Forget Intel Broadwell, Skylake On the Way

Thanks for the reply. About the RAM and storage subsytem, yeah definitely useful. Copy'd & pasted.

Yeah, I get that realistically and logically I should just buy something and get it over with. Hell, I won't even be using it for any heavy duty stuff. No hi-res games. No video editing. Just some work for school, painting and maybe a few casual games.

But every time I think about Skylake, it make me want more! I'll.. I'll deal with that eventually.

Thanks, again.

 

Some features of Skylake graphics architecture

 

^Seriously considering going from a Haswell to a Broadwell to a Skylake MacBook Pro.

I planned on going from Sandy to Broadwell, but then I spilt water all over my Sandy, forcing me to buy a 13" MBP Haswell. I need a 15" though, and don't want to wait until Skylake, but everything fantastic is coming with Skylake.

Edit: Also, Maxwell am cry. Dedicated GPUs are going to be for the drivers only for many users.

 

Classic Intel, by the time they release Skylake DX12 will be here and it'll be (planned) obsolete. whoops

~Aeny

 

DirectX 11.2 but OpenGL 5.0 and OpenCL 2.1?! Wat?

 

So is this new architecture Intel's tick or tock?

 

It wouldn't surprise me if in typical Intel fashion, they released an updated driver at least partially support DirextX 12 4-6 months into the platform's lifespan. I'm also not sure if we'll actually see games using DirextX 12 in 2014.

 

It'll take a few years for DX12 to reach a point where you NEED it in order to play a game.

For example, BF3 telling DX9 users to go away.

(Though it didn't stop a handful of determined people altering BF3 to run on DX9 or hack-installing DX10 on Windows XP.)

 

Skylake in 2014? first see then believe. Although I'd be happy to see it in 2014 And yes, it's not about the need to HAVE it, it's more like it would have been nice to have this at some point, it's not impossible for the IGP not to have this.
I'm sure they could add this in later with a driver update, but I honestly don't expect Intel to do that, that'll be for the generation after Skylake I'm sure.
Even if we got DX12 on it, by the time DX12 only games are going to be out the IGP will probably be close to worthless to play those games at decent settings anyways, but hey I got no crystal ball

Typical Intel fashion? Care to point out when they added DX support with a driver update because I can't think of any.

TLDR: it's just a number and I like to see the latest numbers when I purchase something with my hard earned money

~Aeny

 

DirectX, I couldn't give you examples, but OpenGL and CL I can.

 

They can likely release *support* for it but many features will require appropriate hardware in the GPU to take advantage of it. But in the past major version level changes (i.e. DX9 to DX10 to DX11) meant that if your GPU didn't have the feature set it would not run that DirectX. Like when many DX11 games came out there were a lot of users that were "stuck" with DX10...

 

I really don't know enough about GPU's to fully understand this at all, but from what I understand there are supposed to be pieces of the GPU that are somewhat programmable, thus they can add a limited amount of features this way I suppose, they could of course opt to completely implement it in software but that would probably be slow as hell By the way I know GPU's can't magically grow extra hardware to support new things that were announced after the hardware is released, to make that clear. Would be great if they could though.

I can certainly believe they bumped openGL or CL this way, perhaps how I described above. But it doesn't mean that they do, if I recall correctly the original Intel HD is stuck one OpenGL release below what it's hardware can handle for example. Sorry Mr. Wonderful that may have sounded a bit mean, I was just interested if they (partly) ever did this and perhaps never announced it.

Finally just to throw it out here, didn't nVidia and AMD already say they were going to bump their latest GPU's to DX12? if they can I'd be surprised that Intel can't on something that isn't released yet? I guess I just want to see better and faster (a lot faster) IGP's where DX12 would be 'closer to the metal' and hopefully faster. One may dream I suppose.

As a sidenote, I should really read up on how GPU's work.. but exams.. must stop posting and study encryption
~Aeny

 

Intel Desktop CPU Roadmap Updated
4.5/15/25W Skylake U, Y series debut around Computex 2015.

 

You can bump your GPU product to DX15 if you like, by... software emulation. (That's what the OpenGL driver has been doing.)

 

So there is really only a 6 month life to Broadwell before Skylake hits. Interesting.

