Intel Clarkdale Linux Graphics Performance

[kresh]

Why only open source drivers?

I don't understand why only open source drivers were benchmarked. I understand the reasoning behind open source software, but not how it applies to video drivers.

The drivers only work for a particular card, and it does not lock out anyone. I could see the need if manufacturers refused to write drivers for a platform, but using OEM video card drivers is not like Microsoft embracing, extending and extinguishing.

If it is just to use open source because of open-source sake, it is putting philosophy before features or performance. Not only that but it is hypocritical because every piece of software ultimately gets translated into a closed, proprietary instruction set (Intel, AMD, ARM. Sparc, Nvideo, ATI - all have proprietary instruction sets).

I am not disparaging open-source, but does it really matter with video cards? Especially if the OEM driver performs much better.

[deanjo]

I don't understand why only open source drivers were benchmarked. I understand the reasoning behind open source software, but not how it applies to video drivers.

The drivers only work for a particular card, and it does not lock out anyone. I could see the need if manufacturers refused to write drivers for a platform, but using OEM video card drivers is not like Microsoft embracing, extending and extinguishing.

If it is just to use open source because of open-source sake, it is putting philosophy before features or performance. Not only that but it is hypocritical because every piece of software ultimately gets translated into a closed, proprietary instruction set (Intel, AMD, ARM. Sparc, Nvideo, ATI - all have proprietary instruction sets).

I am not disparaging open-source, but does it really matter with video cards? Especially if the OEM driver performs much better.

You must be new here (and opensource in general). There is a large percentage of opensource users that do place philosophy over features and performance thus why tests such as this one take place.

[Hephasteus]

I'm sorry but putting DX 10.1 on chip like this just seems like a fail formula to me. Not that it won't give good experience on layout but i just can't see this being huge seller.

[gbeauche]

Intel's Linux driver also supports XvMC, which is marginally better than X-Video, where as the ATI Linux drivers do not offer such support. Though it will be nice once VDPAU / VA-API support is available through a Gallium3D driver.

There is already native VA-API support for G45. At this time, it only does MPEG-2 VLD, i.e. full video decode. Intel is working on H.264 support and this should be available by Q2. I don't think there is any H.264 video decoding at Gallium3D level yet, so VDPAU / VA-API support would be useless at this time.

[kresh]

You must be new here (and opensource in general). There is a large percentage of opensource users that do place philosophy over features and performance thus why tests such as this one take place.

Thank you for the reply, it is something that has nagged me for a few months. I whole heartedly agree with and see the advantages of open-source software because of the freedom it brings, just not as it extends to device drivers.

If the philosophy is that important, then why use chips that have proprietary instruction sets? Why has the open-source community designed and built open cpu chips, chipsets, gpu, open variants of PCI, and so on.

I am not trying to be obtuse, but it seems that it is extremely hypocritical to use open-source video drivers, regardless of the performance, yet almost everything inside the computer is proprietary.

Like I said earlier, every single program gets converted to a proprietary instruction set that executes it.

[kresh]

Sorry:

"Why has the open-source community designed" should have been "Why has the open-source community not designed", I missed the one minute edit.

[Wyatt]

Thank you for the reply, it is something that has nagged me for a few months. I whole heartedly agree with and see the advantages of open-source software because of the freedom it brings, just not as it extends to device drivers.

If the philosophy is that important, then why use chips that have proprietary instruction sets? Why has the open-source community designed and built open cpu chips, chipsets, gpu, open variants of PCI, and so on.

I am not trying to be obtuse, but it seems that it is extremely hypocritical to use open-source video drivers, regardless of the performance, yet almost everything inside the computer is proprietary.

Like I said earlier, every single program gets converted to a proprietary instruction set that executes it.

It's impractical.

Let's get one thing out of the way real fast: fabrication plants cost billions of dollars. Low-level hardware design costs millions of dollars. What you propose is maybe in the realm of a certain Mr. Gates, not a community of software developers.

Anyway, the way I see it, hardware is hardware; software is software. How hardware does things on the metal is not important to the end user or probably 99% of the hackers out there. Whether their driver can or will be supported two or three years from now or not is important. Whether the driver has proper mode-setting and OpenGL support is important. And if it's open, someone can fix it. And that's exactly what's happening.

One thing that's now puzzling me: how much more open do you need it to be if you've been given pretty much every bit of documentation regarding it? Or if the company that makes the hardware makes an open driver and actually submits patches to the kernel to make sure it functions. And works with the community to re-architect the entire graphics stack so everything can improve?

[BlackStar]

If the philosophy is that important, then why use chips that have proprietary instruction sets? Why has the open-source community designed and built open cpu chips, chipsets, gpu, open variants of PCI, and so on.

This would be an open-hardware community and, yes, such a thing exists. There are people who collaborate to design their own hardware, releasing the necessary specifications to the world. The Ronja project is a successful example:

Ronja is a free technology project for reliable optical data links with a current range of 1.4km and a communication speed of 10Mbps full duplex.

However, the logistics are so mind-boggingly hard, that it's very very difficult to achieve results on regular PC hardware. Take a look at the Open Graphics project, for instance, which has been in development since 2006 (probably even before, but that's the first reliable reference I could find).

Open-source drivers are at least an order of magnitude easier, which makes for a good compromise.

[bridgman]

If the philosophy is that important, then why use chips that have proprietary instruction sets? Why has the open-source community designed and built open cpu chips, chipsets, gpu, open variants of PCI, and so on.

I am not trying to be obtuse, but it seems that it is extremely hypocritical to use open-source video drivers, regardless of the performance, yet almost everything inside the computer is proprietary.

As others have said, it's primarily a function of development cost. A new GPU generation costs $100M or more to develop and bring to market; a new CPU generation may be even higher. Most of that cost is labour, of course, so in principle a volunteer community could do it for a lot less, but even things like mask charges and test wafers end up running into the millions for a cutting-edge part.

If you wanted to do something that was off the cutting edge, the development costs could be substantially lower. The remaining difficulty seems to be that very few people are willing to pay for something that is not "pretty close" in terms of cost/performance, and without huge volumes the per-unit cost goes up rapidly which makes the cost-performance equation even worse.

The OpenGraphics project attacked that problem by using FPGA parts rather than custom-designed ASICs, which makes small volume production possible without multi-million dollar up front costs, but the downside is that the available processing power is quite a bit less than a regular GPU, maybe 1/50th as much.

Open source software projects can be smaller (one top-notch developer can accomplish a lot on their own) and can get close to (or exceed) the performance of proprietary software, while open source hardware either requires a massive up-front investment or ends up being much further away from proprietary solutions than is the case for software.

There are SPARC processors available (from Fujitsu, I think) which implement a fully open processor design, ie all of the internal logic has been made publicly available and can be modified by "anyone". Combine that with an OpenGraphics card and you would have a good chunk of your system being free-as-in-speech.

The problem is that economies of scale get you, so the result is further from free-as-in-beer than most people are willing to live with.

[kresh]

Whether their driver can or will be supported two or three years from now or not is important. Whether the driver has proper mode-setting and OpenGL support is important. And if it's open, someone can fix it. And that's exactly what's happening.

Ok, that I can understand. Thanks everyone for the responses. I really appreciate it.