A dedicated backtop computer for VR - worth it?

Post by **Chriky** » Sat Jun 23, 2012 5:03 am

It seems like the easiest way to get full movement VR at the moment is with a 'backtop' computer - that is laptop that you wear on your back inside a backpack.

I was wondering if it could be cheaper/better to make a dedicated PC for this, rather than using a laptop. A laptop comes with a screen and keyboard which are not needed, and they are generally designed to be wide and flat. An ideal backtop wouldn't have any built in I/O devices, and would probably be smaller (although it could afford to be thicker as well). The main advantage would be you could upgrade the parts inside like a normal PC, bit by bit, which might be cheaper than getting a new laptop every time.

Is there anything out there that could be used to make something like that? The main problem seems to be powering it.

Here's some things I found in a quick search;
http://www.pcupgrade.co.uk/productdetai ... goryid=298
http://www.pixmania.co.uk/uk/uk/1067567 ... g-min.html
http://www.amazon.co.uk/gp/product/B002 ... B002UPWYPW

Perhaps it would work out more expensive just due to how cheaply major manufacturers can make laptops.

Post by **Okta** » Sat Jun 23, 2012 5:36 am

It depends on if you plan some experiments with movement tracking otherwise a PC with extended cables running though a sprung hook above the head should do just for turning on the spot and some slight shifting.

Post by **brantlew** » Sat Jun 23, 2012 7:27 am

If your range is limited then you can get away with using a video transmitter. But a backtop is required for true ranging. The specs on a couple of those devices (Atom CPU's) are no better than a similarly priced netbook. Given the choice, I would consider a netbook over those units. Another thing you might consider is sturdiness and mobility. Those devices are not designed to be moved around all the time so you might have to worry about cables and cards dislodging. At least with a netbook or notebook you know that it's designed to be shaken around often. I put together a quick checklist when I put together my backtop that you might find helpful.

http://www.mtbs3d.com/phpBB/viewtopic.p ... 000#p74094

Post by **Ameria** » Sat Jun 23, 2012 7:32 am

One thing I'd been thinking about, with an eye on the Rift as a target device, is the Raspbery Pi boards (http://www.raspberrypi.org/) - the GPU isn't the greatest thing ever, but it's decent enough to produce competent graphics (particulary at Rift resolution) and the board itself is both tiny and low-power, whilst still sporting HDMI output and USB ports. Of course, the price tag doesn't hurt either, making the barrier-to-entry pretty low.

Another potential advantage is that being a (mostly) open bit of hardware with a fixed design, it should be perfectly possible to hack around with the drivers/kernel/etc as necessary to reduce the latency as far as humanly possible... the closed source GPU driver being the one big fly in the ointment there, but it's still a better situation than on your average PC setup.

Post by **cybereality** » Sat Jun 23, 2012 12:01 pm

Well a laptop is designed to be mobile, so its a little safer to me moving around with it in a backback. With a custom built PC you can easily run into problems like this, not be mention the weight and size. And the nettops that are small enough will not have enough power to run any modern games, especially not in 3D. I guess you could build a powerful desktop PC and use wireless video, that's one option, but then you still need to wear a battery pack and deal with latency. Not the perfect solution. Really just getting a decent mid-to-high end laptop is probably the most hassle-free option available.

I ended up going with a Lenovo Y570, which I got for around $650. It has an Intel i5 and Nvidia 555M GPU. Honestly not that powerful, but it can run older stuff like L4D in 3D. Was able to play BF3 on all low settings, but only in 2D. And it also doesn't support HDMI 1.4a, which sucks, so I can't use it with my HMZ-T1 or GT750 (projector) unless I use side-by-side. So looking back I kind of wish I spent a little more and got a more powerful gaming laptop that supported 3D over HDMI. Guess I could still sell it and upgrade, but that's kind of annoying.

I seriously doubt you will be playing any remotely modern games on that Raspberry Pi thing, certainly not in 3D. Although if you just wanted a mobile wearable computer, then that could be an interesting use for it. Not that useful for VR though.

Post by **rfurlan** » Sat Jun 23, 2012 12:52 pm

One issue you will face is that most laptops equipped with a medium to high-end GPU will not run on batteries for too long. In some cases performance also takes a huge hit, my Alienware m14x for example throttles down the GT555m so much while running on batteries that most real-time 3D content is unusable - or rather unpleasant to use due to low framerates.

I would go for a low-latency wireless video solution instead

Post by **brantlew** » Sat Jun 23, 2012 1:12 pm

If you wear a separate battery pack, the you can run your laptop + GPU at full power for a couple of hours.

Post by **PalmerTech** » Sat Jun 23, 2012 6:58 pm

I like wireless links the best, but if you want to have full freedom, I think that using a laptop is the best way to go. If you only want it for backtop use, you can find laptops on eBay with broken screens, cracked hinges, etc for several hundred less than a fully working on.

The Rasberry Pi looks cool, but I think you are better off using an iOS or Android device. Easier to develop and share apps, built in tracking, good GPUs, and battery powered! It will be a while before you can render anything really impressive, but I think that 5-10 years from now, all HMDs are going to have the majority of their content generated onboard using mobile phone based chipsets.

