Slashdot Mirror

Not sure if I get what you are asking. But I'll try:

AMD: http://developer.amd.com/resou...
NVidia: https://developer.nvidia.com/n...
https://developer.nvidia.com/o...

The thing is...that market that the WiiU sits in has Android/iOS encroaching on the bottom end (yes, there ARE titles in the Nintendo market space from them) and Sony/Microsoft on the high end. The Wii kept them afloat because of a unique UI model that worked well for the titles they were interested in having on the platform.

That market picture needed something OTHER than what they fielded with the WiiU to keep it going.

One of the things you missed is things like this:

The MadCatz M.O.J.O.

There's something else on the horizon as well- something that may well change the landscape quite a bit.

The NVidia Tegra TK1

It has 4 Cortex A-15 CPUs. It has 192 Kepler Shader cores. (Hint: This is roughly like having a mobile version of the GT630...)

While the titles are "thin" there's enough compelling to have people more than interested in things. Why have a console/mobile console that's limited in what you can do with it when you can have the full monty- being able to read e-books on it, being able to really surf the web on it, being able to play Hulu on it or your own movies if you've got a media server...you get the idea.

You're right, they're not the same market. But the market for the former is MUCH larger than the latter, sir. Nintendo's suffering for missing the mark there.

Old hardware is your best bet. Anything new would be unsupported by the older 2.2/2.4 kernels, PCIe, SATA, chipsets etc.
*Slot 2 Pentium II or III CPU's and Socket 370 CPU's are perfect. If you want multiprocessor, a Tyan or Supermicro dual slot/board is a good bet but stay away from any board with RDRAM using the i820 or i840 chipsets. They did however realize how big a mistake RDRAM was and Intel made SDRAM->RDRAM bridge chips so those chipsets could use PC-100/133 SDRAM. Tyan made a dual processor i840 board with dual slot 1 and SDRAM using the bridge chips.
*At least 256 meg of ram, 512MB - 1GB is ideal. Make sure your board supports the RAM you have.
*An AGP Riva TNT card or better yet, a Geforce 1, 2 or 3 graphics card. 3D support may not be available*
*Sound Blaster Live!, Ensoniq, Turtle Beach or Aureal sound cards should all work. Though the Sound Blaster Live! is probably your best bet.
*You are also going to need an ATA hard disk (2+GB) and CD/DVD rom drive, I am unaware of any P2/3 board that supported USB booting so you need the optical drive.
*If no onboard LAN card is present (most common scenario) you want a PCI 3Com 3c905B/C, or any PCI card based on the DEC Tulip chipset (21040/21041/21140/21142/21143). Many older Netgear FA311 cards also worked flawlessly, based on a well supported National semi chip that I think was a tulip clone)
*Bonus: decent 19"+ Trinitron CRT monitor. I still have a 21" Sun Trinitron.

Stay away from ISA cards as much as you can. I had a hell of a time getting my old ISA Sound Blaster AWE 64 Gold sound card running under Mandrake back in the day. And that was a "plug and play" card without jumpers. As for why to use Pentium 2/3 boards and not a pentium 4, the p4's after socket 432 willamette generation might not run a 2.2 or early 2.4 kernel. Socket 478 gained things like SATA and PCIe so its a crap shoot. Pentium 2/3 is a guarantee.

*Nvidia hardware 3D support does not appear to be supported on 2.2 kernels. I checked the README for the oldest Linux Driver and 2.4 and 2.6 kernels were mentioned. Have a look here: http://www.nvidia.com/object/linux-display-ia32-71.86.15-driver.html and check the hardware issues section in the README!

Have fun kickin it old school.

Correct, but if you're familiar with the low-end steam machines, they're meant to be used to stream the games from a separate system with beefier specs. My assumption is that this will be competing with nVidia's Shield console.

Not necessarily. I mean, the streaming feature will probably work great, but the developers seem to also want to make it a true portable machine.

