AMD Develops New Linux Open-Source Driver Model

An anonymous reader writes "AMD privately shared with Phoronix during GDC2014 that they're developing a new Linux driver model. While there will still be an open (Gallium3D) and closed-source (Catalyst) driver, the Catalyst driver will be much smaller. AMD developers are trying to isolate the closed-source portion of the driver to just user-space while the kernel driver that's in the mainline Linux kernel would also be used by Catalyst. It's not clear if this will ultimately work but they hope it will for reducing code duplication, eliminating fragmentation with different kernels, and allowing open and closed-source driver developers to better collaborate over the AMD Radeon Linux kernel driver."

Comments

AMD

[ArcadeMan]

Better integrated GPUs in lower-cost CPUs. Why choose Intel?

Re:AMD

[epyT-R]

Why choose intel?
1. better performance in just about everything
2. lower power consumption
3. more overclocking headroom
4. chipsets 'just work' and don't tend to have bizarre compatibility problems.

Re:AMD

[aztracker1]

For *most* people, I would go with AMD, imho for general purpose at the mid-range, you get a lot for the money with AMD.. so for my parents, grandparents, and the like I would use AMD. Same goes for a starter kids system... you can start with the integrated APU and throw in a dedicated GPU a year or two later. AMD sockets tend to stick around forever, and can generally get an upgrade in.

On the higher end it really depends. If you are doing software development, then the multi-core support of an 8350 might work better for you than an i7. If you are wanting to do video/image editing on large images/files, then the i7 will be a much better option. Video transcoding can go either way and is dependent on the software you are using. If you want to create a hackintosh, then the i5/i7 is really the only way to go.

Re:AMD

[invictusvoyd]

And better beeetcoin mining .. zzz just kidding .. but nvidia sucks at bitcoin

Re:AMD

[hobarrera]

Why choose intel?
1. better performance in just about everything

If you're willing to pay. Not everyone need the top performing CPU on the market.

2. lower power consumption

A very valid point, but irrelevant on the desktop area (unless you live somewhere where electricity is expensive, maybe?). Quite valid for laptops though.

3. more overclocking headroom

Really? You think that will make people choose Intel? 90+% of the people don't even know what overclocking is.

4. chipsets 'just work' and don't tend to have bizarre compatibility problems.

As do AMD's.

Sure, there's plenty of scenarios where Intel is a better choice. But there's plenty of others where AMD is the best choice.

Re:AMD

[TheRealLifeboy]

This is good news indeed. The more open source drivers we use, the better. At least it is then possible to inspect and improve in principle, whereas with proprietory drivers, it's practically impossible.

Thank you, AMD, for trying to make the world a better place!

Re:Why, oh why?

[Anonymous Coward]

The proprietary driver includes licensed/patented code that they can't legally use in an open source driver.

Good for them

[Ynot_82]

Seems AMD have taken on-board what Nvidia chose to ignore.
Being the advice offered by the Kernel devs

http://lists.linux-foundation....

Re:Good for them

[Anonymous Coward]

Go fuck yourself.

Intel

[Anonymous Coward]

Lower power consumption in better CPUs. Why choose AMD?

Re:Intel

[Rich0]

Lower power consumption in better CPUs. Why choose AMD?

The most recent generation of chips definitely has power issues, but it seems like you get a lot more bang for your buck with AMD. What can you get that is decent from Intel for $120? You can get a fairly decent chip from AMD for that.

Re:Intel

[ArcadeMan]

Hell, a 40$ APU from AMD beats anything Intel has to offer for almost twice as much, not to mention that even at that price the AMD has a decent GPU while the Intel has none at all.

Re:Intel

[Rich0]

Hell, a 40$ APU from AMD beats anything Intel has to offer for almost twice as much, not to mention that even at that price the AMD has a decent GPU while the Intel has none at all.

This is why I think AMD tends to be represented more in the DIY arena. Companies like Dell don't want to sell you a $150 CPU+motherboard upgrade. They want to sell you a $1200 PC. If you're going to throw away your old case, PSU, video card, RAM, hard drive, DVD drive, etc - then you might as well spend an extra $200 on the CPU.

On the other hand, if you're only upgrading CPU+MB, and maybe RAM, then AMD makes a lot more sense. If my options are to spend $500 every 6 years on an Intel CPU+MB, or $150 every other year on an AMD CPU+MB, then I'll take the latter. I'll actually spend less money, and for most of the time I'll have a better system. Sure, the Intel system will outperform the AMD system in years 1-2, but the AMD system will outperform in years 3-6, and by a huge margin in the last two years. A CPU is a rapidly-depreciating asset, so it doesn't make a lot of sense to spend a lot of money going high-end - you're far better off buying something moderate and replacing it more often. Then Moore's Law will work for you, and not against you.

Re:Intel

[K. S. Kyosuke]

Why choose AMD?

Because they're trying to wean us off this IBM PC/x86 nonsense?

Re:Intel

[ArchieBunker]

And a midrange i5 CPU will beat AMD's top of the line FX-8350, so what is your point? Both of those cost the same.

Re:Intel

[epyT-R]

To be fair, intel did try this, too (ia64). It was amd that extended x86 to 64 bit land.

