FrankNFurter writes "It's been a rumour for several weeks, but now it's confirmed: AMD buys ATI. What implications is this merger going to have for the hardware market?" In addition to AMD's release, there's plentyofcoverage out there.
That's complete bollocks. Nothing related to copyrights and patents prevents ATI and nVidia from releasing the specs to the X.org/XFree developers. Nobody's ever said "ATI must release the code to fglrx", not least because we know it's so awful we would want to start from scratch. ATI and nVidia's poor reputation in the Free Software community has to do with their refusal to provide non-NDA-encumbered documentation.
Actually, their competitors are unaffected because:
1. They have large enough staff to decompile and perform clean reverse engineering of NVidia's drivers, e.g., one team analyzes the decompiled code and takes notes (without copying code of course), another team designs improvements and implements based on that analysis
2. Their competitors own electron microscopes, making analysis of the chip internals relatively simple.
Now tell me: why are the likes of NVidia and ATI keeping their products undocumented and their drivers closed?
And to counter your argument: what happens in two years when ATI and NVidia decide your card is too old to support, and yet it still performs very well but you NEED the features in the latest kernel and latest x.org? Go ahead, buy a new video card -- oops, nope, sorry, they changed slot specs again, and PCI Express cards are no longer available because PCI-X finally gained market share in the consumer market and PCI-E ended up as short-lived as VLB did in the VLB vs. PCI war.
(do I expect PCI-E to die? No, it was a hypothetical example showing the potential problem with proprietary drivers)
Now tell me: why are the likes of NVidia and ATI keeping their products undocumented and their drivers closed?
Because, if they DO PROTECT THEIR IP, The OTHER GUY has to waste TONS OF MONEY on reverse-engineering teams and highly-qualified people to reverse-engineer the processor via electron microscopes.
It's not the EQUIPMENT that is expensive, it is the PEOPLE. And, as you Linux zealots know FULL WELL, reverse-engineering is EXPENSIVE in terms of PEOPLE and TIME.
If you publish the specifications of your latest graphics chip for all to see, suddenly your competitors don't have to divert staff from working on next-generation architectures just to reverse-engineer your system. Instead, they can analyze your documentation in a fraction of the time.
It's a two-way street, so stop deluding yourself that there's only one side to the story. Publishing full specs for your graphics chips is like writing your competition a blank check. Intel is the only one who doesn't have an issues doing this because their graphics technology is always following.
And to counter your argument: what happens in two years when ATI and NVidia decide your card is too old to support, and yet it still performs very well but you NEED the features in the latest kernel and latest x.org? Go ahead, buy a new video card.
Yes. There are still many well-supported video cards sold in AGP. In fact, you can still get well-supported video cards in PCI, a fifteen-year-old technology. They're not top-performers, but beggars can't be choosers.
The video card market is transitioning to PCIe with surprising speed precisely because they do not want another VLB fiasco. The PCI -> AGP transition was slow because PCI still had a future for other types of cards, but the AGP -> PCIe transition was rushed to avoid market confusion. You can still buy plenty of AGP cards, but the big players have made it clear: there won't be any more improvements for AGP.
If you publish the specifications of your latest graphics chip for all to see, suddenly your competitors don't have to divert staff from working on next-generation architectures just to reverse-engineer your system. Instead, they can analyze your documentation in a fraction of the time.
Bullshit, sorry. We don't want their beloved silicon blueprints of their latest GPUs, just information on how to make them work. Want to draw a polygon? Send this command to the card. Do a hardware T&L? This other one. You can learn only so much from a driver sourcecode or techincal specifications on how to program a GPU. Don't beleive me? Check the information released by both nVidia and ATI for their older GPUs, and see how much you can infer from them.
You've obviously never tried to reverse engineer a chip using an electron microscope. If so, you wouldn't be saying it is "relatively simple". You obviously don't realize that modern chips have upwards of 8 layers of wiring stacked on top each other, that cover the local interconnects and transister wiring (poly and active area). Now, let's say you can decompose the chip, layer by layer, snapping photos as you go. With today's 65 and 95nm processes, you can only see a few transistors at a time, given the resolution and field of view of the microscope. These chips have millions of transistors on them. How long do you think it would take to cover a sizeable area of the chip, to the point that you could gain some useful information about it? At best, you could probably look at a few latch circuits at a time... That's like looking at the grains of sand on a beach and trying to map out the coastline of Hawaii. By the time these alleged reverse engineers are done, Nvidia would have released two new architectures. Money and time are much better spent creating rather than copying.
IIRC there's a mutual deal between ADM and Intel (part of the setlement of a lawsuit) that allows one company to use the other's technologies. that's what allowed AMD to integrate SSE 1/2/3 in athlons and Intel to integrate AMD64 in pentium4/xeon.
if they both buys graphic chipsets companies, does this means nvidia's technology on ATI GPUs and the other way around ?
or will they shield the newly aquired techs from the setlment ?
As far as I can tell this deal only covers patents made before 2001 (section 2). I could be wrong though, not very good at legalspeak, and didn't read the entire contract. AFAIK they have another cross-licensing agreement as well, but it only covers all x86 extensions and improvements. This is the deal that you're probably talking about as SSE and AMD64 are x86 extensions. So to answer your question: no they would not need to share tech acquired from ATI.