FTA: "SteamBoy is the first device that allows to play Steam games on the go," SteamBoy Machine's press release reads, "you will keep playing your favorite games at the bus, the office, the school or the doctor's waiting room."

Correct, but if you're familiar with the low-end steam machines, they're meant to be used to stream the games from a separate system with beefier specs. My assumption is that this will be competing with nVidia's Shield console.

It's been all this time and multi-monitor (sorry, "TwinView") and SLI is still not supported? I'm starting to give up hope that it ever will be. My options are to either get a new, single, video card; or reconfigure X and lose my extra monitors while I'm in-game, and then do it again when I'm done; or just not use SLI at all. It was supported on Windows Vista, for crying out loud.

http://la.nvidia.com/object/nz...

"The Way It's Meant to be Played"
Nvidia pays you shitload of money for participating in this program, and can additionally guarantee certain sale goals (by bundling your product with their GPUs).
In order to participate you only have to do two things, insert nvidia ad clip at the start of the game, and let nvidia rape your codebase.

On paper Nvidia pays you for joint marketing campaign, but deep down in the paperwork you are letting them decide what your codebase will look like.

Oh look, what a coincident: Watch dogs, game that is crippled on AMD, is a participant
http://www.geforce.com/getwatc...

ftp://download.nvidia.com/XFre... This is an example, notice where it says run MAKE and MAKE INSTALL. And you had to do it multiple times, in different folders... and half the time it broke. And you didn't know why. Absolute madness.

Not even comparable, mantle is an alternative renderer.

No it isn't, Mantle is a graphics API and for the game developers who are writing for it how do you expect nVidia to support that code path without a spec or reference implementation?

Mantle might be closed but it doesn't keep nVidia from looking at the games source.

Neither does GameWorks, just because you call into GameWorks libraries doesn't mean you need to distribute the code for those libraries. For example they even have samples that you can download the source code for that makes use of the GameWorks libraries.

I wouldn't say more so. https://developer.nvidia.com/c... Gaze on the list of proprietary technologies and libraries that lock a developer in from start to finish... Mantle is just an API like any other that has to be supported alongside directx, opengl or whatever.

What version of Vmware View? Doesn't the vSGA scaling depend on which 'profile' you use?

See here: http://www.nvidia.com/object/v...

That link is referencing a citrix only idea, but their general distinctions between user types is apt. As of View 5.3, there is no longer a lockin to NVIDIA products yet nobody has made any yet to my knowledge. Intel & AMD are on the list to produce something sometime.

This might be related to the Nvidia Jetson TK1 that will be released pretty soon.

This isn't particularly new. It's nice tech, but each ~$2000 K1 board supports 4 users. 4. The K2 board supports 2 'power users'. (ref: NVIDIA data sheet: http://www.nvidia.com/content/... )

If I cram 4 K1 boards in a server, I can now support 16 virtual desktops with 3D acceleration for an $8k delta over and above the other expenses of VDI.

Unless you ABSOLUTELY MUST have VDI for 3D workloads, I can't see how this makes sense.

Please read more than a single fucking PDF before speaking in absolutes.

http://www.nvidia.com/object/virtual-gpus.html

This isn't particularly new. It's nice tech, but each ~$2000 K1 board supports 4 users. 4. The K2 board supports 2 'power users'. (ref: NVIDIA data sheet: http://www.nvidia.com/content/... )

If I cram 4 K1 boards in a server, I can now support 16 virtual desktops with 3D acceleration for an $8k delta over and above the other expenses of VDI.

Unless you ABSOLUTELY MUST have VDI for 3D workloads, I can't see how this makes sense.

What PCIe cards are you plugging in again? Graphics cards? You still have yet to demonstrate that it is not a novelty. I have never seen a CAD setup like that. Nor have I heard of a gaming rig that uses a laptop CPU but has an external graphics box. Maybe you're right and it will be all the rage in CAD houses.

What devices are these? Still graphics cards?

http://www.red.com/store/produ...
http://www.blackmagicdesign.co...
http://www.nvidia.com/object/q...
http://eshop.macsales.com/item...
http://www.amazon.com/Apple-Du...

I could go on but really the answer is "Every single PCI-E card that exists." Or "Every single PCI-E card that is important to professional users that just because you don't know about doesn't mean it doesn't exist."

Really, considering the quality of drivers out of nvidia for the last year I'm glad I switched to ATI. I think it started around the nvidia 302.xx series, where the mass lockups began and the nvidia forums(before they were hacked) that had the 480k post thread with 1m+ views for TDR's. Then it was the crashing with firefox, that lasted from the 302's right up to the 320's. It only got worse about the time the 310's or 315's rolled around and the drivers were causing hardlocks across all 400,500,600 series cards. And I think it was right around the 308's where the complaints got so bad that nvidia was willing to pay shipping costs for anyone in the continental US to have their rigs sent to California so they could try to find out why the TDR problem was so rampant.

I haven't heard anything good on the state of nvidia drivers, if I have a complaint about ATI drivers is that some programs are bit more sluggish compared to my nvidia card, but I'll take the stability over the TDR, TDR, TDR, TDR, TDR, TDR. And sadly it wasn't one card(had a 400, and two 560 series cards), and one configuration, or even one power supply or a particular CPU in my case. It was across AMD, Intel, various ram speeds, paired, non-paired, different PSU's, and machines in more than one physical location.

My general policy has been to flip-flop every generation and go nvidia to ati and back again. But the last series of drivers pissed me off to no end that I dumped them for ATI, and Matrox didn't go anywhere they're still making video cards only on the business end though. The problem of course is much like the CPU business right? Remember the days of Cyrix, AMD, Intel? Well it was a case of hardware pushing so fast that not all of the companies could keep up. Same deal happened in the videocard market.

NVIDIA definitely write their own OSX drivers. I'm pretty sure AMD/ATI and Intel write their own OSX drivers too but these days GPU drivers are usually delivered with operating system updates (in a similar way that you can get driver updates through Windows update). Given how squeezing out GPU hardware documentation for Linux has been tough I don't think NVIDIA/AMD would be keen to help someone else write drivers that unlocked full functionality...

46.3.1 Implementing Odd-Even Merge Sort http://http.developer.nvidia.c...

There's a promo for what it's supposed to do from nVidia here. The short of it is that they're trying to replicate what pressure-sensitive active styluses do without requiring you to actually have a pressure-sensitive stylus. Instead it seems to use some kind of pattern-recognition on the input signatures from the passive stylus to figure out what you're intending to do, and does things like vary stroke width with pressure, or treat the back side of a stylus as an eraser, etc. Cool if it works: if you can replicate a more expensive hardware stylus in software, go ahead. But does it work reliably?

I did not look at the actual numbers claimed nor what they are. But a factor of 7 between a GPU and a 32 core intel system is not impossible. My BS alarm trip around a factor of 20 for a 2 processor system.

If you look at state of the art nvidia GPU, you pick a tesla K10, ( http://www.nvidia.com/object/tesla-servers.html ). You get about 4T.5flop/s single precision of performance and a bandwidth of 320GB/s. The flop is realistic for compute intensive (read dense mat mul) and the bandwidth is never reached. Probably 250GB/s is more reasonnable.

On the CPU side, if you peak a Xeon E5 such as this one ( http://ark.intel.com/products/64595/ ), you need 4 of them to get to 32 cores. you get 32core*2.6Ghz*8floatpersimd = 665Gflop/s which is actually realistic for dense kernel such as matmul. and 4*50GB/s bandwidth. But in practice you difficulty reach 30GB/s per processor so 120GB/s aggregated.

So the GPU is about 7.5 times faster floating point wise and 2 times faster bandwidth wise. but here we are talking peak, and practical performance varies a lot from application to application and depending if you can use your architecture properly. But overall for some well chosen kernel a factor of 10 still seems not too unreasonnable.

Call it what you want. nVidia claims to be working closely with Valve, including embedding their own engineers in the project.

