I can only say the same. Bought a mighty expensive Sony netbook only to figure out that this integrated PowerVR graphics chip is a total disaster. In Windows and much more in Linux, where it is 1. a pain to even bring to work and 2. so slow that it's barely usable.
I must say i am really disappointed by Intel, which (in my oppinion) had and have great Linux support in their products. But not in this one.
I got an Atom with Intel graphics and NVIDIA Ion I can switch to. I really like it for gaming. Apart from it being a bit hacky (not grandma-friendly, not even dad-friendly) I am having no problems with it. On the contrary, I am glad I bought it.
Sounds like my own Asus eeePC 1215N. I however never managed to get the NVIDIA chip working in Ubuntu. Whenever I want to use the graphics (games, WebGL, etc..) I have to reboot to windows.
How did you get your system to run, if you don't mind me asking?
In any case the poor experience of this system is what have made me aim for a AMD or at least clean (as in no NVIDIA or other chips) Intel system for my next machine.
ATI has always had far worse Linux support than NVIDIA. I'm not sure if the situation has changed since AMD bought them, but I'd be hesitant to suggest going that route.
I chose the system because I had terrible experiences with amd/ati on Linux.
I use bumblebee, acpi_call and bbswitch. with that I can launches processes that use the nvidia chip by pretending a command, for example "optirun ioquake3". that launches a secondary x server and using virtualgl streams the frames to my primary one. works quite well!
Are you talking about the VAIO P series? I had a VGN-P788K; I purchased it for $300 off ebay, but I recently resold it. It worked pretty well with the Linux 3.2 kernel in Ubuntu 12.04. There are definitely annoyances, but if you want the smallest laptop possible it may be worth it. The specs were pretty nice - 1.4 pounds, 8" 1600x768 display, 1.6 GHz Atom processor, 2GB of RAM, 64GB solid state drive, 802.11n wifi, built-in webcam, and integrated bluetooth.
I wanted to get one of those, but even used they're still expensive. I think Ubuntu 12.04 or Fedora 17 will run acceptably. For Ubuntu, http://ubuntuforums.org/showthread.php?t=1984236 is the place to get help.
you mean, intel, the manufacturer of wifi drivers that only works with binary blobs? or the manufacturer of GPUs that work wonders with linux because they have no feature at all?
i wouldn't count them so happily as linux supporters.
>GPUs that work wonders with linux because they have no feature at all?
That's a little unfair. They might suck at 3-d immersive games, but some of us do not want to play 3-d immersive games, and the GPU integrated into my Core i5 does everything I want it to except perhaps smoothly play 1080p video. (720p playback OTOH works flawlessly.)
I tested your assertion (on a clip with 48 frames/sec) and I could not get the video to skip. So, strike the part of my previous comment about 1080p video. (Maybe Adobe's notorious Flash Player for OS X was the cause of the skipping I observed in the past. If you give VLC a .flv file, does VLC use Adobe's player to play it?)
Indeed, we are in agreement -- because FLV is a simple container format, it indicates that VLC is not using Flash Player, and is in fact notably more efficient in decoding.
Intel was the first vendor to have open source OpenGL 3.0 drivers, and have really driven Mesa development for the last few years. That hardly feels like "no features at all."
Disclaimer: I'm starting at Intel on the open source graphics team next month.
I don't really know anything about those drivers (they're not handled by the open source team). No one on the open source team likes them either.
It has seemed to me that using these PVR chips has always been a stop gap measure for Atom until the in-house graphics were ready (where ready means sufficiently energy efficient or something). This seems to be confirmed by the fact that ValleyView graphics are going to be coming to Atom.
No, i have never had problems with Intel hardware in my laptop.
The graphics drivers are extremely stable and have very few glitches. I can remember fglrx crashing my desktop or having graphics glitches oftne enough, but i can't remember one time where the intel driver wasn't working.
Can't remember having problems with Wifi either.
Also this article is full of praise for the WLAN experience: http://www.phoronix.com/scan.php?page=article&item=intel...
If all linux drivers would work as flawlessly as the intel graphics drivers, that would be awesome :)
(Of course it would be great if the Wifi driver would be completely open source as well!)
Nice that it exists, but I wouldn't want to try that with only an Atom CPU. When you have a low-power CPU like that is exactly when you need graphics acceleration.
I wasn't arguing that you have to like it. The argument was "if it works, then it's at least one answer". Saying it's never the answer is disingenuous.
Which is a shame, because at a hardware level I've always liked Imagination's tech.
Wish Imagination would solve this problem by open-sourcing some proper drivers.
Because most of the performance woes will be bad driver integration, and world will be wowed by a decent driver making the SGX punch, as it does, above its weight.
There used to be a pretty decent Linux driver for PowerVR Kyro and Kyro 2 GPUs back in the day†. I wonder if it could be updated to the current breed of chips.
† I recall playing Quake 3 Arena on it on Linux, with framerate matching or exceeding Windows'.
"It hit the market during the last few weeks." is inaccurate, these chips have been around for a while (although my country just got more stock recently) and Phoronix covered this as it happened.
These chips don't even have gpu drivers for Windows 7 x64. You have to run the 32-bit version, which is strange since the boards all support >4GB ram.
Another interesting thing to note is that the D2700 is already EOL'd [1] - i don't know when this happened exactly, but it can't have had more than 6 months on the market - and Intel have extended embedded production on the previous generation 5xx chips(?).
As far as I've understood this article, the relevant source is available. So what's the problem? Why should Intel spend time and resources to integrate it with the mainstream, esp. if this is a short-lived processor family?
Don't get me wrong. Intel is free to do whatever it sees fit. And Linux users and enthousiasts are free to buy or not to buy Intel-based hardware. As long as they are well-informed, which is the purpose of this post.
The kernel module is GPL, but all (heh, "all" -- it's a very big driver) it does is do things like memory allocation and mapping, power management, mode setting and command transfer.
All the meaningful parts of the 3D engine (shader compiler, GPU code generator, etc...) are in userspace (in prebuilt binaries for MeeGo). The kernel doesn't do anything but take the output from those components and hand it to the hardware to execute.
I've been exposed to the PVR driver stack professionally, so I really can't say anything specific other than that it's really large. I wouldn't suggest looking at the proprietary code at all (unless you're on a 32 bit Linux that matches MeeGo's libraries well enough, in which case you can just go ahead and use it).
I think I remember being told about a project to reverse engineer the SGX, but doubt it's made much progress. Even Nouveau is just barely usable for 3D now, and that's for much more accessible (and desirable) hardware.
Disassembling is incompatible with clean room reverse engineering, so other techniques must be used. Nouveau used a tool (renouveau) that poked at the graphics card and reported its state before and after.
> Disassembling is incompatible with clean room reverse
> engineering
Not necessarily. The idea behind clean room development is that one part of the team uses whatever legal techniques they want to analyse software, and document everything they discover. The documents are checked to make sure they don't contain any copyrightable information from the original (if information is not copyrightable, the methods used to obtain that information, be it decompiling or analysis of behaviour don't matter). A different team, who hasn't seen the original software at all, work off the document that doesn't contain any copyrightable information from the original.
This requires two different teams, while the nouveau/lima "blackbox strategy" (put something in, watch out's out, make conclusions, rinse, repeat) only requires one team to implement clean room.
To date, all Atom SoCs but Pine Trail (which has an Intel GPU) have used the PVR SGX. Cedar Trail is just another member of this family. It's a proprietary IP core with a GPL kernel driver but no free userspace. Intel provides the userspace libraries gratis for a handful of platforms (MeeGo, in this case) but sadly not all of them (my peeve is the lack of any x86_64 support).
So no: Cedar Trail is a poor platform to use for linux if you want accelerated graphics (but there are some cheap boards I'm looking at that seem really attractive for hobby projects). But it's not Intel becoming "more evil", they're just shipping the same stuff.
There will be a Valley View chip out later this year which goes back to Intel Graphics. Is that evidence of Intel becoming "less evil"?
In my opinion reviving meego just to be able to ship unusable and half-baked drivers is evil. I've written this at least 3 times on HN now, I'm not against using PVR chips, I'm against Intel not shipping related consumer ready drivers.
Why are the drivers unusable and half-baked? Have you tried them? You're saying that if they don't build on Ubuntu or Mint or whatever you use that they're ... evil? That's a ridiculous stretch, sorry.