Post by **profvr** » Mon Jun 25, 2012 4:35 pm

I would second using a phone. A hobby project of mine is a VR engine running on an iPhone; if you get the 4S there is sufficient GPU power for interesting demos. There's a HDMI dongle available, and I drive a Sony HMZ-T1 off mine.

Post by **pierreye** » Wed Jul 04, 2012 5:29 am

I prefer wireless HDMI as I want to play the latest PC game in 3D. Also upgraded the GPU would be less expensive in the long run compare to backtop. The only con is you are limited by the range of wireless transmitter. The best I can get with direct line of sight is around 20ft without any video degradation. If you mounted the wireless transmitter overhead, you could get around 40ft.

Post by **Chriky** » Wed Sep 26, 2012 1:56 pm

I just came across this - http://www.hardkernel.com/renewal_2011/ ... 3999328931

It's essentially the guts of a Galaxy S3 with HDMI out for $130. It should be capable of pretty good performace, stuff like this runs very smoothly on S3 - [youtube-hd]http://www.youtube.com/watch?v=nQvzNFf5900[/youtube-hd]

It's got a lot of USB slots but the problem is going to be getting drivers to work on Android or Ubuntu for peripherals.

Post by **brantlew** » Wed Sep 26, 2012 2:04 pm

Very cool device.

Post by **profvr** » Wed Sep 26, 2012 2:30 pm

Here's one of the VR systems we have in the lab. Its an iPhone driving a HMZ-T1 or Glasstron:

Only the video controller on your back (we didn't get round to making one of the mods to the HMZ-T1 to make it portable, but the Sony Glasstron comes with battery power), and you get a multi-touch hand-held controller to play with as an input device. The device is plenty powerful enough to do some interesting demos.

Post by **brantlew** » Wed Sep 26, 2012 2:41 pm

@profvr: They didn't mention it, but I hope Sony changes that ridiculous controller box for the second version of their hardware. Almost single-handedly destroys the portability of this device for these types of applications.

Post by **profvr** » Wed Sep 26, 2012 2:46 pm

@brantlew: absolutely. The controller box is a really PITA, especially since they aren't universal AC. I don't expect battery power, but an external DC converter would be wonderful.

Post by **love2scoot** » Mon Feb 11, 2013 2:39 am

Alright, I’d like to resurrect this thread. The stars are beginning to align on HMDs, trackers, 3D, games and more. With the advent of Nathan’s Half Life VR mod, the next logical step is to move to a full 360deg experience. Several approaches have been suggested in this and other threads and while overhead cable harnesses and wireless HDMI are possible options, I think the low hanging fruit is the backtop. While the backtop is most commonly associated with a backpack mounted laptop, there are additional approaches that could offer different price points and more customizable solutions. None of these solutions are complete, but these forums are great places for ideas to grow.

Requirements:
In order to integrate with a variety of peripherals and be flexible enough to run a variety of games, I think the backtop should have the following features:

-x86 CPU
-high performance GPU
-energy dense battery pack with monitoring
-Display Port or HDMI out
-USB ports
-A total system run time of >3 hours on a single charge

There are certainly other criteria, but I think the above cover the basic requirements.

Laptops:
The laptop approach offers full integration and simplicity, but at a higher price point. On the upside, we have a fully integrated platform with battery monitoring, custom cooling solutions, and drivers fully tested for the installed OS. On the downside, many laptops are still shackled with HDDs and not SSDs, which is certainly not ideal for a computer constantly in motion. Additionally we add the weight and cost of several components that are not needed for a backtop, like optical drive, keyboard, screen, etc. Also consider that laptops with higher end graphics performance are typically associated with larger screens (although exceptions exist) and therefore have a larger form factor.

Until recently, finding a laptop solution that hit on all the requirements above was not an easy task, especially at a more reasonable price point. Higher end GPUs rapidly deplete battery life, so you could play, but not for long without an outlet. Recent advanced in lithography have begun to offer hardware that provides acceptable performance at lower power levels. Taking into account price, form factor, and the above requirements, I found a couple options (although I’m sure there are a lot more):
Acer Aspire V3-571G-6622: Under $650, 4400mAh 6 Cell battery, nVidia GeForce GT 730M GPU
Acer Aspire V3-551G-X419: Under $600, 4400mAh 6 cell battery, AMD Radeon™ HD 7660G + 7670M Dual Graphics with 2GB of dedicated DDR3 VRAM

The above two options should work very well as a backtop, but are certainly expensive if used as a VR specific platform. In an ideal world, a company would offer laptop internals in a compact form factor with some flexibility for configuration. There are two areas of hardware that are emerging as possible solutions in this vein: embedded boards and nettop /small form factor PCs.