Re:Intel

[Rich0]

The FX-6300 is actually faster in multi-threaded cases, which isn't surprising since it has 6 cores vs 2. The single-threaded performance of the Intel processor is clearly better. So, I'd call that one a bit of a toss-up depending on needs.

The Intel chip definitely uses less power though - that is one thing I definitely don't care for in my Phenom II systems - heat management is a real problem.

Re:Intel

[hairyfeet]

What you fail to mention is the fact you could have probably built most if not all of an AMD 6 core system for the cost of the i7-860 alone, and if one were to go one notch down to the quads (which lets face it software just hasn't kept up with hardware and even more triple cores spend more time at idle than under load) then one could have easily built the entire system. The 8 cores were built just to say they could and frankly have never been priced competitively, no different than those 5Ghz chips they released not too long ago.

My oldest boy has an AMD hexacore, 8Gb of RAM, a 1Tb HDD and an HD7750, the whole system ended up costing something like $375 after MIRs. It kicks ass at gaming, transcoding, hell it'll do anything your average user will be able to think up for a PC to do and do so quite well and cost a grand total of $70 more than your i7-860 BY ITSELF without so much as a stick of RAM or anything according to Intel.

So if you want to sit here and argue that you are one of the 1% that actually NEED every MHz of speed you can get, which until we see benches done that are compiled with GCC I wouldn't trust the benches BTW but that is a different story, then fine, do so. But the bang for the buck is so far in AMD's camp right now it isn't even funny, you can choose from several quad cores including fully unlocked for less than a Pentium Dual, and if all you care about is power the AMD Jaguars spank the Intel Atoms on performance while using less than 25w for a quad. AMD is just a crazy deal ATM which is why I've had no complaints when it comes to being AMD exclusive, the customers get great performance at a great price.

Re:Intel

[aztracker1]

Funny, but I changed from an i7-860 to an FX-8350 last year... It was because the motherboard went bad, not because it was too slow... I find that the 8350 does a lot better for the work I do (web application dev, generally with a couple of different DBs in the background) I have 32GB of ram, but it's only seeing 24 for some reason... next time I wipe and re-install, will probably update the bios. That said, it is still leaps and bounds better than the i7 for me. I cannot attest to gaming, as I don't really game much at all.

Re:Intel

[aztracker1]

Well, they seem to be shipping enough GPUs for the likes of the XBone, and the PS4 ... That said, I think if they had the exclusivity with Dell that Intel has had, then they would be able to supply. I see a lot of AMD from most other mfgs, and sales overall are down for everyone, including Intel.

Re:Intel

[Anonymous Coward]

The problem is though that AMD's single core performance is not just lacking, it is appalling. a quad core i7 will easily outperform a hexacore AMD on multi threaded loads and it will absolutely cream it on single threaded/sequencial loads. AMD has it's advantages, but performance an longevity aint it. If you are trying to save a few dollars and bang for buck is most important than AMD does well. But if you want performance then AMD aren't even playing in the same league.

Re:Intel

[Blaskowicz]

AMD's use of Hypertransport makes it a bit easier to create fast low latency interconnects, needed for high performance clusters or supercomputers.

Re:Intel

[Rockoon]

A comparison amongst users in typical rigs, rather than in review-site benchmark rigs.
i7-860 1217 / 5110 (812 samples, single threaded / overall)
FX-8350 1512 / 9049 (3149 samples, single threaded / overall)

The AMD in question is winning against the Intel in question in single threaded, and winning greatly in multi-threaded. However this AMD chip, at $200, is not really what the GP was talking about. He was talking about ~$150 APU's that also saves him money on a video card.

Comparison of the i7-860 vs the A10-6800K

i7-860 1217 / 5110 (812 samples, single threaded / overall)
A10-6800K 1555 / 5006 (205 samples, single threaded / overall)

Re:Intel

[Kjella]

Humm... a site that offers absolutely no actual benchmarks just a mysterious performance number (check their FAQ) with zero ability to reproduce or verify, submitted by users with all kinds of overclocked rigs that's credible. For example it claims the FX-8350 has much better single thread performance (1,512) vs (1,217) which is not supported by any serious review I can find. So either the whole world is in a conspiracy against Passmark, or these numbers are a joke. I wonder how much AMD paid them to invent these numbers?

Re:Intel

[Mashdar]

I would say AMD generally prices their parts competatively. If you are talking about >$300 Intel parts, you are correct that AMD has nothing to offer (I don't count 220W parts as viable, as I'm not in the market for a desk-side-vacuum-cleaner). But at $180 an FX-8350 looks pretty competative vs a $200 i5-4570:
http://www.tomshardware.com/ch...
If you are using efficiently multi-process applications (e.g. video compression), AMD is the clear winner. If you are using mostly-single-process applications (Blizzard games?), Intel is the clear winner.

In my usage, single-process applications tend not to be CPU-bound, or they tend not to be computationally taxing. But YMMV. And some games are obviously highly 1-2 core CPU bound (Blizzard), which is worth considering.

Finally points:
Over clocking: If you are planning on overclocking, the least expensive intel part is $240 (33% over FX-8350). Overclocking won't close the FX-8350 single-threaded performance gap, but it helps.
Heat: The FX-8350 is rated at a TDP of 125W... The i5-4570 is rated at 84w. So AMD is hotter and louder.

