Intel Skylake / Kaby Lake

[SAAA]

So those slides are complete rubbish.?..

But then why should the IPC be accurate, as said there s not a single bench, at least to my knowledge, that has been published by Intel.

If I'm reading that slide correctely it looks like they are pointing out the improvements of Haswell over previous generation in the first point, saying broadwell is based on that, then they added TSX and other minor improvements for that 5.5%.

Btw it's actually less than that in some applications, around 3% on average:
http://www.sweclockers.com/artikel/...mot-haswell-prestanda-vid-samma-klockfrekvens

Broadwell-K gets more only for the larger cache so it's harder to use it as a comparison in clock for clock basis.
Funny how the L4 alone is almost equal to a tock increase... it's better they implement it on all the lineup someday. :hmm:

 

[Sweepr]

If I'm reading that slide correctely it looks like they are pointing out the improvements of Haswell over previous generation in the first point, saying broadwell is based on that, then they added TSX and other minor improvements for that 5.5%.

Btw it's actually less than that in some applications, around 3% on average:
http://www.sweclockers.com/artikel/...mot-haswell-prestanda-vid-samma-klockfrekvens

According to NotebookCheck it's around 5% on average.

Apart from per-MHz performance, which is about 5% higher in Broadwell...

www.notebookcheck.net/Broadwell-vs-Haswell-Duel-of-the-ULV-Processors.138010.0.html

I did a quick comparison of the individual scores, int and float subtests, using those and other scores, they present a much better picture than the GOPS value:

So a lot faster in integer and about the same in floating operations. AVX3 should adress that, on Xeons at least...

Nice job.

 

[Walter E Kurtz]

How different are the days of Broadwell and Skylake release, compared to the first Core 2 generations where leaks hit you in the face left and right. Even now after Broadwell is listed in webshops and the official reviews have been published, we're still waiting for overclocking numbers. I will assume that both 14nm architectures are overclocking lemons until proven otherwise.

 

[Dave2150]

How different are the days of Broadwell and Skylake release, compared to the first Core 2 generations where leaks hit you in the face left and right. Even now after Broadwell is listed in webshops and the official reviews have been published, we're still waiting for overclocking numbers. I will assume that both 14nm architectures are overclocking lemons until proven otherwise.

We've seen that the 5775C can do 4.2Ghz stable on 1.281V - though the fact that we haven't seen any stable overclocks past this is indeed fishy.

 

[Olecki]

We've seen that the 5775C can do 4.2Ghz stable on 1.281V - though the fact that we haven't seen any stable overclocks past this is indeed fishy.

On one tech portal from my country person, who visited Computex2015, says that 4.2 GHz is max. It's not about temeratures, but some complicated enginering stuff and there wouldn't be stable CPUs over 4.2GHz. So probably that's why nobody publishing OC tests . It should be fixed in next product line - Skylake.
Of couse this is kind of 'rumor', but usually this person seems to be well informed.

 

[BigDaveX]

How different are the days of Broadwell and Skylake release, compared to the first Core 2 generations where leaks hit you in the face left and right.

Different circumstances. Intel needed those leaks back in 2006, as their product line (with the arguable exception of the Core 1) was just getting annihilated across the board by AMD, and investor confidence in the company was the lowest it had been in over two decades. Nowadays, those sorts of leaks wouldn't achieve much of anything.

 

[Fjodor2001]

Different circumstances. Intel needed those leaks back in 2006, as their product line (with the arguable exception of the Core 1) was just getting annihilated across the board by AMD, and investor confidence in the company was the lowest it had been in over two decades. Nowadays, those sorts of leaks wouldn't achieve much of anything.

So you are suggesting the leaks were intentional and originated from Intel itself?

 

[Dave2150]

Nope.

That picture is complete nonsense - Haswell was not a 10% jump over Ivybridge - just take a look at the 4770k review on Anandtech - it's 5-6% at best.

 

[Lee Saxon]

So you are suggesting the leaks were intentional and originated from Intel itself?

Almost any product release, celebrity gossip, political, or any other kind of "leak" you see is an intentional PR move.

 

[mikk]

That picture is complete nonsense - Haswell was not a 10% jump over Ivybridge - just take a look at the 4770k review on Anandtech - it's 5-6% at best.