This is because x86-64 is an open standard. AMD released it as open when they announced it, because it was the only way to gain industry acceptance.
Once AMD got Microsoft's cooperation building support for x86-64 into Windows, they hardped on about the open standard. This protected AMD from Intel, who were already secretly working on their own implementation of x86-64. Normally, once Intel realized how potentially powerful x86-64 was, they were sure to create their own incompatible version (ala SSE and 3DNOW!) to try and derail AMD.
But the open standard stopped Intel from doing this. Microsoft pointed to the open standard, and told Intel flat-out that they were not going to support two versions of 64-bit x86.
x86-64 is an open standard. AMD's copyrighted implementation of x86-64 is called AMD64. Inte;'s copyrighted implementation of x86-64 is EMT64.
There's a slight chance that AMD might be smart and release hardware specs for ATI cards, or make the drivers Free Software. If either of those things happen, this would be a very good thing, in my opinion.
Well, that could be a very good thing. Those specifications will also help driver-makers a lot. It might also help to get the linux drivers which are pretty poor for ATi at the moment.
The AMD-fans/nerds are more linux-minded then Intel (IMHO), and AMD probably knows this. They can really make a business-blow by releasing this, in the mind of open-source.
this could be real good if AMD's acquisition of ATI allows them to produce full chipsets in the same fashion Intel has with its Centrino line. let the competition begin!
also, not official yet, as government regulatory bodies need to approve it.
this could be real good if AMD's acquisition of ATI allows them to produce full chipsets in the same fashion Intel has with its Centrino line. let the competition begin!
Yeah, the part that really sweetens the deal for us end consumers is that ATI will now get to benefit from the research that AMD inherits from IBM [nytimes.com] for chipsets. Hopefully ATI can make some better video cards with all the research that the other two have benefited off of. I hope that the same chipmaking technologies AMD has been using can now be used to improve ATI's GPUs and chipsets.
Since (in my opinion) NVidia has taken the lead in GPUs, I hope that ATI will be boosted back into a competitive state and price wars ensue.
Again, to me this is nothing but great news for the end-consumer.
The linux crowd (or at least a vocal minority of them), don't want drivers, they just want documentation for the card, they'll make their own drivers.
On the other hand, releasing either open source drivers, or a combination of binary drivers, along with documentation (so those who want to write their own CAN), would certainly be the best of both worlds.
As I understand it, the secrets GPU companies want to protect are not the driver itself, but what low level functions are provided by the card. Open sourcing the driver would expose those functions. IMHO, GPU companies should export a higher level interface from the card - say OpenGL, via firmware on the card. They could have OpenGL and DirectX ROMs for the card. Or even ship with DirectX by default, and offer an "upgrade" to OpenGL. (Not as nice for Linux folk like me, but M$ have dominant market share.) The OS driver would be a very thin layer. This would keep all their hardware secrets in the card where they belong, but allow full use under open source operating systems.
The drawback would be a lockout for experimental 3D APIs. But it would be no worse than the binary driver situation we have now.
I'm getting the 'gist' of why this transaction needs to happen. AMD needs GPU functionality on the CPU. I think everyone kinda expected that to happen at some point. The Inq. then takes a left turn in the plot and mentions 'mini-cores' which are multi-cores with massive amount of threads. Sort of but not really like Intels' hyperthreading times 32x. Shitloads of threads.
Bottom line?
ATI will work on AMD's new cores. I don't know if they'll work on something that'll plug into a PCIe slot still like nVidia.
nVidia will still be around making graphic cards for AMD. Just won't necessarily be anything remotely similar to what's out on the stores today. AMD doesn't like closed technology like Intel does. So it'll be an open platform still which is a 'good thing' (tm).
Forget about GPU's and chipsets. The main innovation has to come from these new GCPU's.
ATI was going to lose its Intel chipset business anyway with or without this takeover. So no big loss here.
Intel has about a year lead on this tech and probably be first out to market with it.
CPU cores change radically every 5 years or so. With GCPU's, think more in terms of GPU's and radical changes every year to 18 months. Crazy shit.
Plenty of space at FAB 36 to build the new cores and the recently announced plant they are building in New York. So no more costly production runs in Taiwan.
If AMD didn't do this, they'd be out of business in 5 years. Period.
See the entry in the Hacker's Dictionary / Jargon File for "Wheel of reincarnation [catb.org]":
wheel of reincarnation: [1968] Term used to refer to a well-known effect whereby function in a computing system family is migrated out to special-purpose peripheral hardware for speed, then the peripheral evolves toward more computing power as it does its job, then somebody notices that it is inefficient to support two asymmetrical processors in the architecture and folds the function back into the main CPU, at which point the cycle begins again.
Several iterations of this cycle have been observed in graphics-processor design, and at least one or two in communications and floating-point processors. [...]
Ok, time to blow off the moderations that I've made so far as you're missing something fairly obvious. Take a look at the processor on the 7800GTX, with 300 million transistors it is currently the most complex chip being shipped (I don't how big Cell is). In exchange the peak processing power is 320Gflop/s (40 in the vertex processors, 280 in the fragment shaders). For comparison the floating point performance in a CPU is ~ 8Gflop/s. That's a whopping 40 cores to break even on graphics processing.
