GMA600 is used on many Windows 8 tablets, Fujitsu Q550, KUPA X11, and more.
According to specs it should support DirectX 10, and judging by it's performance on iPads (same PowerVR core) the chip itself delivers nice performance.
The current Windows 7 driver (3077) is hardly providing any HW acceleration. Most apps crash or display a blank screen. And those that don't, have abysmal performance.
So, the question is, if and when we will get proper Windows 8 support?
Win8 driver for GMA600 is what all GMA600 users want. But the intel driver support is just ********
I dont know whether they cant or they wont support it... In my opinion they dont care about us.. they hope we'll throws our Windows tablets in the garbage and buy new i5 tablets for thousands of $/€.
iam ****** of .. this ****** intel support sucks... my next harware is AMD for sure...
Developer Support Engineer, PowerVR Graphics Imagination Technologies:
"We supply Intel with our reference drivers which they use as a basis for their own drivers. We have no control over when they will release drivers and we cannot release drivers directly to end users for their platforms.
It's our customer's responsibility to support drivers on any platforms that use our GPUs. You should contact Intel to ask for an updated driver."
Microsoft and Fujitsu say also that intel has to provide new drivers.
so please react!
The driver situation and support for the GMA 500 and 600 GPUs used in many netbooks, MIDs, and tablets is completely unacceptable.
Current Windows 7 drivers fail to use a large portion of hardware acceleration. This is especially true in DirectX and any advanced form of video. The basic Windows 7 drivers for both GPU platforms should be able to at least decode standard H.264 video at 720P, which currently causes much stuttering, audio latency and dropped video frames. The same PowerVR SGX 535 core is used in various iOS and Android devices has no issues playing video including 1080P material via MDL or HDMI (iPad, iPhone 4, Galaxy S series, Galaxy Tab, etc.). Clearly we're not getting driver optimizations other vendors are willing to invest in for their platforms.
Even Windows WEI scores and other benchmarks point to poorly optimized drivers. A Z540 with a GMA 500 @ 200 MHz scores 4.6 Aero Desktop and 2.5 in Gaming Graphics performance with v2208 drivers built from IEGD kits (stock Intel 2048 drivers are around 2.3/2.5 as I recall). Meanwhile, the GMA 600 @ 400 MHz scores 2.9 in Aero Desktop and 3.0 in Gaming Graphics with v3077 drivers. The inconsistency of these scores reflects clearly the break in Intel's support.
Further, current release drivers for Windows 7 for both GPUs (v2048 & v3077) demonstrate a variety of graphics corruption even with programs as common as MS Office 2010 (flashing taskbar and titlebar bits).
Flash and HTML5 video from YouTube stutters graphically and audio even at 480P resolution. Rolling a system back to the Microsoft Software Graphics will actually improve video quality at the expense of forcing everything through the CPU.
It would be one thing if these were significantly old GPUs which Intel never claimed any hardware acceleration support, but this isn't the case. The GMA 600 GPU is about 2 years old and used nearly exclusively in hardware where it cannot be replaced. The GMA 500, while old enough to potentially warrant twilight status, has gotten better (though mediocre) support. Better support for one platform in this case largely results in better support for both platforms due to the amount of similarities.
Below is what Intel claimed with the GMA 600 series, only about half of that slide is remotely reasonable to say true, even if one ignores the lack of proper hardware acceleration in favor of poorly optimized OpenGL and D3D functions.
Now we're getting told that Intel will not produce Windows 8 drivers for GMA 600 / 500 platforms. This is even after 6 months ago, multiple OEMs had apparently gotten authorization to say GMA 600 products would receive driver support from Intel (Fujitsu - Q550, HP - Slate 2, Kupa - X11, Dell - various netbooks -- some of these devices are barely a year old from release).
The same applies to Linux and Android, where with the GMA 500 for example, one of Intel's own employees was able to produce an OpenGL capable and accelerated driver 2 years ago from source code sitting in Intel's repositories (Viliv S5 Android Port Well Under Way, Could Pave Way for Android x86 Devices). (I hope I am not going to cause him any trouble by drudging that back up). Further, people have been forced to use IEGD drivers they compile themselves to get performance in Windows 7 where one would expect at least decent support.
The sad part is that merely fixing the Windows 7 drivers in their current state would solve most Windows 8 acceleration problems. Plenty of people are using Windows 8 builds (Dev RP, CP, RTM) just fine with Windows 7 drivers for their graphics subsystem. The fact that Windows 8 does not work with GMA 500 / 600 drivers for Windows 7 indicates just how broken they are at the moment.
If Intel is not going to support these platforms, please work with ImgTec to release some degree of source code and development resources or at least hand them off under NDAs to OEMs and Microsoft so they may produce competent drivers.
Other information about Windows 8 and GMA 500 / 600 devices. Note their links to Notebookcheck.com which indicates better Directx (10.1 SM 4.1) on the GMA 500, and only a 40-50% performance boost in synthetic 3D benchmarks for the GMA 600 despite being 100% faster clocked:
More commentary regarding Windows drivers:
More Intel presentation slides and claims of 1080P video support on Z6xx / GMA 600 (still unfulfilled):
Intel's own demos and claims regarding GMA 600 performance. (Dx 10.1, nope, OpenGL 2.1, nope, OpenVG, nope) Can we at least get the prototype higher performance drivers used here:
At this point, one must wonder if a class action against Intel for misrepresenting their product would be in order.
