long as it is decently overclocked and most importantly has sufficient core count. In this case, the upgrade path to 5820 or 4790K is not as simple since if one doesn't plan on upgrading every 2-3 years, the gamer is betting that in 2016-2019 there will be almost no games that will take advantage of 6 core CPU. I'd even say an i7 920 @ 4.0ghz is not much slower in games than i7 4790K @ 4.6ghz for systems with any single-die GPU and most of that is related to that fact that it has 4 cores and a solid IPC. However, in the next 3-6 years I think core count will become much more important than the small difference in clock speed on a 4 vs. 6 core Haswell CPU. It's not as if $340 for a quad-core i7 is that cheap either to not consider the X99 platform with a 5820 for $200-250 more. Over 3-4 years, $200-250 for a total CPU platform is not a lot if one thinks some major games from 2015-2018 will utilize a 6 core CPU.

Ignoring the 4 vs. 6 core discussion, most disappointing is that 4790K appears to have been overhyped by Intel. From HardOCP:

"Truth is, it’s (5GHz) only possible on few samples and that the CPUs will throttle if used with air coolers at the voltages needed to get to the desktop. On average, the TIM change and extra decoupling on the rear of the CPUs has resulted in around 200MHz headroom over first gen Haswell in our testing so far.

From information shared with me currently I would expect no more than ~5% of Devil's Canyon processors to reach 5GHz, and that would be on water cooling."

One can say there is little competition and that Intel is still improving but had they not overhyped all the improvements they made to this processor, there wouldn't be a big deal. This looks like a great chip for those who don't overclock much since it has a much higher base clock than 4770K, but from overclocking, it's only approaching IVB levels. Hopefully at 14nm, Intel's Skylake can consistently hit 5.0-5.2Ghz on air.

This chip is perfect for me - a high base clock with fixed thermals, and large clock increase over the 2500K so we don't have to rely on IPC or threading to get a gain.

If you waited this long with 2500K, and skipped 4770K, I don't see any point at all to buy what is dead-end socket and a slightly faster 4790K. Why not skip this one as well and just go straight to Skylake? At least there will be next generation features - 14nm, PCIe 4.0, and DDR4.

Also, after all these years, I still don't understand how someone as experienced as you doesn't overclock CPUs. 2500K overclocks to 4.3-4.4Ghz with a cheap after-market cooler. This should easily hold you over until Skylake.

 

[AtenRa]

Well we should be honest with ourselves and acknowledge that if Intel could have clocked their version of the FX-9590 to 5GHz turbo as a stock config then they would have...which leaves us with really only one conclusion in terms of where our OC'ing expectations ought to be - the chip has maybe 5% OC headroom left in it after Intel is done binning and "optimizing the TIM".

All I'm interested in now is delidding and bare-die with WC results (to highlight why the extra caps are on the package). Everything else we already know about from delidded 4770Ks.

What i would really like to know is if there is also a gap between the CPU die and the IHS in Devils Canyon. If you happen to test the DC and remove the IHS please check if the gap remains. I have a theory about the gap and i would like to know if it is also present in Devils Canyon.

 

[BFG10K]

If you waited this long with 2500K, and skipped 4770K, I don't see any point at all to buy what is dead-end socket and a slightly faster 4790K.

Dead-end? This should last about three years for me like SB did. And Broadwell should work on Z97 too.

Why not skip this one as well and just go straight to Skylake? At least there will be next generation features - 14nm, PCIe 4.0, and DDR4.

Modern cards barely saturate PCIe 2.0 x8. What do I need PCIe 4.0 for? As for DDR4, it'll be years before it's price competitive with DDR3. Probably at least 3 years, which is probably when I'll be looking to upgrade again.

Also, after all these years, I still don't understand how someone as experienced as you doesn't overclock CPUs. 2500K overclocks to 4.3-4.4Ghz with a cheap after-market cooler. This should easily hold you over until Skylake.

Wikipedia sums it up rather well:

As an overclocked component operates outside of the manufacturer's recommended operating conditions, it may function incorrectly, leading to system instability. Another risk is silent data corruption by undetected errors. Such failures might never be correctly diagnosed and may instead be incorrectly attributed to software bugs in applications, device drivers, or the operating system. Overclocked use may permanently damage components enough to cause them to misbehave (even under normal operating conditions) without becoming totally unusable.

