Intel Broadwell Thread

Page 69 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

dahorns

Senior member
Sep 13, 2013
550
83
91
So I am really confused by this. Idle power time increases by a very respectable 37.8% and load increases by 18.1%, which is good, but wifi browsing decreases? Maybe there is some issue with this model laptop

Otherwise, the increases seem pretty good with battery life. Performance may not have increased much at all, but if we remove the outlying wifi results the increase in battery life is pretty good (18-37.8%).

Edit:
The increase is only 37.8%, not 40%. 721min/523min = 1.378585. The other figure comes from 78min/66min = 1.181818


Still very confused.

Read the numbers again. WIFI surfing time increases by 16 minutes. The only scenario runtime that decreases is the load test, which makes sense given that the Broadwell system uses more power at full load. However, you need to consider that the Broadwell system also performs better at full load than the Haswell system. In particular, at full load the Broadwell system maintains max turbo of 2.7 ghz for the CPU and 900mhz for the GPU. The Haswell system throttles the CPU down to 1.4 ghz, with the GPU at 1 ghz.

To me, it seems that the improvements in Broadwell are somewhat hidden by the clever design of Haswell (benchmarks are short enough to run at full turbo in Haswell, even if it cannot sustain that performance). Whether due to an inability to raise clocks in Broadwell or due to a desire to maintain product segmentation, Intel did not raise the turbo in Broadwell. This means that in most benchmarks you'll only see the nominal 5% IPC gain.

Perhaps, raising the turbo requires architectural changes designed for the 14nm process. Thus, you may see bigger jumps in benchmarks numbers with Skylake.
 

Qwertilot

Golden Member
Nov 28, 2013
1,604
257
126
I guess they just can't have it both ways with the benchmarks CoreM is their big new thing this time round and that really is dependent on short term turbo to to look good in benchmarks.
 

ninaholic37

Golden Member
Apr 13, 2012
1,883
31
91
With the same TDP it has lower perfs, the two exemples above HW is consuming up to 24W while BDW is close to 29W, that is at SoC level.

And both are supposed to be 15W chips...
I think it's designed to blow their TDP using Turbo until the CPU hits 100 degrees, then throttle down back to non-turbo frequency (15W)? If true, your numbers seem to make sense.
 

coercitiv

Diamond Member
Jan 24, 2014
6,403
12,864
136
If you want to go in the details to grab a few watts i can point that there s only 8MB on the laptop as well as AC/DC and DC/DC conversion being better than 90% efficency at power like 40W, and so on and we ll get back to twice the chip official TDP, the 29W figure is close to reality, perhaps 27-28W but not less.
Well, if you go into the details, grab a few watts more and bring that down to about 25W, I might be able to give you an interesting explanation for that figure: both CPUs in these Latitudes appear to be operating in a cTDP up state. Haswell U chips that support cTDP up can operate at 25W TDP. All they need is enough cooling and some software support from both OEMs and Intel. (doc)

Configurable TDP can be enabled using Intel's DPTF driver or through HW/EC firmware. Enabling cTDP using the DPTF driver is recommended as Intel does not provide specific application or EC source code.
And the link for the DPTF driver for Dell Latitude E5550 is right here (under Chipsets). This screenshot from the HW review also shows some power values that seem to agree with the 25W value.

Take this review for the Inspiron 13 as a counter example of what happens when a Haswell U abides by its nominal 15W TDP: maximum power usage under load is 26W, quite in line with the 24W from the Beema equipped Pavilion 13.
 

Fjodor2001

Diamond Member
Feb 6, 2010
3,938
408
126
Don't worry, at some point these folks will slowly, but surely, start to realize that there is a darned good reason why Intel has delayed and delayed their full rollout of 14nm broadwell SKUs...and in the meantime have gone silent on 10nm.

Meanwhile the data we do have are rather telling, as you rightly point out.

Cryptic... :sneaky:

Are you saying 14 nm is flawed in the sense that it will not perform as originally expected? And if so, can that really be fixed by an uArch change in Skylake?

And do you mean we'll have to wait for 10 nm before we see any improvement, or are you perhaps suspecting 10 nm may be even more problematic?

I'd really appreciate it if you could unlock some of the mysteries here...
 

Abwx

Lifer
Apr 2, 2011
11,172
3,869
136
I think it's designed to blow their TDP using Turbo until the CPU hits 100 degrees, then throttle down back to non-turbo frequency (15W)? If true, your numbers seem to make sense.