10%: http://www.hardware.fr/articles/897-25/gains-moyennes-cpu.html
8%: http://www.computerbase.de/2013-06/...or-fuer-desktop-pcs-test/14/#abschnitt_gesamt

It depends on the applications tested. In general, Haswell is 8-9% better clock per clock. It can be more or less depending on the application tested. Usually CPU gaming tests from Anandtech are nonsense because it's all GPU bound.

 

[MountainKing]

Whooaaa...Now to get to know about pricing!

 

[ShintaiDK]

Whooaaa...Now to get to know about pricing!

Just look at current pricing and add inflation.

 

[imaheadcase]

I wonder if we will ever have a product %50 faster than last gens, or if marketing will limit that from even happening.

 

[ShintaiDK]

I wonder if we will ever have a product %50 faster than last gens, or if marketing will limit that from even happening.

Define 50% faster. And we get whats possible within what the 99% crowd wants.

 

[MountainKing]

Just look at current pricing and add inflation.

Maybe the end of the road is finally coming for my E6600. Probably will be i5 6600 IF the motherboards and RAM is properly priced (unlike 5820K boards).
The i7 6700 might sell for $500?

 

[ShintaiDK]

The i7 6700 might sell for $500?

No. It will sell around 350$.

 

[MountainKing]

Looks good. Need to figure out board and ram costs

 

[Sweepr]

The iGPU of Skylake, according to the public record currently available, should be part of the lot of GPU compatible with the Tier 2 (resource-binding model) of DX12, and will exploit the Feature Level 12.0.

This not only puts the iGPU of Skylake on another level than the GPU currently integrated in Haswell and Broadwell (Tier 1 and FL 11.1), but considerably closer to the GPU Intel GPU Maxwell 2.0 of NVIDIA in terms of functionality, while overcoming Maxwell 1.0 and Kepler GPUs always the house of the chameleon.

Tier 1: Intel Haswell and Broadwell, NVIDIA Fermi
Tier 2: NVIDIA Kepler, Maxwell 1.0 and Maxwell 2.0, Skylake
Tier 3: AMD GCN 1.0, 1.1 and GCN GCN 1.2

Feature level 11.0: NVIDIA Fermi, Kepler, Maxwell 1.0
Feature level 11.1: AMD GCN 1.0, Intel Haswell and Broadwell
Feature level 12.0: AMD GCN GCN 1.1 and 1.2, Skylake
Feature level 12.1: Maxwell NVIDIA 2.0

www.bitsandchips.it/hardware/9-hard...e-supportera-il-feature-level-12-0-delle-dx12

 

[ame132]

actually, intel kind of retracted its slides about skylake and feature levels
i hope this means skylake is 12+, not 11+
https://forum.beyond3d.com/threads/direct3d-feature-levels-discussion.56575/page-14#post-1849354

 

[imaheadcase]

Define 50% faster. And we get whats possible within what the 99% crowd wants.

Umm, it defines itself. lol

 

[ShintaiDK]

Umm, it defines itself. lol

50% faster IGP? 50% faster quicksync, 50% faster memory, 50% faster cores?

 

[kimmel]

50% faster IGP? 50% faster quicksync, 50% faster memory, 50% faster cores?

50% more magic is what is usually meant by those asking for 50% faster... without providing units of measurement.

 

[IEC]

50% faster IGP? 50% faster quicksync, 50% faster memory, 50% faster cores?

Probably means a benchmark suite like AT runs for CPU reviews. As in, composite performance index.

 

[knifethesky77]

Have there been any leaks about skylake-H? I plan on upgrading my little beast Nehalem 720QM to the next skylake-H counterpart when it finally arrives and I was wondering what kind of performance gains I could expect?

 

[Enigmoid]

Have there been any leaks about skylake-H? I plan on upgrading my little beast Nehalem 720QM to the next skylake-H counterpart when it finally arrives and I was wondering what kind of performance gains I could expect?

Upgrading to haswell is > 2x already.

i7 720qm is around 2.9 points in CB 11.5 Multi. (current U level chips can do better)

Most haswell chips (4700m) score around ~7.