Embedded Boards:
At this point these systems are almost certainly under powered for use with VR, but there is still potential here. Take for example the recently release Gismo Boards from AMD. These boards are 4in square and pack a 1GHz APU with Radeon 6250 Graphics (probably Desna). AMD is following this up with the AMD Embedded R-Series Platform which will likely be based on Hondo, with 7000 series graphics. While the latter may be enough to play very basic 3d games, this space should see a big jump during this year’s transition to 28nm lithography. For real gaming, we need slightly higher performance hardware.

Nettop / SFF PCs:
The nettop has found a sweet spot in recent years as an ideal Home Theatre PC. With full HD video decoding engines built into even lower power GPUs, these little boxes can decode and view any media you throw at them. With more recent advances in GPU power, these little boxes are poised to give us laptop power in a lower cost and smaller form factor. Take for example the ZOTAC ZBOX nano AD12. This tiny little PC packs an E2-1800 APU, with a dual core 1.7Ghz CPU and Radeon HD 7340 Graphics, all for ~$220 barebones. Or take the Giada i35G mini PC with an Atom D2500 dual core 1.86GHz CPU with nVidia GT610 Graphics, all for ~$260 loaded. For the real deal however, we now have the Sapphire Edge VS8. This is basically an A8-4555M APU with Radeon 7600G graphics built into a nettop. There’s an excellent review of it here: http://www.kitguru.net/desktop-pc/zardo ... pc-review/. It’s also reported that consumers will have the option to purchase this system barebones, which means SSD integration could be done outright. No word yet on pricing, but this little box seems to deliver on the laptop without the laptop concept. Finally, we have the much hyped Xi3 Piston, one of the first “steamboxes” out the door. Pricing for this machine may be quite high, but we’ll see.

Caveat: battery supply and battery metering
You may notice that all the options under both embedded and nettops use barrel connectors for power. There are a number of products sold at Amazon and similar retailers that are basically big Lithium Ion battery packs that come with multiple barrel connectors for running your laptop. These products can contain a great deal more power than is available in a typical laptop battery and would help supply power to the HMD being used as well. Marry one of these battery packs with the VS8 above and we are really starting to get to a viable, customized backtop solution. Metering is a bit trickier, as tracking battery charge would involve checking the physical battery charge meter. That said, with the additional power available as part of the external battery, you may not even need to watch the meter.

Looking forward:
In January 2013, AMD released details regarding their next generation Steamroller based APUs that will replace the current Trinity line (like the A8 and A10 products mentioned above). There are expecting a 20% - 40% improvement in graphics performance over the current Trinity line.

At CES 2013, Valve’s Gabe Newell seemed to imply that the “better” category of steambox hardware would be centered around the $300 price point, and speculation is that these boxes will use AMD APUs. Essentially this would give us a new market segment of hardware (along the lines of the VS8 and Piston) that could be an excellent basis for a backtop solution.

Post by **STRZ** » Mon Feb 11, 2013 5:32 am

Ouya + Battery pack moddet into a new case could be intersting in the long run, their current Tegra3 is very weak, but they recently announced to uprade every year, so that in 2014 when the consumer Rift comes out, there would be a Tegra4 based Ouya supposedly, which is a lot more powerful matching current gen consoles, and would be a nice platform for VR indie games. People who use Unity3d and UDK, which supports the Rift, could export their games to Ouya/Android. It's very attractive because 100$ it's very inexpensive for a gaming "console", matching the philosophy of the Rift itself.

X86 Phones and especially pure mobile Linux platforms like Ubuntu Phones/Tablets could be very interesting too, because they won't be restricted to dedicated Phone software like Windows Phone, WinRT or iOS, giving you the possibility to acess the Linux Steam client. Not having the Java layer like Ouya/Android means better performance. And Unity3d already supports Linux (hoping that Epic with their UDK too in the future)

Btw, Valve have announced to open up their Steam service even more, so that in future you won't need to get through the Greenlight voting system. Every VR indie game developer could bring his stuff to Steam, and we can create dedicated Steam integrated online stores for VR compatible games.

Post by **brantlew** » Mon Feb 11, 2013 8:39 am

love2scoot wrote:many laptops are still shackled with HDDs and not SSDs, which is certainly not ideal for a computer constantly in motion

This is an absolute requirement for a backtop. Unless the entire game can stay resident in RAM, the game will stutter and often just freeze when using a HDD. Disabling drop protection can help this but I think it's a really bad idea.

Don't forget this little machine as a backtop. http://www.pcper.com/reviews/Mobile/Ori ... ing-Throne Expensive, yes - but it's got an ideal form factor and a lot of power.

I've got a lot of experience putting together backtops and I will try to post some photos soon of my system as a reference. Unfortunately, the concept of the backtop and the reality of using one are two different things. I'm not going to mince words here - backtops are a massive pain in the ass to setup, configure, and use. I am a huge advocate for untethered VR, but backtops are just a stop-gap and I think the way to go forward is either wireless video transmission or mobile platforms (smart-phone). I would love to see some small developers create VR games for the iPhone or Android platform. Being able to create a self-contained VR "helmet" with display, sensors, and CPU in a single package would offer a very simple and approachable experience - and you don't need hyper-realistic rendering to create immersion. The amount of immersion you gain by being untethered FAR outweighs the graphical fidelity you lose.

