Intel 35W processors vs 17W

I have some questions (asked at the bottom of this post) about the benefits of the 35W processors.

Just for reference, I picked the processors from the 13" Macbook Air and 13" Macbook Pro.

Specs taken from Intel website.

17W: i5-3427U - CPU Speed: 1.8 GHz, Turbo: 2.8GHz. GPU Speed: 350MHz, Turbo: 1.15GHz
35W: i5-3210M - CPU Speed: 2.5 GHz, Turbo: 3.1GHz. GPU Speed: 650MHz, Turbo: 1.1GHz

Based on reviews, well built laptops can easily maintain speeds near full CPU Turbo when under load. This means that the CPU performance difference between these two this is a measly 10-15%? GPU performance is about the same.

What happens to this extra 18W of power? Is it simply there to provide more headroom when CPU and GPU Turbo are being used simultaneously (e.g. gaming)? The 35W processor seems to be totally pointless for a "typical" user - this includes a lot of business users.

Am I missing something? What are the real world benefits of the 35W processor? Can somebody explain why the 35W processor doesn't seem to offer a solid performance gain over the 17W one?

Thanks!

 

Here you have review of 35W, 45W and 17W together

In Review: Intel Ivy Bridge Dual-Core CPUs - Notebookcheck.net Reviews

Here is 35W vs 45W and it doesn`t seem like the 45W Quads is doing any better than 35W in gaming. Sadly no 17Ws though
Intel Ivy Bridge Guide for Gamers - Notebookcheck.net Reviews

 

Have a look at the passmark scores for the two processors.

PassMark - CPU Benchmarks - List of Benchmarked CPUs

Also do a compare side by side.

ARK | Compare Intel® Products

John.

 

vPRO
the CPU die size
PCI-E revision

and speed on the actual CPU itself as well as the integrated GPU HD4000 (where most of the power goes).

 

Passmark CPU Scores:

3427u: 3552
3210m: 3902

Only a 10% higher score. Again my original question, what is the benefit of that extra 18W?

So that extra power is being consumed by the GPU? So the 35W chip would really only be a benefit for casual gamers (since serious gamers would want a seperate GPU)? Comically enough, vPRO is activated on the 17W processor, but not the 35W I listed.

 

the 3427u is an ultralow voltage cpu, most likely soldered to the board that saves a bit of power but 18w is still a lot.

John.

 

i´d rather say what are the advantages of the 17W. That review is indeed very interesting and i read it some month ago myself too. You see power consumption is always almost the same. I wouldnt say that will justify the 30 preformance gain from the i5 3210m to the i5 3317u.

Furthermore the turbo speed will only be reached if the cooling is sufficient. Most thin laptops dont really have good cooling so the max frequency might not even be achieved. My MSI clocked the 3210m at 3.0997GHz if i recal correctly due to good cooling it was able to use all the extra turbo Hz the processor has to offer.

 

the i5 3210m should be compared to the 3317u since they have the same price. The 3427u is likely to be more expensive. Note that my i5 scored 4140 points with 1333MHz RAM. With 1600MHz it might even get higher. Scores can differ quite a bit.

On 3d mark, another bencmark program the i5 3210m scores almost 3986 points the i5 3317u scores 3322 points that is a 20% gain quite a bit i´d say. The i5 3317u is roughly as good as the normal 35w i3´s

 

That review is great and was really helpful. So it seems the CPU performance increase is actually 20%-30% in ideal conditions.

I'm actually amazed by the these 17W processors. The performance difference is probably not noticeable in everyday use, which helps me better understand why people like laptops with them.

 

Something to note is that the 17W/35W numbers are TDP specs (Thermal Design Power). All TDP specifies is the amount of thermal energy the heatsink/cooling system paired with the chip must be able to dissipate in order to keep the chip within operating temperature ranges. There is no direct translation of TDP into actual power consumption; there is just a weak correlation which means next to nothing.

It's safe to say the low voltage chips run cooler and consume less power but how much is the question. I'd wager it is probably not half of the 35W chip as the TDP spec leads you to believe. Ideally for this test you would take two notebooks identical save for the CPU and perform load testing while monitoring the power draw and CPU temperatures. Then you could form a tentative conclusion.

 

The term TDP seems to be used more liberally than that though- Intel Extreme Tuning Utility displays monitoring data such as "CPU Total TDP: 7W, IA core TDP: 3W". Apparently those are instantaneous values and yet still referred to as TDP..

It's like how words like ROM can no longer be interpreted literally, for eg. with respect to custom smartphone ROMs.

 

The difference is that 17W TDP processors are put into tiny under-designed (cooling-wise) systems that will never give the sustained performance they are theoretically capable of.

35W TDP processors will be put into larger, more capable (cooling-wise) systems and can be used at (or much closer to) their theoretical potential.

Both will draw very close to the same power for the same workload - the 35W TDP processor will simply be able to sustain that performance indefinitely, while the 17W TDP processor will more than likely throttle itself to stay within it's design limits.


Conclusion: on paper - very, very similar (within ~10% as stated already...) in the real world - 17W systems are thinner, hotter, less powerful and ultimately more expensive.


17W will come into it's own with Haswell - until then, all available systems with 17W processors are barely able to keep my 'net surfing workload up to proper speeds.