Intel's community support on the SGX is no better or worse than Apple's or Google's or Samsung's. They support the OSes they want, which in this case means Windows and Linux i686 ("MeeGo" libraries are compatible with basically any modern distro -- try them). They're not free drivers, because no drivers for this part are free.
It's not evil, it's just lazy. Atom is a low-cost chip and they can't spend zillions developing it. They needed a small GPU and it was much cheaper to license PowerVR than to design a new GPU. (I guess Intel's GPU is not that area-efficient or something.)
I agree that it's most likely incompetence rather than malice, although it casts Intel in kind of a poor light. They're a chip designer who have been trying to make inroads on graphics recently, and who have produced mediocre but usable graphics chips in the past, and they can't figure out how to stuff something in there themselves?
I'm not against them to license a simple, low power consumption chip. But The evil is creating proprietary and unusable software drivers. If it was only laziness the could just license enough driver code to be open sourced. The Linux ecosystem is capable enough to hand over development from there.
No, the Linux ecosystem is not capable. Look at the open source Radeon and Noveau drivers; they're way behind their proprietary counterparts. Also, Imagination refuses to publish PowerVR documentation and it's not clear that they'd even allow open source developers to access the docs under NDA.
That wouldn't have helped in this case. The Poulsbo chipset launched in 2008, and (unaccelerated) GMA500 support was only merged into Linux in the 3.3 kernel in May 2012.
This just goes to show how companies only support Linux with half heart and as long as it somehow fulfills their agenda.
Intel which is supposed to be a good citizen does not think twice when it comes to betray Linux support, and the same can be said for any other company.
Which is who's fault? If everybody has trouble with the time, effort, expense of supporting linux, perhaps its linux that is the problem.
I'm not a linux user, ever since my introduction to it several years ago. Bought up-to-date computers for my team. Linux fan suggested we develop on linux. Tried it - disaster. Didn't support video chip, raid controller, even all the memory. Open office tools crashed. First doc I imported displayed in upper-left corner of the screen, every character in the file in one big spiderey blob.
We expunged linux, booted Windows(!), it asked for drivers ("Have disk?" Yes!) and everything up and running in minutes.
And we were developing new Infiniband code FOR LINUX! For testing we ran a virtual machine booting linux - that worked of course.
Moral: until Linux has a "Have disk" button, I don't see it deploying successfully on production machines. In a hobbyists garage, sure. But not mainstream. For all the reasons listed in other posts here.
This would work only if you could get Linux drivers on a disk. There's a better solution that both Linux and Windows have had for years: put the drivers in a repo. Vista and 7 will grab drivers from their repository and install them automatically. Linux does the same. But only if the manufacturer has drivers for it.
Bought up-to-date computers for my team. Linux fan suggested we develop on linux. Tried it - disaster. Didn't support video chip, raid controller, even all the memory. Open office tools crashed. First doc I imported displayed in upper-left corner of the screen, every character in the file in one big spiderey blob.
That's the problem, there was no driver support. That's not the fault of Linux, that's the fault of companies not making drivers for Linux. If the hardware maker supplies Linux drivers, they're installed automatically while the system is installing, seamlessly.
I don't see it deploying successfully on production machines
Then you're blind, because it's on millions of production machines around the world. We're a Windows shop and even we have hundreds of Linux servers and desktops.
Sure, servers have an extremely limited ecosystem of apps/services, you can get a handle on installation and support, even hardware. But try to corral a bunch of developers or home users. The interoperability matrix explodes.
Compare your 'millions' with the what, half a billion Windows machines, and you make the point very well.
I think you make my point just as well, you're just directing blame at the wrong people. The problem you and I both see is a lack of Linux drivers. Rather than blaming the Linux developers though, I'm blaming the people responsible for making their drivers Windows-only, as well as making them closed-source binaries so Linux developers cannot make it work on their own.
The problem is that developing drivers for Linux AND Windows will always take more time/resource than developing drivers for Windows alone. This is hardly Linux's fault.
And, well, if you buy hardware from shoddy vendors, don't expect it to work with every OS. For me personally Linux has always worked decently out-of-the-box, and outperformed Windows with a bit of tinkering.
(I admit I never bothered to do tinkering in Windows though. I don't know if you can, and how much can be done.)
I must say i am really disappointed by Intel, which (in my oppinion) had and have great Linux support in their products. But not in this one.
Stay away from Atoms with non-Intel Graphics.