![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
"Intel Could Shut Down AMD's CPU Production Completely in Two Months"
- Thread starter Matthew Kane
- Start date
maybe after this AMD will pop out and go "booyah x86 killa right up in hrrrrr!!!" and it'll be the fastest cpu ever, so enticing that porting windows and other software from x86 is a given, finally moving us away from the x86 chains that are the bane of our existence, making all the drama a non-issue. then they'll license it to intel and intel will come up with something faster a few years down the road and we'll be back to healthy competition.
okay sorry that was my right before i go to bed already half-dreamin gpost
okay sorry that was my right before i go to bed already half-dreamin gpost
no2censorship
Limp Gawd
- Joined
- Oct 20, 2008
- Messages
- 393
True.
I don't like intel because they censored amd's sales. I don't buy products from a company if someone else satisifies my needs fine.
I don't like amd better because their name is amd.
And i'm pretty sure that intel is going to win the legal case, but I think that intel is going to be stuck with an x86 architecture; basically, they're obviously having trouble getting larrabee, and that's because 20 x86 cores suck for gaming. Or they're not efficient in any if they 20 are good enough,
And I don' think amd really cares if they lose this; people are underestimating their deals with GFC; keep in mind they'll probably have 32nm tech by years end thru ibm.
And, intel's greater wealth is decent evidence that they benefitted from not being in a free market economy. I'm not saying that's the truth, but no man can prove it.
Obligations to stock holders, and the fact that intel being so wealthy that they could never be dismantled, doesn't necessarily add up though. If they have a few stock holders who wouldn't remain loyal, of many, yet Intel is extremely wealthy, and have an excellent legal council, all of which are facts, means obligations to stockholders might not matter as much if intel sacrifices pisses of a few of many off. Intel is mainly doing this because they're running scared; they've seen that amd has gained back some market share, and that intel themselves is obviously having some trouble making a gpu--with x86 cores, an ancient architecture, that needs replacement badly.
Not likely. Even if they did, anything that can't run x86 code as well as what we have now will fail in the market place. Without the software base that x86 has, anything new would flop right away no matter how fast it was. The only way around this would be an x86 emulator, but it would have to be a damned good one that was capable of running with all the stability and performance of existing x86 systems. It would have to have no compatibility issues and no drawbacks. This seems highly unlikely.
Also bear in mind, that the CPUs we have today aren't really x86 CPUs anymore. At least, not in the traditional sense.
We don't know a damn thing about Larrabee's performance yet. It is extremely premature to make those kinds of predictions, especially because you don't know anything about Larrabee's design and you also don't have a very good understanding of how a processor works (educated assumption on my part based on your statements about your Kuma CPU from that other thread you made). I'm not saying that you're necessarily wrong; only that we have no way of knowing one way or the other, so it is impossible to make any sort of accurate prediction.
You don't seem to get it. If AMD loses x86, they're finished. End of story. Their entire business depends on CPU production, and without that, they'll fold like a house of cards. Even if ATI were to start doing phenomenally, that still wouldn't be enough to keep them afloat. They'd have to spin ATI off into a separate company again and declare bankruptcy, otherwise they'd drag ATI along with them. 32nm tech or not, without the ability to produce x86 CPUs, they don't have a product.
I don't know if it's possible to prove, but Intel's size has nothing to do with them overpowering AMD. It has to do with the fact that they've been around for much longer, have a much more diverse business, and the fact that they essentially invented the modern CPU.
That's not how it works with corporations. If Intel doesn't take every possible step to maximise their profits and protect their IP, their large shareholders can sue, and would probably win. Why would Intel risk losing a lawsuit and by extension a lot of money, when instead they can just do what they're supposed to and go after AMD? It's a win-win situation for them, and they have nothing to gain from doing anything else. In fact, Intel has been quite lenient with AMD already, since they gave AMD 60 days to settle the dispute before filing legal action instead of the 30 which is required by the cross-licensing agreement.
What are you talking about? Intel isn't running scared at all. They have significantly more market share than AMD, they're making cash by the truckload, and they've got AMD on the ropes since their current CPU lineup is faster than AMD's in every single performance metric. And where do you get the idea that they're having trouble making a GPU? There's no information anywhere to support that claim. By all accounts, Intel is perfectly on schedule with Larrabee.
Oh boy, this is getting good. An anti-corporate rebel.
Its N2C, we got so sick of his bullshit he was kicked out of genmay for starting, ahem, bullshit, in the soapbox. I didn't even think he was on the forums as all, to be honest.
no2censorship
Limp Gawd
- Joined
- Oct 20, 2008
- Messages
- 393
Good points, I agree with everything you said. The only 2 things I kind of differ on, but could very well be wrong on, are:
1. the death of the x86 architecture, even if it's evolved heavily, seems kind of inevitable to me, so that makes me think that amd could survive without it. They'd possibly make something a little better. and they'd have beat intel to succeeding the x86 architecture, even though intel gave them their start when they licensed it.
But, secretive as amd is, they may have some top-secret x86-beater already 3/4 finished, with the help of ati. If you think about it, ati's used the same old basic gpu architecture since 2007. I know it won't run old software, but it there could always be new software for it. Or, it could be meant to be a decent x86 emulator from the get go.
2. I say intel's not on schedule with larrabee, even though they've met their own deadline so far, because a company as large as intel could've devoted more budget to have had it released and 100% satisifactorily. I know it takes a long time to develop a revolution, but I'm still not entirely convinced that Larrabee was the smartest way for intel to go. I'll know once the first benchmarks are released and the first reports.
It's not inevitable to me, considering every consumer PC in the world uses x86. To shift to a completely different instruction set would be an incredibly monumental task, such that I'm not even sure it would be possible logistically. Plus, there is really no incentive to make the transition at the moment. And even if x86 dies off, it'll certainly happen fairly far into the future, and if AMD ends up not being allowed to use x86 at the end of all this, there's no way that they'll be able to make that transition and survive.
Doubtful. Extremely doubtful. And even if they did have an x86-killer ready, AMD doesn't have the market power to instigate that kind of transition. It would become a niche product and slowly die off due to the lack of compatible OSes and software and Intel's larger marketshare.
And that's not even taking into consideration the fact that a GPU-based CPU would be very inefficient at performing many of the general-purpose functions that current CPUs do well. GPUs excel at tasks involving many simple instructions, but they aren't ready to take over for CPUs.
If you really believe that, then you don't know very much about R600 and RV770. It's true that there are similarities, but there are also a massive number of design tweaks in RV770 that make it a significantly different design from R600 and RV670. But that isn't the discussion here.
Lack of native software would kill it right off the bat. And any x86 emulation would inherently be slower than running code natively on Intel x86 CPUs, so there would be no reason for people to switch since all their software would run much better on Intel's products anyway.
How is it not satisfactory? You're making no sense. And you can't just throw money at a design team to make it go faster. Not that Intel would want to anyway, since they're perfectly on schedule.
Intel hasn't even outlined specifically what they intend Larrabee to be used for. You have no way of knowing whether or not it was a smart way to go. In fact, I doubt you have any significant amount of knowledge about it altogether. So you have no basis on which to criticize it since you know little or nothing about it. And if you do know a lot about it, then go ahead and enlighten me. But considering you haven't given a single specific reason why you think it isn't a good design, I think it's a safe bet that I'm right about this.
Matthew Kane
Supreme [H]ardness
- Joined
- Dec 1, 2007
- Messages
- 4,233
Agreed to that one
well certainly it was a bit of wishful thinking. but afterwards i thought about it a little more and i think may be possible....
what if they licensed PowerPC? they gain the developer base involved with PowerPC Apple products and linux, they can make deals with IBM providing chips for the next gen of p-series servers/blades.
and the apps we're talking about....photoshop...firefox....office...i'm sure there's more work to port them to PowerPC than a simple recompile (actually theres already ppc firefox, just not for windows obviously)...but i also dont imagine there would be all that many changes necessary.
i think what's really locking everyone into x86 is microsoft, and their non-support for all other architectures.
but if its good enough that microsoft needs to support it to compete in the server market...then we could start seeing it supported in the mainstream...with individual developers following suit.
or...Windows dies in a fire and this whole x86 lock-in illusion completely disappears and everyone will realize that all the good non-windows software out there will run on pretty much anything already. maybe AMD can play the end game, betting their odds on linux/FOSS while competing in the server market for initial revenue? i cant imagine them not having a very strong backing from IBM if their PPC implementation is that good.
PowerPC and Apple's PowerPC products? haahha are you trying to be funny? apple has moved to x86 years ago. I highly doubt companies developing software for current are willing to invest time and money to port their software to PowerPC considering the market share. if they need to port their software it might as well be to x86!
Apple dumped PowerPC because IBM couldn't produce fast enough chips within reasonable power envelopes. It wouldn't make any sense for AMD to switch to PPC production since they'd run into the same problems and would end up producing inferior products. Not to mention the fact that there is practically no software support left in the consumer market for PPC-compatible OSX software, and none at all for Windows on PPC (and without Windows support, there is no way that they will get any marketshare since Apple won't license OSX to non-Apple computers).
holy christ man.
all im saying is that the developer base for PowerPC Apple products is still there, they didnt jump off a bridge as soon as they switched to x86.
THUS,
IF...and this is a BIG if, AMD came out with a PowerPC cpu that was so unbelievably badass that the niche markets using Apple PCs for stuff like video editing, graphics, created enough demand for that type of support on a platform, like, say...linux...or even older versions of MacOS, or AIX, or whatever other OS are or would be available on PowerPC, then they have an existing technical base to tap into to do those types of ports.
IBM's problems wouldnt be an issue. PowerPC is a great architecture/ISA that IBM happens to own licensing rights to. just like with x86/intel, they wouldnt have to be the ones producing the chips. AMD would do it all in house, and the fact that IBM isnt really on the bleeding edge as far as fabrication goes and isnt all that willing to invest that much into it is all the more reason AMD might benefit from a good implementation of PowerPC...IBM might tap them to produce chips for their AIX/linux/p-series stuff, and that might carry over to the desktop/workstation market...
IF...it was THAT good...which, as ive said...it very wishful thinking. But if AMD is forced to leave behind the x86 market i think PowerPC is the next best thing for them to do (short of scaling down to very niche markets like they were in their early days). that's all im saying
The existence of some aging and increasingly irrelevant MacOS-on-PPC developers is not a good business reason for creating a new processor architecture.
Someone earlier was right in saying that x86 is largely Microsoft fueled. For a long time there was only one other company with the clout to present a non-x86 platform a widespread consumer audience and that company (Apple) switched to x86. Others like Sun missed the boat altogether: in this thread-heavy, media-rich world, why didn't I get a SPARC desktop? I know, Sun's been walking with a limp ever since that dot-com bubble burst, but for a while they really were in a unique position to threaten x86 should they have wanted to. That day has some and gone.
One of their visions may still take hold. What's a platform that could replace x86 and is in widespread use? Java. Java needs what it's never really had: an OS core to feed it basic services, some usability-oriented APIs, and a hardware platform optimized to Java. BEA had something like this (but based on x86) and it was fantastic.... basically the JVM -was- the operating system, and it was blinding fast, but that's all gone now that Oracle bought them and has its own OS-and-VM strategy. So there's a door open there. I'd love to see someone put out "Java engine" chip given modern manufacturing and muscle by AMD (well maybe just GlobalFoundries!) that runs a full-featured JVM and includes a chintzy window manager, directly on hardware with no Windows or MacOS in the middle to mess things up. Then, all you'd need is driver support and you'd have a x86-alternative with legs.
When looking for an x86 alternative, look for companies or technologies with an incentive to get away from Intel / x86. Java is one. GPU is another, and we see progress on that front as well. (Good job nVidia with CUDA to get ths party started.) This IMO is the real reason behind Larrabee. nVidia and ATI (AMD) -do- have a lot of market clout, and graphics are driving PC evolution in a lot of ways. Also, both companies have incentives to develop x86-alternative technologies. Intel sees this and is developing an alternative so that in the future if ATI and/or nVidia were to introduce a new computing architecture and declare that will receive their majority attention (vis-a-vis drivers, new hardware, etc.) Intel will be in a position to say "that's okay, you can leave, more market share for us."
The other thing I don't see discussed here is how the CPU is now part of a much wider ecosystem of hardware services that exist inside your typical desktop PC. The days of it simply being a matter of a CPU are long gone, now there's a support chipset coming along for the ride, and that comes with little things like PCIe, SATA controllers, memory buses, etc., all of which present not only huge technical hurdles to replace or embrace, but most of which also require licensing from standards bodies or individual IP holders, Intel definitely among them. And once you do that, don't forget you need to rewrite even basic things like your BIOS. x86 is dominant because I can get a cheap AMI BIOS or PHOENIX BIOS from anywhere for a few bucks per unit and it'll manage all my device and chipset configuration. That's "x86" code too, don't forget that.
And another problem: driver support. Drivers are piece of code too, like any other application, and some are built in a similar way to any other application, using regular ol' portable C code or some such. But we're after speed here, and many of the "bottleneck" drivers (I/O, disk, chipset, GPU) do have embedded x86 (aka: assembly language / machine code) to improve performance, and none of that will be portable either. So you have to enable that entire ecosystem to write non-portable, high-performance drivers for a platform that has no user community.
Someone earlier was right in saying that x86 is largely Microsoft fueled. For a long time there was only one other company with the clout to present a non-x86 platform a widespread consumer audience and that company (Apple) switched to x86. Others like Sun missed the boat altogether: in this thread-heavy, media-rich world, why didn't I get a SPARC desktop? I know, Sun's been walking with a limp ever since that dot-com bubble burst, but for a while they really were in a unique position to threaten x86 should they have wanted to. That day has some and gone.
One of their visions may still take hold. What's a platform that could replace x86 and is in widespread use? Java. Java needs what it's never really had: an OS core to feed it basic services, some usability-oriented APIs, and a hardware platform optimized to Java. BEA had something like this (but based on x86) and it was fantastic.... basically the JVM -was- the operating system, and it was blinding fast, but that's all gone now that Oracle bought them and has its own OS-and-VM strategy. So there's a door open there. I'd love to see someone put out "Java engine" chip given modern manufacturing and muscle by AMD (well maybe just GlobalFoundries!) that runs a full-featured JVM and includes a chintzy window manager, directly on hardware with no Windows or MacOS in the middle to mess things up. Then, all you'd need is driver support and you'd have a x86-alternative with legs.
When looking for an x86 alternative, look for companies or technologies with an incentive to get away from Intel / x86. Java is one. GPU is another, and we see progress on that front as well. (Good job nVidia with CUDA to get ths party started.) This IMO is the real reason behind Larrabee. nVidia and ATI (AMD) -do- have a lot of market clout, and graphics are driving PC evolution in a lot of ways. Also, both companies have incentives to develop x86-alternative technologies. Intel sees this and is developing an alternative so that in the future if ATI and/or nVidia were to introduce a new computing architecture and declare that will receive their majority attention (vis-a-vis drivers, new hardware, etc.) Intel will be in a position to say "that's okay, you can leave, more market share for us."
The other thing I don't see discussed here is how the CPU is now part of a much wider ecosystem of hardware services that exist inside your typical desktop PC. The days of it simply being a matter of a CPU are long gone, now there's a support chipset coming along for the ride, and that comes with little things like PCIe, SATA controllers, memory buses, etc., all of which present not only huge technical hurdles to replace or embrace, but most of which also require licensing from standards bodies or individual IP holders, Intel definitely among them. And once you do that, don't forget you need to rewrite even basic things like your BIOS. x86 is dominant because I can get a cheap AMI BIOS or PHOENIX BIOS from anywhere for a few bucks per unit and it'll manage all my device and chipset configuration. That's "x86" code too, don't forget that.
