Intel is releasing their Windows tablet variant of Medfield today code named Clover Trail. We are not sure what to make of it because for the most part, Intel refuses to say anything specific.
The silicon itself is a variant of Penwell, the phone chip Intel release in January. The differences between the two are a second core, four threads total, and the GPU is now an Imagination SGX545, upgraded from the older SGX540. The video encoders and decoders have been upgraded in ways that Intel will not talk about, but they now support Microsoft specific formats that the rest of the industry shuns. There is also a power management controller that is Microsoft specific, but Intel won’t even admit it, much less say anything about the functionality.
How fast is the chip? 1.8GHz max turbo with a TDP of 1.7W, but Intel has been seriously downplaying that number. Why? They want to talk about the lowest power mode, and it will be interesting to see how many sites blindly parrot back that absurdly low but otherwise meaningless number. On a more humorous note, Intel is on its third generation Atom DX10 hardware, but still can’t get it working, Clover Trail is still DX9.x. SIGH. The competition is all DX10 or DX11, and has been for years.
We would love to say more about the chip, but Intel refuses to say anything more technical than what we put above. Really. They had a technical press and analyst briefing at IDF, and the entire talk was about how wonderful Windows 8 was, less that 1/4 of the slides mentioned an Intel chip, much less said anything useful. Sad doesn’t begin to describe that debacle.
Luckily, there were lots of pictures of the chip to show off, you know how Intel is proud of their silicon manufacturing capabilities. Just kidding, Intel refused to show the chip in their briefing, afterwards, and in a Twilight Zone turn of events, would not even hand out a press picture to print. Really? Is this what the company has sunk to? Are they that afraid the financial press will realize they can’t compete with ARM on anything much less die area? To put a humorous twist on matters, Intel did hand out a rendering of the chip, basically a 3D model not to scale, but we won’t waste the bandwidth on it.
So, once again, Intel won’t talk about silicon, but is really proud of their silicon. They are better than the competition, but won’t say how or why. They are better suited for the job, but won’t tell you anything specific. Anyone get the idea that this chip abjectly sucks? If you did get that impression, you would be right.
Why? It is huge. Seriously, it is massive. The little brother Penwell is 82mm^2, bigger than most of its ARM competition and on par with a full Core iSomethingmeaningless. Unfortunately, it is also larger than its ARM competition while being notably lower performing. To compound the problem, Intel’s Atom parts are on a 32nm process, the competition is on 40nm. Intel is larger, less efficient, immensely more costly to make, and incompatible with everything that matters.
Back to Clover Trail, this joke of a chip is bigger yet. Double the core count and massively increase the GPU size, and what do you end up with? Increasing the size of the two largest units on the chip may not double the area, but it won’t be far off that mark. Clover Trail is massive, so big that it is not economically viable in the markets that they are fighting for. That promise from the last financial analyst day of good margins on Atom in the phone/tablet market would go up in smoke if Intel released the die size, and they know it. Intel can’t compete in tablets and phones.
To make matters worse for this chip, it is running Windows 8. If you recall, the little brother Penwell ran Android at a snappy pace with half the cores, likely half the GPU power, and far less memory. Clover Trail runs Windows 8 at barely tolerable speeds even with the massive hardware hacks Intel put in place to fake performance. Bloat the die size, add in vastly more DRAM and storage because Windows needs at least 10x what Android does, and suck far more power to do so, and this is somehow a viable product? You need a bigger battery to simply attain parity on power driving up the BoM cost yet more.
Then there is the software costs. Microsoft threw Intel under a bus with WART, if you buy the ARM version of Windows for about 2x the cost, you get Office for ‘free’. If you buy the x86 version of Windows 8, you get the OS for a little less than half the price, but without office. You then have to buy the full version of Office to put on it for $125 and up, way up if you want Outlook.
That puts any Clover Trail machine at $170 for the software alone before the bloated hardware costs to store and run everything. How big is the Windows base install again? How big is the Office install again? How big is an Android install with an office suite? How much does Google Docs add to the mix? What is the cost on the added hardware and storage? You can buy a full Nexus 7 for $30 more than what Microsoft gets for the software on a Clover Trail tablet, and that is before the added hardware costs. The Nexus works better, has better battery life, and is not a security nightmare either.
This chip is simply not competitive, and Intel knows it. Microsoft knows it too, if you recall, they announced two tablets called Surface. These were designed based on essentially unlimited budgets, numbers that no OEM would sanely consider if the line was ever meant to be profitable. Given the best of everything, what did Microsoft choose? Not Clover Trail, not even any Atom, they picked a Core iSomethingmeaningless for the x86 surface, that should be a red flag to onlookers.
But it gets far worse. This chip is a locked down nightmare. It has been said that Intel is locking out Linux, and all other OSes for that matter. Intel denied it, and said it was a driver issue and the like. They are being flat out dishonest, Intel representatives have directly told SemiAccurate that this SoC is Windows only. There are Microsoft specific hardware features, and there never will be drivers released for any other OS period. Nor is there any intention to do so. There will be other CPUs with similar code names that may someday get non-Windows drivers, but not this one.
To make matters worse, it will have a locked bootloader. Microsoft controls what you can put on this device, what drivers you can load, and what software you can load. Intel could have stopped this monstrosity, but didn’t even tepidly condemn it. Internally, they know it will fail, SemiAccurate has seen their sales estimates last summer, but won’t do anything to make the devices bearing it even marginally palatable. Intel does not want to sell Clover Trials because of what it will do to their margins.
You can see this in their policies. How do you make a product successful? You get people to develop for it. You get it out there and hope the next big thing that comes out of nowhere has your chip in it. It is a standard x86 CPU with a standard Imagination GPU, so this is a good start. You just buy a development board, dive in to the documentation for a weekend, and start banging out code. With Clover Trail, there are no development boards, nor is there any way to actually get one without a multinational corporation behind you. There is no documentation either without the same backing. Neither are going to change, like we said, Intel does not want to sell Clover Trail, they are desperately afraid of it.
In the end, Clover Trail is a dog. It is tied to a millstone OS that bloats the software cost, the BoM cost, and power requirements. It has neither the CPU performance of the smaller die Core iSomethingmeaningless nor the power savings of ARM CPUs. It is vastly more expensive to manufacture, and no one wants it. There is no up side for Clover Trail, and that explains Intel’s silence perfectly.S|A
Latest posts by Charlie Demerjian (see all)
- SemiAccurate has Skylake-SP die shots, sizes, and more - Jul 21, 2017
- Skylake-SP has a diverged core - Jul 19, 2017
- Intel’s Purley platform architecture is a step forward - Jul 13, 2017
- Intel launches Purley aka metal Xeons - Jul 11, 2017
- AMD’s Epyc has lots of connections - Jun 26, 2017