Depends on the feature set. Past DX updates have required features in the hardware to utilize the latest DirectX. I'm sure you could emulate them but at a significant hit to performance.

 

DX12's main features is allowing closer to metal programming (AMD stated Mantle API can be ported to DX12 due to programming similarities) and better multithread support.

From what I've read, those two features can't be emulated. Only the ones pertaining to graphics.

 

The performance gain can't be, but the API can. Actually a reference implementation of a graphical API used purely for testing/debugging purpose during the beginning of the API life cycle usually involves no graphical hardware at all. It's a pure CPU implementation.

OpenGL drivers always involve some hardware support and some software emulation. The Quadro/FirePro ones just have more on the hardware side. I'm not familiar with D3D stuff, but I won't be surprised if they do the same thing on this API, just to a less extent.

 

More details on Skylake processors

 

The dual-core chip with GT3 graphics with 64MB of eDRAM will be an interesting part. It'll be great for high-end ultra books, or make ultra books great for "Prosumer" photo work, etc.

The lack of DDR4 in the lower series is still disappointing.

All the same, I think we have a good idea of what processors are going to end up in which MacBooks.

 

Intel shipping Broadwell, but next-gen Skylake chip could slip | PCWorld

Intel keeps Skylake PC processors on schedule for next year - Computerworld

 

AnandTech | IDF 2014: Intel Demonstrates Skylake, Due H2 2015

Intel demos Skylake silicon; production expected in 2H 2015

 

Still not convinced that Intel will release Skylake till 2016... Sure it may be working but they said the same thing about Broadwell last year and it won't be out till Q1 2015...

 

Sure, the proof will be in the delivery (time), but the teething problems will not need to be rehashed with Skylake as Broadwell is now 'done'.

We'll see what back to school looks like next year for thin, light and all day computing with Intel platforms.

 

Skylake-Y CPUs and motherboards spotted

 

Not really sure what that link tells us?

Could you decipher it a little for me? Ty.

 

It basically states that there are motherboards for Skylake-based systems being manufactured and tested for validation in India & Malaysia.

 

Yeah exactly, basically an "FYI Skylake is being tested" post.

 

Intel Launching 17 Broadwell-U 14nm Processors at CES 2015 - Will Feature Intel Iris Graphics 6100 GPU

Looking at the speeds, there is no point to the "M"s anymore.

 

are you stating just for the dual cores or also the quad cores? I wouldn't mind a 4.5GHz dual core. That would be pretty cool and it would only require 30-35 watts...granted that also doesn't give any room for the GPU to function but their is much room for higher freqs and TDPs

 

The way the GTe is going I can't help wonder if it's going to become a GPU with integrated CPU lol.

 

No quad cores. :\

Though I suppose most people probably only need a dual-core laptop with hyper threading and a great iGPU.

 

That's what AMD APU's are pretty much.. The new ones have absurd amounts of GPU power but the CPU cores are junk.. For Intel it would probably become the other way around

 

Well right now it it the other way around, but looking at the relative improvements of the CPU cores vs the IGP from Sandy Bridge to Haswell, it's not hard to see where Intel's focus is currently.

 

That sounds very NVidia,

 

so your saying there focus is on CPU right? because everyone on here expected the igpus to be much larger by now

 

I'm very sure he mentioned that Intel is focused on making a "good enough" IGP to counter AMD's APU.

Have you seen Broadwell M's IGP? It's bigger than the two CPU cores and memory controller combined.

 

Sort of, what I was really getting at is that Intel knows it's light years ahead of AMD in terms of CPU performance, so they can afford to devote more resources and focus on making a better IGP. Hence improvements in CPU performance have been miniscule compared to the huge leaps Intel has made from the IGP of Sandy Bridge to Haswell.

Hell at this point Intel could probably stop R&D activities of their CPU division for 2 years and AMD still wouldn't be able to achieve parity.

 

Current observations indicate that Intel is focusing on improving the IGP, not the CPU.
The CPU has undergone minor revisions at best... the subsequent improvement in performance increased by roughly 5% (at best) every 2 years since Sandy Bridge debuted.
So, we are looking at about 15% improvement from Sandy Bridge to Haswell for the most part.
Broadwell will result in another 5% increase in performance on the CPU side.

