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

[ehume]

Pretty much.
And I have a friend who has a 5GHz Haswell 24/7. See how useless this is?

Your likeliness to accept an anecdote as true or false is relies on what you hold to be true. Seeing that you don't think 5GHz 24/7 HSW on air is possible, you'll undoubtedly write off my claim as false.

Seeing as I believe that there's a nonzero chance of the original Haswell series being able to be 24/7 stable at 5GHz on air... I'm going to be more than happy to dismiss your claim. Even if the chance is one in a million, or a trillion, or whatever... rarely is anything ever 100%. Considering you don't think it's possible even if "Jesus blessed it," I'm going to guess you're "100% sure" it could never happen.

Regardless, it doesn't matter. It can't really be proven either way, and it wasn't even the point I was trying to make, which is that you can't write something like this off with two points of data. I was hoping I could prove that it wasn't "impossible" to hit 5GHz on air at some reasonable level of stability by way of proving that even current Haswells can do that on rare occasion, but I wasn't successful in that endeavor. I'm still not convinced that it can't happen, but it doesn't matter at this point.

So, can Devil's canyon hit 5GHz on air? I still maintain that it's too early to say.

Data is data. When theory doesn't fit the data, you know what goes out the window.

 

[rtsurfer]

Snip.

Interesting.
Does your friend have a 4670K or a 4770K.?
Should be harder in the latter, which was the CPU I had in my mind when I was arguing.

For a 4670K I think even BallaTheFeared had a 4670K that was 5Ghz Linpack stable.?

Here is a more real world sample that I still dont fully believe because it is self-reported with no proof required to be submitted.
http://www.overclock.net/t/1411077/haswell-overclocking-guide-with-statistics
Still I would take this over HWBOT charts.

There are a couple of people more their than a few months back when I looked at it.
None of the 5Ghz one could be run on air, water maybe.
I'll stop arguing now,we will have for users to get their hands on.

 

[StinkyPinky]

I suspect we will get answers on Monday. It's only a day away....patience.

 

[Ajay]

Data is data. When theory doesn't fit the data, you know what goes out the window.

In the CPU overclocking world - there are no systematic tests of a large number of CPUs (AT isn't going to buy 100 CPUs over a range of lots and then run exhaustive and controlled tests). Instead we have a bunch of individual results from reviews sites and users under a wide range of circumstances. Then we can finally say, you have a decent chance of hitting X+/- GHz for a given CPU/stepping/chipset and that's going to be as good as it gets.

Bottom line, we aren't going to know what sort of realistic ranges the new 4790 overclocks to for a few months. The reviews on Monday will give us a ball park range.

 

[witeken]

FPiednoel:

"If you don't get 5Ghz on air one devil's canyon Haswell parts, I got to give u some training ... I got 4.6Ghz on air WITHOUT Fan ... Demoed"

Seems legit.

 

[24601]

FPiednoel:

"If you don't get 5Ghz on air one devil's canyon Haswell parts, I got to give u some training ... I got 4.6Ghz on air WITHOUT Fan ... Demoed"

Seems legit.

If Review site's cherry samples are not getting 5ghz on water, theres no way in hell that you should expect an average OC of 5ghz from the average store-bought CPU on ambient air.

 

[witeken]

If Review site's cherry samples are not getting 5ghz on water, theres no way in hell that you should expect an average OC of 5ghz from the average store-bought CPU on ambient air.

He also says:

"telling how to do it is spoiling the OverClockers pleasure of looking for it, let's give them some time"

"it takes more than 30 min to get above 4.7ghz ... it needs skill."

 

[24601]

He also says:

"telling how to do it is spoiling the OverClockers pleasure of looking for it, let's give them some time"

"it takes more than 30 min to get above 4.7ghz ... it needs skill."

If we go by the HWbot example, that "skill" is simply excluding testing of anything that would usually crash the computer.

I wasted months and months trying to get honest results from people when Ivy Bridge came out, and then wasted months and months trying to get honest results from people when Haswell came out.

These truths were hard-won for me.