Post by **RoadKillGrill** » Mon Feb 11, 2013 12:24 pm

I have gotten the YEI 3-Space sensor working on my Ouya via usb and Bluetooth, so pretty much all models can work with Android 4 devices capable of USB hosting. And the Bluetooth works on all Android devices so far.

I also found a battery pack that just out of the box works with the Ouya accidentally, I got it as a mobile convention charge station and while unpacking it realized it can plug into the Ouya and power it.
This battery can also power the USB at the same time.
http://www.newegg.com/Product/Product.a ... 6882021228

Post by **Direlight** » Mon Feb 11, 2013 6:18 pm

I think it's bad idea, risk of severe electric shock on a non-factory sealed device. Regular laptop has battery that can explode all ready. Better just to remove laptop screen. Assuming that's safe and still boots up.

[youtube]http://www.youtube.com/watch?v=Gc2v6FbqS2Q[/youtube]

Fast & Custom laptops using SSD

http://www.powernotebooks.com/Custom-La ... t-185.html

Post by **love2scoot** » Mon Feb 11, 2013 7:13 pm

Thanks for the perspective on SSD vs HDD. So the requirements become:

-x86 CPU
-high performance GPU
-energy dense battery pack with monitoring
-Display Port or HDMI out
-USB ports
-SSD based storage
-A total system run time of >3 hours on a single charge

Don't forget this little machine as a backtop. http://www.pcper.com/reviews/Mobile/Ori ... ing-Throne Expensive, yes - but it's got an ideal form factor and a lot of power.

I remember reading that review when it came out. Mighty impressive machine, but quite pricey. I left out this guy:
http://www.newegg.com/Product/Product.a ... 6834152372. It's $1100, which I think is still beyond a reasonable budget, but it has some serious GPU muscle.

Re: mobile phones, etc. Yes, miniaturization will certainly happen over time, but games seem to be keeping up (for the most part) with advances in hardware. Lightweight games like Angry Birds and the like are certainly a growing market segment, but I assert that games leveraging VR capabilities will require higher end hardware for the foreseeable future.

Another way of looking at this might be power envelope. The Sapphire VS8 was measured in the review I linked to at ~35W during peak load. The APU inside was designed with a 19W TDP and this hardware is fast enough to run some games at a decent performance level. A 35W A10-A4600 APU + Radeon 7XXX dual graphics will soak up closer to 60W (76W if you assume the same overhead as the Sapphire system) but would give sufficient performance to run most modern games at high settings. So assuming that game engines keep up with hardware advances, can we design a system with a total power consumption between 35W and 75W that would have both a small form factor and sufficient performance for a backtop? If manufacturers can produce systems like the Sapphire VS8 and the Xi3 Piston, then I think the answer is yes.

Re: batteries
I was certainly looking for a system that could power both the nettop and 2A USB (HMD power) simultaneously. That Rosewill product looks promising.

Brantlew: I look forward to the pictures of the backtops you’ve created. Ever since I saw the untethered Skyrim video you posted, I’ve wondered as to the subtleties behind the integration of all those components. Do you think that any of the difficulties in backtops would be mitigated with the more compact form factors available in the nettop market segment?

Post by **adoral84** » Tue Feb 12, 2013 10:11 am

I have an asus G74 that I'm not actively using that I might try this out with (need to drop in an ssd though).

It's very capable for a laptop with a 560m although it's pretty huge and I think it's more around 1 hour of actually gaming play but I'll take that for the solid 60 fps at rift dev kit resolutions in most games.

Post by **love2scoot** » Thu Feb 14, 2013 5:30 pm

So assuming that the VS8 or steambox (or other SFF PC) becomes available, why drop this into a backpack? The mesh backpacks I've looked at recently look like they may be comfortable for a limited time, but would not provide any padding or distribution of the weight of the hardware over a larger area.

Here's what I'm thinking for a solution:
http://www.motorcyclegear.com/street/ar ... _the_body/

On closeout, some of these can be had for around $50. I think there are many benefits here:
* These vests are highly adjustable to accommodate several different body types.
* They are form fitting, and several offer support across the chest, waist, shoulders, and back.
* The back armor is articulating, but each piece is rigid. This would allow mounting of hardware directly to one or maybe two pieces of the armor and would distribute the load across the whole vest.
* Having hardware mounted externally would offer slightly better cooling (at least for the PC).
* Loops could be mounted in the waist straps to secure headphones and HMD while fitting the vest.

Moving from laptops to smaller form factor PCs opens up many options. Using articulated back armor as a hardware mounting platform could dramatically improve comfort and ease of use for untethered VR. Thoughts?

Post by **mAchiNE** » Tue Feb 26, 2013 11:00 pm

@love2scoot
That is a great idea for a finished system (i.e. you already have all parts you want working how you want).
A back pack allows for easy adding/removing of devices you want to test out with your setup, without any modifications to mounting setups etc you might have to do when using articulated back armor as a hardware mounting platform, so I guess both options are good for different reasons.

If the aim of all this is a great VR experience then people interested in FPS type experiences might like to use the TN Games 3rd Space Gaming vest as the platform to build your backtop mounting system on, it is also fully adjustable so can fit a range of body sizes snugly so the equipment mounted to the back (possibly in a back pack stitched onto the vest?) shouldn't move around to much and the weight should be a bit better distributed than with a normal backpack, not to mention the haptic feed back of getting shot. Will only work with a setup using an external 12V battery but any system that's designed to play for more then 30mins will probably need that anyway.

Just my 2c

Post by **love2scoot** » Wed Feb 27, 2013 1:42 am

The 3rd space gaming vest certainly does look like an appealing option, but in another thread I thought that some users were complaining about limited support & spotty performance. If it worked it would certainly add to the immersion.

In trying to build some specifics up around a couple of my requirements, I’ve spent some time trying to find hardware that would be a good match for an untethered setup (in addition to the hardware I mention above). The results have been lackluster, but it’s worth mentioning:

Arctic MC101: http://www.arctic.ac/en/p/living/office ... tml?c=2296
It’s the only device that I could find which uses an A10-4600m APU and is not a laptop. It’s very expensive at over $500 and the form factor is awfully large

Next I looked into mini ITX boards with an FM2 socket. Looking at the A10-5700 APU (65W TDP) at full CPU+GPU load puts this platform around 100W. Match this with an m350 mini ITX enclosure and a wide input pico PSU and you could manage a decent rig in a small form factor. The problem I see here is the power load is probably too great: the proportional increase in batteries would probably be too great for a comfortable setup.

I also ran across a whole bunch of creative solutions that are not appropriate for this scenario, like:
An express card -> external PCIe1x -> PCIe16x slot with external power
PCIe16x ribbon extension cables for more creative placement of a higher end video card
A lot of vaporware on external thunderbolt GPUs

Unless some PC company gets creative and releases an A10 gaming nettop with dual graphics, it looks like we’re going to have to wait until Haswell or Richland (or more likely Kaveri) until we see compact performance for untethered gaming. In the mean time maybe I’ll try to track down a Zotac Zbox or Intel NUC and try things out on HL2-VR.

Post by **colocolo** » Wed Feb 27, 2013 4:18 am

that could work but wont be very cheap. You would need a LiFePo4 battery 12V 25Ah which you can buy for around 400$ and weighsonly 4 kilo or 7,6lb THen you'll buy a voltage regulator which you can buy for cars, so that you can use 110 or 220V. This cost another 50-100$. So ultimately you would have a 3,8 Pounds battery pack with 300Wh.
If you then put a decent gaming rig onto your back you have around 16 lbs on your back. Not very lightweighted but for a hardcore gamer could be something.

Post by **love2scoot** » Wed Feb 27, 2013 10:39 am

Whoa! That sounds like it may be more involved than I was anticipating.

Looking at your numbers, you are spot on, but there’s some wiggle room here I think. Although the A10-5700 hit max at ~ 100W, typical loads were closer to 66-75W. So in order to run this system for 3 hours, we would need ~ 225Wh battery capacity or more.

My initial target battery was this:
http://www.amazon.com/Capacity-33600mAh ... ttery+pack

At 125Wh it’s really about half of what we need. That said, this battery should be sufficient for all the other hardware that I mention previously, including the VS8 which was running @ ~35W.

There are a LOT of battery options out there, and in short order I came across this:
http://www.bixnet.com/bp300.html

It is in line with what you suggested, although it uses a different lithium chemistry. At 300Wh it should provide sufficient capacity and with a discharge current of 12A, it would provide sufficient current even as it drops close to the 10V cutoff voltage.

Although this does not include onboard voltage regulation, the wide input pico PSU I mentioned above should handle this. I was thinking this would be that appropriate part:
http://www.logicsupply.com/products/m3_atx

With 125W capacity, it should provide ample capacity, and provides 6A on the 12V rail down to 11V and 5A on the 12V rail between 8V and 11V. If the battery cuts out at 10V we should be good to go here. Correct me if I’m wrong but this PSU should also eliminate the need for voltage regulation as well.

On the downside, we’ve now moved from ~$130 to ~$400, which is a big jump. From a capacity / dollar perspective, we’re a little better off with the Pink Pack of Power above:
Pink: 125Wh / $130 = 0.9615 Wh / $
BP300: 300Wh / $400 = 0.75 Wh / $

In an ideal case we could simply use that value and extrapolate how much a 225Wh capacity battery should cost:
0.9615 Wh / $ = $1.04 / Wh
1.04 x 225Wh = $234

So to run a system at a nominal 75W load for three hours it should cost ~$234 for batteries, assuming we can find something with the same value as the PPoP above.

Post by **love2scoot** » Wed Feb 27, 2013 3:39 pm

Really the ITX route is a worst case scenario. There is a tangible gap right now between the VS8 and an A10-5700 in an ITX board. The performance we are looking for here is really focused on gaming. To that point, we’re disproportionately bound by GPU vs CPU power when it comes to game performance. While APUs from AMD are certainly higher performance from an integrated GPU perspective, we need to tip the scales further.

