Intel Broadwell Thread

[Sweepr]

The Surface Pro 3 in that chart is running 3DMark Ice Storm Extreme. The 5Y70 benchmarks are using Ice Storm Unlimited.

The graphics score for the Surface Pro 3 (i5) in Ice Storm Unlimited is 54,000~55,000.

Source: Futuremark (Link)

Thanks for the correction. Still impressive for a 4.5W TDP chip. They succesfully brought 15W Core i5 Haswell-U's level of CPU & GPU performance (in fact slightly better graphics performance) to a tablet/convertible chip.

 

[Burpo]

 

[North01]

http://www.legitreviews.com/intel-core-m-5y70-broadwell-y-benchmark-numbers-run-at-idf-2014_150217

3DMark Cloud Gate:

Core M 5Y70

• Physics - 2679
• Graphics - 5737
• Total - 4576

Surface Pro 3 (i5)

• Physics - 2701
• Graphics - 5927
• Total - 4683

 

[Roland00Address]

So my mom's laptop broke so we got her a new one this week. It will be a lenovo yoga 11s with the i5 4210y, 4gb, 128gb ssd for $430 (it is a scratch and dent but it has the same warranty.) The big things we were looking for was price though I was pushing heavy for a ssd but that is so hard to get in this price range.

My mom does not need more cpu power her use is very basic, but part of me wish it was broadwell with these numbers.

 

[Enigmoid]

http://www.legitreviews.com/intel-core-m-5y70-broadwell-y-benchmark-numbers-run-at-idf-2014_150217

Intel said that they are about two times faster on gaming and three times faster on most of the other applications than the Qualcomm Snapdragon 801 or 805 processors. Intel also noted that the 14nm Core M processors will offer 50% more CPU performance as well as 40% faster graphics performance when compared to last year’s Intel Core i5-4302Y. Intel feels strongly that they have ARM, AMD and Qualcomm beat for raw processor performance
Read more at http://www.legitreviews.com/intel-c...rs-run-at-idf-2014_150217#KyoiDFvKffiIHlJ1.99

Core M really looks like a huge win for intel. I can't help feeling that there is some catch somewhere.

 

[Abwx]

Thanks for the correction. Still impressive for a 4.5W TDP chip. They succesfully brought 15W Core i5 Haswell-U's level of CPU & GPU performance (in fact slightly better graphics performance) to a tablet/convertible chip.

All this to dissipate a mere few watts..??.

This copper plate makes direct contact with the aluminum housing of the tablet, so essentially the table housing itself is one gigantic heatsink.

Casted aluminium, quite a massive piece of metal with a thermal inertia that has nothing to do with what be used in commercial designs, notice how it was machined to get the heavier possible heatsink :

http://www.legitreviews.com/intel-c...rs-run-at-idf-2014_150217#KyoiDFvKffiIHlJ1.99

 

[IntelUser2000]

Laws of physics are the same for anybody, including intel.

So if the sole limiter is Laws of Physics why does some CPU companies do so much better than others?

Hint: There's enough room to do better within the Laws of Physics.

My IVY BRIDGE 17W i7 uses only 12W core power running Cinebench R11.5 at 2.8GHz.

Why wouldn't a chip with a) on 14nm b) even better binning c) further process tricks to lower power d) 2.4-2.6GHz(lower than what my chip ran at, and according to you should be 20-30% lower power) be able to do that at 40% of the power?

It's time to face it. Core M kicks ass.

 

[liahos1]

All this to dissipate a mere few watts..??.

Casted aluminium, quite a massive piece of metal with a thermal inertia that has nothing to do with what be used in commercial designs, notice how it was machined to get the heavier possible heatsink :

http://www.legitreviews.com/intel-c...rs-run-at-idf-2014_150217#KyoiDFvKffiIHlJ1.99

isn't the iPad aluminum?

 

[IntelUser2000]

All this to dissipate a mere few watts..??.

Casted aluminium, quite a massive piece of metal with a thermal inertia that has nothing to do with what be used in commercial designs, notice how it was machined to get the heavier possible heatsink :

Oh good, maybe we won't see this: http://images.anandtech.com/reviews/tablets/apple/ipadair/throttlesm.png

 

[Sweepr]

given apples claims of 50% better gpu performance and 25% better cpu performance in a8 vs a7. the A8 gets TROUNCED here. i wonder how hard it would be for apple to move ios to x86 compatibility and get the future successors of these chips in ipads.

Its crazy to see the amazing performance.

Its going to be hard for Apple and others to justify the usage of ARM in their highend tablets as time goes.

Indeed. Imagine a ~12'' iPad Pro or a larger Nexus tablet based on one of them. I'm sure some people would be willing to pay more for top notch Haswell-U like performance in that form factors. Premium Windows convertibles are just one side of the equation, there's so many possibilities.