And another problem: driver support. Drivers are piece of code too, like any other application, and some are built in a similar way to any other application, using regular ol' portable C code or some such. But we're after speed here, and many of the "bottleneck" drivers (I/O, disk, chipset, GPU) do have embedded x86 (aka: assembly language / machine code) to improve performance, and none of that will be portable either. So you have to enable that entire ecosystem to write non-portable, high-performance drivers for a platform that has no user community.
and that's the tough part. which is why i think PowerPC is a more realistic route, if there's to be any attempt by AMD of retaining what they'd lose from the x86 market. tapping the PPC/Apple developer base was a weak point (but not an invalid one), really the words flashing in my head are IBM, AIX, and ppc linux developers. a bit of kernel work and bam, AMD's brand-spanking new bleeding-edge non-x86 processor already has support for basically every piece of hardware out there.
but i dont think it will be long before we see something completely new hit the scene, i just dont think AMD, bleeding money as it is, really has an opportunity to make a gamble that big. who knows, maybe as they see it, they have nothing to lose?
no2censorship
Limp Gawd
- Joined
- Oct 20, 2008
- Messages
- 393
I kind of sound like a hypocrite by changing my views so radically, but I've become an intel fanboy.
For a few reasons:
as someone pointed out, maybe intel was kind of protecting fry's and dell buy not allowing them to sell something amd may not have been able to supply them enough of.
amd makes lousy products now; my kuma 7750 be was inexpensive (i wish it was 45 nm, but it's old skool 65 nm=(, but not that great of a value. no cpu fan fail incase my ac f64pro were to die. their bundled cpu coolers are the dickens to take out, they die earlier than intel's do. don't intel's cor i7 coolers allow for some pretty mean overclocks anyway? plus, i heard they made the push-pins on the i7 cooler not a giant PITA anymore.
When i eventually get an i7 920 and the things I'll need for it, all I'll have to do is use the as5 remover kit i have on the stock cooler, clean it off, then put my AC mx-2 on the i7 chip=)
the core i7 920 isn't really more expensive than the inferior pII 940 (runs hotter, slower, and included cooler sucks.)
and i've never liked ati--> [being one of the few out their who would rather have a gf fx 5900 ultra over a 9800XT anyday, as I hated the 9700 pro i had and it's drivers, while i didn't even use the 2x AA mode with my gainward 5900ultra the whole ~3/4 year that I had it, and I have no idea why nvidia didn't say that the 2x mode was RGSS; only thing that sucked with the gf fx was the 50 series drivers, but I just used the 40 series anyway] -->and amd doesn't encourage them to make products that aren't designed ridiculously, so I no longer care if amd goes under.
i don't know who will win the legal case, but it doesn't really sound like amd has one.
For a few reasons:
as someone pointed out, maybe intel was kind of protecting fry's and dell buy not allowing them to sell something amd may not have been able to supply them enough of.
amd makes lousy products now; my kuma 7750 be was inexpensive (i wish it was 45 nm, but it's old skool 65 nm=(, but not that great of a value. no cpu fan fail incase my ac f64pro were to die. their bundled cpu coolers are the dickens to take out, they die earlier than intel's do. don't intel's cor i7 coolers allow for some pretty mean overclocks anyway? plus, i heard they made the push-pins on the i7 cooler not a giant PITA anymore.
When i eventually get an i7 920 and the things I'll need for it, all I'll have to do is use the as5 remover kit i have on the stock cooler, clean it off, then put my AC mx-2 on the i7 chip=)
the core i7 920 isn't really more expensive than the inferior pII 940 (runs hotter, slower, and included cooler sucks.)
and i've never liked ati--> [being one of the few out their who would rather have a gf fx 5900 ultra over a 9800XT anyday, as I hated the 9700 pro i had and it's drivers, while i didn't even use the 2x AA mode with my gainward 5900ultra the whole ~3/4 year that I had it, and I have no idea why nvidia didn't say that the 2x mode was RGSS; only thing that sucked with the gf fx was the 50 series drivers, but I just used the 40 series anyway] -->and amd doesn't encourage them to make products that aren't designed ridiculously, so I no longer care if amd goes under.
i don't know who will win the legal case, but it doesn't really sound like amd has one.