Intel Broadwell Thread

[AtenRa]

Broadwell vs Haswell at the same performance,

http://www.youtube.com/watch?v=KeDtXucTwRI#t=35

Broadwell has up to 30% lower power.

 

[IntelUser2000]

Unfortunately the hinge has a little too much flex in it so using the touch screen is mushy and it's not well balanced, too much weight in the touch screen., if you touch the screen the whole laptop tips easily.

I don't know how many people do this, but I doubt most are using the touchscreen akin to how the alien and the human interaction happened in the E.T. movie.

My XPS 12 is pretty top heavy too. But as long as the screen is open less than 100 degrees its not a huge issue. Plus I find that I mostly scroll with part of my hand holding the screen. Yes, during pressing buttons I need to do what you described but my other hand would also be on the keyboard.

I wish that were the *only* big issue that Laptops have.

About the only one existing haswell ultrabook that really gets the touchscreen thin enough with a good hinge is the dell xps 11 but the touch keyboard on that design is wretched or the fiercely expensive HP 1040.

XPS 11 I considered for a while but it sucks in MANY accounts. Flat keyboards, serious throttling despite having the Y chip and NOT being fanless. Pricing is outrageous. Pen support is practically nonexistent.

Broadwell has up to 30% lower power.

By the way, you can get that 30% reduction by straight shrink to 14nm alone.

 

[Dufus]

Mostly correct, but you're only accounting for the switching power.

Right, perhaps due to savings in the package itself the available power for cores would have been a fair bit higher than for the crippled Haswell at 4.5W TDP. To many variables missing IMHO such as PL1,PL2,PL3, primary and secondary plane powers, temperatures...

 

[bullzz]

@AtenRa - you missed the part where he said "without any performance tuning". so I think we can expect more than 30%

 

[Flogger]

Gentlemen, I'm just having a discussion on notebookreview.com (with the new 980M GTX release) where everyone seems to believe that Broadwell will be hotter than Haswell so better to snap up Haswells while stocks last. Is this complete misunderstanding?

My Haswell (4700MQ) is running at aver 100'C on relatively light stuff like X-Plane 10 and Rome II, so I dread to think what an even hotter enthusiast CPU would do in terms of throttling and system life.

 

[kimmel]

Gentlemen, I'm just having a discussion on notebookreview.com (with the new 980M GTX release) where everyone seems to believe that Broadwell will be hotter than Haswell so better to snap up Haswells while stocks last. Is this complete misunderstanding?

My Haswell (4700MQ) is running at aver 100'C on relatively light stuff like X-Plane 10 and Rome II, so I dread to think what an even hotter enthusiast CPU would do in terms of throttling and system life.

People make things up on the Internets. Nothing new here. In fact I think I hear "get the old product while they last" every generation.

In fact. Look at the TJ max on the compare. Now, if you are speculating on random unreleased products I have no idea and nor do they.
http://ark.intel.com/compare/75459,83612

 

[krumme]

Gentlemen, I'm just having a discussion on notebookreview.com (with the new 980M GTX release) where everyone seems to believe that Broadwell will be hotter than Haswell so better to snap up Haswells while stocks last. Is this complete misunderstanding?

My Haswell (4700MQ) is running at aver 100'C on relatively light stuff like X-Plane 10 and Rome II, so I dread to think what an even hotter enthusiast CPU would do in terms of throttling and system life.

Whats the problem of beeing hotter?
If the cpu/gpu is designed to run hotter, as they always is, its no problem. New tech.
We will see more dynamic adaption of voltage/freq/temp whatever. Great

 

[Flogger]

The problem with hotter is throttling (and possibly lifespan). I get throttling on current gen games, so if we get more poorly optimized console ports going forward, the latest 980M will be fine (a welcome change from the 880M) but CPU throttling will become more common. If Broadwell is weeks rather than months away then why not get something a bit cooler rather than rushing to get Haswell, as so many are.

 

[Enigmoid]

My Haswell (4700MQ) is running at aver 100'C on relatively light stuff like X-Plane 10 and Rome II, so I dread to think what an even hotter enthusiast CPU would do in terms of throttling and system life.

I don't think that has anything to do with the heat given off by the CPU, merely a crappy cooling system (laptop CPUs do not have the solder or whatever that the desktop chips have).

A good mobile cooling system will keep the CPU below 80 no matter what you do.

 

[Flogger]

This was across the board - all the way from Alienware to MSI to Asus ROG.

I run a Clevo whose chassis and cooling is unchanged since 2010. In 2011 running a HM67 chipset in it with i5-2430M and 580M GTX I was getting Witcher 2 High without ever going over 70 on either. Same chassis and cooling (+/- refresh tweaks) but i7-4700MQ and 880M GTX, same game and with Haswell and Kepler it's nearer 90. I don't know if the thermal paste is the same or whether I kept exactly the same video settings across both systems. Perhaps some of the refresh tweaks cut costs in the cooling dept...