Once you can fab a processor large enough to contain 40 functional cores - how big a GPU do you think you could fab on the same process? The simple fact is that a GPU is completely crippled compared to a CPU. There are huge tradeoffs in the design to get that kind of performance. Stream processing is very limited compared to a von Neumann architecture if you care about latency in the slightest. But for graphics - it's perfect. Throwing completely independent parallel chunks of data through an array of vector processors is a much simpler challenge than attempting to extract parallelism from sequential code. The sequential code has pesky things like control-flow that is missing in the gfx shaders, and I don't mean the rubbish that ATi/Nvidia are selling as control-flow in their current designs. That is sheer marketing given the size of the shader batches and the depth of the pipelines.
So I don't think the big 'ol wheel of reincarnation is going to move rendering back into software anytime soon. But what people forget is that AMD is not really a processor company. They are a fab company that just happens to design some kick-ass processors. Their main business is silicon, and buying ATi is the biggest chunk of vertical integration you can imagine...
Seems highly unlikely to me that they'd stick a GPU into the CPU. Modern GPUs are a similar size to CPUs (if not larger) and need much higher memory bandwidth... so you'd be doubling the size of your CPU and you'd need a 256-bit 1GHz+ memory interface. And then the 'high end' users would just go and buy a PCI-Express card when the next generation came out, making the whole thing a total waste.
I could see perhaps that they'd stick a cheap and crappy GPU into a cheap and crappy CPU for the low end of the market, but with Vista coming out with all its eye-candy that may not even be viable for rendering the Vista desktop, let alone games.
"Hello? AMD chips have an integrated memory controller, with the AM2s having support for DDR2 memory, running at 667 MHz. So unbuffered memory, running in dual channel mode runs at a speed of approximately 10.7GB/sec. In comparison current graphics technology runs at about 6.4GB/sec."
Hello? My 7800GS card has a memory bandwidth of 40GB/second from on-board RAM. It would be utterly crippled by a measly 10GB/second shared with the CPU.
"AMD doesn't like closed technology like Intel does. So it'll be an open platform still which is a 'good thing' (tm). " Actually Intel has been a big supporter of OSS. They helped port Linux the Itanium and have provided all the documentation to their video chips. I think you are confusing Intel with Microsoft. Intel has been one of the most open hardware companies. AMD has also been very good. ATI like nVidia.... Well let's say not so good. I really don't get this. AMD could use some good chip-sets but they have made their own for the Opteron so I don't see what they gain from ATI. AMD could use a good low end integrated video solution for low end desktops and servers. Yes it is true but servers almost never use nVidia or ATI graphics cards. When I set up a server I only plug the monitor in when I do the install and if something really bad happens. I have to think this comes down to laptops. AMD has not done well in that market and a one stop shop for a laptop solution like Intel offers might be a good solution. I wouldn't hold my breath on the good open source ATI drivers for Linux. Of course if it happens I might dump my nVidia based motherboard and Video card. I have been buying nVidia just because of their better Linux support for years.
Not at all a good idea, nor where things will be heading except in the embedded market perhaps. Not only would it reduce yields enormously because of the larger die size, but it would also put two points of failure into one chip, AND make it much harder to upgrade just one component.
I could see this perhaps in the mobile/embedded market, but not in the server/workstation space. At least not for a LONG time. It's just not a good idea.
I can't see this being good for customers. As we all know, ATI's products tend to be miserably supported, though this hasn't been the case for AMD thus far. How will this affect the nForce line of chipsets? Given ATI's past I'd much rather have an nForce than whatever ATI kicks out.
On the other hand, perhaps AMD will drag ATI out of it's rut, but I think it's just as probable that ATI will drag AMD down, and that's good for nobody.
Today: AMD has integrated memory controllers to get good memory performance.
Tomorrow: AMD has integrated video controllers to get good 3D performance.
OK, so not very close to reality considering what would be involved. AMD bought into ATI because it wants to focus on CPUs, not chipsets.
However, it does make for an interesting point of interest: the three primary components of PC architecture today are the CPU, GPU and chipset that bind the two together. AMD had two parts of the equation, and ATI has two parts as well, though one of these parts overlap. Now AMD is one company that has end-to-end solutions? There's got to be something interesting coming out of that marriage.
One thing that Intel has always done better than AMD is provide the "whole package".
What I can buy from Intel:
Server chassis + power supply Motherboard CPU(s) NIC RAID
What I can buy from AMD:
CPU(s)
Small-medium OEMs are going to like Intel because it gives them one point of support for most of their major components. It also gives them a single "partner" with which to negotiate pricing; the larger volume of product means they can get overall better pricing.
Taking on ATI might be AMDs move to start fixing that shortfall in their business model. If they put a solid OEM-friendly motherboard on the market, it will be a huge step in the right direction. With Conroe presently beating the pants off AMD's offerings, this is well-timed.
"It's been a rumour for several weeks, but now it's confirmed: AMD buys ATI. What implications is this merger going to have for the hardware market?"
I wonder if this means no more ATI cards in Macintosh computers, seeing as how Apple uses Intel now? Or, even more interesting, could it mean Apple switching over to AMD?