Disclaimer: My next system is going to be Intel, primarily because I want the machine to be near-silent, and 125W is hard to work around.
Note: All prices based on Newegg at the time of writing.

Re:Intel

[ArcadeMan]

As opposed to all the "official" benchmarks that have been compiled to give Intel an edge against AMD?

Re:Intel

[Rich0]

And remember, this is comparing against a previous generation, lower end than needs be Intel chip.

Looking on Newegg I don't see much else at the arbitrary $120 price-point.

Of course, if I were actually buying an Intel chip I doubt I'd buy one at that price. That was basically my point - the Intel chips are much better when you get up there in price, but at the lower end the AMD chips are fairly competitive. In the current generation Intel has more of an edge - 10 years ago AMD was much better in this space.

Somebody else argued that with the right optimization AMD does much better than the benchmarks suggest. I don't have any data to support that one way or the other, but my main amd64 box is running linux with everything built -march=amdfam10, so if that is the case then I'm better off for it...

Re:Intel

[TheRaven64]

I chose AMD because Intel is very aggressive about market segmentation. I wanted a MiniITX board with a low-power CPU and (at least) 4 SATA ports (so I could have RAID-Z and an optical drive). Intel Atom boards all restricted the number of SATA ports (a quick look now seems to show that they all come with 2 ports). If I wanted Intel and 4 SATA ports, I was pushed to the more expensive i3 range.

Re:Intel

[zippthorne]

Genuinely curious here.. what kind of salary range do you think is appropriate for a person to be spending $200+ on CPU alone?

Re:Intel

[Rockoon]

For example it claims the FX-8350 has much better single thread performance (1,512) vs (1,217) which is not supported by any serious review I can find.

Its supported by the one you linked to, whose results you either clearly misunderstood or completely misrepresented on purpose.

You claimed that the Intel won 9 benchmarks while the AMD only won 4, but in actuality when someone carefully looks at the results its the Intel that only won a single benchmark, not the 9 you claimed, while the AMD won the other 12. Note where it says "lower is better" and "higher is better" you are supposed to use your brain in order to interpret the results rather than misrepresent them.

So you are either a complete fool that nobody should pay attention to, or a liar that nobody should pay attention to.

I tried to give you a pass on it. You didn't want that... so now here is my wrath you ignorant lying fuck. For all the other people reading, here is the benchmark this dumb fuck linked to and then immediately lied about.

This lying fuck claimed that the Intel won 9 of the 13 benchmarks in the link, but the Intel actually only won 1 of the 13. Right there in plain site, from the ignorant lying fuck Intel fanboy.

Re:Intel

[Bengie]

Except AMD's "TDP" is what you should see in the real world and Intel's "TDP" is worst case and should almost never see. Almost every comaprison that I have seen of AMD's CPUs from the past 3 years have all consumed about 2x the power for the same amount of work, no matter the performance.

On idle vs load, AMD is running about 25watts per hour more idle and about 60watts per hour more under load. Assuming 22 hours per day idle and 2 hours full load, that's about an extra 244 KWHs per year extra for AMD. For a cool climate, that's fine, but if you have to AC any of that, now you've at least doubled your power, so about 500KWH more, or about $50/year assuming $0.10/KWH, all in the attempt to save $50 on your computer cost. In the end, Intel pays itself back in 1-2 years, not to mention faster in most cases, many times nearly 100% faster.

Re:Intel

[hairyfeet]

Sorry but I have to call bull. I have an AMD hexa (Phenom II 1035T if it matters) and idling according to IES (the Asrock OCing tool) my board is using around 8-16w simply depending on when little background tasks kick up one of the cores and under full bore I have yet to be able to hit 80w and that was with a 6 core WMV to AVI video transcode which is pretty damned CPU intensive if you have ever worked with WMV files.

And where do you get $50 in computer costs? as I said I built a complete hexacore system for $370, for just $70 more than the i7-860 BY ITSELF. When you put in parts that are equal feature wise, board that has 4 RAM slots and dual PCIe for dual GPUs, 1Tb HDD, 8Gb of 1333Mhz and a DVD burner along with a nice case and decent 500w PSU? You are talking DOUBLE what the system cost NOT $50!

I'm sorry but you can twist the figures all you want but bang for the buck is in the AMD camp by a very large margin and has been since the release of the Athlon X2 back in early 09. The only thing you are gonna get at a similar price from Intel is severely gimped chips, no support for ECC, limited to no VM support, locked multipliers, gimped cache, lower speed, lack of turbo, their low to mainstream offerings are frankly lousy except for office boxes and those only playing single thread games.

Re:Intel

[TheRaven64]

I bought an AMD E-350 motherboard and a Chenbro case with 4 removable drive bays. It takes a slimline optical drive (which actually has a BD-RE drive, because they weren't much more expensive than DVD drives, although in the 3 years since I bought it I've yet to burn a single BD).

It's running FreeBSD, booting from a RAID-Z array (so cheap snapshots, block-level checksums, single drive failure recovery, and all of that good stuff). The GPU is now supported and so it runs XBMC connected to my projector for video playback too.

There's also an e-SATA port that I should be using for external backups, but am not currently...