I know that the current Clevo with 980M can max out Far Cry at over 60fps without getting temps above 70, but Haswell CPU still stubbornly over 90.

 

[DrMrLordX]

I see a problem if we have, for instance, a G3258 user that bought an Z97 board, thinking Broadwell-K might be a nice upgrade option, only to find that it will not work. Can we confirm what the original English of the VR-Zone article says, or not?

That is an oft-cited upgrade path. So how many Z97 boards support VCCST? All of them? Some of them? I'm guessing that any Z97 board that has Haswell-refresh CPUs on its compatibility list should have what it takes to run Broadwell-K, but that's just a guess?

 

[AtenRa]

@AtenRa - you missed the part where he said "without any performance tuning". so I think we can expect more than 30%

And the fact is that Haswell in that test has been brought down to Broadwell performance at 4-5W. Now, Broadwell due to 14nm and especially due to Low Power 14nm process they currently use on the Core-M SKUs, it has way better performance at lower power than what Haswell can achieve. I dont expect more than 30% lower power for Desktop and higher TDP Mobile parts. So that 30% lower Power should be the optimal they can achieve.

 

[Abwx]

Abwx,

its comparing to traditional scaling. So compared to regular 14nm design its 38% lower because Core M gets specific optimizations. So you still need to take into account what 14nm itself without the optimizations brings.

If you consider that average per process gen reduction in power is usually 30% then the total you get 0.4375x. If you reverse calculate that from 4.5W you get 10.3, not too far from 11.5W on the Y parts.

.

Well i thought 10.9W, so your estimation is not that far from mine, that said i m skeptic that it will have significantly higher perfs/watt than either BT or Mullins/Beema, case is that it should sustain the same perf/Watt up to higher absolute perf level but i dont see it as a revolutionary product, they got back to IB perf/watt ratio but with the benefit of a smaller node, Haswell was less power hungry than IB but that was due to the rest of the plateform more than anything else.

Given Core M transistor count it s an honorable result though.

 

[Roland00Address]

And the fact is that Haswell in that test has been brought down to Broadwell performance at 4-5W. Now, Broadwell due to 14nm and especially due to Low Power 14nm process they currently use on the Core-M SKUs, it has way better performance at lower power than what Haswell can achieve. I dont expect more than 30% lower power for Desktop and higher TDP Mobile parts. So that 30% lower Power should be the optimal they can achieve.

To my understanding haswell ULT (u series) and ULX (y series) were the same die from intel and same pch and even the same bga socket.

To my understanding intel is creating a whole different die for broadwell core m compared to the rest of broadwell (this is what Ryan Smith said on his core m preview). And they are going to use these "feature improvements" to get better efficiency thus Intel is trading die space for better power use his means you get .63x scaling instead of .51x according to intel's slide (24% increase). Intel also to my understanding you will not use all these "feature improvements" on their higher end skus such as the desktop skus since they are wasted die space and their benefit in desktop is of little use.

Thus you may only get a typical 30% lower power shrink with desktop and mobile for intel is not doing anything special for those parts. But with broadwell core m they are going to nurture their favorite child to make sure it gets as much performance per watt as possible to allow them to charge a higher price per chip.

 

[crashtech]

That is an oft-cited upgrade path. So how many Z97 boards support VCCST? All of them? Some of them? I'm guessing that any Z97 board that has Haswell-refresh CPUs on its compatibility list should have what it takes to run Broadwell-K, but that's just a guess?

Didn't Z97 come out right at the same time as Haswell Refresh? I didn't think there was a 9-series that didn't support the Refresh. The graphic previously posted by mikk seems to confirm it:

http://www.bit-tech.net/hardware/motherboards/2014/05/11/the-intel-9-series-chipsets-examined/1

 

[witeken]

To my understanding haswell ULT (u series) and ULX (y series) were the same die from intel and same pch and even the same bga socket.

To my understanding intel is creating a whole different die for broadwell core m compared to the rest of broadwell (this is what Ryan Smith said on his core m preview). And they are going to use these "feature improvements" to get better efficiency thus Intel is trading die space for better power use his means you get .63x scaling instead of .51x according to intel's slide (24% increase). Intel also to my understanding you will not use all these "feature improvements" on their higher end skus such as the desktop skus since they are wasted die space and their benefit in desktop is of little use.

Thus you may only get a typical 30% lower power shrink with desktop and mobile for intel is not doing anything special for those parts. But with broadwell core m they are going to nurture their favorite child to make sure it gets as much performance per watt as possible to allow them to charge a higher price per chip.

To be clear, It's a 0.51x scaling according to him, but because they added more transistors the die is 0.63x the size of HSW-Y.

 

[Kallogan]

Gentlemen, I'm just having a discussion on notebookreview.com (with the new 980M GTX release) where everyone seems to believe that Broadwell will be hotter than Haswell so better to snap up Haswells while stocks last. Is this complete misunderstanding?