NVIDIAs response. Will NVIDIA no longer support AMD processors to the same level? The shocking thing to me about this announcement is that nForce chipsets are the best chipsets for AMD64. Also, NVIDIA driver quality across the board is better than ATI.
So, we'll see how this shakes out. If, as others have said, AMD forces ATI to produce better drivers, and good Linux drivers, that may be a good outcome...
The other interesting aspect is (as it often is) Apple. Now AMD gets an instant slice of the Apple pie (sorry) since ATI makes most current Apple graphics chips. Interesting development there... Intel can't be happy.
I suspect the tension level just notched up at NVIDIAs headquarters as well.
I think the marketplace has been very well-served by the two dualities that existed before this move: ATI and NVidia beat each other's brains out, as did Intel and AMD. This new dynamic with 3 players does not seem, to me, to promise anywhere near as many benefits for us, the customers. Will ATI become more AMD-centric? Undoubtably. Will NVidia (which has been a great AMD booster) become less supportive of AMD processors? Probably. As this plays out, it seems to me that NVidia will basically be an Intel graphics house (including Macs), and ATI will melt into AMD, becoming mostly an internal chipset house. In the end we lose a very healthy competition between NVidia and ATI. We gain, perhaps, a stronger AMD to keep Intel honest.
AMD is covering the remaining $2.5b of the deal with a commitment letter from Morgan Stanley Senior Funding, with the debt secured by "a pledge of the capital stock of certain material units of the company, accounts receivable and proceeds from any sale by Advanced Micro of its equity interest in Spansion Inc." The CFO is overly optimistic that the company can get rid of that debt "quickly," without layoffs, and with savings of $75m and $125m over the next two years. DJ Newswires says ATYT will no longer work with Intel, and the execs say that they can make up the lost sales with the severing of Intel-ATI ties. Pretty lofty goals, I'd say.
Taking into account all the fanboi anguish, let me point out the very simple fact that now ATI no longer directly competes with NVidia. You could say that the competition would be between AMD and NVidia now, but that's not quite right either. The fact is that the market has become so diverse that all these companies were already competing with each other, despite partnerships and deals.
AMD, ATI, NVidia and Intel *all* make motherboard chipsets. ATI, Nvidia and Intel all make video processors. So do SIS, S3, and VIA.
Yet they all work (relatively) well with each other.
This isn't about marketshare, it's about technology. ATI does something that AMD wants, so AMD is acquiring the company for the tech. The market won't feel a thing, I promise you. Competition will continue, just like it did when Micron acquired Rendition (wipes a tear for his Verite v2200) and when NVidia bought out 3dfx (wipes another for his Banshee).
Since everyone's got their prognosticator's caps on today, I'm going to come out and say that, within 5 years, we'll be seeing GPU processors integrated into the motherboard, accessable to both ATI and NVidia (and Matrox, and S3, and...) The power and bandwidth demands for next gen GPUs are becoming more than expansion boards can handle. Instruction sets are becoming extremely CPU-like. Since the whole universe seems to be moving into Multi-processor designs anyway, perhaps we'll even see some kind of GPU-MMX style expansion of the x86 instruction set (call it v86 for now).
I think we're seeing a move back to specialization. We've already got separate Audio chips, separate networking chips, even chips to handle I/O for RAID and such. With the new market for Physics co-processors, I'm sure we'll only see more for tasks such as AI, and when the next big UI design is unleashed (either some kind of brain-reading technology, or a true 3d input system -- the WII is just the tip of the iceberg!) another co-processor will be made to handle that. With AMD's focus on integrating external processors with technologies such as HyperTransport, undoubtedly they'll be able to compete for a long time.
And the best part is, we get to choose from strong market competitors. As long as there is innovation, we win.
There is one obvious reason for the purchase, already stated by others. I'm just reiterating.
Next year, AMD will be shipping quad core Athlons and Opterons. But, if they wanted to they could replace one CPU with a GPU and have video on die. And if they wanted to they could replace a second CPU with sound, USB, SATA, Gigabit, wireless etc etc etc, and have an entire computer on a chip.
VIA has been trying to do this for years. AMD has the fab capacity to pull it off.
AMD could be the first company to enable the $150.00 PC to exist (by saybe 2009). Smaller than a mac-mini, dual core, and all you need to get it to run is slap some flash memory on board for a hard drive substitute, some DDR2, a cheap DVD drive and Voila! Instant computer.
Imagine a Dual Core Athy with a gig of ram, 20GB flash disk all in the form factor of about twice the size of an IPOD.
Oh you could put a screen on it too, DGMS.
This could be a great thing. My only advice for AMD / ATI is: Dedicate some resources to drivers, or better yet, open source the GPU API.
I was thinking the exact same thing. But I think that, if done correctly, this could really solidify both AMD and ATI as market leaders.
If AMD pressures their new acquisition to create a half-decent set of Linux drivers, then this will all be worth the hassle. Or even if they convince ATI to open up the specs, that would be ok too.
right.. because as we all know.. Linux support == market leader.
They may become a market leader for Linux desktops (GPU's aren't needed in servers where Linux is popular).. but Linux desktops are only 1-2 percent of the desktop market...
so even if they gain all of it.. they still won't be a market leader in GPUs.
