It is, but actually it was on the roadmap for a H2'16 release for a long time, it's just that people have forgotten about Intel's mobile efforts after they quit last year.
Spreadtrum is also working on a 10nm (Intel Custom Foundry) SoC.
This is a waste, i never buy, let alone let a friend or colleague buy atom based products, performance is soo piss poor for 2017, its a joke. Its slower than core 2 duo from 10 years back. Intel has to stop this nonsense, and just make a lower mhz kaby lake like 1ghz, and it will still be faster than these atoms
I'm quite happy with the performance of Atom based CPUs in tablets. I own a Dell Venue 8 Pro running Windows 10. I run CAD programs and microcontroller development IDEs on it to view CAD models in the field and flash/debug MCUs in the field. The performance is absolutely excellent.
Well, I suppose something as ancient as autocad 2007 would fly on it. Modern cad software would run very poorly.
"microcontroller development IDEs" that's pretty much a glorified text editor. Not exactly demanding to run, also targeting MC platforms isn't exactly demanding when it comes to the actual compilation.
Try the latest visual studio and see how long you will take it before you smash that poor venue in anger. Atoms are seriously slow. In fact even core m is too slow for the vast majority of professional work. I pin my hopes on a zen apus making their way to tablets and convertibles, finally producing some useful and reasonably priced ultra portables.
Even if AMD has a competitive chip for ultra portables, which right now they don't, what makes you think anyone would allow them to actually produce a good device that could be sold in great numbers to consumers? We've been down this road before.
That being said, you do sound like an absolute joy of a person to be around. :)
I don't care about devices selling in great numbers. It has been a while since I bought one of those. My needs and tastes rarely coincide with the mainstream.
What I'd like to have is a 2 cm or 3/4 inch thick tablet device, with a notebook grade quad core in around 25 watts TDP budget and a huge ass battery. Fast storage for data acquisition, 32 gigs of ram, good stylus support.
Certainly not something that would sell in great numbers, even considering its usability and current uniqueness.
In other words you'd like a unicorn that can drag your private jet and stows itself in your carry-on luggage during flight. Pretty sure if that could exist it'd hit mainstream extraordinarily fast.
Think the biggest problem with Atom is the lack of RAM and storage, that messed up Win 8.1 on my HP Stream 7 . But I only paid 30 € including 24 % vat, so for that money its good enough.
So it's a suped-up Moorefield? Double the cores and upgraded GPU (at least, generationally an upgrade. I have no idea how a PowerVR GT7200 compares to a the iGPU of Moorefield SoC). Oh, and one a newer process node. Those all sound like good upgrades, but it's still a three year old architecture that wasn't exactly a game changer when it came out.
I guess being x86 gives it a bit of a selling point, but I'm pretty sure if this were a truly viable path, Intel wouldn't have dropped out of the smartphone SoC space.
The real issue is what, EXACTLY, Spreadtrum gets from Intel. Do they get masks or something so level it's useless for understanding, or high level designs + simulation code so that they can tweak the design as they wish?
The point is, without knowing what's being sold, one can't adequately judge the deal. IF high level design is being transferred, the Chinese govt (which is basically the power behind Tsinghua Unigroup) may consider how many chips get sold to be irrelevant and unimportant; what's really being bought is a pool of technology to kick China's home-grown CPU design skills up a notch.
I'm not commenting on whether that's good or bad (for China, for the US, for Intel, for the world). Just pointing out that the visible part of this deal may be practically irrelevant to the real content of the deal.
I get it, the US is trying to sabotage the already pretty successful Chinese chip design by allowing intel to give them the lousy atom arc, hopefully luring them into adopting that and scrapping their own design.
I think you're missing my point. The issue is not that China is buying a lousy Atom core, we all agree on that. The issue is that are they ALSO buying the simulation programs, the testing suites, and the design infrastructure that can be used (by those skilled in the art) to design much better CPUs?
That sort of infrastructure is not sexy, but it takes an immense amount of time and man-hours to build; so if you can buy it rather than build it you would do well to do so...
Writing a digital logic simulator is easier than writing a basic 2d game, and I mean it.