It consume 43W, the laptop, even if run one hour like Notebookcheck did.

Take this review for the Inspiron 13 as a counter example of what happens when a Haswell U abides by its nominal 15W TDP: maximum power usage under load is 26W, quite in line with the 24W from the Beema equipped Pavilion 13.

Surely but then scores are no more the same than when pumping 25W.
 

Abwx

Lifer
Apr 2, 2011
11,172
3,869
136
Cryptic... :sneaky:

Are you saying 14 nm is flawed in the sense that it will not perform as originally expected? And if so, can that really be fixed by an uArch change in Skylake?

And do you mean we'll have to wait for 10 nm before we see any improvement, or are you perhaps suspecting 10 nm may be even more problematic?

I'd really appreciate it if you could unlock some of the mysteries here...

The CPU Z screens show that supply voltage at 2.7 is in the 1.050-1.100V range, to give a comparison a i7 is binned for 3.5 at 1.0V while the 4670K is at 1.050V at same frequency, that is 20% frequency deficit at thev same voltage, this say that perf/watt will be inferior to previous gen.
 

Fjodor2001

Diamond Member
Feb 6, 2010
3,938
408
126
The CPU Z screens show that supply voltage at 2.7 is in the 1.050-1.100V range, to give a comparison a i7 is binned for 3.5 at 1.0V while the 4670K is at 1.050V at same frequency, that is 20% frequency deficit at thev same voltage, this say that perf/watt will be inferior to previous gen.

So essentially we'll have to wait until 10 nm (i.e. 2017) until we see better perf/watt from Intel compared to Haswell? Or do you think 10 nm will be even worse in that regard?
 

Sweepr

Diamond Member
May 12, 2006
5,148
1,142
131
Gotta love the troll posts hidden as genuine questions. Readers be careful, the same person also states Haswell was a regression in perf/watt (probably claimed the same about IB vs SB too and so on...).

According to AnandTech Broadwell-U improves power-efficiency (Gigabyte BRIX):

The slightly higher base clocks in the Core i7-5500U (compared to the Core i7-4500U) are probably the reason for the Haswell-based unit appearing more power efficient than the Broadwell counterpart - but, make no mistake here - the Broadwell unit wins the performance per watt test quite easily.
 
Last edited:

Fjodor2001

Diamond Member
Feb 6, 2010
3,938
408
126
Gotta love the troll posts hidden as genuine questions. Readers be careful, the same person also states Haswell was a regression in perf/watt (probably claimed the same about IB vs SB too and so on...).

According to AnandTech Broadwell-U improves power-efficiency (Gigabyte BRIX):

Well, Abwx said that "perf/watt will be inferior to previous gen", implying that Broadwell is a regression in perf/watt compared to Haswell. How is it trolling to ask for a confirmation to that statement?
 

SAAA

Senior member
May 14, 2014
541
126
116

So they are scrapping the higher powered Broadwell entirely? That sounds both bad news (for 14nm transition plus all those delays) but possibly good if they are just skipping to Skylake (finally!).

I'm curious if this could mean something for the -e line too, or are they going to release a Broadwell server variant + derived enthusiast products? Rumors rumors, make one think too far...
 

Fjodor2001

Diamond Member
Feb 6, 2010
3,938
408
126
So they are scrapping the higher powered Broadwell entirely? That sounds both bad news (for 14nm transition plus all those delays) but possibly good if they are just skipping to Skylake (finally!).

But if Intel's 14 nm process is the problem, can Skylake solve that? It's on 14 nm too, just a new uArch.
 

SAAA

Senior member
May 14, 2014
541
126
116
But if Intel's 14 nm process is the problem, can Skylake solve that? It's on 14 nm too, just a new uArch.

I'm not sure 14nm is the problem, not looking at core-M at least: you can spin it how much you want but they definitely increased perf/W with that and the process has been checked by external labs in both sizes and electric charateristics.

I'm pretty much certain that if you had a Broadwell sample right now you could test it at the same clocks of an Haswell equivalent and find it consumes less power at any value; aka better frequencies at same voltage... the problem is another entirely: heat!

If you halve die size (what they did with logic, don't mind iGPU) and reduce power consumption of anything less than half you end up with more heat: more W/mm2, that given the already sky-high temperatures of Haswell means only problems for Broadwell at higher frequiencies.
Also higher temperature could lead to even higher power consumption hence incrasing the problem even further.