Getting ANY of their stuff working under Linux is painful.
I'm running OpenSuSE 10.1 on my Thinkpad R51 with a pretty standard ATI Mobility Radeon and can I get the ATI drivers working? Can I hell. Always "no device for screen" or some such error. So I'm stuck with the OSS drivers which although are great for 2D, don't perform well enough for anything other than TuxCart.
On the other hand, the NVidia FX5900 in my desktop machine (also running OpenSuSE 10.1) was a breeze. Drop to run level 3, run installer, reboot, job done.
On the other hand, the NVidia FX5900 in my desktop machine (also running OpenSuSE 10.1) was
a breeze. Drop to run level 3, run installer, reboot, job done.
NVidia seems to make better blobs than ATI, but it is still
a blob [openbsd.org]:
Blobs are expedient. Many other open source operating systems cheerfully incorporate them; in fact their users demand them.
Why are posters so fond of the anti-open source hardware vendor
NVidia?
Probably because most Slashdotters are not driver hackers nor OSS purists, they are developers, gamers, and power users -- and Nvidia's hardware (and driver support for the hardware) is phenomenal.
Your gripe is not baseless, though: would it kill Nvidia to open up a bit? Perhaps the renewed competition will encourage them to do so, although it's equally likely that they will take the opposite tack and circle their wagons ever more tightly. As long as they provide excellent binary drivers for Linux, I doubt that they will feel much incentive to go Open Source...
nVidia CAN'T open up their drivers IIRC. nVidia was founded by a bunch of SGI engineers, and once they started producing products the folks over at SGI found some of their technology in the nVidia products. As part of the settlement, nVidia can't release the code since I think they had to license it from SGI in the end.
I could swear that's they way that it is, but I can't find any definitive reference to the settlement.
We, your faithful processor purchasers (yes, we have many), have long been forced to buy nVidia hardware because of ATI's poor quality drivers under Linux. Please work the same magic you did with the AMD64 and give us something we can be proud of.
The problem is that Intel's integrated graphics is utterly incapable of running games, it doesn't even properly support DX9. On the other hand both Ati's and Nvidia's IGPs are DX9 compliant and lightyears faster than Intel's IGP. Thus we come to the conclusion that Intel is hurting gaming with it's crappy IGP, because you can't buy a similarly crippled IGP for AMD even if you wanted to. Luckily the next version of Intel's IGP will at least support the DX9 spec properly, as it wouldn't be able to display Vista's Aero UI otherwise..
Tomorrow (Score:3, Funny)
Re:Tomorrow (Score:3, Insightful)
*shudder*
Re:Tomorrow (Score:4, Insightful)
Parent
Re:Tomorrow (Score:5, Insightful)
Parent
Re:Tomorrow (Score:5, Insightful)
1. They have large enough staff to decompile and perform clean reverse engineering of NVidia's drivers, e.g., one team analyzes the decompiled code and takes notes (without copying code of course), another team designs improvements and implements based on that analysis
2. Their competitors own electron microscopes, making analysis of the chip internals relatively simple.
Now tell me: why are the likes of NVidia and ATI keeping their products undocumented and their drivers closed?
And to counter your argument: what happens in two years when ATI and NVidia decide your card is too old to support, and yet it still performs very well but you NEED the features in the latest kernel and latest x.org? Go ahead, buy a new video card -- oops, nope, sorry, they changed slot specs again, and PCI Express cards are no longer available because PCI-X finally gained market share in the consumer market and PCI-E ended up as short-lived as VLB did in the VLB vs. PCI war.
(do I expect PCI-E to die? No, it was a hypothetical example showing the potential problem with proprietary drivers)
Parent
Think about what you just said (Score:5, Insightful)
Because, if they DO PROTECT THEIR IP, The OTHER GUY has to waste TONS OF MONEY on reverse-engineering teams and highly-qualified people to reverse-engineer the processor via electron microscopes.
It's not the EQUIPMENT that is expensive, it is the PEOPLE. And, as you Linux zealots know FULL WELL, reverse-engineering is EXPENSIVE in terms of PEOPLE and TIME.
If you publish the specifications of your latest graphics chip for all to see, suddenly your competitors don't have to divert staff from working on next-generation architectures just to reverse-engineer your system. Instead, they can analyze your documentation in a fraction of the time.
It's a two-way street, so stop deluding yourself that there's only one side to the story. Publishing full specs for your graphics chips is like writing your competition a blank check. Intel is the only one who doesn't have an issues doing this because their graphics technology is always following.
And to counter your argument: what happens in two years when ATI and NVidia decide your card is too old to support, and yet it still performs very well but you NEED the features in the latest kernel and latest x.org? Go ahead, buy a new video card.
Yes. There are still many well-supported video cards sold in AGP. In fact, you can still get well-supported video cards in PCI, a fifteen-year-old technology. They're not top-performers, but beggars can't be choosers.
The video card market is transitioning to PCIe with surprising speed precisely because they do not want another VLB fiasco. The PCI -> AGP transition was slow because PCI still had a future for other types of cards, but the AGP -> PCIe transition was rushed to avoid market confusion. You can still buy plenty of AGP cards, but the big players have made it clear: there won't be any more improvements for AGP.
