Intel Broadwell Thread

[Sweepr]

I agree with III-V, this is overall performance, which likely means the higher base clock increase has a non-negligible impact.

I doubt the Core i5 4300U/4310U notebooks from Dell are running the single-threads tests at anything less than Turbo speeds most of the time (just like the new Core i5 5300U version).

Taking a look at the single-thread scores they delivered on their promise of >5% better IPC (Latitude E5550 Haswell vs Broadwell). That's probably the best comparison till someone locks a desktop Broadwell-K at equal clocks as Haswell later this year.

Cinebench 11.5 Single-Core
Core i5 5300U (2.9GHz Turbo): 1.37
Core i5 4310U (3GHz Turbo): 1.3 (+5,3%)
Core i5 4300U (2.9GHz Turbo): 1.2 (+14%)

Cinebench 10 Single-Core
Core i5 5300U (2.9GHz Turbo): 4498
Core i5 4300U (2.9GHz Turbo): 4016 (+5%)

Mozilla Kraken (Lower is better)
Core i5 5300U (2.9GHz Turbo): 2389
Core i5 4310U (3GHz Turbo): 2586 (+8%)

Sunspider (Lower is better)
Core i5 5300U (2.9GHz Turbo): 101.2
Core i5 4310U (3GHz Turbo): 109 (+8%)

 

[dahorns]

I know that they did it only for Core M, but what about when the power delta of a laptop is say 23-24W while the CPU is rated at 11.5W TDP.?.

It doesnt take a lot of brain cells to deduct that the 11.5W official spec is a scam, indeed the SoC power usage can be extracted from the delta with enough precision, in this case we dont even need to do calculations since it s obvious that the official spec is largely exceeded.

Aren't they just measuring power from the wall? Particularly when you're talking about a unit with a battery, doesn't that tell us more about the power brick and the battery than device itself?

 

[videogames101]

I know that they did it only for Core M, but what about when the power delta of a laptop is say 23-24W while the CPU is rated at 11.5W TDP.?.

It doesnt take a lot of brain cells to deduct that the 11.5W official spec is a scam, indeed the SoC power usage can be extracted from the delta with enough precision, in this case we dont even need to do calculations since it s obvious that the official spec is largely exceeded.

A scam for who? You think the guys designing products were scammed into using Core M in products not designed for the real power draw? Obviously not, thye had plenty of samples and would have known the power and thermal limits of the CPUs they were making products with. Apparently you think YOU are being scammed, but the reality is Intel isn't marketing TDP numbers to you, it's marketing battery life and real-world performance feel.

 

[bullzz]

@Abwx - do you think a product that you will find in many fanless products a scam or a huge Discovery tablet with a chip that never saw light of the day

the one thing OEMs know is if a chips power profile would fit into their design
we can do without the hyperboles

 

[AtenRa]

The really sad part is the Core i5 4310U in the Dell E5450 review was single channel as well.

Yea Broawell GT2 doesnt bring the performance people were expecting.

 

[witeken]

Yea Broawell GT2 doesnt bring the performance people were expecting.

It seems an Intel architect's definition of

On-die graphics is improving quite a bit as well. If you're into energy efficiency or even more graphics, Broadwell. I think the tech community will be very pleasantly surprised with Broadwell. But I'm biased, so we'll just going to have to prove it the hard way.

or

We are allocating more die area to graphics. So expect to see a definite improvement from Ivy Bridge to Haswell. Then expect a bigger leap from Haswell to Broadwell. My guess is we will leapfrog ATI on-die graphics with Broadwell.

is different than most other people's. I think Intel still could be doing better in graphics, but of course they don't really have an incentive to go after the mid and high-end of the market.

 

[frozentundra123456]

I doubt the Core i5 4300U/4310U notebooks from Dell are running the single-threads tests at anything less than Turbo speeds most of the time (just like the new Core i5 5300U version).

Taking a look at the single-thread scores they delivered on their promise of >5% better IPC (Latitude E5550 Haswell vs Broadwell). That's probably the best comparison till someone locks a desktop Broadwell-K at equal clocks as Haswell later this year.

Cinebench 11.5 Single-Core
Core i5 5300U (2.9GHz Turbo): 1.37
Core i5 4310U (3GHz Turbo): 1.3 (+5,3%)
Core i5 4300U (2.9GHz Turbo): 1.2 (+14%)

Cinebench 10 Single-Core
Core i5 5300U (2.9GHz Turbo): 4498
Core i5 4300U (2.9GHz Turbo): 4016 (+5%)

Mozilla Kraken (Lower is better)
Core i5 5300U (2.9GHz Turbo): 2389
Core i5 4310U (3GHz Turbo): 2586 (+8%)

Sunspider (Lower is better)
Core i5 5300U (2.9GHz Turbo): 101.2
Core i5 4310U (3GHz Turbo): 109 (+8%)

That is true, but didnt they say 30% better performance per watt as well? A five to ten percent performance improvement is decent, but the TDP is the same. If they had achieved this with a lower TDP, that would have been a bigger accomplishment. I suppose we have to see real life power consumption and battery life tests though to be sure. Real world battery life is the final metric. Perhaps that is improved even at the same rated TDP.

 

[witeken]

That is true, but didnt they say 30% better performance per watt as well? A five to ten percent performance improvement is decent, but the TDP is the same. If they had achieved this with a lower TDP, that would have been a bigger accomplishment. I suppose we have to see real life power consumption and battery life tests though to be sure. Real world battery life is the final metric. Perhaps that is improved even at the same rated TDP.

TDP is pretty arbitrary. Intel can set TDP whatever they want.

Also remember how AMD used that silly 2X performance/TDP rating for their Puma+ cores?

If you want to know energy per instruction, you have to measure those quantities, not something else, which Intel has done with pre-qualified silicon:

http://www.anandtech.com/show/7318/intel-demos-14nm-broadwell-up-to-30-lower-power-than-haswell

 

[AtenRa]

That is true, but didnt they say 30% better performance per watt as well? A five to ten percent performance improvement is decent, but the TDP is the same. If they had achieved this with a lower TDP, that would have been a bigger accomplishment. I suppose we have to see real life power consumption and battery life tests though to be sure. Real world battery life is the final metric. Perhaps that is improved even at the same rated TDP.

I dont have good news for you, you only get higher idle :thumbsdown:

Battery runtime

Dell Latitude E5550 Core i5 5300U Broadwell
Idle (without WLAN, min brightness) = 12h 01min
WiFi Surfing = 6h 21min
Big Buck Bunny H.264 1080p = 6h 27min
Load (maximum brightness) = 1h 06min


Dell Latitude E5550 Core i5 4310U Haswell
Idle (without WLAN, min brightness) = 8h 43min
WiFi Surfing = 6h 05min
Big Buck Bunny H.264 1080p = 6h 22min
Load (maximum brightness) = 1h 18min

 

[Sweepr]

That is true, but didnt they say 30% better performance per watt as well? A five to ten percent performance improvement is decent, but the TDP is the same. If they had achieved this with a lower TDP, that would have been a bigger accomplishment. I suppose we have to see real life power consumption and battery life tests though to be sure. Real world battery life is the final metric. Perhaps that is improved even at the same rated TDP.

IMHO Turbo clocks for both CPU cores and graphics were a bit disappointing, I wonder if that has something to do with their 14nm process. Example: HD4400 Turbo 1100MHz vs HD5500's 900MHz; Core i5 4310U Turbo 3GHz vs Core i5 5300U 2.9GHz. The IPC improvement is there but 5-10% higher Turbo clocks would help too. It's still a solid CPU/GPU bump from Haswell-U in my book and I look forward to Skylake-U later this year.