 

[Abwx]

My IVY BRIDGE 17W i7 uses only 12W core power running Cinebench R11.5 at 2.8GHz.

Quite fair but i m sure that you are aware that IB has intrinsicaly better perf/watt than HW, this latter is a favourable comparison for Core M.

Why wouldn't a chip with a) on 14nm b) even better binning c) further process tricks to lower power d) 2.4-2.6GHz(lower than what my chip ran at, and according to you should be 20-30% lower power) be able to do that at 40% of the power?

It's time to face it. Core M kicks ass.

Compared to the current lines and processes it do quite well but really nothing exceptional in respect of perf/watt, you had noticed that Intel doesnt speak a lot about perf/watt but are rather focusing on absolute perfs for their marketing, once the competitors get to their next node it may well morph as Core Meh...

Oh good, maybe we won't see this: http://images.anandtech.com/reviews/tablets/apple/ipadair/throttlesm.png

Given the recent technical tendencies i think that in the future we ll se a lot more of theses, unfortunately..

 

[IntelUser2000]

Quite fair but i m sure that you are aware that IB has intrinsicaly better perf/watt than HW, this latter is a favourable comparison for Core M.

Not at the U level.

Core i7 3687U: 2.1/3.3GHz, 17W + 3W PCH
Core i7 4600U: 2.1/3.3GHz, 15W combined

once the competitors get to their next node it may well morph as Core Meh...

The 20nm parts coming aren't impressive either. The 20nm Exynos coming from Samsung is talking about 100-200MHz frequency increase. And we already know A8 went under lot of peoples' expectations.

 

[Abwx]

isn't the iPad aluminum?

Dont know but i would have thought that tablets are builts with relatively light but rigid parts, of course a lot of aluminium, casted preferably for the cover, is a good thing quality wise but it s also aknowledgment that the devices are quite large thermal source, if there s no heating issues lighter materials will be prefered as lower weight is also a desirable feature.

 

[IntelUser2000]

isn't the iPad aluminum?

http://forums.macrumors.com/showthread.php?t=855475

When Steve Jobs mentions the Environmental factors in the Keynote (just finished watching it) ... he says the Aluminum Case and Glass are "highly recycle-able" so yes it is aluminum.

 

[erunion]

Apple has been using aluminum backs for heat dissipation since the first ipod in 2001.

 

[Abwx]

Not at the U level.

Core i7 3687U: 2.1/3.3GHz, 17W + 3W PCH
Core i7 4600U: 2.1/3.3GHz, 15W combined

This say nothing , it could be due to a more conservative timing wise turbo for HW, also the PCH got its TDP more than halved...

Core i7 3687U
3300 MHz (1 core)
3100 MHz (2 cores)

The 20nm parts coming aren't impressive either. The 20nm Exynos coming from Samsung is talking about 100-200MHz frequency increase. And we already know A8 went under lot of peoples' expectations.

About Samsung and Apple i cant give a valuable opinion, i dont know exactly how their products perform and what are TSMC s 20nm process caracteristics, for the time the main competitors , although on a lower segment, are AMD s Mullins/Beema and Intel s own Bay Trail, the next iterations of those chips wont be a problem ST wise but on MT they will both beat handily Intel s 2C flagship while Skybridge (Puma/A57) will be competitive even in the GPU department.

 

[Dufus]

But running even longer than that it will bring it back to PL1, which is TDP.

PL1 is programmable as are the rest, at least so before being locked, so should not be assumed to be set at TDP.

I've never heard something so absurd. Going from say 1.6 ghz to 2.4 ghz does NOT require more than 2.5x the power.

It's not so absurd. For core power the square of the voltage needs to be taken into account. For instance on my HSW running defaults core voltage will increase from 0.774 to 0.876 from 1.6GHz to 2.4GHz. This will mean an increase with same execution load of 1.92x the power before even taking into account the capacitance. Likewise from 2.0GHz to 3.0GHz is 2.25x the power.

Now the CPU runs more than just the cores these days so there is also an amount power for the rest of the package and if measuring overall package power (including cores) then results will be skewed, especially at the lower frequency end whereby increase in power may even seem to be less than the direct ratio of multiplier increase. Once core powers are sufficiently high the rest of the package power becomes less influential on the overall result.

 

[Roland00Address]

All this to dissipate a mere few watts..??.

Casted aluminium, quite a massive piece of metal with a thermal inertia that has nothing to do with what be used in commercial designs, notice how it was machined to get the heavier possible heatsink :

If a company is going to spend a $100+ dollars buying an intel cpu they are going to spend the extra money to get a good heatsink profile if they are marketing the tablet as an extremely thin one. If they do not care about absolute thinness they can do other tricks to keep cost down by making the tablet thicker.

Aluminium is also very cheap to make, the raw metal is super cheap. The problem is not the cost of the raw metal itself but the actual machinery and time that costs money.