Parent
Re:Think about what you just said (Score:5, Insightful)
Bullshit, sorry. We don't want their beloved silicon blueprints of their latest GPUs, just information on how to make them work. Want to draw a polygon? Send this command to the card. Do a hardware T&L? This other one. You can learn only so much from a driver sourcecode or techincal specifications on how to program a GPU. Don't beleive me? Check the information released by both nVidia and ATI for their older GPUs, and see how much you can infer from them.
Parent
Re:Tomorrow (Score:5, Informative)
Parent
Re:Tomorrow (Score:5, Interesting)
if they both buys graphic chipsets companies, does this means nvidia's technology on ATI GPUs and the other way around ?
or will they shield the newly aquired techs from the setlment ?
Parent
Re:Tomorrow (Score:5, Informative)
As far as I can tell this deal only covers patents made before 2001 (section 2). I could be wrong though, not very good at legalspeak, and didn't read the entire contract. AFAIK they have another cross-licensing agreement as well, but it only covers all x86 extensions and improvements. This is the deal that you're probably talking about as SSE and AMD64 are x86 extensions. So to answer your question: no they would not need to share tech acquired from ATI.
Parent
x86-64 is not part of the IP sharing (Score:5, Informative)
Once AMD got Microsoft's cooperation building support for x86-64 into Windows, they hardped on about the open standard. This protected AMD from Intel, who were already secretly working on their own implementation of x86-64. Normally, once Intel realized how potentially powerful x86-64 was, they were sure to create their own incompatible version (ala SSE and 3DNOW!) to try and derail AMD.
But the open standard stopped Intel from doing this. Microsoft pointed to the open standard, and told Intel flat-out that they were not going to support two versions of 64-bit x86.
x86-64 is an open standard. AMD's copyrighted implementation of x86-64 is called AMD64. Inte;'s copyrighted implementation of x86-64 is EMT64.
Parent
Don't really know.. (Score:5, Interesting)
But on the other hand, this could split the market and get things like todays uncompatible browsers. (Which is VERY annoying somethimes)
And we have a psychic [slashdot.org]
Re:Don't really know.. (Score:4, Interesting)
Parent
Re:Don't really know.. (Score:5, Insightful)
The AMD-fans/nerds are more linux-minded then Intel (IMHO), and AMD probably knows this. They can really make a business-blow by releasing this, in the mind of open-source.
Parent
Re:Don't really know.. (Score:4, Interesting)
Parent
could be good.. (Score:5, Interesting)
also, not official yet, as government regulatory bodies need to approve it.
It WILL Be Good! (Score:5, Insightful)
Since (in my opinion) NVidia has taken the lead in GPUs, I hope that ATI will be boosted back into a competitive state and price wars ensue.
Again, to me this is nothing but great news for the end-consumer.
Parent
Probably Not Going To Happen... (Score:5, Funny)
Nah, weren't you reading yesterday? ... (Score:4, Interesting)
On the other hand, releasing either open source drivers, or a combination of binary drivers, along with documentation (so those who want to write their own CAN), would certainly be the best of both worlds.
Parent
Driver code not the issue (Score:4, Interesting)
The drawback would be a lockout for experimental 3D APIs. But it would be no worse than the binary driver situation we have now.
Parent
I'd say I told you so.... (Score:5, Funny)
http://www.theinquirer.net/default.aspx?article=3
-Charlie
AMD & ATI (Score:5, Funny)
DAAMIT!
This is a very good thing. (Score:5, Interesting)
*head asploded*
I'm getting the 'gist' of why this transaction needs to happen. AMD needs GPU functionality on the CPU. I think everyone kinda expected that to happen at some point. The Inq. then takes a left turn in the plot and mentions 'mini-cores' which are multi-cores with massive amount of threads. Sort of but not really like Intels' hyperthreading times 32x. Shitloads of threads.
Bottom line?
ATI will work on AMD's new cores. I don't know if they'll work on something that'll plug into a PCIe slot still like nVidia.
nVidia will still be around making graphic cards for AMD. Just won't necessarily be anything remotely similar to what's out on the stores today. AMD doesn't like closed technology like Intel does. So it'll be an open platform still which is a 'good thing' (tm).
Forget about GPU's and chipsets. The main innovation has to come from these new GCPU's.
ATI was going to lose its Intel chipset business anyway with or without this takeover. So no big loss here.
Intel has about a year lead on this tech and probably be first out to market with it.
CPU cores change radically every 5 years or so. With GCPU's, think more in terms of GPU's and radical changes every year to 18 months. Crazy shit.
Plenty of space at FAB 36 to build the new cores and the recently announced plant they are building in New York. So no more costly production runs in Taiwan.
If AMD didn't do this, they'd be out of business in 5 years. Period.