 

[frozentundra123456]

I dont have good news for you, you only get higher idle :thumbsdown:

Battery runtime

Dell Latitude E5550 Core i5 5300U Broadwell
Idle (without WLAN, min brightness) = 12h 01min
WiFi Surfing = 6h 21min
Big Buck Bunny H.264 1080p = 6h 27min
Load (maximum brightness) = 1h 06min


Dell Latitude E5550 Core i5 4310U Haswell
Idle (without WLAN, min brightness) = 8h 43min
WiFi Surfing = 6h 05min
Big Buck Bunny H.264 1080p = 6h 22min
Load (maximum brightness) = 1h 18min

Actually 40% (edit) increase in idle time seems fantastic to me. I am thinking of a student who goes from class to class all day, and uses the comp only briefly between classes, but wants the charge to last all day. Similar case for me. I carry a tablet to work, it sits idle while I am working, but I do use it to check email or surf during lunch and on breaks, and for a couple of hours after I get home. So I would be able to charge in the morning and have the laptop last all day at idle with occasional light use. I would think a lot of other use cases fit this type of pattern as well-- mainly idle, with occasional periods of light use.

Now obviously longer battery life under heavier use would be desired, especially WiFi surfing, but increased idle does not seem insignificant.

 

[Enigmoid]

IMHO Turbo clocks for both CPU cores and graphics were a bit disappointing, I wonder if that has something to do with their 14nm process. Example: HD4400 Turbo 1100MHz vs HD5500's 900MHz; Core i5 4310U Turbo 3GHz vs Core i5 5300U 2.9GHz. The IPC improvement is there but 5-10% higher Turbo clocks would help too. It's still a solid CPU/GPU bump from Haswell-U in my book and I look forward to Skylake-U later this year.

Yeah, if you look at the stress test BW 15W has no trouble running at high clocks.

Prime 95 + Furmark

i5-4310U
1.4 ghz CPU + 1000 mhz igp

i5-5300U
2.7 ghz CPU + 900 mhz igp

Very impressive load numbers.

 

[Qwertilot]

I wouldn't be at all sure if that gain in idle time is down to Broadwell vs Haswell - Didn't Haswell already have a very low power idle states? Just not always used right.

Not much to gain from a lower CPU power draw if so, as everything else in the machine will dominate power use. Quite possible just Dell simply sorting out their implementation of idle power draw.

Have to say, looking at that furmark measure, I do think Intel might well have done quite nicely in terms of improvement but been hoist by their own petard in terms of benchmarks
(vs their earlier things which could boost really fast for just a short benchmark but then couldn't maintain it.).

 

[coercitiv]

Very impressive load numbers.

Under load, the Broadwell equipped Latitude has a 7W higher power consumption than the Haswell model. It's also hotter and louder under load. It may keep that 2.7Ghz tubo, but it does so for a price.

Broadwell was supposed to bring 30% drop in power usage at normalized performance. In fact, more enthusiastic forum members considered that to be a minimum gain, with a possible positive surprise.

I would say at LEAST 30% efficiency improvements on final silicon.

I know we're talking about two entire systems with their respective components, but these are quite similar in construction, in fact they are as close as we can get when changing from a CPU gen to another.

I guess we'll get a better picture as more reviews come in.

 

[NTMBK]

TDP is pretty arbitrary. Intel can set TDP whatever they want.

Also remember how AMD used that silly 2X performance/TDP rating for their Puma+ cores?

If you want to know energy per instruction, you have to measure those quantities, not something else, which Intel has done with pre-qualified silicon:

http://www.anandtech.com/show/7318/intel-demos-14nm-broadwell-up-to-30-lower-power-than-haswell

All that demo showed is that Intel got one golden sample back from the fab. It's pretty meaningless if they can't get good enough yields at that voltage.

EDIT: Though to be fair, that was with the Y-series Broadwell, which is made on the SoC process variant- I would genuinely expect that to have lower power draw than Haswell-Y on the non-SoC variant. The U-series that people were arguing about above is on the non-SoC variant.

 

[Arachnotronic]

All that demo showed is that Intel got one golden sample back from the fab. It's pretty meaningless if they can't get good enough yields at that voltage.

EDIT: Though to be fair, that was with the Y-series Broadwell, which is made on the SoC process variant- I would genuinely expect that to have lower power draw than Haswell-Y on the non-SoC variant. The U-series that people were arguing about above is on the non-SoC variant.

The only difference between BDW-Y and BDW-U is the packaging, AFAIK.

 

[NTMBK]

The only difference between BDW-Y and BDW-U is the packaging, AFAIK.

Nah, it's actually on a different process:

But even that’s still not enough, and for Core M Intel went so far as to give Broadwell-Y its own die and design a low-power optimized version of their 14nm process just for it. This variant is designed to further reduce power consumption by optimizing the resulting transistors for lower power, lower voltage, lower clockspeed operation. By doing this Intel was able to further reduce power consumption in all of the major areas over what would be a traditional 14nm Intel process.

http://www.anandtech.com/show/8355/intel-broadwell-architecture-preview/4

See also slide 47 here: http://www.intel.com/content/dam/ww...advancing-moores-law-in-2014-presentation.pdf

EDIT: There was another slide somewhere detailing 4 different 14nm flavors... can't find the darn thing right now though.

EDIT 2: Bingo! Found it, slide 21 of this: http://www.semiwiki.com/forum/files/Intel Manufacturing Slides 2013.pdf

 

[witeken]

Nah, it's actually on a different process:



http://www.anandtech.com/show/8355/intel-broadwell-architecture-preview/4

See also slide 47 here: http://www.intel.com/content/dam/ww...advancing-moores-law-in-2014-presentation.pdf

EDIT: There was another slide somewhere detailing 4 different 14nm flavors... can't find the darn thing right now though.

EDIT 2: Bingo! Found it, slide 21 of this: http://www.semiwiki.com/forum/files/Intel Manufacturing Slides 2013.pdf

Sure, but back then BDW-U wasn't announced, so they didn't disclose anything about that. Since it's the same 2+2 die, one would assume that Intel didn't bother creating 2 version of the same silicon. Why would they do that? Maybe that's the reason BDW-U doesn't clock higher in boost.

 

[dahorns]

Under load, the Broadwell equipped Latitude has a 7W higher power consumption than the Haswell model. It's also hotter and louder under load. It may keep that 2.7Ghz tubo, but it does so for a price.

I mean, I guess, but the price doesn't seem that high. You're talking about a ~20% increase in consumption under load, but a nearly 100% increase in clock speed.

Broadwell was supposed to bring 30% drop in power usage at normalized performance. In fact, more enthusiastic forum members considered that to be a minimum gain, with a possible positive surprise.

It kind of seems like it delivered? Benchmark numbers aren't noticeably different because most benchmarks are short enough to run at max turbo, which didn't really change. But, the new processors can now run at sustained full turbo--apparently indefinitely (at least for this particular model).

If you wanted to get a normalized number, you'd have to down clock the broadwell CPU to 1.4 GHZ

 

[Arachnotronic]

Nah, it's actually on a different process:



http://www.anandtech.com/show/8355/intel-broadwell-architecture-preview/4

See also slide 47 here: http://www.intel.com/content/dam/ww...advancing-moores-law-in-2014-presentation.pdf

EDIT: There was another slide somewhere detailing 4 different 14nm flavors... can't find the darn thing right now though.

EDIT 2: Bingo! Found it, slide 21 of this: http://www.semiwiki.com/forum/files/Intel Manufacturing Slides 2013.pdf

I would bet very serious money that Broadwell-U 2+2 and Broadwell-Y are identical dice.

 

[III-V]

I would bet very serious money that Broadwell-U 2+2 and Broadwell-Y are identical dice.

Broadwell-Y has the VR inductors under the primary package, though. Does BDW-U have that? The primary PCB/whatever-you-call-it may be the same, though.

 

[Arachnotronic]

Broadwell-Y has the VR inductors under the primary package, though. Does BDW-U have that? The primary PCB/whatever-you-call-it may be the same, though.

Like I said, packaging for the BDW-Y is different, but the actual silicon die is identical.

 

[III-V]

Like I said, packaging for the BDW-Y is different, but the actual silicon die is identical.

Oh, missed your earlier comment.

 

[III-V]

I dont have good news for you, you only get higher idle :thumbsdown:

I'd hope more mature bios revisions would bump that up. That idle time is really nice... but hopefully Skylake makes progress on load battery life.

Bah, double post.

 

[AtenRa]

How come idle battery runtime is so important ?? I though Laptops switch to sleep state when idled after a few minutes.