Intel Skylake / Kaby Lake

[Abwx]

You then throw around a lot of technical jargon that supposedly "proves" that Intel is all marketing fluff and no technical substance, but IMO your arguments don't actually make any sense because you basically hint at stuff and hide a lot behind "oh here's a fancy formula, if you were an EE you'd get it."

I m here to discuss technical matters in technical terms, now if all Intel is providing are numbers on marketing slides so good but then dont come stating that i m saying that they are lying..

Either they provide numbers to support their claims or else they are not that honnest for me, at some point the enginering and tech crowd wants measurements in a controled environment, that s all enginering is about, not reviews like this one :

http://anandtech.com/show/9117/analyzing-intel-core-m-performance

Where are the power measurements if it s really an analysis of Core M..???.

Or was it a deliberate requirement from Intel..?.

Indeed the scores and boost frequencies tell us that these are not performed at 4.5W, and by a long shot, yet all is made to make people think that it s the case, so how do you call this if not deliberate misleading of the public..?

 

[crashtech]

I guess all that's left is to argue minutiae, when AMD's latest apparently can't score a victory at ANY wattage.

 

[Abwx]

I guess all that's left is to argue minutiae, when AMD's latest apparently can't score a victory at ANY wattage.

But that s all the problem, the competition has no wattage other than the ones at Arkintel, so how can they only claim that they are better..?.

At the end it s rather a huge marketing machine that keep repeating the same slogans just to make sure that people really think that they are really better.

Otherwise how do you explain that the Core M "analysis" is done without the slightest power measurement yet the same review state that perf/Watt is much better than previous gen, you think that it s serious analysis...?...

 

[CHADBOGA]

But that s all the problem, the competition has no wattage other than the ones at Arkintel, so how can they only claim that they are better..?.

At the end it s rather a huge marketing machine that keep repeating the same slogans just to make sure that people really think that they are really better.

Otherwise how do you explain that the Core M "analysis" is done without the slightest power measurement yet the same review state that perf/Watt is much better than previous gen, you think that it s serious analysis...?...

Look at the form factors the OEM's fit the CPU's into and how long their battery life is under various usage models.

 

[coercitiv]

Either they provide numbers to support their claims or else they are not that honnest for me, at some point the enginering and tech crowd wants measurements in a controled environment, that s all enginering is about

What stopped you until now to get a Haswell U platform, measure power usage under load and publish your findings? This is not a car emission test, the budget, tools and knowledge needed to make the necessary measurements are well within your reach, and any other person with decent technical training.

You know what the biggest problem with your theory is? Everyone must be in on it. The whole wide world! Even AMD employees have to conspire with Intel in order to keep the ugly truth under wraps. Moreover, when you insist with your theory even in the face of proof from other forum members, you are insulting everyone of us since we must be either liars or incompetent to produce evidence that contradicts your theory.

Make your measurements. Publish your findings. Prove me wrong.

 

[AtenRa]

What stopped you until now to get a Haswell U platform, measure power usage under load and publish your findings? This is not a car emission test, the budget, tools and knowledge needed to make the necessary measurements are well within your reach, and any other person with decent technical training.

You know what the biggest problem with your theory is? Everyone must be in on it. The whole wide world! Even AMD employees have to conspire with Intel in order to keep the ugly truth under wraps. Moreover, when you insist with your theory even in the face of proof from other forum members, you are insulting everyone of us since we must be either liars or incompetent to produce evidence that contradicts your theory.

Make your measurements. Publish your findings. Prove me wrong.

Something better, why dont WE ALL ask AT to make the same review with Intel parts ??

 

[coercitiv]

Something better, why dont WE ALL ask AT to make the same review with Intel parts ??

I would love to see a similar review with Intel parts, some Skylake and one Haswell / Broadwell for comparison. In fact, I would love to see more reviews like this in general, since this is the kind of material I expect and appreciate from Anandtech.

Problem is, the conflict we're having here cannot be resolved this way. Abwx already made up his mind about the competence of Ian Cutress as a tech reviewer:

Anyway it s always worrisome to read that some reviewer who has weak knowledge about electronics come to misrepresent a product due to his own incompetence in the matter, said reviewer actualy should apologise for such misrepresentation of AMD s APU.

This is why I decided to stop invoking 3rd party sources as proof, I've stopped invoking my own experience and measurements as proof, and I find myself with only one option left: ask for measurements from Abwx. I'm not doing this to provoke, I'm doing this because I genuinely believe this is the only way he may discover the truth. Either I'm incompetent or he's deluding himself.

PS: and don't get me wrong - I respect the technical expertise and the value of Abwx as a forum member but I also believe anyone of us can get blindsided sometimes and we need to face the facts sooner or later. I have stood by him when he was right, I find myself in total opposition in this situation.

 

[zir_blazer]

For the uninformed, Intel GPUs since haswell can be pass-thru'd into the VMs without separate monitors thanks to GVT-g. This technology is currently impemented on linux as XenGT and KVMGT.

XenGT and KVMGT (iGVT-g) are for GPU virtualization, not Passthrough. Intel calls IGP Passthrough GVT-d, and at least on Xen is supported since 4.6 I think.

 

[Rngwn]

XenGT and KVMGT (iGVT-g) are for GPU virtualization, not Passthrough. Intel calls IGP Passthrough GVT-d, and at least on Xen is supported since 4.6 I think.

The iGVT-g was called "mediated pass-trough" at some point, so I keep calling it that. For GVT-d, you need a dedicated card and extra monitor to make it work.

 

[zir_blazer]

GVT-d doesn't need a dedicated card and extra monitor. You can use it to pass the IGP (And its attached monitor) to a VM, and may access the host via SSH or something like that. Basically, you have GPU and monitor on VM, and a headless host.
Intel IGP passthrough required some extra patches compared to a discrete Video Card since it works a bit differently, as it is directly under the PCIe Root if I recall correctly.