It's definitely possible that a new architecture fixes that: both by reducing clockspeed needed for the same performance (higher IPC...) but also by changing the layout itself.
Right now the cores are very close (and smaller at each node), make them spread apart, pull out the FIVR and other tweaks that all lead to lower power densities with same performance for example. And this is coming from someone with close to zero knowledge in cpu design, imagine what they could do!
 

Fjodor2001

Diamond Member
Feb 6, 2010
3,938
408
126
I'm pretty much certain that if you had a Broadwell sample right now you could test it at the same clocks of an Haswell equivalent and find it consumes less power at any value; aka better frequencies at same voltage... the problem is another entirely: heat!

If you halve die size (what they did with logic, don't mind iGPU) and reduce power consumption of anything less than half you end up with more heat: more W/mm2, that given the already sky-high temperatures of Haswell means only problems for Broadwell at higher frequiencies.
Also higher temperature could lead to even higher power consumption hence incrasing the problem even further.
In that case it spells trouble for everyone in the business (Intel/GF/TSCM/Samsung). Will future node shrinks make things even worse, I don't see how we can get higher performing chips at later nodes.
It's definitely possible that a new architecture fixes that: both by reducing clockspeed needed for the same performance (higher IPC...) but also by changing the layout itself.
Yes, in theory. But lately new Intel uArch has not brought much higher IPC. Certainly not enough to compensate for that.
Right now the cores are very close (and smaller at each node), make them spread apart, pull out the FIVR and other tweaks that all lead to lower power densities with same performance for example. And this is coming from someone with close to zero knowledge in cpu design, imagine what they could do!
In that case, how come they didn't do that already on Haswell? They ought to know that would improve things going from IB->Haswell on 22 nm too.
 

jpiniero

Lifer
Oct 1, 2010
14,845
5,457
136
It's a tradeoff for sure. If anything I'm sure Intel is going to increase the density as much as they can, since it helps what matters (tablets, laptops, servers not to mention costs) even if it hurts the max clock speed. Removing the FIVR will help with the heat but I can't imagine it would make a huge difference.
 

coercitiv

Diamond Member
Jan 24, 2014
6,403
12,864
136
Right now the cores are very close (and smaller at each node), make them spread apart, pull out the FIVR and other tweaks that all lead to lower power densities with same performance for example. And this is coming from someone with close to zero knowledge in cpu design, imagine what they could do!
If they are experts and you are just a beginner, imagine what a total stranger in this field could come up with!

Grandpa!
 

Abwx

Lifer
Apr 2, 2011
11,172
3,869
136
So essentially we'll have to wait until 10 nm (i.e. 2017) until we see better perf/watt from Intel compared to Haswell? Or do you think 10 nm will be even worse in that regard?

There will be no 10nm in 2017, at least not on mass scale production, it is already 12-15 months that 14nm is ramped and output is marginal, next DT chips are launched mid year and 2 years is not too much to amortize this process, likely that 3 years starting from June will be necessary given the slow ramp.


Gotta love the troll posts hidden as genuine questions. Readers be careful, the same person also states Haswell was a regression in perf/watt (probably claimed the same about IB vs SB too and so on...).

What i bolded is effectively trolling, and even lying since i never made such a statement, but for sure HW has lower perf/watt than IB.


I'm not sure 14nm is the problem, not looking at core-M at least: you can spin it how much you want but they definitely increased perf/W with that and the process has been checked by external labs in both sizes and electric charateristics.

I'm pretty much certain that if you had a Broadwell sample right now you could test it at the same clocks of an Haswell equivalent and find it consumes less power at any value; aka better frequencies at same voltage... the problem is another entirely: heat!

Where are the published numbers from thoses labs that would say that the thing has better perf/Watt.?.

So far it consume 20% more than HW in Notebookcheck review.

As i pointed it supply voltage is higher for 14nm than for 22nm, how could this result in lower power comsumption.?
 
Last edited:

NTMBK

Lifer
Nov 14, 2011
10,269
5,134
136
Haswell 30W system load:

http://www.notebookcheck.net/Review-HP-EliteBook-840-G1-H5G28ET-Ultrabook.114717.0.html

/snip

Broadwell 30W system load:

http://www.notebookcheck.net/Dell-Latitude-14-E7450-Ultrabook-Review.135484.0.html

/snip

No problems here. Looks like you're barking up another cTDP up tree like that time with the GTX 970/GTX 980s.

The max temps have crept up slightly, from 79C to 82C- though without knowing if the two systems have comparable cooling systems it's tricky to read much into it.
 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |