Slashdot Mirror
The Truth About OpenGL Driver Quality
rcht148 (2872453) writes "Rich Geldreich (game/graphics programmer) has made a blog post on the quality of different OpenGL Drivers. Using anonymous titles (Vendor A: Nvidia; Vendor B: AMD; Vendor C: Intel), he plots the landscape of game development using OpenGL. Vendor A, jovially known as 'Graphics Mafia' concentrates heavily on performance but won't share its specifications, thus blocking any open source driver implementations as much as possible. Vendor B has the most flaky drivers. They have good technical know-how on OpenGL but due to an extremely small team (money woes), they have shoddy drivers. Vendor C is extremely rich. It had not taken graphics seriously until a few years ago. They support open source specifications/drivers wholeheartedly but it will be few years before their drivers come to par with market standards. He concludes that using OpenGL is extremely difficult and without the blessings of these vendors, it's nearly impossible to ship a major gaming title."
wanted an ATI card. Better performance and Image Quality for less money, but I just don't have time to be screwing around with making games work :(. I miss the hey-day of my 1650. $90 bucks, rock solid stable and fast. Just couldn't keep up.
Hi! I make Firefox Plug-ins. Check 'em out @ https://addons.mozilla.org/en-US/firefox/addon/youtube-mp3-podcaster/
I work with some high end graphics software
The Mesa drivers don't work.
A prolonged and nasty install of the Nvidia drivers is req. for the software to work.
Not a good user experience on Linux
The article seems to mention Windows/Linux (or Linux/Window). What about OpenGL/GLES drivers on other platforms, such as Mac OS X, Android, iOS, ?
ATI blows equally. Intel is known to have a little better drivers but have software worts to encourage them to CPU bound for obvious reasons. Or was the case 6 years ago when I worked for a famous game company.
Windows 8/8.1 blows on Nvidia with the latest drivers if you do not have the latest cards. Ask any owner as the majority of the 8.1 update 1 failures were NVidia related.
My ATI 7850 also craps out requiring a re-image with any .4 drivers. 12.4 and 13.4 I avoid even though they are WHQ.
The situation with the graphics markers are like the ISPs with broadband or the major telecoms when picking a cell phone. Not a monopoloy but an oligopoly run by a few. Boy I miss PowerVR, S3, 3DFX Vodoo, and Matrox.
You can bet if they were still around competing toe to toe with Nvidia and ATI everyone would benefit regardless of which side you pick. To me I view them as picking AOL vs RealPlayer. Yuck.
For the record I was an nvidia fanboy at one time too before owning ATI cards.
http://saveie6.com/
It wouldn't matter if it was... nobody else would be able to freely use the information.
File under 'M' for 'Manic ranting'
Why not? I thought clean-room reverse engineering was perfectly legal?
My guess. I didn't try hard. But the important question to everyone, is this A B C in the right order?
Priest: "Universe from nothing, no laws of physics, sped up time"+ huge discrepancies. Creationism? No. Big Bang Theory
The King wears no clothes.
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...
sure just reverse enginner a GPU and all its subsystems, its the button next to "fuck off you moron noob"
Vendor B has the most flaky drivers. They have good technical know-how on OpenGL but due to an extremely small team (money woes), they have shoddy drivers.
Is that the excuse? So uh, ATI has always had money woes? Since time was time? That should have been a sign.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
It is, assuming you don't stumble over any patents in the process.
Actually, it's microsoft that doesn't play well with much else. Since the driver development consists of closed and open teams, the openness of the code isn't the issue here (though it would be nice to have).
OT: "Geldreich" is a German compound of Money (Geld) and rich/plentyful (reich). So if he's called Rich Geldreich, that could be written as Rich Rich... Yeah, I know: no one knows Richie Rich today.
Computer simulation made easy -- LibGeoDecomp
There's a comment at the bottom of the article by David Poole that links to a post talking about OpenGL driver quality on desktop Linux and mobile Linux. The summary from that blog post is:
Boy I miss PowerVR, S3, 3DFX Vodoo, and Matrox.
I don't, having to code for OpenGL, Direct3D, Glide, Rendition and MSI to optimally support all the different vendors on the market was a huge PITA. Though I do agree that the competition was so fierce that technology was bounding forward at a brilliant pace! ...and that part I do miss.
Presumably such work wouldn't really qualify as "clean room" following the above poster's recommendation... hence the requirement for anonymity.
File under 'M' for 'Manic ranting'
It's about time for someone to host a Github clone as a Tor hidden service for the explicit purpose of allowing people to share source code without having to worry about being punished by the imaginary property police.
ATI is catching up and are competitive. NVidia lowered the price and made their Quadro turn into the Titan serious to counter the ATi 290x.
Good for consumers. However, their drivers are shit. ATI's drivers have improved then had issues again with frame pacing and mantle on older AMD chipsets. Nvidia had some questionable hardware and now worse drivers which are unstable and Windows 8/8.1 HATE. They do not even support all of directX 11.1 which is the cause of the crashes.
Part of me feels ATI and Nvidia are doing this on purpose so they can sell the remarked gamer cards as FirePRo's and Quadro's for real professional work yada yada at an expensive price. I mean if it is so bad even for 2d Adobe apps you need a $2,000 card just so video artificats do not pop up you know you have trouble.
Or maybe I am cynical to think of a conspiracy to sell professional grade cards more with real opengl of course.
http://saveie6.com/
I didn't want to comment, because I've had a love hate relationship with some of these drivers. Two video cards ago when I was building my current computer, I wanted ATI because I was tired of the NVIDIA post-kernel-install. NVIDIA got lucky because the ATI card was DOA. So I went with a 9600GT. And I kept using it till a few months ago, when I replaced it with a 630 (about the same speed, but doesn't have the 'single monitor only' problem when using the Nouveau drivers. But the Nouveau drivers proved to be Unstable! , including kernel error messages. They are also painfully slow compared to the proprietary drivers. I have given up on Nouveau for the time being. If they get stability down, and get clock and cooling down, I will go back to Nouveau.
The situation with the graphics markers are like the ISPs with broadband or the major telecoms when picking a cell phone. Not a monopoloy but an oligopoly run by a few. Boy I miss PowerVR, S3, 3DFX Vodoo, and Matrox.
Ask the people stuck with Poulsbo how they feel about PowerVR graphics, they are one of the few who suck worse than nVidia for driver support. 3DFX with their Glide API was king of proprietary solutions. S3 was the patent champion, even today their patented S3 Texture Compression causes trouble for open source. And Matrox made Intel's 3D performance look stellar. YMMV but I feel the competition in the graphics market is still working fairly well, at least a lot better than on the CPU side. It's just that the primary focus is who can push the most FPS in the latest games using the most bleeding edge drivers, that's what drives sales. But if you think it was any different back then, it's time to take off those rose colored glasses. The only thing that used to be really stable was Intel's server drivers, practically zero performance but it didn't bring your server down.
Live today, because you never know what tomorrow brings
Slashdot has become an advertising site. Intel is always the best. Any article which compares Intel with AMD or Nvidia is a piece of crap. Intel 20 years behind in graphics.
They really aren't that far behind anymore for an enormous amount of uses, some pretty graphics-intensive. From the HD integrated graphics onwards Intel has been making great strides with every iX generation, catching up in most ways that matter in all but the most demanding areas (high end games and GPGPU). And I'm not saying that because I love these guys, I spent a decade telling customer after customer that Intel just straight up lied (as did the driver) about the graphics capabilities of the 9XX series of integrated graphics, that our software would never, ever work properly on these cards, and they should have read the minimum specs that clearly stated that we didn't support these chipset (this is software that was an order of magnitude more expensive than the laptops they were buying to run it on, but customers can be a silly bunch). Having that same argument with management every two years "because thousands of people have these chipsets!" (they'd usually shut up after a realistic time estimate of the work to support these cards, along with a table of probable performance and visual quality).
