Intel Broadwell Thread

[positivedoppler]

I actually heard somewhere on the radio this morning advising people that Broadwell makes no sense whatsoever and we should just wait for Skylake because Skylake is right around the corner. Not sure what Skylake has to offer but something about wireless charging.

 

[Idontcare]

I actually heard somewhere on the radio this morning advising people that Broadwell makes no sense whatsoever and we should just wait for Skylake because Skylake is right around the corner. Not sure what Skylake has to offer but something about wireless charging.

It is true, Broadwell makes no sense provided you aren't interested in improved performance (performance/clock, performance/watt, performance/dollar, etc.).

Provided none of those metrics are relevant or of interest to you (the royal you), you are better off waiting for something that won't be released until a later date.

I will probably skip both and hold out for something 10nm or later.

 

[frozentundra123456]

I actually heard somewhere on the radio this morning advising people that Broadwell makes no sense whatsoever and we should just wait for Skylake because Skylake is right around the corner. Not sure what Skylake has to offer but something about wireless charging.

"Somewhere on the radio", I will have to remember that source. Sounds really well documented.

 

[positivedoppler]

"Somewhere on the radio", I will have to remember that source. Sounds really well documented.

Lol, I know. But hey, at least I admit I heard it ambiguously on a radio. I was driving on my way to work and going through radio stations so I'm not even sure who was speaking. But he claim since Broadwell is available for the most part in First 1/2 of 2015 and SKylake will be in the 2nd half of 2015, it makes very little since to purchase Broadwell. Plus Skylake is suppose to be a "game changer" of some sort.

 

[frozentundra123456]

Lol, I know. But hey, at least I admit I heard it ambiguously on a radio. I was driving on my way to work and going through radio stations so I'm not even sure who was speaking. But he claim since Broadwell is available for the most part in First 1/2 of 2015 and SKylake will be in the 2nd half of 2015, it makes very little since to purchase Broadwell. Plus Skylake is suppose to be a "game changer" of some sort.

I think it depends on the application. Broadwell is supposed to be a good advance for mobile, but will have very limited availability on the desktop. Hopefully for us desktop diehards, Skylake perhaps will be more focused on increased cpu performance instead of just gpu and low power. It is all speculation at this point though. With all the delays and sort of disappointing performance increases lately, I would just buy what I want and best suits my needs now and not count on anything in the future, unless it is better igpu, which does not really interest me.

 

[jpiniero]

I still expect the Skylake locked quads to be slower than the 4790K. Broadwell-K is a bit of a wildcard due to the edram and it's TDP rumored to be 65W.

 

[positivedoppler]

I think it depends on the application. Broadwell is supposed to be a good advance for mobile, but will have very limited availability on the desktop. Hopefully for us desktop diehards, Skylake perhaps will be more focused on increased cpu performance instead of just gpu and low power. It is all speculation at this point though. With all the delays and sort of disappointing performance increases lately, I would just buy what I want and best suits my needs now and not count on anything in the future, unless it is better igpu, which does not really interest me.

So far I have heard there is no delay in Skylake. If that is true, then the time gap between Broadwell and Skylake does shrink. What will Intel do? If Intel launch Skylake on schedule, then people might shy away from Broadwell. If you delay Skylake, then competition might catch up. Though I am not sure who is Intel's competition in Broadwell and Skylake. Besides Apple and Samsung who manufactur their own chip for tablets, there really isn't much of a high end tablet market outside of Microsoft surface.
Convertibles and high end low power laptops?...Intel has no competition right now in either market. It makes them no sense to push out Skylake at the expense of Broadwell.

 

[Roland00Address]

These devices are confusing me. On one hand they are offering better battery life, but on the other they certainly are not as powerful as the haswell U series and are not going to be much cheaper either. It's painful to own an expensive device and then knowing there is a much more powerful one for a couple hundred more. I kept myself from buying any Haswell U series device just because of that. I think I'll wait for the mainstream broadwell U-series. Hope they will be closer to the Haswell M series in processing while consuming way less power.