Logic to process is a little more complex, however, such software sells freely in the western world, and it would be trivial for the Chinese to acquire, crack and run in.
Process software for 14 nm finfet is not really considered a trading secret, that is now pretty much in the public domain.
Furthermore, even without external help or piracy, the Chinese won't take long to get in the game. This is not the China from 10 years ago. China has long exceeded the scientific output of countries like S. Korea and Japan, and is far ahead of Taiwan - all those big chip designers and makers. The Chinese aren't dumb, they can certainly do it on their own, and they pour an immense amount of resources into achieving independence, competitiveness and ultimately even domination in the semiconductor business.
And intel is aware of that, thus they do what would at least win them some money and market rather than giving them Chinese the cold shoulder, which wouldn't hurt them more than it would hurt intel to miss on that deal.
Hm. Was not expecting that. Then again, that thing looks _huge_ for a smartphone. I suppose that's some sort of ES chip, and that it can be integrated into a smaller (DRAM-stackable) package?
I'm guessing the chip would not require the packaging, but would instead be soldered on to the system board. The die itself doesn't look too unreasonable.
The chip looks 50% smaller than a thumb nail, possibly a Chinese person's thumb (i.e. possibly rather small). Pretty much in line with smartphone socs.
What high-end chipset is ever going to use a PowerVR GT7200 in 2017? Are they going to clock the GPU at 2GHz?
Also, why use the old airmont instead of the not-so-new-anymore goldmont architecture? Doesn't this mean they will be limited to LPDDR3 where the high-end devices have been using the 2x faster LPDDR4 for years?
4K30 HEVC means it's also well behind all the dirt-cheap chipsets supporting 4K60 decoding for netflix and the like, meaning it's not even going to take their place.
This is looking like SoFia all over again. Way too old components coming up way too late and expecting those to somehow compete in a cut-throat market.
Well bottom tier would make more sense, but the article clearly states "high-end handhelds":
"Despite this, we were surprised to hear that At MWC 2017, Intel’s partner Spreadtrum introduced a brand new application processor for high-end handsets"
Well this would've made the Zenfone 2 more power efficient, although it was not the CPU or SOC that was a power hog. The modem was the real problem with power efficiency.
Sounds like an interesting chip for tablets and mini PCs, too. Especially if they're not asking for typical Intel prices. Which is quite possible, since they don't have to invest in the fab R&D.
If you are smart, you use a portable application development framework. The code I write runs on every major and most of the minor platforms. No modifications required whatsoever.
I use Qt for the front end, pure C++ for the core logic. Qt sucks, but is still without competition, so depending on your experience with it, you might call it either "the best" or "the least sucky". For me it is the latter, I seem to have a gift for finding bugs and limitations in it, that even when reported remain unaddressed for years.
Quite late for a high-end since this will be competing with SD820 to Sd830s. It will be awesome for a mid-range with large batteries, phablet or tablet.
It would be great for cheap Windows tablets because those haven't seen a new chip since the Z38xx two years ago. I'm a big fan of those because a lot of software still needs x86, especially Windows and Linux stuff. There's still a market for tablets cheaper and smaller than Surfaces with their pricey Core M chips.
This most likely isn't PC-compatible, just like the Moorefield/Merrifield and SoFIA chips before it.
And do you mean Z8xxx? In any case, cheap Windows tablets are likely going to be on ARM chips from here on out. It can hypothetically run on any AArch64 platform with UEFI and ACPI support - even as low end as a Raspberry Pi 3 (although it's below the system requirements for 64-bit Windows due to insufficient RAM), if the UEFI developed for Windows 10 IoT Core for RPi 2 is ported to 64-bit.
On paper that should be a reasonable performer. IIRC Medfield held its own against high end QCOM SoCs back in the day (albeit without an integrated baseband). And Cherry Trail on my Surface 3 can run Skyrim perfectly well (albeit the Intel graphics obviously plays a part and the TDP will be higher).