Most of the improvements were in the IGP and TDP overall.
Broadwell is supposed to feature a more powerful IGP... but how this reflects on their CPU as a result is a question.

While AMD's CPU might not exactly 'compare' to Intel, they do have a rather powerful IGP as a result.
This could be one of the reasons which contribute to CPU weakness (At least in single core tasks)... but if you compare HSA enables programs compared to standard programs which mainly rely on the CPU... then AMD Kaveri for instance (Even the weakest component) heavily outperforms Intel.

Essentially, Intel is a brute force solution, while AMD is shifting in a different direction.
Problem is... there's practically little to no support for HSA at the moment on a large scale.

It will be interesting to see what Broadwell brings to the table though...
But if current news are any indication, the quad core mobile solutions will be severely delayed.
Any idea on when they will follow?

 

BS for one....we keep getting people making these numbers up -_- was 5-10% each year now 5% every 2 years rofl

second bold is inconsistent with the rest and makes no sense at all.

I wouldn't dare say intel is focusing on GPU personally. It is a joke compared to what they can do if they choose to do so but they rather slide by with an under powered GPU.

 

I disagree that Intel's graphics are "a joke" in their current state. It's true that they used to be that way, but Intel has been working hard the last few years to get their graphics closer to equal to AMD & Nvidia's low end, with their Iris graphics close to midrange. Intel's biggest priority now is energy efficiency, and the limiting factor for both processor and graphics is that.

 

Look I don't want to argue, and that "5% every 2 years" statement is certainly wrong, BUT there's no denying that we've been seeing progressively less IPC improvements over the years if you compare the jumps made from Core to Nehalem, then Nehalem to Sandy Bridge, and finally Sandy Bridge to Haswell.

 

Lets not forget improvements like AVX, Skylake bringing 512b AVX registers to the architecture, although unfortunately software takes several years to catch up.

The reason for the "integrated CPU" post was from some older data showing current limits for Skylake. The 95W unit with GTe having a TDC of 65A max 95A for IA cores while the GT cores where TDC 95A max 156A. Not sure how the voltage scales but on my HSW 4700MQ the iGD will consume a little over 30W on heavy load at 1.050V 1400MHz (~30A) while the IA cores will go to 70W at 3.5GHz with 1.000V (~70A). The iGD thermals are worse for whatever reason, that is GT cores are hotter Watt for Watt vs IA cores.

So looks like the GTe cores can pull substantially more power than the IA cores.

Still not sure about the FIVR, whether it's being ditched or if there are going to be IVR's for each domain.

 

i just hate numbers being made up and someone reading this and thinking they are real :/ bad info is a B to kill

 

How do you sleep at night? lol

 

No rest for the wicked

 

Get a hot GF/wife.

 

yea but if they are a B...it doesn't work out the way it was planned :/

<----this guy

i myself have spread a ton of crap info over the years because of not fact checking and i try not to do that at all and try to prevent that. I have to fact check nearly everything because people use axioms and made up crap like its cool.

its better post a "it isn't much" vs "5% (made up)" because...well you know why.

 

Back in the frequency war era statements like "+5%" could be kind of true. Old habits die hard.

 

I would say every time you spread crap info back then 2 more people spread it for 4 generations leading to 40 people using crap info...

JK

I would argue it fueled the fire though.

 

When I wrote that reply, it was late.
'Every 2 years' might have been an exaggeration and incorrect... however, it would be more accurate to say 'about 5% increase every year' (and even then under highly specific circumstances - more is possible, but again, under 'best case scenario').

Point remains that Intel focused a lot more on the IGP instead of the CPU.

EDIT:
Checking some other numbers seems to support the idea of a very low difference in IPC between Sandy Bridge and Haswell CPU performance:
http://www.pcper.com/reviews/Proces...erformance-and-Architecture/Clock-Clock-Sandy

Best case scenario was 20%... most other cases (majority) seems to indicate a difference of 5%, maybe up to 10%.

Difference: SB release date 2011... Haswell release date 2013?

That's 2 years of difference... which roughly amounts to my initial post being very close to accurate.

The numbers aren't imaginary/made up... they are what I understand, 'direct observations'.