(Now where I am getting these estimates between broadwell 5 watt and broadwell 15 watt is comparing the current haswell u series which is 15 watt tdp vs the haswell y series which is 4.5 watt sdp)

Do not worry about the performance difference between broadwell 5 watt vs broadwell 15 watt for it looks like on cpu performance they are going to be very close. We are talking within 75 to 85% of each other. In normal every day terms you are not going to notice the difference between something taking 4 seconds vs something taking 5 seconds, especially since broadwell y will get you similar cpu performance as haswell u.

The reason why all this occurs is due to how turbo works, in short bursts they have practically identical performance it is the sustained usage where the u series processors will be better. You are not going to on a tablet, run cinebench, some video conversion, rendering whatever on these devices and if you do it will still do it just as well in quality it will just take an up to 33% longer.

Now the gpu performance between broadwell 5 watt vs broadwell 15 watt will probably be huge for that is very much thermally bound. Hell you see this being a problem where you go from 5 watts to 65 watt with iris pro in each situation give the chip more tdp and it will perform substantially better (comparing the macbook pro retinas vs the imacs)

That said we are looking at some very nice graphics in the y series for a 5 watt tdp chip, we are talking (if the 3d mark scores are accurate)

• between the intel 4400 (haswell 15w) and 4600 graphics (haswell 37 to 47w).
• close but still not there yet of nvidia's gt630m/gt720m level of performance
• close but still not there yet with amd apusin the r6 apus / 35w tdp range. The r6 is the graphics of the 28nm A10s, in laptops FX are now the best 28nm apus.

 

[Roland00Address]

I actually heard somewhere on the radio this morning advising people that Broadwell makes no sense whatsoever and we should just wait for Skylake because Skylake is right around the corner. Not sure what Skylake has to offer but something about wireless charging.

It makes sense for some people depending on what type of mobile computer you are going to buy, let me try to simply explain.

Core M is the very low tdp version of broadwell that you will find in many tablets or other devices that want to go fanless. These will have an integrated pch and will be a different chip than what you see with the 35 ish watt + processors. Those chips will not have an integrated pch and will have a seperate chip on the board for that. In effect if you want a socketed laptop cpu you have to wait much longer but if the chip is bga you can get it prior to christmas.

This next entire paragraph is Speculation. Why is intel delaying the release of mainstream broadwell laptop chips is due to either intel is having yield problems on 14nm or they can't produce enough chips at 14nm and thus they are going to release the chips with the most profit and wait for longer on the more mainstream cpus. Haswell in the 37 w / 47 w range is already a good chip and the improvements you will see on these chips via going to broadwell will be in the graphics and you will not see as much as battery gains on these chips.

According to Intel We will see the more "mainstream" Broadwell chips not this year but sometime next year. We will also expect to see "mainstream" Skylake sometime at the very end of 2015 so in other words christmas.

-----

Now Skylake from a design point of view is more or less finished. They could in theory drop just abandon Broadwell and just print Skylake cpus right now for the most part and have very little delay.

Here is why they won't do it.

Both are 14nm chips but Broadwell is the larger of the two so unless you can get a higher price why would you release the bigger die if you are already capacity constrained due to poor yields and/or just not having enough 14nm wafers coming out of the fabs.

-----

For simplicity sake think of Core M and normal Broadwell core as different as laptops 47 watt quad cores and the i7 unlocked desktop quad cores with 88w tdp. They may have the same architecture but they are different with a different socket and everything. They have different pin counts and everything one has 946 pins the other 1150.

To my understanding we also won't see many broadwell desktop chips with the ones we will see being marketed with better graphics for things such as All in Ones. This is going to happen for they are going to use the 14nm capacity for core m, laptops, and atom and keeping a lot of the desktops on 22nm until they have enough capacity to do everything on 14nm.