Suspect the challenge will be more on the software side (viz Geekbench results) - are Android apps set up to take advantage of four cores? Also power drain will always be an issue.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
47 Comments
Back to Article
K_Space - Wednesday, March 15, 2017 - link
Now that is a surprise!witeken - Wednesday, March 15, 2017 - link
It is, but actually it was on the roadmap for a H2'16 release for a long time, it's just that people have forgotten about Intel's mobile efforts after they quit last year.Spreadtrum is also working on a 10nm (Intel Custom Foundry) SoC.
edlee - Wednesday, March 15, 2017 - link
This is a waste, i never buy, let alone let a friend or colleague buy atom based products, performance is soo piss poor for 2017, its a joke. Its slower than core 2 duo from 10 years back. Intel has to stop this nonsense, and just make a lower mhz kaby lake like 1ghz, and it will still be faster than these atomsDarkknight512 - Wednesday, March 15, 2017 - link
I'm quite happy with the performance of Atom based CPUs in tablets. I own a Dell Venue 8 Pro running Windows 10. I run CAD programs and microcontroller development IDEs on it to view CAD models in the field and flash/debug MCUs in the field. The performance is absolutely excellent.Ro_Ja - Wednesday, March 15, 2017 - link
CAD Programs aren't heavy?ddriver - Wednesday, March 15, 2017 - link
Well, I suppose something as ancient as autocad 2007 would fly on it. Modern cad software would run very poorly."microcontroller development IDEs" that's pretty much a glorified text editor. Not exactly demanding to run, also targeting MC platforms isn't exactly demanding when it comes to the actual compilation.
Try the latest visual studio and see how long you will take it before you smash that poor venue in anger. Atoms are seriously slow. In fact even core m is too slow for the vast majority of professional work. I pin my hopes on a zen apus making their way to tablets and convertibles, finally producing some useful and reasonably priced ultra portables.
niva - Thursday, March 16, 2017 - link
Even if AMD has a competitive chip for ultra portables, which right now they don't, what makes you think anyone would allow them to actually produce a good device that could be sold in great numbers to consumers? We've been down this road before.That being said, you do sound like an absolute joy of a person to be around. :)
ddriver - Thursday, March 16, 2017 - link
I don't care about devices selling in great numbers. It has been a while since I bought one of those. My needs and tastes rarely coincide with the mainstream.What I'd like to have is a 2 cm or 3/4 inch thick tablet device, with a notebook grade quad core in around 25 watts TDP budget and a huge ass battery. Fast storage for data acquisition, 32 gigs of ram, good stylus support.
Certainly not something that would sell in great numbers, even considering its usability and current uniqueness.
lmcd - Thursday, March 16, 2017 - link
In other words you'd like a unicorn that can drag your private jet and stows itself in your carry-on luggage during flight. Pretty sure if that could exist it'd hit mainstream extraordinarily fast.Alexvrb - Thursday, March 16, 2017 - link
But think of all the incentive companies have to build something hardly anyone would buy!!Ariknowsbest - Friday, March 17, 2017 - link
Think the biggest problem with Atom is the lack of RAM and storage, that messed up Win 8.1 on my HP Stream 7 . But I only paid 30 € including 24 % vat, so for that money its good enough.ddriver - Wednesday, March 15, 2017 - link
"Spreadtrum" - now that's an awful name for a brand. Sounds like spread rectum. Seriously!jordanclock - Wednesday, March 15, 2017 - link
So it's a suped-up Moorefield? Double the cores and upgraded GPU (at least, generationally an upgrade. I have no idea how a PowerVR GT7200 compares to a the iGPU of Moorefield SoC). Oh, and one a newer process node. Those all sound like good upgrades, but it's still a three year old architecture that wasn't exactly a game changer when it came out.I guess being x86 gives it a bit of a selling point, but I'm pretty sure if this were a truly viable path, Intel wouldn't have dropped out of the smartphone SoC space.
name99 - Wednesday, March 15, 2017 - link
The real issue is what, EXACTLY, Spreadtrum gets from Intel. Do they get masks or something so level it's useless for understanding, or high level designs + simulation code so that they can tweak the design as they wish?The point is, without knowing what's being sold, one can't adequately judge the deal. IF high level design is being transferred, the Chinese govt (which is basically the power behind Tsinghua Unigroup) may consider how many chips get sold to be irrelevant and unimportant; what's really being bought is a pool of technology to kick China's home-grown CPU design skills up a notch.