 

[ams23]

Tegra K1 seems to be a lot more competitive @ GFXBench than 3DMark Unlimited Graphics relative to SP3's Haswell-U scores though.

3DMark Ice Storm Unlimited is CPU limited and not particularly stressful on the latest and greatest ultra mobile GPU's (the framerate is > 200fps in game test 1 and > 100fps in game test 2). GFXBench 3.0 on the other hand is GPU limited.

 

[bullzz]

@Abwx - "Casted aluminium, quite a massive piece of metal with a thermal inertia that has nothing to do with what be used in commercial designs"

thats a load of crap. Lenovo U430P - a friend bought this last month for $500. this is a mid range laptop and has aluminum chasis
all ipads and macbook airs from 2012 have aluminum chasis

 

[Burpo]

In an Entertainment Dock w/fan
"able to be overclocked.. up to 40%"
https://www.youtube.com/watch?v=znKgvPk3-uc

 

[Khato]

3DMark Ice Storm Unlimited is CPU limited and not particularly stressful on the latest and greatest ultra mobile GPU's (the framerate is > 200fps in game test 1 and > 100fps in game test 2). GFXBench 3.0 on the other hand is GPU limited.

Okay, though now we also have 3DMark Cloud Gate results for the Core M tablet - surely that doesn't also fall under the category of not being adequately stressful? Sadly we can't make direct comparisons, but we easily can compared Core M to, say, a GTX 750M and then compare that to K1 Gfxbench. Namely, Core M obtains a Cloud Gate graphics score of 5737 versus an average, according to notebookcheck, of 11088 for a GTX 750M. So it's safe to say that Core M is roughly half the performance of a GTX 750M in that benchmark. Then we can easily compare a GTX 750M against the Shield Tablet under GfxBench - 92.2 fps vs 30.9 fps for Manhatten and 159.5 fps vs 64.7 fps for T-Rex. So 3x faster for Manhatten and 2.5x faster for T-Rex. Also, just to complete the comparisons to see how Intel's Gen7.5 at least translates performance between 3DMark Cloud Gate and Gfxbench, Iris Pro 5200 is basically equal to a GTX 750M on Cloud Gate but falls notably behind on Manhatten, 76.5 fps vs 92.2 fps, and slightly behind on T-Rex, 152.1 vs 159.5. If that trend continued then the end result would be that Core M would be about 1.2x faster than K1 in Gfxbench.

Which is the long way of saying that the information thus far points to Core M being quite competent in terms of graphics. Not as far ahead as Intel would probably like of course, but Intel has long since proven that they can't forecast graphics performance targets very well.

 

[USER8000]

Ice Storm Unlimited is rendered off-screen.

https://www.youtube.com/watch?v=XGEzGTTl1j8

So basically no one knows whether the demo was properly rendered at all??

Whats the point then?? We have see graphics tricks from all vendors(including Intel,AMD and Nvidia when it comes to textures and filtering) to improve scores in 3DMark. This is why we need independent test runs so people can make sure that all is being run fine.

Even then we need actual games run too - being good in 3DMark does not mean we actually see as big gains in games.

I don't know why Intel did not run GRID2 as it has optimisations for them,and runs relative well on their IGPs.

 

[Idontcare]

All this to dissipate a mere few watts..??.

Casted aluminium, quite a massive piece of metal with a thermal inertia that has nothing to do with what be used in commercial designs, notice how it was machined to get the heavier possible heatsink :

http://www.legitreviews.com/intel-c...rs-run-at-idf-2014_150217#KyoiDFvKffiIHlJ1.99

It is not clear to me why this is something to complain about.

Don't we want smart engineers working smarter at giving us better performing products? Isn't the above just one example of that happening?

If I already need a sturdy rigid case to frame my tablet/phone/laptop/etc then why would I care if that sturdy rigid frame is optimized to be multi-purpose and helps keep the internal stuff cooler or running faster?

This convo is sliding into the "complaining for the sake of complaining" category IMO.

 

[USER8000]

It is not clear to me why this is something to complain about.

Don't we want smart engineers working smarter at giving us better performing products? Isn't the above just one example of that happening?

If I already need a sturdy rigid case to frame my tablet/phone/laptop/etc then why would I care if that sturdy rigid frame is optimized to be multi-purpose and helps keep the internal stuff cooler or running faster?

This convo is sliding into the "complaining for the sake of complaining" category IMO.

It depends if actual production Core M tablets have as hefty heatsink/cases though - if they do there is no problem. However it adds to the weight and size of the tablet too and most companies are aiming for as light as possible.

If they don't it increases the chance you will not see the higher end benchmark scores due to thermal limitations,and you also need to consider most tablets have 10.1" screens and under in size not nearly 13" screens like the pre-production Core M tablet.

Look at the iPad MK5:

http://cdn.macrumors.com/article-new/2013/08/fanaticfone_ipad_5_shell_inside.jpg