Re:Intel

[hobarrera]

With Intel I get 12hs out of my laptop (MacBook Air 2013). I'd never go AMD on a laptop.

Then again, you're right that you get more bang for buck: my desktop is AMD. Eight 3.0Ghz cores for a very low price. But price isn't always the only factor.

Re:Intel

[hobarrera]

Look like you've missed the last two generations of Intel HD video. An Intel HD 5000 is a VERY GOOD video card, even for some gaming (though not sure about last-gen gaming).

Re:Why, oh why?

[EvanED]

In addition to what other people have said, GPU drives contain shader compilers and probably other kinds of optimization routines which actually give a "competitive advantage".

This approach has been tried before

[Black Art]

The X community has said specifically that this sort of end-run around the GPL is strictly forbidden. I expect yet another flame war over this at any moment.

Re:This approach has been tried before

[Ynot_82]

The Radeon driver is under MIT, not GPL.

Re:This approach has been tried before

[Rich0]

The X community has said specifically that this sort of end-run around the GPL is strictly forbidden.

Ultimately, only a court can say what is or isn't forbidden. I don't buy the whole can't-link-to-GPL-code argument. If AMD isn't distributing the kernel, then the kernel's license is irrelevant. The fact that AMD's code references kernel symbols doesn't make it a derivative work, unless you buy into the SCO argument that things like #defines are copyrightable. I could see the argument that after the linker resolves the symbols you end up with a derivative work in RAM (maybe), but that image isn't distributed anywhere (certainly not by AMD).

As far as I'm aware no court has ruled on this either way. It would be interesting to see how this goes.

Phoronix must have juicy drivers

[Anonymous Coward]

if they state Catalyst exposes OpenGL 4.4 instead of the 4.3 that is the reality for the rest of us. Minor nit but...

It only makes sense.

[Anonymous Coward]

Linux has roughly 10% of the desktop market in the United States. You'd think that this kind of story wouldn't be so uncommon. As soon as SteamOS is released M$ won't have a chance. Windows 9 will likely be the last version of Windows.

Re:It only makes sense.

[ArcadeMan]

Linux has roughly 10% of the desktop market in the United States.

Do you like to pull numbers from your ass?

If Apple only has around 10% of the desktop marketshare in the USA, Linux has 0.1% at best. In some European countries such as The Netherlands, however, I'm pretty sure Apple's marketshare is much lower and Linux is much higher.

Re:It only makes sense.

[anon mouse-cow-aard]

The actual desktop share for Linux in the US is around 2% by most measures... so the while grandfather might have been off 5x, you were off by 10x... http://gs.statcounter.com/#os-...

Re:It only makes sense.

[zippthorne]

SteamOS handles MS games by streaming them from another machine running Windows...

Re:It only makes sense.

[ArcadeMan]

Given that page counters usage is down and Google Analytics usage is up, I wouldn't trust the numbers from statcounter.

Re:It only makes sense.

[anon mouse-cow-aard]

I wrote by most measures... this one like shows 1.47% share world-wide... 8x higher than your number. http://www.netmarketshare.com/... You accuse someone of pulling numbers out of a body cavity, then do exactly the same thing? When presented with some kind of data you just disparage it... How about some data that backs up your claim?

Re:It only makes sense.

[ArcadeMan]

Google doesn't publish global stats, so I can't quote anything.

Re:Does AMD still matter?

[cbhacking]

Well, both of the high-performance current-gen game consoles use AMD. On that basis alone, they're quite unlikely to go anywhere.

Re: Does AMD still matter?

[Jezisheck]

Of course it matters. For instance, I've been using AMD chips in all my computing devices since like 2005 after just horrible experience with my first Celeron (single-core, 2.0 GHz). AMD has proven to me over the time that they can make competitive CPUs (heck even GPUs, I got Radeon instead of nVidia too) and ask far less money for it than Intel would. And that's what counts for me cause thing like lifespan or power efficiency I can't really measure by myself.

Re:Does AMD still matter?

[wjcofkc]

AMD still has roughly 20% market share, so I would say yes, AMD still matters. It's a pretty big market for what can be considered only two-players. AMD certainly has it's struggles, and with the prospect of an NVIDIA\Intel APU alliance on the horizon competition is going to get tougher. But even then they will probably be the "most bang for your buck" option which is a large but not exclusive part of what is keeping them alive. I use AMD APU's, but as a Linux user the first thing I do is disable the Radeon portion in the BIOS on pop in an NVIDIA card. My current frequency-unlocked quad core is no slouch and I hope to be getting an 8-core AMD in a couple of months.

Re:Does AMD still matter?

[drinkypoo]

Honestly I have tried Cyrix, I really tried with AMD but at the end I have just gone back to intel.

All my PCs have had AMD processors for ages now, but I'm with you finally. I am tired of the poor power management support in the Linux kernel for AMD, and I don't care whose fault it is.

This new move from AMD is the only thing that could possibly make them relevant in my eyes in graphics, but only if they actually follow through and succeed...

Same'ol firmware separated from kernel module appr

[sgt scrub]

It doesn't sound all that different. Everyone wanting to stay proprietary has a separate firmware to load by the module; but, the proprietary module has a few more whistles than the one from the community. Is this not the same thing?

