Intel's Medfield SoC: Blazing!

[podspi]

I imagine the cost of design is as low as ARM, perhaps even lower since Intel has provided a complete(?) reference platform.

 

[Red Storm]

Would be funny to see Intel become the next Nexus partner, although it makes more sense for Google to use Moto since they own them now.

 

[alent1234]

why would google dump TI? i've already read that anyone who has used qualcomm CPU's and is doing an ICS upgrade is going to put a performance penalty on current hardware. why would google do the same thing again? it makes no sense

 

[Fire&Blood]

Andy Rubin has said in the past that for the Nexus, they start from scratch each year, selecting the SoC and the OEM in Q1 each year, so there would be no dumping of TI or anyone else since there are no long term commitments to begin with. Both Intel and Qualcomm have an equal shot at being selected as TI does.

 

[alent1234]

but that's the whole point, why go with a new SoC every year? Sony already said that ICS being based on TI means they have to code a new HAL for their phone which is going to incur performance issues. Same will probably go for Samsung and others.

It's like google is doing it on purpose to break compatibility with older hardware

 

[quest55720]

but that's the whole point, why go with a new SoC every year? Sony already said that ICS being based on TI means they have to code a new HAL for their phone which is going to incur performance issues. Same will probably go for Samsung and others.

It's like google is doing it on purpose to break compatibility with older hardware

I think samsung and others like it that way. It forces people to upgrade their phones frequently. The biggest reason I will never go with an android based phone.

 

[alent1234]

ok, i keep thinking US market but i guess everywhere else where you buy a phone outright it makes sense to drop support as fast as possible.

 

[Dulanic]

but that's the whole point, why go with a new SoC every year? Sony already said that ICS being based on TI means they have to code a new HAL for their phone which is going to incur performance issues. Same will probably go for Samsung and others.

It's like google is doing it on purpose to break compatibility with older hardware

I don't get your point. Are you implying that Google is lowering performance for SoC's they don't currently use? Any manufacturer is free to chose whatever SoC they wish. All they need to create is drivers for different hardware...

 

[MrX8503]

I didn't think Intel could do it, but it looks like they just did. Very impressive from Intel and I bet they've been working on this since the beginning boom of smartphones 5 or so years ago.

 

[alent1234]

I don't get your point. Are you implying that Google is lowering performance for SoC's they don't currently use? Any manufacturer is free to chose whatever SoC they wish. All they need to create is drivers for different hardware...

ICS is optimized for OMAP and Sony has already come out and said that they have to do a lot of custom coding of the HAL to get their current qualcomm phones upgraded to ICS. and there will probably be a performance penalty. Same with Samsung, anything not built around OMAP will probably perform worse than OMAP hardware.

if this is the case with ICS then I bet any phone with Medfield will perform worse than OMAP phones. and with medfield already having mediocre specs why would anyone buy it once all the new ARM CPU's hit the market? the GPU on medfield is already worse than the SGS 2 and that's almost oboslete

 

[kaerflog]

First Medfield phone already here.
http://www.engadget.com/2012/01/10/lenovo-k800-intel-medfield-smartphone-hands-on/

 

[Patranus]

So this is going to hit in the second half of 2012 and the iPad 3 will be here (with Apple's next generation chips) in 2 months.

While the results are impressive for Intel's first shot out of the gate, by the time it hits the market, it will (probably) be already behind. If intel's recent track record with desktop/laptop CPU/chipset in any indication, they will have a hard time to keep pace int eh mobile phone market.

Apple would have to post a ~50% gain in performance of their next generation platform to keep pace with Intel. The iPhone 4S showed a 100% increase in performance (int he same test) from the iPhone 4.

 

[WelshBloke]

So this is going to hit in the second half of 2012 and the iPad 3 will be here (with Apple's next generation chips) in 2 months.

While the results are impressive for Intel's first shot out of the gate, by the time it hits the market, it will (probably) be already behind. If intel's recent track record with desktop/laptop CPU/chipset in any indication, they will have a hard time to keep pace int eh mobile phone market.

Apple would have to post a ~50% gain in performance of their next generation platform to keep pace with Intel. The iPhone 4S showed a 100% increase in performance (int he same test) from the iPhone 4.

Out of curiosity who's making them this time round? Are they sticking with Samsung?

 

[Dulanic]

Out of curiosity who's making them this time round? Are they sticking with Samsung?

It's not confirmed, but rumors are still Samsung. Apple was rumored to use TSMC but that was back in July and since there have been rumors it is still Samsung as TMSC doesn't have the capacity. Still haven't seen confirmation either way for sure.

http://www.eetimes.com/electronics-news/4229790/Samsung-ramping-Apple-A6-chips
http://www.intomobile.com/2011/10/1...not-tsmc-build-future-quad-core-a6-processor/

 

[Patranus]

Out of curiosity who's making them this time round? Are they sticking with Samsung?

Apple (P.A. Semi)/TSMC

 

[McWatt]

Apple would have to post a ~50% gain in performance of their next generation platform to keep pace with Intel. The iPhone 4S showed a 100% increase in performance (int he same test) from the iPhone 4.

In the last iPhone (N)s to iPhone (N+1) upgrade, Rightware Browsermark went up by 39% and Sunspider js speed (1/benchmark time) by 34%. In the most recent (N) to (N)s upgrade, Rightware improved by 70% and Sunspider by 58%.