Engineers from Valve and NVIDIA have spent a lot of time collaborating on a common goal for SteamOS: to deliver an open-platform gaming experience with superior performance and uncompromising visuals directly on the big screen.

NVIDIA engineers embedded at Valve collaborated on improving driver performance for OpenGL; optimizing performance on NVIDIA GPUs; and helping to port Valve’s award-winning content library to SteamOS; and tuning SteamOS to lower latency, or lag, between the controller and onscreen action.

The collaboration makes sense as both companies strongly believe in the importance of open-platform innovation, and both companies are committed to providing gamers with a cutting-edge visual experience.

Valve will deliver a great, open-platform gaming experience, and NVIDIA will continue to be the best choice for gaming on any open platform or operating system, including SteamOS.

nVidia has already taken notice because they are one of the companies working with Valve: https://developer.nvidia.com/sites/default/files/akamai/gamedev/docs/Porting%20Source%20to%20Linux.pdf

I'm surprised no one has mentioned this more generic solution that's already in production:

NVIDIA's GeForce Grid

Really? GPU's are being used more and more for more than just graphics processing. Many interesting parallel processing problems are being off loaded to GPU's where they are number crunching on hundreds of cores much faster than can be done on your main CPU. See http://www.nvidia.com/object/cuda_home_new.html for one such set of libraries for Nvidia cards.

So WHO CARES if you cannot see the difference in what gets displayed. There is a LOT more going on.

You have the library of games published by Sony and Nintendo licensees. It's historically been far less expensive for an indie developer to get a game onto Google Play Store than onto Sony's and Nintendo's store. For instance, Nintendo wouldn't let Robert Pelloni develop Bob's Game for the DS because his business was home-based.

Most people who would buy a Shield already own a smartphone where they can get that game if they really want it.

In what sense do you mean "not anywhere near the level"? Please clarify.

Here's an example 3DS game that will be out for the platform soon. There's nothing on Android that's going to compete with that. Not even iOS gets much in the way of larger games, and most of the ones that they do get are ports of old console games. Most of what's available for phones are $.99 apps that are designed to be an easy way to kill ten minutes. There's nothing wrong with that, but we haven't seen very many games that are more sophisticated than that, and most of the ones that have come out are console ports.

Because your good gaming PC is a desktop PC stuck on a desk in a different room of your home, and a Shield thin client is cheaper than a gaming laptop.

And a 50 ft. HDMI cable and wireless mouse/keyboard are even cheaper. Besides, the people who want to game on their TV in the living room likely already have a console. Would they really want to drop another $300 just to be able to stream from their PC assuming that they even have a good gaming PC or an NV graphics card?

I'm not sure what you meant by "most". What would a mouse do for a platformer like Mega Man or Castlevania series or a fighting game like Street Fighter or Mortal Kombat series? And if player 1 is using a mouse and keyboard, what do players 2-4 use?

There are certainly genres of games where a controller works better, but those types of games usually aren't the popular PC titles that most people play and many of those games aren't released for PC. Of the PC titles listed as supported by Shield, it looks like almost all of them are also available for console and many are FPS games where a keyboard/mouse combination works better. The types of games that would work best on Shield usually aren't ported to the PC and even the Android games for the device are either ports of old console games or available for other Android devices.

The Nvidia Shield is a solution looking for a problem. To use half of the features requires a gaming PC with Nvidia hardware that's already going to provide a better experience for the titles available for it than using a controller. The Android game library is nowhere near large enough to justify a $300 purchase price, especially when most of those games can already be played on your phone. It does a lot of things, but it does them all worse than other solutions making the market for the device rather limited.

>Not totally true. Stroke/path/fill rasterization work is not supported by current 3D rendering APIs (and thus not accelerated by 3d hardware).

Incorrect. It's there, developers just aren't using it for some reason.
https://developer.nvidia.com/nv-path-rendering

Nvidia has been doing that for a while. They hired several vector-graphics programmers a few years ago and had them add that functionality to their cards. The problem is, no developers use this stuff.
https://developer.nvidia.com/nv-path-rendering

Actually, I misread that. They provided specific feedback to NVIDIA to get them to fix the drivers.

(Part of me wanted to make a joke about Black Mesa, but I realised it was too dumb.)

1. Nobody is required to offer a warranty on any product.
2. If there is a warranty, it must be either a full warranty or a limited warranty (as per MMWA it must be conspicuously stated if it is a full or limited waranty)
3. Almost nobody offers a full warranty on their products, only a limited warranty so most of the MMWA provisions applicable to a full waranty or implied waranty are moot.
4. According to MWA, a company can pretty much limit the warranty any way they want if it is a limited waranty.

Shield (as with most actual products in the universe) only offers a limited warranty and conspicuously states this.

This article seems to be talking about newer hardware and the NVIDIA binary blob driver. If you're stuck with Nouveau and an older NVIDIA card, your performance is going to be much worse than Windoze.

How old are you talking? NVIDIA has been providing blob drivers for FreeBSD for a long time. Even your GeForce MX 440 has drivers available.

Reading the post two steps above mine or whatever.

Though "dual monitor" seem like something which one would more likely search for than "sli", sli is even totally unrelated.

Tried a new search with: nvidia x dual monitor
Second hit there is: ftp://download.nvidia.com/XFree86/Linux-x86/304.32/README/configtwinview.html

Which make more sense.

Third hit is: https://help.ubuntu.com/community/NvidiaMultiMonitors

But none of those mention what he was suggested to use?

Anyway.. Whatever. Good to know where to find it later on if ever needed.

If you use an NVidia card, yesterday, it seems: https://developer.nvidia.com/opengl-driver

On the other hand, AMD only publised catalyst drivers for OpenGL 4.3 today.

The easiest on-ramp to speeding up image/video processing is probably the npp library https://developer.nvidia.com/npp [nvidia.com] It has functionality and syntax similar to Intel's ipp library but uses an NVIDIA cuda-capable GPU to accelerate the operations.

If you want to dig in deeper you could explore OpenACC http://www.openacc-standard.org/ [openacc-standard.org] OpenACC is a directives based approach to accelerator programming. You comment or mark up your code with OpenACC directives that provide additional information that the compiler can use to generate parallel code.

Finally, you can learn CUDA C, or OpenCL, or CUDA Fortran, or NumbaPro, or one of the other programming languages that are supported on the GPU hardware of your choice. NVIDIA's CUDA C compiler is based on LLVM and the IR changes have been upstreamed to LLVM.org, There are several languages and projects in development that are leveraging the LLVM infrastructure to add GPU/parallel support.

[disclaimer: I work for NVIDIA, but the words above are my own.]

Agreed 100% about CUDA and OpenMP! Already invented a new multi-core string searching algorithm and having a load of fun playing around with my GTX Titan combing CUDA + OpenMP. You can even do printf() from the GPU. :-)

The most _painless_ way to learn CUDA is to install CUDA on a Linux (Ubuntu) box or Windows box.
https://developer.nvidia.com/cuda-downloads

On Linux, at the command line fire up 'nsight' open the CUDA SDK samples and start exploring! And by exploring I mean single-stepping through the code. The NSight IDE is pretty darn good considering it is free.

Another really good doc is the CUDA C Programming Guide.
http://docs.nvidia.com/cuda/cuda-c-programming-guide/

Oh and don't pay attention to the Intel Propaganda - there are numerous inaccuracies:
Debunking the 100X GPU vs CPU Myth: An Evaluation of Throughput Computing on CPU and GPU
http://pcl.intel-research.net/publications/isca319-lee.pdf

Agreed 100% about CUDA and OpenMP! Already invented a new multi-core string searching algorithm and having a load of fun playing around with my GTX Titan combing CUDA + OpenMP. You can even do printf() from the GPU. :-)

The most _painless_ way to learn CUDA is to install CUDA on a Linux (Ubuntu) box or Windows box.
https://developer.nvidia.com/cuda-downloads

On Linux, at the command line fire up 'nsight' open the CUDA SDK samples and start exploring! And by exploring I mean single-stepping through the code. The NSight IDE is pretty darn good considering it is free.