My Haswell (4700MQ) is running at aver 100'C on relatively light stuff like X-Plane 10 and Rome II, so I dread to think what an even hotter enthusiast CPU would do in terms of throttling and system life.

Best thing about haswell is that it did (re)introduce easy undervolting via Intel XTU. That why i bought into Haswell cause other than that, no reasons really to buy over Ivy. Undervolting is making such a huge difference in laptops...You can run at full turbo forever or at least at higher clocks. So many tend to throttle at stock voltage. If undervolting is still on with Broadwell, np.

 

[Roland00Address]

To be clear, It's a 0.51x scaling according to him, but because they added more transistors the die is 0.63x the size of HSW-Y.

Yes exactly but with one addeum

Broadwell which is Desktop, Laptop, Ultrabook != "Improved" Broadwell which is Core M.

Broadwell may be overall faster due to tdp and more cores.
Broadwell Core M is more efficient even if overall it is slower.

We just need to be on the same page for some people talk about Broadwell, some about Core M, and some switch between the two for the architecture design is practically the same Intel just did some tricks to make it more efficient, cut down leaker, and to improve power consumption.

It is like having 3 moats around a city vs 1. Those extra 2 moats do nothing about the normal everyday use of the city/town/castle but 3 moats slow down a certain type of problem (in this analogy invaders, in the broadwell situation power consumption, voltage, and leakage). Thus if you have a certain type of problem you may want to "waste resources" to prevent that type of problem if its likely.

Lastly those extra features are not better graphics but things that improve efficiency according to that intel slide and thus different dies of "normal" broadwell will be smaller.




And if I am buying a device that is $500 awesome tablet/productivity device I want intel to spend a little more on the SOC if it will make it better. On a $100 to $200 tablet it is less important and atom and die size/price is more important.

 

[Enigmoid]

This was across the board - all the way from Alienware to MSI to Asus ROG.

I run a Clevo whose chassis and cooling is unchanged since 2010. In 2011 running a HM67 chipset in it with i5-2430M and 580M GTX I was getting Witcher 2 High without ever going over 70 on either. Same chassis and cooling (+/- refresh tweaks) but i7-4700MQ and 880M GTX, same game and with Haswell and Kepler it's nearer 90. I don't know if the thermal paste is the same or whether I kept exactly the same video settings across both systems. Perhaps some of the refresh tweaks cut costs in the cooling dept...

I know that the current Clevo with 980M can max out Far Cry at over 60fps without getting temps above 70, but Haswell CPU still stubbornly over 90.

What is the TDP of those chips? You went from a 35W dual to a 47W quad.

 

[witeken]

Since there is no earnings thread, Intel (CEO BK) reported in the earnings call that 14nm yields are still behind expectations. At this point, I wouldn't be optimistic about 10nm.

 

[DrMrLordX]

Didn't Z97 come out right at the same time as Haswell Refresh? I didn't think there was a 9-series that didn't support the Refresh.

It would appear that any board that supports Haswell refresh out-of-the box (Z97/H97) should also support all the features necessary to handle Broadwell-K. So perhaps the notion that Z97 will have compatibility problems is due to a misunderstanding.

 

[crashtech]

It would appear that any board that supports Haswell refresh out-of-the box (Z97/H97) should also support all the features necessary to handle Broadwell-K. So perhaps the notion that Z97 will have compatibility problems is due to a misunderstanding.

Yeah, that's what I think. So far it seems safe to recommend 9-series paths that could include Broadwell-K at some point. Next question is how the performance of such is expected to compare to Devil's Canyon.

 

[Flogger]

Best thing about haswell is that it did (re)introduce easy undervolting via Intel XTU. That why i bought into Haswell cause other than that, no reasons really to buy over Ivy. Undervolting is making such a huge difference in laptops...You can run at full turbo forever or at least at higher clocks. So many tend to throttle at stock voltage. If undervolting is still on with Broadwell, np.

Thank you! If I went that route, then getting a 4910MQ over 4710MQ would make no sense temp or performance -wise, if I understand correctly (last newbie question, promise!! then I'm off to click the 'order' button)

 

[witeken]

Skylake seems to be on track for a launch in 2015.

http://seekingalpha.com/article/2563875-intel-skylake-development-appears-to-be-on-schedule

 

[Dufus]

Thank you! If I went that route, then getting a 4910MQ over 4710MQ would make no sense temp or performance -wise, if I understand correctly (last newbie question, promise!! then I'm off to click the 'order' button)

Well one would expect the 4910 to be a better binned chip and require a little less voltage than the 4710 clock for clock plus the 4910 has an extra 6 overclock bins while the 4710 has 2. There's also supposed to be support for coarse bclk straps too. However the overruling factor would be the hardware power delivery and cooling. Not much point having the extra speed if the rest of the hardware cannot cope with it.

I think it would be interesting to see how "efficiently" Broadwell overclocks, will the power savings be enough to compensate for the higher power density of the die and increased thermals? Is the move towards BGA an attempt to supplement cooling?