 

[dark zero]

I guess all that's left is to argue minutiae, when AMD's latest apparently can't score a victory at ANY wattage.

And that's a big problem.. and even when BD was decent on Piledriver, it just went worse with Steamroller and Excavator.... AMD failed to do anything good in that 2 generations!

 

[mikk]

Iris Pro 6200 claim was that it was 80% faster than Haswell. Haswell what? Well 6200 is only 20% faster than 5200 so its against GT2 Haswell, HD 4600 graphics.

Why do you have to lie? Intel never claimed 80% faster than Haswell GT3e.

You can't be sure. I have looked at all the slides before I posted that. Have you? Up to 50% on a marketing slide means best case.

You can't be sure either. Up to in Intels test case benchmark, in the past it was based on some 3dmark benchmark. Broadwell "up to" was based on 3D Mark (Cloud Gate I think) in Intel slides. Up to 20% or 25% for BDW ULV over HSW ULV, it wasn't a best case. 3dmark11 was in a +50% range. There are much more examples, all games with 2xMSAA support for sure.

Skylake GT2 is twice as fast as Haswell GT2 at HW Geometry Instancing from Caps Viewer at the same clock speed. Even without 2xMSAA. Far away from any up to claims from Intel. The link you gave me says 3dmark11 and Vantage, means no best case most likely.

 

[IntelUser2000]

Why do you have to lie? Intel never claimed 80% faster than Haswell GT3e.

Why are you so offended?

You don't remember what happened do you? It's about Intel, not you. Anyway couple of websites reported that Broadwell GT3e is 80% faster than Haswell GT3e. Well, they later retracted the statement saying that its over GT2. Because the original source must have not been clear, saying 80% is in general Broadwell over Haswell, and people thought "ah it must be BDW GT3e vs HSW GT3e."

I remember somebody saying that Broadwell GT3e can't be only 20% better than Haswell GT3e. I said that 80% number is way too high nevertheless unless they have done something significant, and how it might plausibly against GT2 parts and how it might make sense as well. And that's not a lie.

This 80% number is the same thing. But its merely a comment.

Up to 20% or 25% for BDW ULV over HSW ULV, it wasn't a best case. 3dmark11 was in a +50% range. There are much more examples, all games with 2xMSAA support for sure.

All marketing numbers are taken with a massive dose of salt. Either they are flat out lying, or the system they are testing are reference systems which don't come out in reality, or they only do tests which show the chips in the best light. I can think of Carrizo and Core M which turned out to be a massive difference from what they claimed. And no, I don't look at "HW Geometry Instancing" benchmarks and such other low-level tests unless I want to analyze the architecture. Those numbers are absolutely irrelevant to figuring out how the unknown part will perform.

 

[Dresdenboy]

Sorry, if already posted:
SKL breaks 7GHz at 4V:
http://www.pcper.com/news/Processors/Extreme-Overclocking-Skylake-702566-GHz

 

[VirtualLarry]

4V! Obviously LN2, and not air.

 

[skipsneeky2]

What's a good i3 6100,good motherboard and 8gb ram kit?Got a buddy who wants me to configure a build and its either going to be based around this cpu or the i5-6500.

I guess with the i3 6100,it only makes sense to get a board that overclocks?Are non overclocking boards considerably cheaper or getting one that overclocks really not much more?

 

[inf64]

What's a good i3 6100,good motherboard and 8gb ram kit?Got a buddy who wants me to configure a build and its either going to be based around this cpu or the i5-6500.

I guess with the i3 6100,it only makes sense to get a board that overclocks?Are non overclocking boards considerably cheaper or getting one that overclocks really not much more?

He should just buy i5-6500.

 

[ShintaiDK]

And Z170, barely cost more. The price difference is also in components.

 

[skipsneeky2]

And Z170, barely cost more. The price difference is also in components.

Z170 is the chipset that overclocks both the i3 and the i5 right?Still budgeting and configuring the build.

As much as i would like picking the i5 6500,overall build budget may limit to the i3 6100.I heard such good praise about this cpu.

 

[ShintaiDK]

Z170 is the chipset that overclocks both the i3 and the i5 right?Still budgeting and configuring the build.

As much as i would like picking the i5 6500,overall build budget may limit to the i3 6100.I heard such good praise about this cpu.

Z170 also allows memory speed above 2133Mhz. Something Skylake likes a lot.

 

[skipsneeky2]

Z170 also allows memory speed above 2133Mhz. Something Skylake likes a lot.

I was checking on that,found this where with 2666 memory it just flys.

http://www.eurogamer.net/articles/digitalfoundry-2015-intel-core-i3-6100-review

 

[Burpo]

Z170 is the chipset that overclocks both the i3 and the i5 right?Still budgeting and configuring the build.

As much as i would like picking the i5 6500,overall build budget may limit to the i3 6100.I heard such good praise about this cpu.

http://www.superbiiz.com/detail.php...pxLuRBmX2ETyAbErX_kvOOcV11h4ccGwogaAkBi8P8HAQ

 

[skipsneeky2]

http://www.superbiiz.com/detail.php...pxLuRBmX2ETyAbErX_kvOOcV11h4ccGwogaAkBi8P8HAQ

Looking into squeezing this into the build,if possible.

 

[DrMrLordX]

Be careful, the non-k overclocking is starting to go away. You may have to use older UEFI revisions to accomplish the feat. Though, if you have read this entire thread, you would know that already.

 

[crashtech]

Pretty sure my BCLK OC is hosed, and that after I was foolish enough to buy a non-K 6700!