First, let’s look at the relative performance between the A10-4600 and the A10-5800. From this review: http://semiaccurate.com/2012/11/12/lapt ... -a10-apus/
The second to last paragraph lays it out well:

On the CPU side of this equation we found that the 4600M offered about 2/3rds of the power that a 5800K could muster, and on the GPU side the 4600M fared quite a bit better where it was able to provide 4/5ths of the average frame rate of a 5800K. To put this in perspective, despite using three times less power, the A10-4600M provides at least 2/3rds of the performance that the 5800K can muster. This is really quite a feat, and it’s pretty clear what kind of chip Trinity was designed to be when you look at these numbers.

So our ITX approach, while offering higher performance, provides less performance per Watt, which in our case is a critical factor.
Now, we don’t want to rest there. While the 4600m is quite capable on its own, extending GPU performance further would give us a nice performance just for gaming. There are two options here: run a discrete GPU or go with dual graphics. The former solution should essentially eclipse in the integrated GPU and the latter should work in conjunction with the integrated GPU.

Solution 1: Discrete GPU
An interesting example of this approach is the MSI GX60 gaming laptop. Using a review at Kitguru: http://www.kitguru.net/laptops/zardon/m ... 0mhd7970m/ we can derive a couple interesting points:
Page 2 states that the laptop comes with an 82Wh battery
Page 20 shows that in gaming mode the laptop lasted 92 minutes from charge (that graph should show 1:32 not 1.32)
With some simple math:
92 mins = 1.53 hours
82Wh / 1.53 hours = ~53.48Watts system power consumption while gaming

It’s worth pointing out that in the notebookcheck.net review (http://www.notebookcheck.net/Review-MSI ... 283.0.html) of this laptop they were able to deplete the battery in exactly one hour at max load. It’s good to know the maximum discharge rate, but I’m going to follow the assumption that in a typical gaming scenario we’ll see usage closer to the 90 minutes from the kitguru review than the 60 minutes max depletion rate that notebookcheck measured.
To put some numbers to this, let’s compare the 7660G GPU present in the A10-4600M, the 7660D GPU present in the A10-5700, and the 7970m Discrete GPU. Since the 7660D GPU is not present on notebookcheck’s site, we’ll use futuremark to give us a general idea of this GPU comparison:
http://community.futuremark.com/hardwar ... 00M/review
http://community.futuremark.com/hardwar ... 700/review
http://community.futuremark.com/hardwar ... 70M/review

The 7970m provides a big performance increase over either APU alone. The 53.48W may also be conservative as the GPU can overclock itself based on its current thermal level, so a well vented chassis may prompt it to suck additional power. Rounding up a bit, even a 65W consumption is below our ITX solution. In an ideal world however, we would like to realize some of that performance with less of a jump in consumption. There are certainly lower performance (and therefore lower consumption) discrete mobile GPUs available from AMD, but another option would be use of their dual graphics technology.

Solution 2: Dual Graphics
AMD’s dual graphics solution is to keep the integrated GPU running and then to add more graphics horsepower by teaming up a mobile GPU so that both can be used in tandem. Tom’s Hardware provides an easy to read table on the options available:
http://www.tomshardware.com/gallery/05_ ... -jpg-.html

So dropping the 7660D GPU (A10-5700) from the list, we can make a three way comparison between the 7660G, the 7660G + 7670M, and the 7970.
http://www.notebookcheck.net/AMD-Radeon ... 830.0.html
http://www.notebookcheck.net/AMD-Radeon ... 173.0.html
http://www.notebookcheck.net/AMD-Radeon ... 675.0.html

We can certainly see things settling out here. The 7670M provides a serious boost in GPU horsepower to the 7660G, but notebookcheck makes sure to emphasize the stuttering problems that can be associated with poor driver support for dual graphics solutions. I’m not totally up on this, but I thought the most recent driver releases from AMD dealt with some of these dual graphics performance issues.
Looking into consumption, notebookcheck reviewed (4) laptops with this dual graphics configuration and average consumption at MAX load was 50-56W. Taking our above example of the MSI, gaming use should be less than the MAX dissipation by around 20%, putting this solution in the 40-44W category.

One thing to note, all consumption estimates on these laptops included power used by the screen. I’m going to go out on a limb here and guess that HMD power consumption will be less than your typical laptop screen. This I think offsets any level of optimistic rounding I’ve done above.

Conclusion:
Here’s our missing piece. Power wise both the discrete and dual graphics approaches fill the gap between the VS8 and the ITX solution. The dual graphics solution would fit us into that power envelope where it becomes practical to use the Pink Pack of Power I mention above. So even though we’re waiting for this GPU power to arrive with next gen IGP solutions, it should be possible to make a compact gaming nettop with hardware that is currently available.

Ok, I'm done research for awhile (whew!)