 

Notebook check has power consumptions figures on the MB Air 13 and MB Pro 13. Other than the "Max Load," power consumption is similar on both laptops.

I'm assuming "Max Load" is CPU+GPU as I've tested my Laptop (V131) and maximum power consumption only occurs when both are loaded. I suspect this supports my idea that a solid benefit will only been seen when you have CPU+GPU loaded scenario, and that is where that extra 18W headroom will shine.

 

I think the main differences are while the peak speeds are similar, the i5-3427U draws vastly less power at idle since it is a specially binned low leakage chip. IIRC, the GPUs on the ULV chips can also idle to 350mhz while the 35W chips idle at 600mhz. The lower leakage also means that voltage required to reach 2.8ghz will be less thus peak heat output is also reduced.
The problem that people often have with ULV chips is that the manufacturer often takes advantage of the lower heat output to couple the chip with a crappy cooling system so the chip spends little time at 2.8ghz. Oh, and another important difference is that the ULV chips are BGA (i.e. soldered to motherboard) while most of the 35W chips use a socket. This means you can generally salvage the 35W chip if the mobo dies.

 

The difference is that Intel sells the 17W "lower power consumption" chips for a premium. It is almost a scam. Traditionally, faster, more powerful processors have always sold for more money. But the marketing department at Intel figured out a way to make the processors slower, yet sell them for more money!

 

The scam sells because all the average buyers are oblivious to information regarding their technology.... it's genius

 

I previously wanted to buy 3612QM to switch with the 3610QM due to the fact that its a lot cooler than 3610QM and use a little less energy, but it perform equal in games.
3612QM is a 35W CPU and 3610QM is 45W. There is a difference there between them.

But I didn`t go through with it since my GT70 is more than capable of removing the heat anyways. But it would have been nice if I went through with the fan control software on my GT70.

 

The best part of all is that the 3612QM was/is more expensive than the 3610QM.. and is clocked less

Brilliant Intel, brilliant.

 

It's possible they wanted to charge more due to it being a quad core in a 35W 'envelope' - if so, it's sneaky, and could be easily seen as misleading.

 

It would be nice if we could "flash" our CPUs to whatever speed we wanted. Then Intel could charge whatever they wanted

 

Forget 17W CPUs. Now its 13W/10W that matters

 

Clocks and Turbos look really slow!

 

kto, that is why they're Mobile Y processors... why would you buy them.

 

Hilarious!

 

It's also about the way they measure the "watt" TDP. It's not the power-consumption, but the amount of energy the chip-package will disperse. So depending on how you measure it, or how often you enable turbo, etc... But yes, marketing.

Having these processors (or just the bios and state tweaks) with an on average lower heat envelope would have benefitted a very large amount of 13' devices rolled out over the last year, though... Kind of surprised they actually have the guts to go out with this now, without presenting newer architecture improvements, or... something. Anything..

 

Guts?

This is how Intel rolls; get the new architecture (to a stable/mature state), get the new process mode (to a stable/mature state) then combine the proven technologies and rinse and repeat for the newer tech/processes being tried.

Nothing surprising or gutsy about it. Just a well oiled machine that marches to it's own drummer.

 

I guess so..

I'm just imagining that presenting it to OEMs now would sound a bit like this: "Thank you, Gentlemen. Yes, we /could/ just release new power-profiles and instantly increase the value of all the items in your stock, while lowering your support department costs, freeing up resources to adapt new models at what we will now treat as a baseline power-envelope for all processor types.

But we're not going to do that, and just sell you new numbers for higher prices! Hahahahahaha! MMWWWAHAHAHAHHAHAH! I have become God, and you His loyal subjects! Fear me and tremble! *thhhbblbllblblblblbthhh*".

*shrug*

I wouldn't have been able to say that. I wouldn't dare to.

 

nipsen, I think part of the 'well oiled machine' is the OEM's - they don't want any surprises or mass failures either...

No God-like power trip: just common sense.

(Intel is diligently following the 'just because you can do something, doesn't mean you should' theorem. Simply stated: proceed slowly but surely and always from a position of power).

 

I see it the other way around. Just because Intel can, due to lackluster competition from AMD, they slow down the release schedule as long as they can and keep trying to find new ways to ship and market the same old chips. Also, why would OEM's like this? If Intel doesn't roll out new and exciting processors quickly enough, there is a lot less incentive for end users to buy the next new thing or for OEM's to make new laptop models based on the same existing processor.

 

Qing Dao, I'm sure there is a little of that at play too - however; you don't get to be a giant in technology by rushing things out before they're 'cooked' to perfection.

When you're at the forefront of the bleeding edge - does it make sense to potentially stumble and fall (and lose your #1 position) just to essentially show off a little early? I would think not.

I know... AMD was in that position over half a dozen years ago and they DID lose that #1 spot - but not because they stopped 'showing off early', but rather because they became lax and overconfident that their way was the only way (Intel had a jarring wakeup call for them around 2006; named Core).

Because most customers buy what they need (or have locally and/or can afford spend) they don't do their due diligence and buy the 'best' match for their needs nor do they buy the 'best' for the amount of $$$$ invested...

They buy to get work done, play and enjoy what is available.

Even if Intel or any company gave us the 'next big thing' earlier than the present cadence... we'd still be having these conversations of why we don't have 2020 technology NOW.