ARMdroid

[Uvaman]

I apologize if am placing this in the wrong place.
Just wanted to give general feedback on the website.
I think its top of the line, I genuinely believe the reviews here are industry's best.
HardOCP is first by a mile in my PC info surfing.
I know the focus is x86, but wouldn't it be time to expand into Android in a enthusiast way?
When I was shopping for my android mini-PC I had to rely on people's 'reviews', this is not as reliable.
I know the focus with Android/ARM is phones, and there is tons of (mostly poor) reviews on those.. but how about the mini-pcs one can get with ARMdroid on them, I know these are mostly from China with the Rockchip stuff, but even the ones I have which are pretty much low end are cool.
Maybe pick-up some higher ends and do some reviews on those, tips and tricks.
Don't get me wrong, I can see the hard-work that the reviews are, I would bet the people behind it would not settle for crap if reviewing some of the higher power Android boxes.. but I think it might be time to throw some in the computing-review mix.
In a way might be a pre-amble to AMD's ARM servers and such, get some real reviewing/benching methodology nailed down early in the process.
I do think the ARM tech is here to stay, to scale up and up.

 

[Red Falcon]

I hear you on this, and while ARM processors and platforms are becoming more mainstream, the devices you are speaking about are hardly mainstream.
While the Raspberry Pi is a great learning tool, the CPU in it is basically a 2003 design, and was intended for very low power projects and/or a programming learning tool.

The other devices you speak of, basically ARM development platforms and single board computers (SBC), are really neat, but again, not very mainstream.
A nearly complete list can be found here: http://en.wikipedia.org/wiki/Comparison_of_single-board_computers

Many of these can run Droid/Linux, or other flavors of GNU/Linux (armel/armhf) and BSD.
I too would love to see [H]ard make some real reviews on these devices, but I don't think they really have an audience to make it worth their while, at least not as of this posting.

It is kind of the same reason we never saw any reviews of PowerPC-based equipment, even though that was the ISA which stood toe-to-toe with x86 throughout the mid-1990s to the mid-2000s, as small of a ratio as it was in terms of competition.
As far as "enthusiast" ARM systems go, there aren't many.

The 32-bit A15 is about as high-end as one can go with development boards.
I haven't seen any A17 SBCs out, and the A53 and A57 CPUs are even tougher to find outside of tablets.

Your heart is in the right place, but in my opinion, the market is too fragmented to make sense of any system in comparison.
The best thing I can recommend to you, and it is a synthetic benchmark site, so take it with a grain of salt, are these:

http://www.androidbenchmark.net/
http://www.iphonebenchmark.net/

 

[defaultluser]

I do think the ARM tech is here to stay, to scale up and up.

The problem is buddy, you're the only one

ARM Linux USB/single-board computers (let's be realistic here, nobody runs Android on these things) made some headway because they were dirt-cheap and offered many of the same features as a fully-fledged HTPC for a fraction of the cost. They made good project computers for people who have more time than money, and didn't mind a little customization of Linux to add the missing ARM features they desire.

But the reality is: once Bay Trail was released, these systems were dead-in-the-water. High-performance with x86-compatibility means no-more compromises, and that combined with Intel's great Ivy Bridge-class video support (Linux display drivers are often a nightmare on those cheap ARM chips) means there's no reason for Linux people to bother tinkering with ARM anymore. And the $90 Linux price-point is only a little higher than those cheap Chinese sticks.

Oh yeah, did we mention Windows being given away for free on low-end devices now? That combined with WIMBoot significantly reducing the size required by Windows installs makes them suddenly palatable on 32GB systems like this.

Here's a great example: my buddy is using the Pi to build a MAME cabinet, and even he admitted to me that he didn't like it. There's not enough processing power for late 80/early 90s titles, and since there's not a massive library of software available, you have to take what you can get (for one, he hates the front-end of MAME for Pi,). But because he is hard-headed, instead of spending a little more for an x86 stick he's just going to "make it work." Meanwhile, he spent more on the monitor than he would on the difference between an ARM and x86 project board, so cost is definitely not an excuse.

Most people don't have the patience or tenacity to do that sort of thing, and that's why ARM USB sticks/single-board computers are dead to all except the really hardcore. And most of us on the [H] would rather spend time tweaking gaming PCs than tweaking/writing kernel modules.

 

[Red Falcon]

^ I have to agree, those are all valid points.
While I would have liked ARM to take off, it just isn't there outside of smartphones and tablets for mainstream use, even for tech-level mainstream use.

It is a very specialized/niche area of technology, and yes, even the [H]ardest individuals tend to stick with x86, and for good reasons: support, wide distribution, legacy programs, compatibility, price/performance/TDP, compatible hardware, etc.
ARM has almost none of that outside of smartphones and tablets, and again, everything is very custom.

While these systems are a lot of fun, I have personally pushed them to their limits, and let me tell you, they don't hold a candle to even very low-end x86 systems such as Atoms or even Celeron-J platforms.
As for the ARM-based servers, they are all but dead.

AMD, amazingly, is still rolling with their "Opteron" ARM servers, but they are highly proprietary, super expensive, and honestly, one can get better performance from an x86 Xeon or Opteron system by a wide margin for a fraction of the cost (not even counting licensing fees).
Calxeda servers have been dead since late 2013, so ARM 64-bit on servers is looking less and less feasible and realistic at this point, sadly.

x86, and primarily Intel, is dominating in this area.
Had the current ARM processors been out a decade ago, they would have had a chance, but not in this era of x86.