Another really good doc is the CUDA C Programming Guide.
http://docs.nvidia.com/cuda/cuda-c-programming-guide/

Oh and don't pay attention to the Intel Propaganda - there are numerous inaccuracies:
Debunking the 100X GPU vs CPU Myth: An Evaluation of Throughput Computing on CPU and GPU
http://pcl.intel-research.net/publications/isca319-lee.pdf

Like in all attemps at getting stuff faster, you should first wonder what kind of performance you are already getting out of CPU implementation. Provided you seem to believe it is actually possible to get performance out of a VB like langage, I assume that your base implementation heavily sucks.

Putting stuff on a GPU has for only goal to make things faster but it is mostly difficult to write and non portable. Having a good CPU implementation might just be what you need. It also might be easier for you to write.

If you really need a GPU, then you need to start learning how GPU works, because a simple copy paste is unlikely to give you any significant performance. A good start at: https://developer.nvidia.com/cuda-education-training

I never properly learned opencl, but it is essentially similar. Except you have access to less low level details on nvidia architecture. Of course, cuda is pretty much nvidia only.

The heavy lifting has mostly already been done for you. There are CUDA wrappers out there that, with a few changes to your code, run it as close to optimally as possible using the card's cores. We had a Nvidia guy come by and give a talk just to show off how relatively painless it is (similar to OpenMPI, in my opinion). If you've got a couple extra people around consider reaching out to Nvidia to have someone show everyone a few of the options.

The best part is the Ouya can work as a thin client for your Steam box. nVidia supports this natively via their Shield device and a gtx 650 or higher video card (those video cards have a built in x264 encoder), it's rumored that this functionality will be extended to Tegra 3 devices (which would include the Ouya).

For older video cards (and until nVidia expands the capability), there's Kainy. Which will allow you to do the same.

You might want to read / view these slides:An Introduction to Modern GPU Architecture Especially slide 42.

Modern GPUs are massively parallel in their execution. Yes they work "only" on one image, but when rendering one scene the sharers work in parallel. For example a fragment (aka per pixel) shader will be run in parallel for each pixel, limited by the number of available shader units (aka core). THIS is why you get the awesome performance: small, self contained programs running in parallel.

Except that they do.

Nvidia publishes both drivers and CUDA for Mac OS X. AMD makes the ATI drivers that are in Mac OS X, but lets Apple do the distribution in point releases.

What else you got?

Nvidia does:
http://www.nvidia.com/object/macosx-313.01.01f03-driver.html
http://www.nvidia.com/object/mac-driver-archive.html

Nvidia does:
http://www.nvidia.com/object/macosx-313.01.01f03-driver.html
http://www.nvidia.com/object/mac-driver-archive.html

There are still very real problems with Win8.

1. Multimonitor is worse than it was before. The stupid tile display, even though it's designed to scroll horizontally (which would coincide with how the overwhelming majority of multi-display setups are) will only work on one screen. The task bar thing is mirrored on both displays, which wastes screen real estate showing you the exact same thing in two places simultaneously. When the tile display is up, the other display still shows your desktop; but if you click on it, it jerks the other display out of the tile thing back to the desktop. You cannot interact with a desktop app on one display, and Metro / Modern / WTF ever on the other simultaneously; which would be the only reason I can think of for this single-display tile scheme, unless they are just lazy and didn't want to deal with screen scaling and positioning.

2. There are real problems in hardware drivers. Windows Update tries to force an ancient version of Nvidia display drivers down over newer versions of drivers, which install without you knowing about it, and blowing up your video config until you either disable the update from ever running, or perform a system restore. See: http://us.download.nvidia.com/Windows/310.54/310.54-win8-win7-winvista-desktop-release-notes.pdf (buried on page 29). I have two different Nvidia cards in my box, which works great on Win7 and other OS families; but I can't load the drivers on both cards at the same time without a blue screen in Windows 8. So, the one I only use for desktop stuff has to use a "Standard VGA Display Device" driver (a.k.a. frame buffer out) in order to be stable. Again, works great in Windows 7 with no hassle whatsoever.