 

[witeken]

I actually heard somewhere on the radio this morning advising people that Broadwell makes no sense whatsoever and we should just wait for Skylake because Skylake is right around the corner. Not sure what Skylake has to offer but something about wireless charging.

Skylake-Y will launch in ~Q4'15, so how does Broadwell not make sense?

 

[NTMBK]

Now Skylake from a design point of view is more or less finished. They could in theory drop just abandon Broadwell and just print Skylake cpus right now for the most part and have very little delay.

Here is why they won't do it.

Both are 14nm chips but Broadwell is the larger of the two so unless you can get a higher price why would you release the bigger die if you are already capacity constrained due to poor yields and/or just not having enough 14nm wafers coming out of the fabs.

It's probably more about not screwing their partners. They promised a 14nm pin-compatible shrink from Haswell, and they have been designing platforms around that. Dell, Lenovo etc. will all have Broadwell prototypes finished and ready, with designs that they started working on when the first Broadwell specsheet became available. Cancelling the line 3 months before launch would seriously burn them, and leave them frantically trying to put together a Skylake design that they can get out the door.

 

[NTMBK]

Skylake-Y will launch in ~Q4'15, so how does Broadwell not make sense?

So the Tick-Tock slip is permanent? Damn. I was hoping that Skylake would bring things back on track.

 

[Idontcare]

So the Tick-Tock slip is permanent? Damn. I was hoping that Skylake would bring things back on track.

Economically, business-wise, it makes absolutely no sense for Intel to not let the slip be permanent.

Devil's Canyon and the Haswell refresh exist for a reason.

 

[ShintaiDK]

Yep. Next Step is Skylake-S in Q2/Q3 since thats the next upgrade step in the cycle to keep the money flowing in.

 

[witeken]

So the Tick-Tock slip is permanent? Damn. I was hoping that Skylake would bring things back on track.

Here's some of BK's statements paraphrased. It typically takes about 18 months (1.5 years) to launch a new generation across all product lines. They choose to release Broadwell-Y first because it was the best way to show off their 14nm process.

14nm is about 6 months late due to yield issues, but they haven't made changes to the 10nm schedule (yet?). Because they already have a 14nm product for fanless and low-power products (U/Y series), the desktop will be next on the priority list (Haswell will then be 2+ years old); Skylake-S will launch in H2 (some people suggest early Q3, I think mid-Q4 is a better bet).

In the months after Skylake-S, Intel will release updates of Broadwell-Y/U/K.

Cannonlake should then launch 12 months after Skylake. If we take Ivy Bridge as reference, the average will be 26 months per node or 13 months per Tick/Tock (if Cannonlake manages to launch in September, as promised).

 

[witeken]

Economically, business-wise, it makes absolutely no sense for Intel to not let the slip be permanent.

Devil's Canyon and the Haswell refresh exist for a reason.

Double negation sentences are difficult :s. If I understand you correctly, I disagree. Business-wise, it makes a ton of sense to keep a 2-year schedule to get multiple process nodes ahead of the competition to make an easy inroad in the mobile space.

 

[liahos1]

http://blogs.barrons.com/techtrader...rtcomings-says-semi-advisors/#comment-7233503

this is laughable. i hope someone from intel contacts these guys and sets the record straight.

 

[Ajay]

14nm is about 6 months late due to yield issues, but they haven't made changes to the 10nm schedule (yet?). Because they already have a 14nm product for fanless and low-power products (U/Y series), the desktop will be next on the priority list (Haswell will then be 2+ years old); Skylake-S will launch in H2 (some people suggest early Q3, I think mid-Q4 is a better bet).

Retailers will likely want systems in stores, and online, sometime in October to take full advantage of the Holiday selling cycle. Given how well Skylake is doing according to Intel, I don't see any reason for them to miss such an easy and lucrative target.

If partners are want Skylake in time for the "back to school" market, it will be early Q3.

I can see a scenario of having Skylake-s out sometime in the summer and the 'K' series available for the holiday season to haul in enthusiasts and folks who like to buy the "latest thing".