I'm not commenting on whether that's good or bad (for China, for the US, for Intel, for the world). Just pointing out that the visible part of this deal may be practically irrelevant to the real content of the deal.
ddriver - Wednesday, March 15, 2017 - link
I get it, the US is trying to sabotage the already pretty successful Chinese chip design by allowing intel to give them the lousy atom arc, hopefully luring them into adopting that and scrapping their own design.name99 - Thursday, March 16, 2017 - link
I think you're missing my point. The issue is not that China is buying a lousy Atom core, we all agree on that. The issue is that are they ALSO buying the simulation programs, the testing suites, and the design infrastructure that can be used (by those skilled in the art) to design much better CPUs?That sort of infrastructure is not sexy, but it takes an immense amount of time and man-hours to build; so if you can buy it rather than build it you would do well to do so...
ddriver - Thursday, March 16, 2017 - link
Writing a digital logic simulator is easier than writing a basic 2d game, and I mean it.Logic to process is a little more complex, however, such software sells freely in the western world, and it would be trivial for the Chinese to acquire, crack and run in.
Process software for 14 nm finfet is not really considered a trading secret, that is now pretty much in the public domain.
Furthermore, even without external help or piracy, the Chinese won't take long to get in the game. This is not the China from 10 years ago. China has long exceeded the scientific output of countries like S. Korea and Japan, and is far ahead of Taiwan - all those big chip designers and makers. The Chinese aren't dumb, they can certainly do it on their own, and they pour an immense amount of resources into achieving independence, competitiveness and ultimately even domination in the semiconductor business.
And intel is aware of that, thus they do what would at least win them some money and market rather than giving them Chinese the cold shoulder, which wouldn't hurt them more than it would hurt intel to miss on that deal.
lmcd - Thursday, March 16, 2017 - link
We don't agree at all. Intel's Atom cores are excellent ever since Silvermont.ddriver - Friday, March 17, 2017 - link
Yeah, they excel at being mediocre and sucky :)fanofanand - Tuesday, March 21, 2017 - link
@ddriver - I got a chuckle out of that one, thank you :)Valantar - Wednesday, March 15, 2017 - link
Hm. Was not expecting that. Then again, that thing looks _huge_ for a smartphone. I suppose that's some sort of ES chip, and that it can be integrated into a smaller (DRAM-stackable) package?MonkeyPaw - Wednesday, March 15, 2017 - link
I'm guessing the chip would not require the packaging, but would instead be soldered on to the system board. The die itself doesn't look too unreasonable.ddriver - Wednesday, March 15, 2017 - link
The chip looks 50% smaller than a thumb nail, possibly a Chinese person's thumb (i.e. possibly rather small). Pretty much in line with smartphone socs.Valantar - Thursday, March 16, 2017 - link
The chip looks small. The package does not. Hence my speculation whether it might be an ES, and the package just for testing purposes.ddriver - Thursday, March 16, 2017 - link
The package is a convenience and / or testing thing. You don't need to use that particular package to use the chip.But even as it is, IMO the height of the thing is the more detrimental aspect rather than its footprint.
ToTTenTranz - Wednesday, March 15, 2017 - link
What high-end chipset is ever going to use a PowerVR GT7200 in 2017? Are they going to clock the GPU at 2GHz?Also, why use the old airmont instead of the not-so-new-anymore goldmont architecture? Doesn't this mean they will be limited to LPDDR3 where the high-end devices have been using the 2x faster LPDDR4 for years?
4K30 HEVC means it's also well behind all the dirt-cheap chipsets supporting 4K60 decoding for netflix and the like, meaning it's not even going to take their place.
This is looking like SoFia all over again. Way too old components coming up way too late and expecting those to somehow compete in a cut-throat market.
