Potential X86 (Non ARM) SoC Alternatives


Grench

Forum Addict!
Joined
Oct 3, 2008
Messages
6,629
There has been a lot of fear and controversy surrounding the recent presentation and consideration of the Intel X86 based Z3770 as a potential alternative for a Pandora successor.

This is the thread for that conversation.  Previous, relatively generic threads have degenerated into irrelevance.  Please, let's not do that to this or the corresponding ARM threads?

For this thread, let's discuss part availability, capabilities, performance, performance per watt for the X86 alternatives.

Later, after one or more ARM and X86 champions are found, in another thread, we can discuss the relative strengths and weaknesses of all viable potential solutions.

I am not an expert on any SoC.  I have done some research into the Intel Z3770 X86 SoC - but I don't claim expertise.  This thread does not have to be restricted to the Intel Z3000 series.  Any X86 can be discussed to determine viability.

So, X86 proponents, what's out there that we should be looking at?
 
Reading materials on the Z3770:

Intel's listing for the part:

http://ark.intel.com/products/76760

Intel's published technical datasheet:

http://www.intel.com/content/www/us/en/processors/atom/atom-z36xxx-z37xxx-datasheet-vol-1.html

What look to be media day benchmarks performed on an Intel demonstration tablet in Intel's lab:

http://us.hardware.info/reviews/4792/intel-atom-z3770-lbay-trailr-review-strong-competition-for-arm

Linux support information and speculation:

http://www.phoronix.com/scan.php?page=news_item&px=MTQ2MDI
 
Does anyone know of an X86 alternative other than the Intel Z3000 series (Z3770 flagship)? Or is the Z3770 the only non-ARM alternative that's actually viable?
 
In the x86 world there are three players: Intel, AMD, and VIA.  VIA is the one that traditionally has served this space and their Nano line is probably what one would be looking at.  However as far as I'm aware they haven't refreshed their product line to respect processes newer then 40nm yet.
 
In the x86 world there are three players: Intel, AMD, and VIA.  VIA is the one that traditionally has served this space and their Nano line is probably what one would be looking at.  However as far as I'm aware they haven't refreshed their product line to respect processes newer then 40nm yet.
VIA is coming out with a 28nm product at some point but nothing they've done or plan to do will be suitable for something like Pandora. Far from low power enough. This is intrinsic to the CPU design itself to an extent (3-way issue x86 for one thing, something neither Intel nor AMD are using in this space), but another big problem is that they're badly behind on integration. I'm not sure about their next generation stuff but their current generation still uses a separate IC for northbridge/memory controller and GPU. That's never going to fly on something of Pandora's dimensions. I don't think they were ever really designed for this space anyway, they've mainly targeted stuff like cheap all-in-ones, embedded, and maybe a netbook or two.

To give you an idea of the power consumption for current gen, look here: http://en.wikipedia.org/wiki/List_of_VIA_Nano_microprocessors The 40nm processor uses 6W for two cores at 800MHz - and that's pretty much just the processor and an FSB link. As you bump up clock speed at all it climbs a lot. 2x800MHz Nano is way too weak (maybe a little higher perf/MHz than Cortex-A15, so compare to a 900MHz-1GHz one or so), and already uses way too much power. 28nm will help but not THAT much.

There isn't that much data on the upcoming 28nm variants called CN-R - one interesting tidbit is they'll have AVX2 (but not FMA) instructions, but that just goes further to suggest they're not really targeting super low power. http://translate.google.com/translate?sl=auto&tl=en&js=n&prev=_t&hl=en&ie=UTF-8&u=http%3A%2F%2Fdiit.cz%2Fclanek%2Fvia-chysta-28nm-ctyrjadrovy-monolit I can't find any information on projected power consumption. It will still be FSB-based with an external memory controller. All signs point to this serving the same markets VIA has been.

AMD's Temash is closer to useful than VIA Nano but still not what I'd consider competitive in terms of perf/W. The only SKU that even starts to fit in a Pandora-like power budget is the 2x1GHz one, which is significantly slower than what other SoCs can give you for similar power (even when you consider AMD's higher perf/MHz vs Cortex-A15, which I consider to be around 20-25%, varying). Yeah the GPU would be on the higher end but I don't think that alone is worth it. Z3770 and even some lower bins in the Z3xxx series are way more attractive. Hell, I'd probably sooner take an up to 2GHz CloverTrail+ SoC.
 