Don't take my keeping you guys honest for granted, if I wasn't doing this tirelessly we would still have people coming in and saying 5ghz is an average air OC on Ivy Bridge and that 4.8ghz was an average air OC on Haswell.

 

[voodoo7817]

I'm strongly considering upgrading to Devil's Canyon from my i7-860 build. Would I be able to re-use my 8gb of DDR3-1333 ram? I think this article, http://www.anandtech.com/show/7364/memory-scaling-on-haswell, implies that I won't actually lose too much performance compared to faster ram, but I thought it would be best to ask here explicitly rather than assume anything.

If I'm able to use my DDR3-1333 ram, what exactly do I lose by not upgrading my RAM to 1600+? ~5-10% (at most) performance? Additional overclocking ability?

It would be nice to save a little money if I can, as it will encourage the new build. Even if it just lets me wait for a great ram sale. Thanks.

 

[Pheesh]

From what I've seen most reviewers are lucky to get any samples, let alone specially selected samples. There is no binning that occurs once parts have a sku designation. Warehouse pulls from their available supply and direct ships to end customer, which may be an ODM, may be a reviewer, or may be a motherboard manufacturer doing validation. Reviewers get what they get, there's not some special secret super binned supply for reviewers.

 

[24601]

From what I've seen most reviewers are lucky to get any samples, let alone specially selected samples. There is no binning that occurs once parts have a sku designation. Warehouse pulls from their available supply and direct ships to end customer, which may be an ODM, may be a reviewer, or may be a motherboard manufacturer doing validation. Reviewers get what they get, there's not some special secret super binned supply for reviewers.

Engineering Samples don't come from the general supply.

I shouldn't have to explain this. I'm stopping now.

 

[tolis626]

I'm strongly considering upgrading to Devil's Canyon from my i7-860 build. Would I be able to re-use my 8gb of DDR3-1333 ram? I think this article, http://www.anandtech.com/show/7364/memory-scaling-on-haswell, implies that I won't actually lose too much performance compared to faster ram, but I thought it would be best to ask here explicitly rather than assume anything.

If I'm able to use my DDR3-1333 ram, what exactly do I lose by not upgrading my RAM to 1600+? ~5-10% (at most) performance? Additional overclocking ability?

It would be nice to save a little money if I can, as it will encourage the new build. Even if it just lets me wait for a great ram sale. Thanks.

Well, it shouldn't be too bad. Certainly not bad enough to be unbearable. And you can just add a new set of RAM sticks later. Although I don't understand why you wouldn't just overclock your memory to something like at least 1600MHz CL9. It's quite easy and risk-free most of the time.

Anyway, I refuse to believe that there will be anyone who will run a 4790k on even their Noctua NH-D15 at 5GHz. Is it possible? Sure, for a suicide run. And there may be someone somewhere that has a really golden chip that can actually do it. I wouldn't really count it as "possible", but it's still not entirely impossible, so I see what Homeles is saying.

With that said, if my CPU can do 5GHz at <1.3V on water, I'd be happy and would maybe consider running that 24/7 (120Hz gaming and all).

Honestly though, I don't understand what people were expecting. 5GHz wasn't common even on SB. Yes, there were people who could run 24/7 5GHz+ on their chips, but some make it sound as is every other SB CPU could be overclocked that high without breaking a sweat.

 

[John Connor]

Was thinking about getting a 4790k, but it looks like you can't even get to 5 GHz. http://www.pcper.com/reviews/Proces...view-and-Overclocking/Overclocking-Experience

 

[MisterMac]

Since piednoel is boasting....

Is it possible some of those FIVR things, need to be seriously tweaked\balanced and OC'ing won't just be multipliers like Sandy\IVY?


If normal ES samples were just clocked up and some idiot said "OMFG NOT STABLE" there might be some credence to Piednoel.


...that you need to find your "right settings" of many variables.

 

[cytg111]

From what I've seen most reviewers are lucky to get any samples, let alone specially selected samples. There is no binning that occurs once parts have a sku designation. Warehouse pulls from their available supply and direct ships to end customer, which may be an ODM, may be a reviewer, or may be a motherboard manufacturer doing validation. Reviewers get what they get, there's not some special secret super binned supply for reviewers.

Well well, reviewers got ES haswells to review .. and those ES clocked significantly higher than retail proved to do. Call it special binning, different wafer, whatever, the end-result is the same, in this case, review samples != retail samples.


A side note.
We know now that it would not *only* be the thermal interface that made the ES haswells clock higher.. so what else was different? Why did those ES'es clock so much better ?

 

[Pheesh]

Engineering Samples don't come from the general supply.

I shouldn't have to explain this. I'm stopping now.

Seems you are assuming. I am talking only about engineering samples- ODM's, OEM's, reviewers all pull from the same pool of engineering parts. There's no specially binned parts for reviewers, the part they get is up to the luck of the draw along with all other ES/QS requestors.

cytg111 some QS samples overclocked well, some didn't. There was no vast difference.

 

[Mondozei]

Reading Intel's Francois Piednoel's statement about how you need to "unlock" 5 GHz OC by "skill" is making me laugh.

This is the same guy who insisted that Bay Trail was the fastest SoC out there, then Anad benched it and the A7 crushed it, leaving him with eggs on his face.

He then spent his time ranting on Twitter how "supposedly" BT was still faster by uploading shaky YT videos with low res.
Im grabbing the popcorn for round two.

 

[inf64]

His tweets are confusing. It's easy and yet it takes skill to get past 4.7Ghz. Yet 4.6Ghz is done "without a fan". In 1st tweet posted on this page he's claiming it's easy to get to 5Ghz with this thing on air. How stable is it? Nobody needs a chip that can only idle at that OC... After previous PR blunders I'm not taking him seriously until I see reviews.

 

[Borealis7]

maybe i missed it, but i didn't see OCed benchmark in PCPER's review. did they only OC and not run the same test suite on the OCed chip?
i wanted to see how performance scaled in the 4.7GHz region. because if it doesn't, this is a complete waste of time and money.

 

[Idontcare]

His tweets are confusing. It's easy and yet it takes skill to get past 4.7Ghz. Yet 4.6Ghz is done "without a fan". In 1st tweet posted on this page he's claiming it's easy to get to 5Ghz with this thing on air. How stable is it? Nobody needs a chip that can only idle at that OC... After previous PR blunders I'm not taking him seriously until I see reviews.

Yep. Somebody doesn't know when they've jumped the shark IMO.

After previous PR blunders I'm not taking him seriously until I see reviews.

Me too!

 

[Racan]

Looks like Intel has enabled TSX on Devil's Canyon

http://ark.intel.com/compare/80811,80807

Let's hope Intel will enable all instructions with future K chips and never see a repeat of the crap it pulled with it's first unlocked Haswell chips.

 

[Bman123]

Well its Monday morning, let's see if any legit reviews show up for these today

 

[JoeRambo]

Looks like Intel has enabled TSX on Devil's Canyon

http://ark.intel.com/compare/80811,80807

Let's hope Intel will enable all instructions with future K chips and never see a repeat of the crap it pulled with it's first unlocked Haswell chips.

AMAZING find! I really hope this means that Intel marketing guy who was responsible for this absurd segmentation has been promoted out of his job (politically correct version of being hit by truck).

I really wanted to test some of our code and latest stuff in JDK's, but Intel somehow managed to bar me from doing it on 4770K, amazing way to woo developers.

 

[TerryMathews]

There goes the apologist's explanation that those instructions were disabled "because overclocking."

 

[witeken]

Reading Intel's Francois Piednoel's statement about how you need to "unlock" 5 GHz OC by "skill" is making me laugh.

This is the same guy who insisted that Bay Trail was the fastest SoC out there, then Anad benched it and the A7 crushed it, leaving him with eggs on his face.

He then spent his time ranting on Twitter how "supposedly" BT was still faster by uploading shaky YT videos with low res.
Im grabbing the popcorn for round two.

No one knew about A7, so those claims were valid. A7 launched only 1 day before IDF. And unless you want to be vendor locked-in by Apple, Bay Trail has still the fastest CPU today. But I agree that you should take his claims with a grain of salt.