Post by **STRZ** » Thu Feb 28, 2013 5:10 am

Wouldn't a 160w. Pico PSU + 12v. battery pack work to power a low power itx mobo + discrete graphics card (max 4pin psu power input) ?

http://www.mini-box.com/picoPSU-160-XT
http://www.megabatteries.com/item_detai ... 1&uid=1318 or
http://www.megabatteries.com/item_detai ... 1&uid=1318 (haven't looked for other sources yet)
http://www.newegg.com/Product/Product.a ... 6814102980 (passive)

You'd need to change/mod the battery connector with a 4pin one to fit the Pico PSU, i did that in the past, not with a battery but with a laboratory PSU for a Audio PC project.

Then you put it all into a good ventilated backpack, without computer case of course.

Post by **love2scoot** » Thu Feb 28, 2013 9:53 am

The only issue with using those standard Pico PSUs (the ones with the yellow ATX connector) is that they expect a regulated 12v input. While those packs are listed at 12V, they will likely hit much higher and lower than that during their discharge cycle. In fact the 240WH pack states 14.5 V peak. That pack would however work with the wide input PSU I listed above or we could put voltage regulation inline.

That NiMH pack looks like it is simply wired in series. I know larger Lithium packs should have a balancing circuit for each cell but I'm not familiar enough with larger NiMH packs to know if this is a requirement for that chemistry. Also, I'm concerned that the relatively high self discharge rate inherent to the NiMH chemistry would not really fit this application as I don't expect to be using this every day.

With the original 100W estimate (peak) for the ITX board + APU, that gives us a little spare room to use for a PCI Express Card (assuming we can produce closer to the 160W peak power allowed by the yellow Pico PSUs. Looking at current draw of a standard PCIe card (borrowed from the Wikipedia article):

Power
PCI Express cards are allowed a maximum power consumption of 25 W (×1: 10 W for power-up). Low profile cards are limited to 10 W (×16 to 25 W). PCI Express Graphics 1.0 (PEG) cards may increase power (from slot) to 75 W after configuration (3.3 V/3 A + 12 V/5.5 A).[9] Optional connectors add 75 W (6-pin) and/or 150 W (8-pin) power for up to 525 W total (75 W + 3×150 W).[10]

So we would have to look into the card specifically to ensure we stayed within the power budget (especially the current it pulls from the 12V rail). Another approach would be to go with an i3-3220 or 3240 CPU with a 35W TDP, which would certainly reduce the power requirements for the ITX board portion, and then try to find a modestly powered PCIe card. In that case the flexible PCIe cable that I mentioned previously could certainly help with custom GPU card placement for optimal cooling. I'd probably still want to surface mount these components on the armor for more optimized cooling.

I'll have to start looking into the 3220/3240 route regarding consumption. This approach would probably allow better CPU performance vs the lower power APU as well. Hmmmmm......

Post by **bobv5** » Thu Feb 28, 2013 12:22 pm

NiMh cells are fine wired in series for most applications. Only if you need maximum performance or reliability would you do any thing different, for example medical equipment, satelites, that sort of stuff. The reason lithium batterys need to be treated different is because they are so unstable.

It would be possible to use some sort of smart charger with NiMh, to keep the battery permenantly topped up when not in use.

Post by **STRZ** » Thu Feb 28, 2013 1:19 pm

love2scoot wrote: I'll have to start looking into the 3220/3240 route regarding consumption. This approach would probably allow better CPU performance vs the lower power APU as well. Hmmmmm......

MSI boards tend to consume the less amount of power and often have good options for undervolting the CPU too. Could be a winner with one of those 3xxx CPU's.

Post by **love2scoot** » Thu Feb 28, 2013 4:12 pm

Thanks for the info on the NiMH cells, that's good to know.

I'll look into the MSI boards, the undervoltage option is a great idea as well. With some tweaking we may be able to squeeze this all into a nice low power envelope

Post by **STRZ** » Thu Feb 28, 2013 6:59 pm

It's possible to flash a modded bios on some Radeon cards (mostly reference models), and this way undervolt it permanently. Otherwise within Windows there's the MSI Afterburner tool where you can undervolt GPU's.

I hope that it's still possible with the upcoming 8xxx models, they should consume a lot less. It could be possible to get something like a 7850 or even 7870 performance class as comparable 8xxx model fitting it into the pico PSU limits.

Nvidia Maxwell cards are rumored to double the power efficiency per watt to kepler, but those won't appear until 2014.

Lurking around in some german power saving enthusiast communities this board seems to be king atm: http://www.asrock.com/mb/overview.de.asp?Model=B75M-ITX

Has also very nice Undervolting options it seems.

Post by **love2scoot** » Thu Feb 28, 2013 9:22 pm

How did I not hear about the thin miniITX form factor?! After some research on silent system sites (mostly silentpcreview.com) I ran across some interesting information. It appears there is a standard called thin miniITX. This standard, along with specifying the exact location of the CPU socket, mandates a 19V DC input onboard for power. Looking back to earlier in this discussion this connector should work out of the box with the Pink Pack of Power we found at Amazon. This simplifies things quite a bit power-wise.