One last edit, can we get someone at Intel to acknowledge Apple's role in conjunction with Imagination Technology as a stumbling block in bringing drivers forward for the GMA 500 / 600 series under Windows, Linux, and Android. Everyone close to the semiconductor industry is well aware of Apple's predatory IP policies, and I suspect they are the only ones licensed to directly incorporate ImgTec / PowerVR source code directly in their drivers. Several chip foundries and developers have turned Apple down in the last few years specifically because of Apple's demands (TI comes to mind). There isn't really another justification as to why demo show drivers have such a huge performance delta against release drivers other than the source code can't even be used in Intel's release drivers. Further, this would explain why Intel has been unable to release even reasonable binary blob (pre-compiled) drivers for Linux and Android when it has been demonstrated otherwise by your own employees and tech shows.
Be honest, Intel got into bed with a bad partner in Imagination, and now has been shafted between Imagination and Tungsten Graphics with the unenviable job of reverse engineering drivers for the SGX 435 series from start to finish. It wouldn't even surprise me if specific features were covered under license which Intel could not use. Why else would some GMA 500 devices come with dedicated Broadcom chips for video decode when the PowerVR SGX 435 can do it natively? Apple under Steve Jobs specifically avoided direct competition at all costs, and it would be a huge coup if the various netbooks and tablets based on Atom Z5xx and Z6xx were able to compete against the iPad as full products under Windows or Android. Further, the release of Intel powered smartphones exclusively outside North America and Europe further bolsters this assumption. Also, any legitimate driver for the SGX 435 under Android or Windows RT would potentially enable hackers to run those OSes on iProducts.
Message was edited by: Aaron
Just want to give a shout out to Aaron for a thoroughly researched thread. While some of things Aaron wrote could never be able to be verified from the outside, there is a lot of truth in what he said about Apple's competitive behaviors. A lot of people admire and praise Apple for its creativity in product development and marketing area. But not many outside of the business know that Apple is extremely ruthless and efficient when dealing with IP, supply line management, and information control issues. Those are areas that the less glamours yet equally important to a tech company's success.
back to the topic. What are we going to do about this GMA 600, the crimped could have been? GMA 600 shares the same infamous video core as the older GMA 500 aka "Poulsbo" (Italian for piece of sh*t?) running at higher frequency. It seems that GMA 500 has a driver ver 2030 that runs decent in Windows 8 when installed in compatibility mode and the user of that chip are at least content with than user of GMA 600/3600 chip. So maybe there is hope if someone more technical inclined and has a GMA 500 device like HP Slate to check it out first and See if that GMA 500 driver supports openGL 2.1 in W8 (Gma 600 official driver doesn't in w8 and I and few suspects it is the main culpult for less than faster Metro App performance).
HP Slate with GMA 500, Win 8
If it works, perhaps we could somehow mod that GMA 500 driver for our use. (which would be quite ironic considering that there are GMA 500 users who want to mod GMA 600 drive to overclock their chip)
I should clarify that my opinions regarding Apple involvement is pure supposition based on available facts.
Many know that Imagination Technologies basically went under after making a foray into consumer 3D graphics with the Matrox M3D and similar in the late 90's. The PowerVR architecture was attempted to be reintroduced in the PC market several times with new iterations, but could never compete with 3Dfx and nVidia at the time for raw performance. They kept at it and moved their work into the mobile segment and were quiet for several years, with lots of restructuring during that time. The iPhone was their big take off back to prominence for PowerVR architecture. The work they did there managed to make them a name again and got them into many segments they would have never even gotten a call back from project management previously.
We know that Imagination Technologies has given driver source code to Intel, but that Intel flatly refuses to release drivers for any platform based on this source code, even in pre-compiled form. The most logical reason becomes that they can't due to licensing agreements with ImgTec.
The alternative is intentional malice on Intel's part to sell Cedar Trail, ULV Ivy Bridge, and Haswell down the line. I choose to not attribute to malice what can be sufficiently explained by incompetence, and in this case incompetence would boil down to poorly negotiated IP licensing.
Apple along with Samsung are likely the largest licensers of PowerVR IP by quite a margin over Intel. Apple likes to make IP deals that reach across the table to other parties, so there is a logic there that Apple's deals with ImgTec would restrict ImgTec's abilities to sell certain portions of IP to other parties or what those licenses may allow. Samsung has a good reputation in the industry as far as not being highly restrictive with their IP much less IP they are licensing from others.
Regarding driver swaps between GMA 500 and 600 platforms, while large chunks of source may be possible, be mindful that the memory interface was moved from 500's dedicated DMA style interface to the shared CPU interface. The core itself was moved from the Northbridge U15 to I *think* the same die as the CPU. While the execution engine itself may be identical, the communications interfaces received very significant overhauls which reduces the likelihood of compatible driver files.
The source code on the other hand should be exceptionally compatible as long as one is using the function subset during compilation for the interface bus appropriate to the platform.
Due to the process size difference, 130nm vs 45nm, for the GMA 500 and 600 respectively, I don't think anyone with a GMA 500 should have any hopes of running their cores at 400 MHz. If anything based on the process sizes, the 500 might be able to get the customary 10-15% overclock headroom, and the 600 more like 25-30%, though more dependent on CPU thermal state.