No, the HD-series integrated chipsets make me quite happy, because now we can at least have minimal support for people who buy these laptops (it's nearly always laptops), and their experience will actually be pretty good.
As an aside, it's easy to tell that TFA is absolutely true by how few major gaming titles ship. Oh, wait.
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.
Om, nomnomnom...
won't share it is specifications?
Confucius say, "Find worm in apple - bad. Find half a worm - worse."
Not being a gamer, I'm all okay with waiting for the rich kid's hardware to catch up. Besides, too much graphics capibility seems to do nothing but encourage developers to force all sorts of ridiculous visual doodads on software that really needs to do nothing but serve as point and click program launchers.
Time is what keeps everything from happening all at once.
"Windows 8/8.1 blows on Nvidia with the latest drivers if you do not have the latest cards. Ask any owner as the majority of the 8.1 update 1 failures were NVidia related."
I know it's an anecdote, but you said ask anyone, so hey, I have 8.1 update 1 and saw no failures with a not the latest card using the latest drivers. Not had the slightest problem, everything worked fine and smooth (well, apart from generally just being Windows 8 - but hey, I like to try before I judge).
A small correction, Nvidia Quadro has not "turned into the Titan". Quadro cards are largely the same hardware as the consumer cards, but with minor changes to enable certain features. The main difference is in the drivers. Consumer drivers err on the side of speed, whereas Quadro drivers will typically have lower performance in a game type situation, but be better suited for CAD / 3D work.
It's kinda sad that no one will reverse engineer the stuff and post all the specs and diagrams it can anonymously. I guess egos are just too important.
My god that's naive. Good luck reverse-engineering a modern GPU. It would be extremely complex and extremely time-consuming. It wouldn't be worth it at all.
Intel 20 years behind in graphics.
The performance per watt compared to similar low-end AMD and NVIDIA chips is very competitive.
I would be happily already gaming under Linux, if the desktops in general weren't so buggy. I will never go to Linux as my main OS until the quality assurance of the desktop reaches a professional level.
It protects you against copyright infringement, but not patent infringement. The famous reverse-engineering of the IBM PC BIOS was back in the days before software patents were considered valid - if exactly the same thing happened today, IBM could certainly have sued and won an injunction and massive damages.
You have probably just used to the glitches.
That's not too different from having to write different code paths for different vendors using OpenGL or D3D for performance reasons.
Mada mada dane.
Well. I wouldn't be so sure about calling that a success. OpenGL ES drivers are know to be even buggier and more terrible than the desktop OpenGL drivers.
In which way was he trolling? You may not agree with him, but at least he calmly rationalized his comment.
I like to rant about OSS, and will continue to, but I also saw a surprising and positive result with the open-source Radeon driver.
On a low-end Radeon 6320, about a year ago Half-Life 2 was extremely choppy on Linux. Of course it might have been a Mesa or compositor problem too, instead of a driver issue.
However, I recently tried it again and the frame rates are now almost as good as under Windows. Nice improvement.
If the article is correct (I didn't follow the link), ATI is in a perfect spot to go full open source. They would have a ton of people helping to build stable and open drivers and would be able to compete with Nvidia without spending a ton of money. Too bad it'll never happen.
This is correct. We had more competition in terms of choosing graphics cards in the past, it didn't mean they were actually competitors or even tried to not do a completely shit job that didnt' help anyone in the long term.
Intel doesn't give a shit even today as far as graphics - good luck getting any launch game to run on any integrated graphics platform on a screen above 1024x768, where even a $50 card from literally anyone else will do better than the extra $50 intel is charging people to have an IGP. Hell, even AMD's hybrid solutions do better for gaming by an order of magnitude. What intel is open sourcing is not a solid 3d background, but a pile of unused 2d renders and basically nothing. It would not be unlike taking mono on linux and saying that Microsoft "gave up the crown jewels for linux" when it's explicitly untrue.
Meanwhile, AMD is budget constrained and not the best, Nvidia is proprietary as fuck (and tries to encourage everyone else to be, look at PhysX's bullshit proprietary nature), and Intel is shit. People don't build gaming rigs with an intel IGP in mind. Our options are just as relatively poor as they were back in the day, just newer hardware/better software.
It's all anecdotes, and anyone who believes personal anecdotes is a fool and deserves to lose their money to a corporation on their foolish decisions. Which is how our economy operates, anyway.
Indeed.
I really DO NOT miss D3D execute buffers. Glide was awesome, and OpenGL 1.2 on IRIX was joyful (if the OS didn't crash on you...)
I remember coming into work one day and my dev manager saying the equivalent of "sorry about your office, but NASA is having trouble with their IR2 at Moffet so we got SGI to lend us one for a few weeks..." and lo and behold next to my desk was a brand spanking new - still had packing materials stuck to it - Onyx IR2 sitting there in all its purple glory. That was my favorite work day ever. This was one of those times when you actually say to yourself "they're paying ME to do this?"
I spent the next week working on multi-pipe multi-process OpenGL issues. Pure nerdgasm...
Those really were the great days of 3D in my opinion. Every week somebody was doing something awesome.
SPEA Fireboards, E&S graphics generators, Lockheed's Real3D, this crazy Hitachi Spherix that sat in my office for months.
DAMN! Nostalgia...
Loading...
They can be the same hardware but binned differently. This isn't a new concept.
I would still like to see OpenGL largely punted in favour of OpenGL ES. As you point out, it is a much smaller API and it reflects the realities of the hardware, as it actually evolved (shader programs and GPU-memory vertex buffers), instead of how the software initially wanted to see it (immediate mode and imperative matrix manipulation).
Hopefully, the inclusion of OpenGL ES as a subset of OpenGL within the 4.x versions will make this more a reality, as well as the use of things like WebGL.
If I had a time machine and I could visit myself in a past life, but it was even more hemmed in than Twitter—say Morse code at one millibaud—my message to self (circa mid to late 1990s) would be this: Screw games.
Yes, I had a blast playing those games. But then I started making "mixed" decisions in how I set up my system to balance the games I liked to play and the development tools I needed to use. In hindsight, that was nothing but bad mojo. The difficulty of achieving a perfect stack is exponential in the number of interacting constraints.
There are many other things I could tell myself, but in most of the other cases I probably had to learn those lessons the hard way. This one is different. I guess I somehow believed I was just chasing a moving carrot I would catch Real Soon Now and that all the fuss to mate the perfect video card to the perfect driver was a temporary growing pain (along with much else at the time). I was wrong. Nearly two decades later, the carrot remains elusive. DRM amounts to a sanity pre-emption field.
My final stop on the video card wagon was a hardened HD5670 (Redwood) with the open source Linux driver, nearly passive heat pipes and Japanese capacitors. If the software doesn't work with my card, screw the software.
I have mucked a bit with OpenCL. Getting the software development stack to work again after each Linux upgrade cycle bears some resemblance to Mine Sweeper. Sometime in the next decade I'll probably spring for a $60 CGN prime plus plus, just so I don't feel left behind.
It goes to a "black screen of death" (due to powermiser settings I've determined). I also get the TDR issue where the display "loses connection to the video card" but, that DOES seem to be able to be offset properly using:
[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\GraphicsDrivers]
"TdrLevelRecover"=dword:00000004
"TdrDdiDelay"=dword:00000007
"TdrDelay"=dword:00000003
&
[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\GraphicsDrivers\Configuration\NOEDID_10DE_06CD_00000002_00000000_2000100^D06938DED0679F9361F1BD7C043BEDA5\00\00]
"TdrDelay"=dword:00000004
"TdrDdiDelay"=dword:00000007
+
[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\GraphicsDrivers\TdrLevel]
"TdrLevelRecover"=dword:00000003
---
* The "Black Screen of Death"'s a widely KNOWN issue.
I've *tried* these settings for a "fix" for that (didn't work, crashed on desktop, not in game OR video online or local - didn't matter IF I used AeroGlass display or classic GDI driven desktop either...):
Windows Registry Editor Version 5.00
[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Video\{0C4EFFE1-F6C9-486D-83A9-21C9F86E8470}\0000]
"PerfLevelSrc"=dword:00000cfa
"PowerMizerEnable"=dword:00000000
"PowermizerLevel"=dword:00000001
"PowermizerLevelAC"=dword:00000001
[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Video\{0C4EFFE1-F6C9-486D-83A9-21C9F86E8470}\0001]
"PerfLevelSrc"=dword:00000cfa
"PowerMizerEnable"=dword:00000000
"PowermizerLevel"=dword:00000001
"PowermizerLevelAC"=dword:00000001
(I PERSONALLY HAVE TO "STEER CLEAR" OF THESE, since they crash me (DirectX Aero display OR classic desktop both do it in minutes...)).
This is with the latest/greatest 335.23 driver mind you (nvlddmkm.sys = OpenGL, & DirectX display are .DLL driven under %Windir%\System32) - funny part is, when I go back to 285.79? I can do DirectX driven local video fullspeed & fullscreen just fine, but NO OpenGL gaming now (due to MS doing those TDR settings above)... when I use the latest driver? I can't DO fullscreen online video, or I get a "crash", everytime (sending me into those powermgt 6++ reboots).
APK
P.S.=> The GIANT pain-in-the-ass, is that it DEFINITELY IS powermgt. (since my flatscreen shows it's dropped into "power saving mode" AFTER these crashes - ONLY WAY OUT TO BE ABLE TO BOOTUP TO WINDOWS AGAIN, MINUS BLACKSCREEN? 6-8 reboots typically (slower each time I've noted - it MUST be "diggging thru" diff. CurrentControlSets backups to get those powermgt. settings, & once they're in place? Folks online said "There's no way to remove them" which is untrue, since a "-" based removal via .reg file can do it, or those multiple bootups can (where you only have a minute or two to remove those powermiser settings & then it goes black - this HAS to be done that many times, otherwise it appears to be "digging into" old CurrentControlSet registry hives to get them BACK again, whether you like it, or not - there a 2 keys worth in each hive x 3 backup CurrentControlSet hives, so the 6++ reboots to normally bootup again, makes sense)... apk
My 7870 performance on Linux has been getting steadily better. The release schedule is WAY faster than it was and I haven't seen any regressions for a long time.
The last 2 releases tripled performance on Portal - it's over 300 FPS now.
Steam on Linux appears to have lit a fire under AMD and real progress is being made. Shit Just Works now.
Want to learn about race cars? Read my Book
Just today installed Fedora 20 (KDE spin) on a HP 2230s laptop. After the initial installation of all system updates I restarted the computer. Now every time I login to my desktop, I'm greeted with "KWin crashed unexpectedly". I cannot start KWin at all and have no desktop effects. Please help.
At the same time I'm personally working with the Intel guys with an issue of backlight flickering on this same laptop under Linux.
I have to deal with problems like this all the time. Open source is garbage!!!
A small correction, Nvidia Quadro has not "turned into the Titan". Quadro cards are largely the same hardware as the consumer cards, but with minor changes to enable certain features. The main difference is in the drivers. Consumer drivers err on the side of speed, whereas Quadro drivers will typically have lower performance in a game type situation, but be better suited for CAD / 3D work.
Not necessarily true anymore for all Quadro's. The titan is a different beast than the other high end cards that Nvidia makes. It has double precession and other on demand hardware features. It is true the drivers crippled double precession floats on it but for cheap engineering cards they are great.
But you are taking a crap-shoot with the drivers.
5 years ago ATI was the suckiest hands down! They have improved slighty and Nvidia has gone down to where they both have their good and bad versions with bugs everywhere where game designers have to put in special code like webmasters did with IE 6 just to not crash or display artifacts.
http://saveie6.com/
Because he finds a workaround and quickly forgets about the issue.