3. Microsoft is clueless when it comes to supporting EFI. Windows Setup will not work if you have any other EFI-based OS installed, or if any of the EFI system partitions that may be present have not-Windows on them. If you physically disconnect those drives, it then works. If you need to use the "automatic restore" and it sees an EFI system partition that has not-Windows on it, it will give you an error saying to need to choose which version of Windows you want to repair, and then not give you any way to do that. Again, physically disconnecting the drives resolves this. Having to disconnect hardware in order to get OS tools to function is complete shit, and the only way it could be worse is if it actually went and ruined those other OS installs while attempting to install / repair Windows 8.

Unless they're confident that Nvidia's Project SHIELD will slide past its Q2 release window, this is a really bad idea.

Would you prefer your Android gaming console permanently tethered to your TV on Tegra 3 or would you prefer a handheld console on Tegra 4 with the option of going mobile?

Unless they're confident that Nvidia's Project SHIELD will slide past its Q2 release window, this is a really bad idea.

Would you prefer your Android gaming console permanently tethered to your TV on Tegra 3 or would you prefer a handheld console on Tegra 4 with the option of going mobile?

Unless they're confident that Nvidia's Project SHIELD will slide past its Q2 release window, this is a really bad idea.

Would you prefer your Android gaming console permanently tethered to your TV on Tegra 3 or would you prefer a handheld console on Tegra 4 with the option of going mobile?

CUDA for Mac: https://developer.nvidia.com/cuda-downloads

You are an idiot, and should stop posting.

And they did say something about that: https://developer.nvidia.com/sites/default/files/akamai/gamedev/docs/Porting%20Source%20to%20Linux.pdf

I understand what you're saying, but I still believe that Roadrunner was unnecessary in this progression.

GPU incorporation in supercomputers had already started when Roadrunner was built. Intel was already demonstrating processors with hundreds of cores in it (later to become MIC... and then finally launch as XeonPhi). I believe that all of that would have happened without the need to buy Roadrunner. Nvidia CUDA already had a lot of traction and OpenCL was already on the horizon.

If anything, I think the best lesson we learned (re-learned!) from Roadrunner was that buying into proprietary accelerators without a decent programming model is a bad idea. Nvidia CUDA walks this line well... the programming model still isn't the best, but at least there is one! Intel MIC is _much_ better in this regard... allowing "normal" programming models to be accelerated.

Hopefully, we've finally learned this lesson and won't make this mistake in the future.

Here's a pretty good writeup on a similar situation in Japan (obviously a little one-sided because of the source):

http://blogs.nvidia.com/2012/02/a-japanese-tale-of-two-supercomputers/

Integrated controller usually means a lot less clutter. For something that's supposed to be portable, that's a huge selling point. I could take my phone and use a Sixaxis controller with it, but that's bulky. I've not heard of a smaller controller designed for it, and most that I have seen are fairly limited, going for the "retro" gaming style with few buttons and no analog sticks.

With an integrated controller, you're looking at something that competes with portable game consoles like the 3DS or PS Vita. That could be a very interesting proposition provided the whole thing isn't too bulky and can still work as a general purpose device.

The problem with Archos is that they usually have good core ideas, but poor execution. Cheap building materials, skimping on components to shave off a few cents (they stuck with resistive touchscreens for way too long), poor ergonomics, you name it. It would appear that this is once more the same kind of deal. It doesn't help that for some reason I can't fathom, Google's not very good at supporting games on Android. They exist, but they have poor visibility and little support, so the Play store usually ends up showcasing cheap freemium games and fucking Angry Birds.

The one combined device that people should keep an eye on is Nvidia's Project Shield. If there's one company out there who's fairly savvy about games and gaming hardware, it's them, and the prospect of Shield looks really interesting.