DanNeely - Wednesday, March 15, 2017 - link
This's a china chip, only needs to compete against bottom tier mediateck, rockchip, and all winner designs.ToTTenTranz - Wednesday, March 15, 2017 - link
Well bottom tier would make more sense, but the article clearly states "high-end handhelds":"Despite this, we were surprised to hear that At MWC 2017, Intel’s partner Spreadtrum introduced a brand new application processor for high-end handsets"
Glock24 - Wednesday, March 15, 2017 - link
Well this would've made the Zenfone 2 more power efficient, although it was not the CPU or SOC that was a power hog. The modem was the real problem with power efficiency.MrSpadge - Wednesday, March 15, 2017 - link
Sounds like an interesting chip for tablets and mini PCs, too. Especially if they're not asking for typical Intel prices. Which is quite possible, since they don't have to invest in the fab R&D.niravvadodara - Wednesday, March 15, 2017 - link
Can you please share Benchmark comparison with other competitors?Ian Cutress - Wednesday, March 15, 2017 - link
If we ever get one. Ask Spreadtrum to send us one :)newbie06 - Thursday, March 16, 2017 - link
The chip has entries in Geekbench database and as expected it's doing quite poorly: http://browser.primatelabs.com/v4/cpu/1993017lefty2 - Wednesday, March 15, 2017 - link
Since Intel left the smartphone market app developers are no longer supporting x86. That's going to be a problem.ddriver - Wednesday, March 15, 2017 - link
If you are smart, you use a portable application development framework. The code I write runs on every major and most of the minor platforms. No modifications required whatsoever.Systab - Thursday, March 16, 2017 - link
What framework do you use?Cross platform development is a notoriously tricky subject for anything more complicated than a simple CRUD. And even CRUDs can be tricky.
Portable application framework -> ie Write Once Run Everywhere has been the holy grail since 1970s and of course popularized by Java in 1990s.
We are still looking ...
lmcd - Thursday, March 16, 2017 - link
"Popularized by Java in the 1990s" and Android in the 2010s, FIFY.ddriver - Friday, March 17, 2017 - link
I use Qt for the front end, pure C++ for the core logic. Qt sucks, but is still without competition, so depending on your experience with it, you might call it either "the best" or "the least sucky". For me it is the latter, I seem to have a gift for finding bugs and limitations in it, that even when reported remain unaddressed for years.bhtooefr - Saturday, March 18, 2017 - link
Although libhoudini's always an option. Not a great option, but an option nonetheless.iwod - Thursday, March 16, 2017 - link
I rather like this in a Chromebook or Laptop.zodiacfml - Thursday, March 16, 2017 - link
Quite late for a high-end since this will be competing with SD820 to Sd830s. It will be awesome for a mid-range with large batteries, phablet or tablet.serendip - Thursday, March 16, 2017 - link
It would be great for cheap Windows tablets because those haven't seen a new chip since the Z38xx two years ago. I'm a big fan of those because a lot of software still needs x86, especially Windows and Linux stuff. There's still a market for tablets cheaper and smaller than Surfaces with their pricey Core M chips.bhtooefr - Saturday, March 18, 2017 - link
This most likely isn't PC-compatible, just like the Moorefield/Merrifield and SoFIA chips before it.And do you mean Z8xxx? In any case, cheap Windows tablets are likely going to be on ARM chips from here on out. It can hypothetically run on any AArch64 platform with UEFI and ACPI support - even as low end as a Raspberry Pi 3 (although it's below the system requirements for 64-bit Windows due to insufficient RAM), if the UEFI developed for Windows 10 IoT Core for RPi 2 is ported to 64-bit.
hojnikb - Thursday, March 16, 2017 - link
Intel should go in way of ARM and just sell those x86 cores as IPsJon Tseng - Thursday, March 16, 2017 - link
On paper that should be a reasonable performer. IIRC Medfield held its own against high end QCOM SoCs back in the day (albeit without an integrated baseband). And Cherry Trail on my Surface 3 can run Skyrim perfectly well (albeit the Intel graphics obviously plays a part and the TDP will be higher).Suspect the challenge will be more on the software side (viz Geekbench results) - are Android apps set up to take advantage of four cores? Also power drain will always be an issue.
lmcd - Thursday, March 16, 2017 - link
Could also potentially be constrained by NAND controller.Smudgeous - Thursday, March 16, 2017 - link
I'd like to see this crammed into something tiny and passively cooled, like the Compulab fitlet. 8 threads would make a nifty little firewall.