Last edited by a moderator:
Additional information on coming uses of the Z3000 series SoCs.

http://arstechnica.com/gadgets/2013/09/intels-atom-cpus-finally-get-serious-with-the-new-bay-trail-architecture/

Not that we should give a whole lot of weight to it, but 2/3 down on page 1 it states:

"On the graphics side, Bay Trail's GPU uses the same Intel execution units (EUs) as Ivy Bridge's HD 4000, which gives them support for most of the same APIs and features—DirectX 11, OpenGL 4.0 (and OpenGL ES 3.0 on mobile, provided the drivers support it), and OpenCL 1.2 support are all present on the API side, and dual-display support for panels up to 2560×1440 (or 1080p over HDMI) as well as via Intel's Wireless Display feature will give multi-monitor fans something to be happy about."

What caught my eye was DirectX 11, OpenGL 4.0 (and OpenGL ES 3.0...). How do we confirm/deny if this SoC can do straight up OpenGL?
 
We know it can do straight-up OpenGL. But that doesn't mean it'll be available in the Linux drivers, per se. It probably will be, because it already is for other Intel GPUs using the same uarch. But who knows. If it's not enabled it's possibly something that can be fixed by a third party.

Proper OpenGL ES 3.0 support in Linux seems like more of an issue. I have no idea what the status is for Intel's Android drivers, and I have no idea how easy it is to migrate those to a more normal Linux if they're sufficiently open.

On an unrelated note, I'm earnestly hoping no one actually brings up Intel's Quark in this thread :p
 
Quark Yikes...

http://www.computerworld.com/s/article/9242931/Intel_supplies_open_source_Galileo_computer_with_Quark_chip_targets_DIY_crowd

I don't think it fits for a Pandora successor device - kind of cool in a twisted way - X86 overhead for control systems devices though... weird bit of sand there. Should be really really cheap...

Even if the Quark were $1 vs the Z3770 at $37, I don't see going to it instead.

The more I read/hear about the Z3770 makes me think that it's the 'SoC to beat' from the X86 side of things. We'll know more as the Dell, Toshiba and Sharp tablets start hitting the reviewer sites. From the specs I'm seeing, though, nobody is giving the system it's full 4GB RAM - which makes sense since they're also all using it in 32bit mode.
 
That wasn't an invitation to open a discussion about this :p I was hoping that people could reason for themselves that Quark wasn't an option at all, not merely a poor one.
 
Hypothetically...

If the Pandora successor were to use a 22nm Bay Trail Z3770 CPU, would that set it up to be ready with minimal changes to move to the 14nm Airmont in early 2015?

I.e. if the Pandora successor were designed for Bay Trail, would it be an easy road to a Pandora successor successor using Airmont?
 
Last edited by a moderator:
I've never even heard of VIA. o.o

-God Ginrai
Was Cyrix, from whom I once owned a fully working 387 math processor clone. Which got obliterated at National Semiconductor, sold to VIA which had also bought another company whose name I can't even remember and basically there's a reason you haven't heard of them which is:

They suck.

Performancewise sucking is actually applauding them: I got one of their older motherboards, released Anno Domini 2003 or so and it struggles to compete with my Pentium II dosbox for crying out loud!

This added with lousy support for just about any environment that was not supported at the time of release (i.e. Windows XP, only, try and put Windows98SE on it and you are SOL), linux "support" that is just that: "support" which only exists because probably some poor bastard bought one of the things thinking he could run Linux on it and then spent his next couple of years writing drivers himself... either that or someone had a terrible urge to reverse engineer on the PC platform and hit the jackpot!

And finally, the (Via manufactured) motherboard has isolation and general hardware quality on the rough level of a radio tower. Abandoned one. From pre WWII. I turn it on and what do I spy on my audio system... a mouse... moving... when I move it... lovely... squirksquirksquirk... click and it goes POP. Yey!

They don't need a die shrink. Their customers need a shrink.

The only time I ever saw one of those things doing something usefull was running MSDOS or derivative in an industrial PC...

Sorry if this came out as a rant... it's just that that VIA board was one of the few computers I've actually bought new at release for a full price, not used. Teached me...
 
^All this talk about VIA makes me wonder what Transmeta are up to these days... oh.. well..nvm
 
Back
Top