Licensing issues with opening the code

[Anonymous Coward]

That point was specifically addressed the the article. It isn't that AMD can't legally open up the code, but that is where they view their secret sauce of optimizations and tweaks. If they were to open it up, they fear that competitors might get a leg up and be able to use the same tricks.

Re:Licensing issues with opening the code

[epyT-R]

that's a stupid place to hide secrets.. I guarantee that nvidia has already reverse engineered the driver. open the code and put it under a license that prevents it being used with anything but a radeon graphics card, which is basically what's going to happen anyway. At least then the code can be built/rebuilt/ported/debugged with the rest of the system when needed.

Re:Licensing issues with opening the code

[TheRaven64]

That doesn't really make sense. The architectures of the three main GPU vendors are sufficiently different that it's highly unlikely that the can benefit much from tricks in the drivers (which are basically compiler optimisations these days, since the performance-critical part of the driver is the shader compiler). nVidia uses something that looks more or less like a scalar CPU (with some SIMD) but with SIMT so that if you run threads in lockstep they just do one decode but execute several steps at once. AMD is a VLIW-inspired design. Intel is a more classical SIMD architecture, but with the tweak that their vector register set is 2D and each operand encodes a start and a stripe size, so you can do vertical or diagonal stripes through the register set and don't have to do vector permutes. Intel and AMD put more of the parallelisation into the compiler (drivers) than nVidia, but the transforms that they do are quite different.

There's more of a chance that AMD drivers would help Broadcom than anyone else, but AMD and Broadcom aren't really in the same markets, especially since AMD sold Qualcomm their mobile GPUs (Radeon became Adreno)

Re:Does AMD still matter?

[zacherynuk]

And the thing is... power management is the ultimate cost, isn't it ? You can't tell a dev, trader or (higher) management to shut their machine down. like you can a 'user' (who we can just policy to shutdown ) but 1000+ rigs running even at trimmings is still $50 / Amp. That shows in finances, bonuses, dolphins and trees.

Phoronix... sigh.

[Anonymous Coward]

As I've been Linux user for a while now, and am well aware of the articles that Phoronix put out on their god-awful website, is there a particular reason that AMD and other companies seem to cater to them on this front? As far as I can tell, they don't do much that's special other than being an announcement portal of sorts. If their claim to fame is just the latest 'bleeding edge' of graphics support for games on the Linux desktop and slapping it up on an advertised site, I guess I get it. If they're target audience is Linux desktop rookies pontificating of about framerates while waiting for the dev's to up the performance, they seem to have hit the mark. I'll admit, there was a time early on, pre in kernel radeon and AMD doc's share that they hit the mark on a few articles, but since then, I don't see them contributing much that's new to the conversation. I just don't see them doing much more than what you would find several thousand Linux users doing with their various dists. to squeeze more out of the graphics front. I guess they just have the name recognition. I can't be the only one who has heard them referred to as Moronix. Am I?

Maybe it's their hideous website. Maybe it's the ho-hum web illusion that they're an analysis site. Maybe it's just me. It just seems like they get special treatment for, well, not being special.

/This isn't a troll, I didn't RTFA, and I'll probably never go to their site again. Just a perspective...

Re:Why, oh why?

[EvanED]

Well, here's the thing... to at least a moderate extent, Intel isn't really competing against AMD or nVidia, because unless something has changed relatively recently, they don't have anything that comes even close to the offerings of the latter in terms of performance. So if AMD or nVidia learns something about how Intel chips works and improves their own a bit as a result, they're not going to take away much business from Intel. On the other hand, if AMD open-sourced the guts of their driver and nVidia learned enough to raise the performance of their own cards by a few percentage points or something, that'd be a somewhat big deal.

The complement to this argument is that because Intel can't win customers based on performance, they have incentive to seek other distinguishing factors. One of those factors would be openness and Linux support.

Re:Does AMD still matter?

[K. S. Kyosuke]

I use AMD APU's, but as a Linux user the first thing I do is disable the Radeon portion in the BIOS on pop in an NVIDIA card.

That would be an extremely dumb thing to do with the current generation of APUs, what with all the extra performance the tightly-integrated cores can give you for anything not related to graphics.

Re:Does AMD still matter?

[edxwelch]

That article for some reason ignores the last 2 quarters where AMD turned a profit. People have being saying AMD is finished for the last ten years, but yet it's still here.

Re:Why, oh why?

[Kremmy]

Intel has decent enough open source drivers, until you get to the chips like the GMA 500 that just don't quite DO that. Intel has done pretty good with Linux support overall, but they definitely haven't solved the problem.

Re:Does AMD still matter?

[cheesybagel]

They were worse off a couple of years back. Now that they have the Xbox One and the Playstation 4 as a steady revenue stream they won't go bankrupt any time soon. Stock ticker prices are also mostly irrelevant. AMD historically has been a company which seldom has had a single profitable year. If you take their financial performance as the sole indicator they should have gone bankrupt decades ago.

I don't know about today but when I bought my AMD CPU a year or two back they had the higher integer performance and the FP performance matched Intel on a price per price basis. i.e. Intel had processors with better FP performance but they cost 2x or 3x more. Not worth the trouble when most FP intensive applications have been changing to use GPGPU acceleration.

Re:Why, oh why?