 

[Ajay]

In the months after Skylake-S, Intel will release updates of Broadwell-Y/U/K.

Has Intel put out a full roadmap for Broadwell yet?

 

[Idontcare]

Double negation sentences are difficult :s. If I understand you correctly, I disagree. Business-wise, it makes a ton of sense to keep a 2-year schedule to get multiple process nodes ahead of the competition to make an easy inroad in the mobile space.

Yeah that was unintentionally a double negative (typo).

It makes no sense for Intel to delay, they risk their revenue stagnating if they leave open any opportunity for the customer to stop upgrading.

Their hardware model is very much akin to the prevailing software model, be it for the OS or for games. You have to have something new in the pipe, and deliver it in timely fashion, or your sales dry up within a year.

 

[jpiniero]

I can see a scenario of having Skylake-s out sometime in the summer and the 'K' series available for the holiday season to haul in enthusiasts and folks who like to buy the "latest thing".

A couple months after Broadwell-K? Not happening. I'm sure Intel would like to get Skylake U and Y out by the 2015 holiday season, but even that's optional.

Their hardware model is very much akin to the prevailing software model, be it for the OS or for games. You have to have something new in the pipe, and deliver it in timely fashion, or your sales dry up within a year.

See, I don't agree with this. The amount of people who buy because a new processor comes out these days - save Apple customers - is tiny. Most of Intel's mainstream line sales come from Corps who have a refresh cycle and just buy whenever the cycle is up, or when they need new machines.

You can see why they are so eager to get Core M out - it's a rare product from Intel that theoretically enables a new segment that didn't exist before.

 

[Ajay]

A couple months after Broadwell-K? Not happening. I'm sure Intel would like to get Skylake U and Y out by the 2015 holiday season, but even that's optional.

Didn't realize that Broadwell=K was going to be that late. Could you please point me to a roadmap - people seem to know the timing of Intel's releases, but I haven't seen a roadmap beyond early Broadwell-Y SKUs. I realize that mobile and server are driving Intel's Core line-up, but it would sure be nice to see an actual Desktop roadmap as well.

 

[jpiniero]

Didn't realize that Broadwell=K was going to be that late. Could you please point me to a roadmap - people seem to know the timing of Intel's releases, but I haven't seen a roadmap beyond early Broadwell-Y SKUs. I realize that mobile and server are driving Intel's Core line-up, but it would sure be nice to see an actual Desktop roadmap as well.

I don't have the link, it was a while back when the second delay happened. I think Broadwell-K plus the 4+3e Xeons are now coming out sometime near June. That and the 4+2 locked Skylake models are the only ones in 2015 that have any kind of rumored release date.

 

[Idontcare]

See, I don't agree with this. The amount of people who buy because a new processor comes out these days - save Apple customers - is tiny. Most of Intel's mainstream line sales come from Corps who have a refresh cycle and just buy whenever the cycle is up, or when they need new machines.

You are making the argument that consumer's don't upgrade or replace their computers anymore? Only corporations do?

Why does Devil's Canyon exist? For corporations? Why did Haswell-E just launch? For corporations? Corporations by K-SKUs and X-SKUs?

 

[witeken]

Yeah that was unintentionally a double negative (typo).

It makes no sense for Intel to delay, they risk their revenue stagnating if they leave open any opportunity for the customer to stop upgrading.

Their hardware model is very much akin to the prevailing software model, be it for the OS or for games. You have to have something new in the pipe, and deliver it in timely fashion, or your sales dry up within a year.

So we agree.

Didn't realize that Broadwell=K was going to be that late. Could you please point me to a roadmap - people seem to know the timing of Intel's releases, but I haven't seen a roadmap beyond early Broadwell-Y SKUs. I realize that mobile and server are driving Intel's Core line-up, but it would sure be nice to see an actual Desktop roadmap as well.

Broadwell-K should be released somewhere in Q2, probably close to 1 year after DC.