A large scale field 2011 study of hardware faults causing a system crash for consumer PCs and laptops showed a 4x to 20x increase (depending on CPU manufacturer) in system crashes due to CPU failure for over-clocked computers over an 8 month period.

In general, overclockers claim that testing can ensure that an overclocked system is stable and functioning correctly. Although software tools are available for testing hardware stability, it is generally impossible for any private individual to thoroughly test the functionality of a processor.

Achieving good fault coverage requires immense engineering effort; even with all of the resources dedicated to validation by manufacturers, faulty components and even design faults are not always detected.

A particular "stress test" can verify only the functionality of the specific instruction sequence used in combination with the data and may not detect faults in those operations. For example, an arithmetic operation may produce the correct result but incorrect flags; if the flags are not checked, the error will go undetected.

To further complicate matters, in process technologies such as silicon on insulator (SOI), devices display hysteresis—a circuit's performance is affected by the events of the past, so without carefully targeted tests it is possible for a particular sequence of state changes to work at overclocked rates in one situation but not another even if the voltage and temperature are the same. Often, an overclocked system which passes stress tests experiences instabilities in other programs.

So Intel with their millions of dollars for testing and world-class labs can't always find faults (e.g. USB 3.0 sleep issue). But we're led to believe that some random anonymous internet guy twiddling BIOS switches and running Prime can do it?

How many times do we see people complaining about something ("buggy this" and "unoptimized that"), then you check their signatures and see they're overclocking. But it must be the software, because Prime is stable, right? I mean it can't possibly be their overclock.</sarcasm>

Overclockers delude themselves into thinking they've achieved 100% stable and correct operation, but the reality is they have no idea. They can't track or measure silent failures, and obvious problems are often blamed on software/OS/drivers, and everything else under the sun instead of themselves.

Heck, I'd wager some of the benchmark results on overclocked systems are wrong, simply because the timers needed to generate and measure the results have to absolutely work properly. But the average overclocker can't even comprehend an idea like this, much less test for it.

 

[StinkyPinky]

Who cares if 1150 is a dead socket (which it isn't anyway). How many people here actually update only the CPU anyway? I never have.

 

[Idontcare]

Wikipedia sums it up rather well:

So Intel with their millions of dollars for testing and world-class labs can't always find faults (e.g. USB 3.0 sleep issue). But we're led to believe that some random anonymous internet guy twiddling BIOS switches and running Prime can do it?

How many times do we see people complaining about something ("buggy this" and "unoptimized that"), then you check their signatures and see they're overclocking. But it must be the software, because Prime is stable, right? I mean it can't possibly be their overclock.</sarcasm>

Overclockers delude themselves into thinking they've achieved 100% stable and correct operation, but the reality is they have no idea. They can't track or measure silent failures, and obvious problems are often blamed on software/OS/drivers, and everything else under the sun instead of themselves.

Heck, I'd wager some of the benchmark results on overclocked systems are wrong, simply because the timers needed to generate and measure the results have to absolutely work properly. But the average overclocker can't even comprehend an idea like this, much less test for it.

+1

This is why I'm personally done with OC'ing on anything that I can't insulate, isolate, or validate on-the-fly.

OC'ing as a hobby is fun, but I refuse to run an OC'ed PC for personal finances or my business.

I also see no harm in OC'ing for gaming and other hobby-like past times. A trashed OS install is easily reversed with backups (made while not OC'ed of course), same with transcoding videos and so on.

Lots of non-critical apps can be ran without concern for silent data corruption. But it is irresponsible to run critical programs (finance, banking, business, etc) on an OC'ed rig.

I'd love an Intel validated 4/4.4 rig. Not sure where the heck I'd get one here though

 

[frozentundra123456]

I'm looking forward to that.

There are a few gaming benches here, although some are... GPU bound :whiste::

http://pclab.pl/art58123.html

Scales pretty nicely in gaming with clockspeed. Those results make obvious just how far AMD is behind in gaming. Even the low end i5 4460 is faster overall than the 9590. Ouch.

 

[Anth Seebel]

nvm

 

[ViRGE]

+1

This is why I'm personally done with OC'ing on anything that I can't insulate, isolate, or validate on-the-fly.

OC'ing as a hobby is fun, but I refuse to run an OC'ed PC for personal finances or my business.

I also see no harm in OC'ing for gaming and other hobby-like past times. A trashed OS install is easily reversed with backups (made while not OC'ed of course), same with transcoding videos and so on.