[TheGavster]

The GMA500 driver might be doing fine, but for some reason they keep licensing third party graphics for the integrated solution on Atom processors, or at least the ones making it into industrialized products. Getting OpenGL up under Linux on PC/104 or other embedded board is a royal pain in my experience.

Re:Does AMD still matter?

[wjcofkc]

I'm not even drinking yet I'm honestly not 100% sure what you are saying. I've tested it and CPU benchmarks are the same with or without using the integrated graphics. I'm not a gamer so I use cheap, lower end cards. I just really really hate dealing with Linux AMD drivers. I also dislike shared memory.

Re:Same'ol firmware separated from kernel module a

[raxx7]

Not sure what you're thinking..

A Linux graphics card driver has 3 components: the kernel module, the X module and the libGL/CL/etc implementation.
There are two AMD driver for Linux -- the proprietary one and the open source one, each with it's own 3 components.

The proprietary one offers better OpenGL/OpenCL performance and features (eg, OpenGL 4.4 instead of 3.1), as well as official certification for a number of applications.
But it also tends to suffer from system integration issues, at the kernel and X level. Sometimes, they work poorly for basic things, they don't work with the latest kernel or X for a while, etc.

So, what looks here is that AMD wants to reduce the proprietary to the libGL/CL component and leverage on the open source for the kernel driver. Maybe X driver too, eventually.

VIA?

[ArcadeMan]

Is VIA still in the race, at all? Last VIA I heard about was the C7, so it's been a hwile.

Re:Why, oh why?

[DarwinSurvivor]

Intel's also does not make powerful graphics cards anywhere near as fast as amd or nvidia.

Re:Why, oh why?

[blackest_k]

they have some shockingly badly supported gpu;s on netbooks. that with cinimon are running the cpu at 250% at idle due to software rendering. with mate its around 5% idle. But no 3d anything and im sure the gpu could work quite well

Re: What?

[Anonymous Coward]

How do you think the open source radeon driver came to be?

Re:Does AMD still matter?

[K. S. Kyosuke]

I'm HSAying what I'm HSAying. ;-)

Re:Why, oh why?

[Kjella]

The GMA500 driver might be doing fine, but for some reason they keep licensing third party graphics for the integrated solution on Atom processors, or at least the ones making it into industrialized products. Getting OpenGL up under Linux on PC/104 or other embedded board is a royal pain in my experience.

Open source support depends on whether they're licensing it from PowerVR or using their own in-house graphics. If you check this page Intel now use their own graphics in "Valleyview" systems, which should be much more open than before.

Re:Does AMD still matter?

[Kjella]

Well, AMD the company is quite different from AMD the Intel-competitor. While they seem to have stopped the downward trend for now, they're doing it by divesting their CPU/APU business and ramping up lots of semi-custom business like consoles and such. It might be a way for AMD to be profitable but large parts of the market will be left to Intel's mercy.

Nice, if now they only fixed their driver's issues

[Agram]

...I would put a lot more trust in this development. Things like hangs and inability to standby (5+ year old problem) or more recently brightness control that worked until approx 14 months ago and since then was never (fully) fixed despite dozens of bug reports. I mean, this is a simple matter of comparing the code for brightness between the version 14+ months ago and the latest one to figure out what is the problem and then fixing it once and for all... Instead, they announce "fix" for it in two consecutive versions, neither of which address the problem in its entirety, and consider it fixed... Yes, some will argue open-sourcing this may help fix things faster. My experience tells me otherwise whenever you have this level if incompetence involved, because after all it is that same incompetence that will drive the separation of open and closed components... Downvote or not, I would love to be proven wrong so that I can finally install a fglrx driver that actually works as it should.

Re:Does AMD still matter?

[Anonymous Coward]

You might have said the same thing about IBM's consumer PowerPC business relatively recently too ;)

Being the chip in current consoles is *far* from a demonstration that you're not going anywhere. In fact, it's a sign that you're willing to give tech away for super cheap in order to get bulk orders in.

Re:Does AMD still matter?

[Billly Gates]

Yep

Typing this on an AMD Phenom II x 6 core system with an ATI 7850 GPU. I got it in 2010 for $599 and it is a freaking 6 core with full virtualization at that price ... with an Asus board!! ... ok the gpu at the time was a ati 5750.

I do vmware workstation for linux and website testing. I need lots of cores!

An icore7 extreme with a non crippled bios would of nearly trippled the freaking price. Nvidia was pricing graphics cards at $1,000 before the latest Radeons x290 outperformed them for half the price. This was before the Crypto minners rose the price up to $700 but still.

Right now ATT/AMD makes great value oriented solutions for cheapskates with the $499 walmart special since the gpu runs candy crush in Metro really well and basic office needs. I use the more powerful cpu;s.

As for shops. Hairyfeet here uses them at his shop exclusively due to value and price as he would ahve to charge more for his users. The vast majority of users are not yuppie gamers who make $70,000 a year nor advanced workstation cad engineers or video editors who want an icore7 extreme. For the secretary or grandma they are the best value.

I do however admit my system is showing it's age :-( Skyrim is not the best on ultra settings even with the recent video card. I would not want to play the latest Crysis either on this. But SWTOR, WOW, office, and VMWare workstation still work fine after I put in SSD pro drives and 16 gigs of ram.Cheap hardware virtualization rocks.