It looks like Apple is improving by roughly 50% per generation in terms of CPU performance for the most common uses of a smartphone. The Medfield article shows that the Medfield reference is currently 33% ahead of the 4S in Rightware and 69% ahead of the 4S in Sunspider. To me that implies that the best guess we can make is that this initial Intel CPU is performing right where we expect to find the ARM CPU chosen for the next iPhone.

The reference has significantly lower power draw than the 4S for most tasks, though, and if you look at the trends in power between iPhone versions, Apple's recently been getting worse in this area, albeit by a small amount. I think it's going to be a challenge for the ARM CPU in the next iPhone to match Apple's recent performance trends while reversing the recent power draw trends, and it would have to do both to match Intel.

refs:
http://www.anandtech.com/show/5365/intels-medfield-atom-z2460-arrive-for-smartphones
http://www.anandtech.com/show/4971/apple-iphone-4s-review-att-verizon/15 (Power Consumption Comparison table)

 

[Mopetar]

I'm somewhat leery of any claims that Intel will gain much traction in this market out of the gate. There are quite a few issues that they will need to deal with.

1) This chip will not be released for several months and is being compared to chips that are fairly long in the tooth at this stage. Other posters have pointed out that the Tegra 3 benchmarks nearly as well, an in my opinion the Tegra 3 isn't terribly impressive. Also Medfield is fabricated on a 32nm process whereas the chips it is being compared to are made with a 40nm process, making the power comparisons quite slanted.

By the time Medfield is released, I don't expect it to have as much of a significant lead in either power or performance. It's even possible that some chips will surpass it.

2) The GPU included with the SoC is already out of date even by today's standards. While an extremely powerful GPU is not necessary for all or even most cases, it is helpful in some and definitely advantageous for devices with higher resolutions. Not having something more powerful may prevent Medfield from being included in any flagship devices. Then again I think we should just be thankful that they didn't use their own graphics in this thing.

3) Getting manufacturers to include the chip in their product may be difficult. Samsung has a strong tendency to use their own Exynos chips, and Apple does their own chip development as well. Those two manufacturers make up a large chunk of the smartphone market that probably won't use Intel. Motorola has had a history of using TI parts and Microsoft has exclusively been using Qualcomm chips in their devices so far so those companies, although not as large as Apple and Samsung, may not use these chips either.

4) Software support will be a major hurdle. Applications that specifically make use of ARM instructions will need to either be updated to support X86 or the instructions will need to be translated by the VM. There are going to be a lot of applications that either don't work or have degraded performance, which could lead to a lot of returned devices. This may also reduce manufacturer interest in using an Intel solution.

Even if they have a good product, it's going to be difficult for Intel to get a foothold in the market.

 

[poofyhairguy]

Wow, Intel did it. I am a little leery of the app compatibility claims, or the overhead of Intel's translation layer.

 

[WelshBloke]

It's not confirmed, but rumors are still Samsung. Apple was rumored to use TSMC but that was back in July and since there have been rumors it is still Samsung as TMSC doesn't have the capacity. Still haven't seen confirmation either way for sure.

http://www.eetimes.com/electronics-news/4229790/Samsung-ramping-Apple-A6-chips
http://www.intomobile.com/2011/10/1...not-tsmc-build-future-quad-core-a6-processor/

Apple (P.A. Semi)/TSMC

So on balance are we going for Samsung?

 

[Medu]

The benchmarks seem to be limited to ones that don't scale well with threads which would give it an advantage over dual/quad core A9's. Still those are good results either way but I expect A15 dual cores to be faster core for core and much faster overall.

 

[pm]

Wow, Intel did it. I am a little leery of the app compatibility claims, or the overhead of Intel's translation layer.

Why is that? One of the big selling points of Java is that it's architecture-independent. There's lots of Java already running on standard Windows-based desktops. Why would there be architecture concerns about app compatibility? And why would Intel's translation layer have a lot of overhead when there's already a pretty good Java bytecode interpreter for x86?

When you look at something like this:
http://en.wikipedia.org/wiki/Java_virtual_machine

and then look at the diagram on the right, you can see that the flows are pretty similar.

And ARM is not like x86 where everything is backwards compatible and new instructions are supersets of prior ones. Various cores have various instructions and some are subsets of previous versions.
http://en.wikipedia.org/wiki/ARM_architecture
You can't run an ARMv6 binary on an ARMv7 core. So new instruction set architectures running Android are not unusual.

 

[MrX8503]

If intel's recent track record with desktop/laptop CPU/chipset in any indication, they will have a hard time to keep pace int eh mobile phone market.\

I don't know what you mean by this. Intel has been ahead in those markets ever since the release of Conroe.

 

[JeffreyLebowski]

And then get demolished by the dual core medfield and 543mp2 thats gonna release by the end of the year.

And that's going to get demolished by the next thing that comes out.

Moores Law is fairly accurate and if live by wanting the fastest thing out, you'll never buy anything because there is always something better around the corner.

 

[JeffreyLebowski]

While the results are impressive for Intel's first shot out of the gate, by the time it hits the market, it will (probably) be already behind. If intel's recent track record with desktop/laptop CPU/chipset in any indication, they will have a hard time to keep pace int eh mobile phone market.

Wait, so Intel has been behind AMD all this time and we didn't know it? If you use that analogy, then Intel is going to dominate the market by leaps and bounds.

 

[WelshBloke]

I don't know what you mean by this. Intel has been ahead in those markets ever since the release of Conroe.

Yeah. ARM are pretty awesome but theres no way I would ever rule Intel out of anything.


Well except maybe desktop graphics cards