Re:This is a very good thing. (Score:4, Interesting)
-Charlie (the author of the Inq article)
Parent
See Hackers Dictionary: "Wheel of reincarnation" (Score:5, Informative)
See the entry in the Hacker's Dictionary / Jargon File for "Wheel of reincarnation [catb.org]":
-Mark
Parent
Re:Graphics in software (Score:4, Interesting)
Once you can fab a processor large enough to contain 40 functional cores - how big a GPU do you think you could fab on the same process? The simple fact is that a GPU is completely crippled compared to a CPU. There are huge tradeoffs in the design to get that kind of performance. Stream processing is very limited compared to a von Neumann architecture if you care about latency in the slightest. But for graphics - it's perfect. Throwing completely independent parallel chunks of data through an array of vector processors is a much simpler challenge than attempting to extract parallelism from sequential code. The sequential code has pesky things like control-flow that is missing in the gfx shaders, and I don't mean the rubbish that ATi/Nvidia are selling as control-flow in their current designs. That is sheer marketing given the size of the shader batches and the depth of the pipelines.
So I don't think the big 'ol wheel of reincarnation is going to move rendering back into software anytime soon. But what people forget is that AMD is not really a processor company. They are a fab company that just happens to design some kick-ass processors. Their main business is silicon, and buying ATi is the biggest chunk of vertical integration you can imagine...
Parent
Don't believe it myself (Score:5, Interesting)
I could see perhaps that they'd stick a cheap and crappy GPU into a cheap and crappy CPU for the low end of the market, but with Vista coming out with all its eye-candy that may not even be viable for rendering the Vista desktop, let alone games.
Parent
Re:Don't believe it myself (Score:4, Informative)
Hello? My 7800GS card has a memory bandwidth of 40GB/second from on-board RAM. It would be utterly crippled by a measly 10GB/second shared with the CPU.
Parent
One correction. (Score:4, Informative)
Actually Intel has been a big supporter of OSS. They helped port Linux the Itanium and have provided all the documentation to their video chips.
I think you are confusing Intel with Microsoft. Intel has been one of the most open hardware companies.
AMD has also been very good. ATI like nVidia.... Well let's say not so good.
I really don't get this.
AMD could use some good chip-sets but they have made their own for the Opteron so I don't see what they gain from ATI.
AMD could use a good low end integrated video solution for low end desktops and servers. Yes it is true but servers almost never use nVidia or ATI graphics cards. When I set up a server I only plug the monitor in when I do the install and if something really bad happens.
I have to think this comes down to laptops. AMD has not done well in that market and a one stop shop for a laptop solution like Intel offers might be a good solution.
I wouldn't hold my breath on the good open source ATI drivers for Linux. Of course if it happens I might dump my nVidia based motherboard and Video card. I have been buying nVidia just because of their better Linux support for years.
Parent
Re:This is a very good thing. (Score:3, Interesting)
I could see this perhaps in the mobile/embedded market, but not in the server/workstation space. At least not for a LONG time. It's just not a good idea.
Makes me uneasy (Score:5, Interesting)
I can't see this being good for customers. As we all know, ATI's products tend to be miserably supported, though this hasn't been the case for AMD thus far. How will this affect the nForce line of chipsets? Given ATI's past I'd much rather have an nForce than whatever ATI kicks out.
On the other hand, perhaps AMD will drag ATI out of it's rut, but I think it's just as probable that ATI will drag AMD down, and that's good for nobody.
AMD designs (Score:5, Insightful)
OK, so not very close to reality considering what would be involved. AMD bought into ATI because it wants to focus on CPUs, not chipsets.
However, it does make for an interesting point of interest: the three primary components of PC architecture today are the CPU, GPU and chipset that bind the two together. AMD had two parts of the equation, and ATI has two parts as well, though one of these parts overlap. Now AMD is one company that has end-to-end solutions? There's got to be something interesting coming out of that marriage.
Re:AMD designs (Score:4, Insightful)
What I can buy from Intel:
Server chassis + power supply
Motherboard
CPU(s)
NIC
RAID
What I can buy from AMD:
CPU(s)
Small-medium OEMs are going to like Intel because it gives them one point of support for most of their major components. It also gives them a single "partner" with which to negotiate pricing; the larger volume of product means they can get overall better pricing.
Taking on ATI might be AMDs move to start fixing that shortfall in their business model. If they put a solid OEM-friendly motherboard on the market, it will be a huge step in the right direction. With Conroe presently beating the pants off AMD's offerings, this is well-timed.
Parent
Goodbye ATI? (Score:5, Interesting)
I wonder if this means no more ATI cards in Macintosh computers, seeing as how Apple uses Intel now? Or, even more interesting, could it mean Apple switching over to AMD?
The interesting thing to watch will be... (Score:5, Interesting)
So, we'll see how this shakes out. If, as others have said, AMD forces ATI to produce better drivers, and good Linux drivers, that may be a good outcome...
The other interesting aspect is (as it often is) Apple. Now AMD gets an instant slice of the Apple pie (sorry) since ATI makes most current Apple graphics chips. Interesting development there... Intel can't be happy.
I suspect the tension level just notched up at NVIDIAs headquarters as well.
Ugggh (Score:5, Insightful)
Execs overly optimistic (Score:5, Interesting)
AMD is covering the remaining $2.5b of the deal with a commitment letter from Morgan Stanley Senior Funding, with the debt secured by "a pledge of the capital stock of certain material units of the company, accounts receivable and proceeds from any sale by Advanced Micro of its equity interest in Spansion Inc." The CFO is overly optimistic that the company can get rid of that debt "quickly," without layoffs, and with savings of $75m and $125m over the next two years. DJ Newswires says ATYT will no longer work with Intel, and the execs say that they can make up the lost sales with the severing of Intel-ATI ties. Pretty lofty goals, I'd say.