These boards have onboard mSATA ports but also appear to be minimalist on features, and therefore appear to have a smaller power draw compared to other ITX systems. User piglover on silentpcreview detailed a build using Intel’s most recent thin miniITX motherboard, the DQ77KB. Using a Kill-a-watt he measured 17W idle and 57W peak with Prime95 maxing things out on an i5-3450T. Here’s the link:

http://www.silentpcreview.com/forums/vi ... 23#p568823

STRZ: The B75TM-ITX certainly does look like the best option on the market right now. I looked through thin mini ITX offerings from Intel, Gigabyte, ASRock, and ECS. The only one that really caught my eye (but appears to out of stock due to its age) was the Gigabyte MSH61QI: http://b2b.gigabyte.com/products/produc ... id=4115#dl. This is a rare bird indeed: a thin miniITX MB with all the standard features + a full size PCIe x16 slot. All current thin miniITX MBs appear to either have no dedicated PCIe slot or a 4x slot (despite having chipset support for all the lanes). To compensate for this, I’m thinking this adapter might come in handy:

http://www.ebay.com/itm/PCI-E-Express-1 ... 27c03fad92

I know that video cards will definitely work with fewer lanes attached, so a couple questions come up:
1) Will this adapter affect signal integrity to the point where stability becomes a concern?
2) Do we have enough power headroom to run a dedicated video card?

While the first one will have to wait for testing, the second is a bit easier to track down. Assuming that we go with a 35W Ivy Bridge CPU (like the one piglover used above) our total system draw should peak around 57W. Just as before, we’re going to assume that gaming will not put as large a power burden on the CPU as does Prime95 and we’ll estimate down to ~45W. Looking into the manual for the B75TM-ITX, it suggests an external 19V power supply with a capacity of 150W. Most of the pull from a video card will be on the 12V rail, so even rounding down we still have plenty of power that can be delivered by the board.

I haven’t done a deep dive into research on Video cards, but what we probably need to look for is something in the ~20W arena for the discrete GPU. Just as an example, the Powercolor AX7750 Go Green card (http://powercolor.com/us/products_features.asp?id=413) should draw ~40W peak. The sweet ticket would be to find a mobile GPU that someone has mounted to a PCIe card.

Assuming we can find a card of ~20W with decent performance, our total system load should be ~65W while gaming. Regrettably this would only give us ~2 hours of play time on a single PPoP battery pack. Comparing this with our previous solutions:

-All of the parts are available now
-It’s cheaper and more available than any of the other solutions
-It has the potential to produce very high performance especially because the CPU outclasses the performance of the AMD APU in other solutions
-With the onboard power and SSD it simplifies the system design a bit, alleviating the need for a PicoPSU and 2.5” SSD mounting
-It has more power draw than the mobile APU solutions proposed, but certainly brings down the power requirements for a full sized ITX solution.

This is good progress!

Post by **STRZ** » Fri Mar 01, 2013 5:38 am

Good mobile GPU cards (MXM modules) are expensive as hell http://stores.ebay.com/upgrademonkey/Mo ... 3098447011

afaik, you can buy adapters to interface those with a desktop board. Honestly, i'd prefer to undervolt and underclock a normal card like a 7750 until you're in the desired power consumption range. ~20w. seems very low though, it could be better to skimp on the CPU instead and go with a AMD Bobcat Board with integrated CPU and add a 7750 to it.

Post by **love2scoot** » Fri Mar 01, 2013 12:41 pm

I actually looked into boards and CPU for the mobile variant of the i3 but at twice the price, this didn't seem like the route to pursue. The bobcat route might work, but we're already starting to get CPU bound with the Trinity APUs. Although, AMD just announced the other day that they will be making a consumer variant of the APU in the PS4, but no word yet on price or availability.

Wow those prices for mobile GPUs are crazy! I think you're right, the underclock/volt route might be the only way forward to drop consumption for dedicated GPUs.

Post by **STRZ** » Sat Mar 02, 2013 12:35 pm

What do you think about hardshell backpacks like those? http://www.youtube.com/watch?v=tKBpVzvd5OE

It could be possible to modify it with a fan on the back, cut some ventilation holes on the bottom and integrate a mini itx system + battery pack into it, more or less like a backtop computer case.

Post by **cybereality** » Sat Mar 02, 2013 2:39 pm

I used a mesh backpack for my backtop setup and it worked pretty well.

I think the most time I spent on it at once was about an hour and I didn't have any problems.

Post by **nanicoar** » Sun Mar 03, 2013 1:22 pm

Some thoughts;

You can remove the LCD panel somewhat easily on most PC laptops. They should boot without their panel just fine, but the antennas are usually integrated into the LCD assy so you might have rogue cables sticking out. You are unlikely to miss its cheap TN panel anyway.

You'll definitely get more bang for your buck with a laptop than with more rare solutions. Hang on to the padding in the packaging it came in; that way you have solved the air circulation and protection problem without any effort worth mentioning.

ED: There even was a video. How could I have missed that?