Re:Does AMD still matter?

[Billly Gates]

I have never seen an AMD system crash except once due to the 13.1 driver issues on my system.

They run just fine and Nvidia seems to have the worse drivers this day and age. Shoot they have bricked some cards actually!

A car analogy.

[gukin]

Intel is akin to a Toyota Camry as AMD is a Scion FR-S. On the Camry side of things, presuming the Camry has a V-6, it'll utterly smoke the FR-S, in the quarter mile and 0-60, it's much more practical, room for six and a big roomy trunk. The FR-S is less expensive but less practical, sits lower, a much smaller trunk a non-existent back seat and has proven to be much less reliable. Then again, the FR-S is a zooty two-door, rear wheel drive, and an utter hoot to drive, while the Camry is . . . a Camry.

Having just purchased an AMD Kaveri A10-7850, I've been having fun playing with the newness of the chip, yeah, sensors don't work, the Radeon (ATI) drive is butt-slow and Catalyst is in beta. Still I'm pretty sure that Kaveri has more under the hood than is initally obvious. For Radeon cards in the 4000,5000,6000 range, the open-source driver is neck and neck with the proprietary Catalyst driver. The 7000-8000 R7, R8 series has a ways to go for now but if those two drivers can start sharing more, everyone wins.

Just to let everyone know, my new Kavei is about as fast as my Intel Core I7-920 in most things and faster in others. As for gaming, I'm a Linux weenie, how many AAA games that really stress a GPU are even available (yet?) for Linux. Yes my gaming system is an Intel I7 and an nVidia 660 TI video card while my play system is my new Kaveri. I enjoy playing with Linux, trying out the bleeding versions of Mesa and the latest x11-driver-video-ati but if I'm going to waste time playing games in Windows, it had better "just work".

Intel needs AMD to keep from being a monopoly so instead of bashing either company, embrace them both, it is a nerd thing after all.

Re: What?

[aztracker1]

I'm sorry, but realistically proprietary GPUs aren't a big issue... the PCIe slot is pretty standard, and they all come on boards that fit into that slot, and beyond that, the odds of a regular person spinning up a fab order for non-proprietary CPU/GPU/board combinations is just plain unlikely. Having open software that works with said hardware is a *much* bigger issue... unless you are planning on using some under-powered GPU that will have trouble with 1080p 3D that is. That said, having open drivers allows for faster and better integrated updates from linux distributions, and more options as a whole.

Re:Does AMD still matter?

[bloodhawk]

The PC Market even it is current state sells 70 million+ units a quarter, The PS and Xbox lifetime combined sales won't even come close to even this years reduced PC sales numbers. If AMD are reliant on the cut price console industry to help them then they are in dire trouble, at best the consoles provide a some pocket money as they will be selling in bulk orders with very low margins.

Re:Because the open source drivers suck so badly

[kenshin33]

? "tainted" is just a flag, nothing more. It's there in case womeone sends them a bug report (a crach for instance), if the kernel is "tainted" their answer will probabaly be : try again without the "tainting" module and see what happens : if the problem goes a way they'll send you crying to the one responsible for THAT module, periode.
There's absolutely NO denying anything to any module.

Re:Does AMD still matter?

[Blaskowicz]

HSA means maybe some limited parts of some applications will be sped up, in an indeterminate future, if code has been specifically written for it and if the linux support is good enough. And it is only for A10 7850, 7700 and 7600, the latter of which isn't available. And distros need to ship code that runs on every CPU.

Re:You probably cannot safely use the GCC compiler

[BanHammor]

I think there were rather understandable exceptions to "tainting everything it touches". Like the Runtime Exception, seen here: http://www.gnu.org/licenses/gc...

Re:Does AMD still matter?

[edxwelch]

The console market is guaranteed cash in the bank, with no competition from any other player. There's few companies that have a deal like that.
And the margin is quite good considering that Microsoft and Sony shared in development costs.

Re:Does AMD still matter?

[ArcadeMan]

I'd be interested in knowing the true numbers because all AMD videocards prices have gone through the roof in the last few months. As an example, the ASUS R9 270X was available for around 240$CAD in january but today it's selling for 289$CAD after a peak of 329$CAD a few weeks ago.

Re:You probably cannot safely use the GCC compiler

[Rich0]

The ones and zeros that comprise your binaries were copied into "your" object files NOT from your source code but rather from Mr. Stallman's compiler and linker. Your binary files can, therefore, be viewed as very complex derivatives of HIS code, or at the very least as having been linked with thousands (millions?) of fragments of his GPL'd compiler and linker programs.

Sure, but if you build the AMD video drivers nothing but the symbol names and parameter types impact what ends up in the compiled binary. It need not even be built on a system running linux - the only thing the compiler needs access to is the kernel header files. That's just a big list of function prototypes, defines, data types, etc. It is entirely about interoperability as well - an area where courts have overridden copyright in the past (it is actually legal to stick a "NintendoTM" logo in a game made for the game boy using a copy of Nintendo's own code to do it, simply because they put code that checks for it in the loader and a judge felt that in doing so they surrendered their copyright and trademark rights).

So, if code only uses the linux headers, how is that different from using the POSIX signal constants or whatever?