The bottom line is, ARM is unfeasible and unrealistic outside of mobile platforms, and impractical outside of the very low-power platform arena.
As much as I like my ODROID-U3, I would never consider using it in a full production environment with even just ten people; even the highest-end models would barely achieve this, barely.

 

[Uvaman]

I don't know, I do see the points of being fragmented, and all that.. but I also think, if ARM doesn't make it out of mobile in a big way is because the focus on the software development never happened (I mean more Google I guess. windows RT for now flopped along with windows 8 in general), not because of the computing performance.
I have used my 2 chewing gum boxes more and more, its really surprising all they can do, and these are quad-core lower end rockchip, I think it a lower end mali chip, its not even a nicer 4k-capable one.

If the software development happens, I can see see ARM scaling up and everywhere.. think about it the play store is enough of a reason to consider you next desktop be a replacement of windows. If Google moves their butt and makes some real software, by that I mean a real 'office suite' (nothing comes close to MS, it just doesn't) and some real photo software, do business with adobe cloud or whatever ect as well as some real considerations for keyboard/mouse.
I don't know much about Atom or how it compares with rockchip/mali.
It might be that all these things hardware-wise are moving in China/ maybe even Russia, not so much here.
I don't know, just talking crazy.. I do understand how reviewing these boxes might be a waste of time for HARDOCP, and I realize its a significant investment to do a proper review (kind of why I would like to see them here)

 

[defaultluser]

windows RT for now flopped along with windows 8 in general), not because of the computing performance.

Actually, Windows RT and the original Surface partially flopped because it was launched on an ancient, outdated Tegra 3 1.3 GHz platform. This is comparable to the Rockchip quad-core A7 chips you adore.

The original Surface could navigate the Windows UI smoothly and render basic webpages without a problem (as you would expect on a Tablet), but when you ask it to do anything even remotely difficult like typing a Word document, the processor is not fast enough to keep-up with Word's real-time pagination and flow rendering, and spellcheck:

http://www.anandtech.com/show/6385/microsoft-surface-review/10

It turns-out that something as "simple" as Word can actually be quite complex, and trying to wedge-in a processor that was slower than the already glacial Intel Atom on the desktop just felt misplaced.

Oh, did I mention that the Cortex A9 was outpaced by the original Intel Atom, an ancient x86 core design from 2008 targeted toward low-power, low-cost computing, based on a processor released in 1993? Yup, that's a whole lot of processing power you got there in your Rockchip...until you actually try to do something interactive with it!

I don't know much about Atom or how it compares with rockchip/mali.

Well, now you do!

Those same Atom chips that ran circles around the 1.3 GHz Cortex A9 were replaced one year later by Bay Trail, which performs usually over 2x faster than the previous models. Now the performance of Atom is good enough to be an x86 desktop replacement (which has fueled the market of Windows convertibles for under $300, competing directly with ARM tablets), and although you can buy comparable-performance ARM A15 cores, WHY WOULD YOU? There's no pressing need for more processing power on Android.

 

[Red Falcon]

^ Very well written.
I agree with the performance explanations, especially when compared to Atom CPUs.

ARM has its place in the market, but for anything serious, it just isn't enough.
This is the exact reason why it isn't taking off in the server market; these CPUs just cannot keep up.

Even by the time many of them are in use (SMP/clustered), the power-efficiency advantage is lost, and a single Xeon E5 can outperform all of them (depending on how many are used) at a fraction of the heat output, power consumption, and physical footprint, not to mention superior IPC.
ARM is great for low-end mobile and ultra mobile devices, and custom/hobbyist projects, but certainly nothing for serious usage where processing power matters.

I should also add, that all ARM CPUs below the A9 are in-order-execution, which severely hampers performance.
Only the A9 and above feature out-of-order execution.

 

[Uvaman]

Thank you all for the informative answers.
I am only an end-user of PCs and devices so I don't go too far chip performance problems/capacities.
Crap, I hope the ARM servers from AMD are not yet another nail on that coffin.
Freaking AMD's coffin is 90% nails by now. (I do hope they make it out of the hole they been in.)

 

[westrock2000]

I am very impressed with cost/performance ratio of Bay-Trail. Intel did a damn fine job on this one. Even the built-in graphics are descent, for desktop duties at least.

 

[Red Falcon]

I am very impressed with cost/performance ratio of Bay-Trail. Intel did a damn fine job on this one. Even the built-in graphics are descent, for desktop duties at least.

For mobile, it is very efficient, but for the desktop?
Honestly, for the cost of it, on the desktop, it just isn't worth it, as there are better x86 options available which have better performance and lower costs.

 

[defaultluser]

For mobile, it is very efficient, but for the desktop?
Honestly, for the cost of it, on the desktop, it just isn't worth it, as there are better x86 options available which have better performance and lower costs.

I think what he means to say is, the GRAPHICS are capable of running a basic 1080p desktop, and for that he's right. You'd be hard-pressed to tell the difference between a powerful Haswell processor and a 2.5 GHz Bay Trail (rest of the system identical) if all you were doing was web browsing, video streaming and email. This is in stark contrast to the Clover Trail core that preceded it, which had piss-poor graphics and slow single-threaded performance.

I do agree that you get a lot more value and flexibility if you go Haswell on the desktop, and even more value if you want to build a small server (as Bay Trail is limited to 4x pcie (usually 2 free), 2x 3Gbps SATA and just one USB 3 port). Considering you can get B85 motherboard plus Haswell Celeron for $100, the Bay Trail integrated boards starting at $70 are not good value.