Lots of non-critical apps can be ran without concern for silent data corruption. But it is irresponsible to run critical programs (finance, banking, business, etc) on an OC'ed rig.

I'd love an Intel validated 4/4.4 rig. Not sure where the heck I'd get one here though

Overclocking isn't nearly as fun as it once was because it's not nearly as productive. While not every chip can be a Celeron 300A or i7-920, these days the clockspeeds attainable on practical cooling setups are becoming smaller and smaller as a percentage of the original clockspeed. It's hard to justify the potential headaches for another 10% over the shipping maximum turbo bin.

Though I'm not quite sure what you mean by a validated Intel 4/4.4 rig? Wouldn't that be 4790K?

 

[crashtech]

Who cares if 1150 is a dead socket (which it isn't anyway). How many people here actually update only the CPU anyway? I never have.

I almost always do, though occasionally it might be a downgrade if the board is being re-purposed. I'm glad CPUs are socketed and like to take advantage of it.

 

[Idontcare]

Overclocking isn't nearly as fun as it once was because it's not nearly as productive. While not every chip can be a Celeron 300A or i7-920, these days the clockspeeds attainable on practical cooling setups are becoming smaller and smaller as a percentage of the original clockspeed. It's hard to justify the potential headaches for another 10% over the shipping maximum turbo bin.

Exactly!

Though I'm not quite sure what you mean by a validated Intel 4/4.4 rig? Wouldn't that be 4790K?

Yes, that is what I mean. I'm glad Intel made such a stock product, validated and everything, at those clocks. That's great performance without the need to mess around with OC'ing IMO.

Intel should release a non-K SKU at the same clocks.

 

[ViRGE]

Intel should release a non-K SKU at the same clocks.

What would be the difference in this case? According to ARK, VT-D and TSX are present on the 4790K, so all you lose is TXT and vPro.

And oddly enough, they are releasing a 4790 non-K. However it has different clockspeeds: 3.6/4 instead of 4/4.4.

 

[campbbri]

I notice in both the PcPer and Toms Hardware reviews, using different motherboards (ASUS and ASRock), that the idle power for the 4790K is significantly higher than it is for the 4770K.

Is this the kind of thing where each board maker needs to release updated firmware or could it be a tradeoff of all those extra capacitors or other changes on the chip itself?

 

[BFG10K]

Is this the kind of thing where each board maker needs to release updated firmware or could it be a tradeoff of all those extra capacitors or other changes on the chip itself?

I believe it's a BIOS issue as Tom's said it dropped to low power states correctly when the CPU was overclocked manually.

Also this review shows lower idle consumption: http://www.vortez.net/articles_pages/intel_devils_canyon_core_i7_4790k_review,1.html

 

[BFG10K]

http://lanoc.org/review/cpus/6877-intel-i7-4790k-devil-s-canyon?showall=&limitstart=

 

[wasabiman123]

Given that the 4790K is 4.0/4.4 boost on stock settings, would that bring about silent data corruption? Or is there no guarantee of that?

I suppose the safest option for gamers who also store important files on their machines would be a non K SKU, or a Xeon like the 1231 V3.

 

[tidy]

.

 

[wasabiman123]

^Might be the massive cache vs mainstream socket? That's a damn huge difference if the benchmark is remotely accurate.

 

[StinkyPinky]

Too many of these sites post these types of benchmarks and don't even bother to explain them very well. That is a massive difference. I don't see how cache can increase performance by that much.

 

[tidy]

.

 

[rtsurfer]

When I read 4.0/4.4 in this thread, I wonder if people are aware that 4.4 is only when a Single Core is loaded.

Its 4.2 when all of them are loaded.:whiste:

 

[wasabiman123]

When I read 4.0/4.4 in this thread, I wonder if people are aware that 4.4 is only when a Single Core is loaded.

Its 4.2 when all of them are loaded.:whiste:

Yeah I know, 4.2 is still pretty good considering it's at stock.

 

[rtsurfer]

Yeah I know, 4.2 is still pretty good considering it's at stock.

I guess.
Whatever floats your boat.

Most 4770K did atleast 4.2 .
This chip is an improvement only for the people whose previous Haswell chip sucked or were on older platform.

Someone who bought Haswell last year & got a decent clocking chip should be very happy as their won't be a replacement on mainstream platform till 2015.

I know I will be, if I can resist Haswell-E that is.