I used to write drivers for hardware a looong time ago (disc drives, UARTS, that kind of thing.) I realize that these graphics cards are way more complicated and trying to squeeze every last ounce of performance out of them can be a lot of effort. (I can remember spending a day trying to save a single instruction inside a device interrupt, and those were relatively simple devices.)
Even so, eventually you can't just kkep adding people to a project. If the concepts are well known then you get some decent programmers to do a workmanlike job of writing the software. If there are still areas of research and black art, then you need people who are initiates in the black art. So, I'm just curious, how many people, and what kind of skills, are involved in creating good drivers for this hardware, and, when a new piece of hardware comes out, how much new stuff is required to make use of it?
In theory, theory and practice are the same; in practice they're different. (Yogi Berra & A. Einstein)
I'm not so sure ATI does blow equally. I have two 660ti's in SLI to power 3 monitors. Anything beyond the ~2 year old 327 series drivers does not work for me. Without surround mode, I at least get three screens with newer drivers... generally I get a stupid amount of slowdown to the point where I feel like I'm running Windows 7 on a 486. If I somehow manage to turn on surround mode, all three screens are recognized, but only 2 of them display anything. The slowdown also gets much, much worse. To date I have not found anything about other people experiencing similar issues, and all of the nVidia documentation shows "this driver improves surround on 600 series and higher chipsets!". 3 fps is better than 2 fps, but still useless.
Replace the proprietary junk with an Octacore DSP as a co-processor and do software rendering assisted with extra instructions. If it works with arm, it should work with Intel.
Like there is nothing to work around in either Windows or OSX. Typically Win7 or Win8 attention seeking dialogs drive me up the wall. They block the whole screen even for tiny minor issues. OSX very often has a frozen dock that doesn't work. Lately I've been unable to add attachment to emails by drag-and-drop in Mail.app
Everyone finds workaround in all systems. I find that the Linux experience has improved immensely in the last couple of years.
That's not too different from having to write different code paths for different vendors using OpenGL or D3D for performance reasons.
Well it is, like I said it's OGL, D3D and 3 more vendor-specific code paths. Targeting 5 is significantly more effort than targeting 2.
Definitely! It was cool to have the 3D accelerator days, a TNT paired with a Voodoo 2 was such a cool combo! But then being able to work on an InifiniteReality the size of a pair of refrigerators with - back then - an enormous amount of computing power was just astounding (from a nerd point of view). I do miss that, and having such a buzzing development community being pulled in all different directions by the latest innovation from one of the many vendors :)
That's not what I meant. Currently, it's not unusual for AAA games to have 2 or 3 different paths (or more, if they explicitly support different generations of hardware) even when using a single API for performance reasons. I was saying that having to support 3 different APIs instead of 3 different code paths using the same API isn't that much more work.
Mada mada dane.
You had the same thing when targeting specific features of the Voodoo 2 as opposed to the Voodoo 1 even when using Glide. Having multiple code paths for a single API to support multiple generations of hardware is not new, it was done back when we were supporting a dozen graphics APIs as well.
And again you're missing my point. I didn't say it was new.
Mada mada dane.
So what is your point? Previously we had to write for many APIs (OpenGL, Direct3D, Glide, Rendition, MSI, etc...), now we have to write for fewer APIs (only OpenGL and Direct3D). In both cases we always had multiple code paths per API so what point are you trying to make?
I was talking about actual bugs. That's more of an UX choice.
Windows generally works just fine these days. It's been the best choice for me.
People seem to still not grok exponential improvements in hardware. Intel GPUs are state-of-the art for integrated graphics, and only about 2-3 years behind the top-end discrete graphic cards at the most. They are good enough for most things now, except top-end demanding recent games.
"It is true the drivers crippled double precession floats on it"
My understanding was that the crippling was done in firmware/hardware. I'd like to see otherwise as it'd save having to buy K10 cards for a current project.