I do agree that in the end it comes down to a what a court rules. I'm pro-GPL in general, but I really don't want to see court rulings that basically say that sticking symbol names in your binary creates a derivative work. Going down that route basically allows OS vendors to require a license just to write software for the OS. Frankly the game console vendors already go way to far in that area, though they generally accomplish it by technical means rather than legal means (mainly because courts have tended to rule against them).

Re:Does AMD still matter? - OS Drivers OK now

[anon mouse-cow-aard]

Hi there, you don't need the proprietary drivers anymore. I use steam games, like portal, and portal2 (beta, but works better than on windos 7, where my kids say it crashes a lot) on an ATI 75xx video card using pure open source drivers (ubuntu 14.04) and it runs really, really well. I don't know what, if anything is missing, The card may run a little hotter and a bit slower than the proprietary drivers, but it works well enough that I am thrilled not to have to bother with them anymore.

Re: What?

[HiThere]

Well, much of the open source community got its start when ESR got upset that he couldn't make a proprietary printer driver do what he wanted it to. Perhaps you should look at your question again. You may still decide that it's reasonable that a corp should decide what capabilities the hardware it sold you should make available, but you also might change your mind.

Re:You probably cannot safely use the GCC compiler

[HiThere]

For that matter, copyright doesn't cover functionally required material. So copyright can't touch it...if you have a good lawyer. Patents, however, and trademarks, are totally different categories of law. Both are applicable, though in this case I don't think there could be any claim made by the person distributing the GPL code. But someone you've never heard of may hold a trademark that it could be ruled to be infringing. And there's NO reasonable protection from patents. Even a good lawyer, a clear case, and a fat wad of cash isn't a guarantee.

Correction:

[HiThere]

Change:

in this case I don't think there could be any claim made by the person distributing the GPL code.

to:
in this case I don't think there could be any claim made under trademark law by the person distributing the GPL code.

Re:Does AMD still matter?

[Loki_1929]

I do vmware workstation for linux and website testing. I need lots of cores!

As someone who manages a decent sized VMware vSphere environment, I can tell you that core counts are not so important as you may think. My AMD-based ESXi servers have triple and quadruple the cores of my Intel-based ESXi servers, yet they experience chronic problems with CPU Ready Time at far lower over-subscription rates and even sometimes while under-subscribed if some oversized VMs are present. It's one of the reasons I'm pushing through a complete shift away from high-core count Opterons (24 and 32 core hosts) and moving toward lower core count (mostly 8 core) Xeon hosts.

And that deeply pains me as someone who's used AMD CPUs since the K6-2 days. Once Intel fell into the P4 clock speed trap, I thought AMD would finally be able to beat them into submission. Sadly, AMD has sat idly by and allowed Intel to dominate nearly every metric of raw performance. As much as I grew up loving AMD as the plucky underdog, I can't ignore reality and I have to make the smartest decisions from a professional standpoint. When licensing costs get added to the mix, Intel absolutely obliterates AMD from a TCO perspective in the server space. I hate it, but it's true, and I'm left with little choice if I want to be honest.

Re:Does AMD still matter?

[Billly Gates]

In a server environment a xeon has a ton of cache and that would be ideal for very wide loads. As a workstation I need to make sure IE 6 renders the site right in one, apache outputs it in another, and a large part of the time is loading and unloading things I am learning. SSD really helps this.

I/O or the lack of it would cause this too. The newer chips at AMD are inferior to the older onces per IPC. Things are true cores and not semi cores with shared cache and fpus. This would cause that issue for sure since the cache would constantly be depleted by the other cores.

Re:You do not seem to understand

[Rich0]

I was referring to the linking argument, not the compiling argument.

The fact that code originating in GCC ends up in the binary has nothing to do with whether it is legal or not to write a non-GPL Linux kernel module that uses arbitrary kernel symbols.

I wasn't really debating anything you said with regard to compilers. Your argument makes sense, though I haven't really given it as much thought. The process of compilation is not creative, but the design of the compiler itself certainly is. I'd have to give some thought to counter-arguments.

Re:Does AMD still matter?

[TheRaven64]

They are popular with FadCoin mining because they are good at GPGPU work. This also makes them attractive to people doing useful and productive work with them, which is (thankfully) a much larger market. Lots of clusters now have AMD GPUs for number crunching.

Re: What?

[AJodock]

No, no, I still don't get it! The hardware is proprietary, what difference does it make if the driver is or isn't as well?

The biggest difference (for me) is card support. I could fire up Kubuntu 13.10 with X server 1.14 on my old AGP ATI X1650 card and would work like it should. If proprietary drivers were the only choice I would be SOL. If I found the old driver that was designed for that card the chances of getting it to work with a modern kernel/X server would be extremely slim. The current Catalyst driver only supports back to the HD 5000 series so my desktops that I use on a daily basis with HD3*** and HD4*** cards are stuck with a Catalyst driver that says "Automated installer and Display Drivers for Xorg 6.9 to Xserver 1.12 and Kernel version up to 3.4", but they run great with the open source drivers.

I don't claim that the open source driver will support ancient cards forever, but AMD is concerned with selling new cards not keeping up old ones, and the open source drivers are much better in that department. If you know how you can always resort to fixing the issue yourself if the driver is broken for your card.