ATI no longer competes directly with NVidia (Score:4, Interesting)
AMD, ATI, NVidia and Intel *all* make motherboard chipsets.
ATI, Nvidia and Intel all make video processors.
So do SIS, S3, and VIA.
Yet they all work (relatively) well with each other.
This isn't about marketshare, it's about technology. ATI does something that AMD wants, so AMD is acquiring the company for the tech. The market won't feel a thing, I promise you. Competition will continue, just like it did when Micron acquired Rendition (wipes a tear for his Verite v2200) and when NVidia bought out 3dfx (wipes another for his Banshee).
Since everyone's got their prognosticator's caps on today, I'm going to come out and say that, within 5 years, we'll be seeing GPU processors integrated into the motherboard, accessable to both ATI and NVidia (and Matrox, and S3, and
I think we're seeing a move back to specialization. We've already got separate Audio chips, separate networking chips, even chips to handle I/O for RAID and such. With the new market for Physics co-processors, I'm sure we'll only see more for tasks such as AI, and when the next big UI design is unleashed (either some kind of brain-reading technology, or a true 3d input system -- the WII is just the tip of the iceberg!) another co-processor will be made to handle that. With AMD's focus on integrating external processors with technologies such as HyperTransport, undoubtedly they'll be able to compete for a long time.
And the best part is, we get to choose from strong market competitors. As long as there is innovation, we win.
Hmmmm, Consoles (Score:4, Insightful)
A.) Xbox, Nintendo
Analysis.....Good move.
One reason why AMD may have bought ATI (Score:5, Interesting)
Next year, AMD will be shipping quad core Athlons and Opterons. But, if they wanted to they could replace one CPU with a GPU and have video on die. And if they wanted to they could replace a second CPU with sound, USB, SATA, Gigabit, wireless etc etc etc, and have an entire computer on a chip.
VIA has been trying to do this for years. AMD has the fab capacity to pull it off.
AMD could be the first company to enable the $150.00 PC to exist (by saybe 2009). Smaller than a mac-mini, dual core, and all you need to get it to run is slap some flash memory on board for a hard drive substitute, some DDR2, a cheap DVD drive and Voila! Instant computer.
Imagine a Dual Core Athy with a gig of ram, 20GB flash disk all in the form factor of about twice the size of an IPOD.
Oh you could put a screen on it too, DGMS.
This could be a great thing. My only advice for AMD / ATI is: Dedicate some resources to drivers, or better yet, open source the GPU API.
Raydude
Re:Linux (Score:5, Interesting)
Parent
Re:Linux (Score:5, Insightful)
They may become a market leader for Linux desktops (GPU's aren't needed in servers where Linux is popular).. but Linux desktops are only 1-2 percent of the desktop market...
so even if they gain all of it.. they still won't be a market leader in GPUs.
Parent
Re:Linux Support ? (Score:4, Informative)
I'm running OpenSuSE 10.1 on my Thinkpad R51 with a pretty standard ATI Mobility Radeon and can I get the ATI drivers working? Can I hell. Always "no device for screen" or some such error. So I'm stuck with the OSS drivers which although are great for 2D, don't perform well enough for anything other than TuxCart.
On the other hand, the NVidia FX5900 in my desktop machine (also running OpenSuSE 10.1) was a breeze. Drop to run level 3, run installer, reboot, job done.
Bob
Parent
Re:Linux Support ? (Score:4, Interesting)
NVidia seems to make better blobs than ATI, but it is still a blob [openbsd.org]:
Parent
Re:Maybe (Score:5, Funny)
A die holding an AMD core and an ATI GPU may be 'neat', 'fab', 'brill' or even 'ace' - but 'cool' - I think not!
Parent
Re:Maybe (Score:5, Insightful)
Probably because most Slashdotters are not driver hackers nor OSS purists, they are developers, gamers, and power users -- and Nvidia's hardware (and driver support for the hardware) is phenomenal.
Your gripe is not baseless, though: would it kill Nvidia to open up a bit? Perhaps the renewed competition will encourage them to do so, although it's equally likely that they will take the opposite tack and circle their wagons ever more tightly. As long as they provide excellent binary drivers for Linux, I doubt that they will feel much incentive to go Open Source...
Parent
Re:Maybe (Score:5, Interesting)
I could swear that's they way that it is, but I can't find any definitive reference to the settlement.
Parent
Re:Maybe (Score:5, Informative)
Because they've supported Linux with binary drivers for a long time, and their drivers work.
ATI is months behind, and half of the time the drivers are too buggy to actually use.
Philosophy of openness aside, that's an important difference.
Parent
Re:Should we welcome our new overlords? :D (Score:3)
Dear AMD:
We, your faithful processor purchasers (yes, we have many), have long been forced to buy nVidia hardware because of ATI's poor quality drivers under Linux. Please work the same magic you did with the AMD64 and give us something we can be proud of.
The undersigned.
Re:Intel has killed gaming...but AMD has restored (Score:4, Informative)
Parent