Page 5 - AMD to actually buy ATI!!!

Let me appoligize if i reiterate points that have already been made but here is how i see it.

First off ATI has no fabs, they out souce the work to TSMC and its subsidaries like Wafertech. This alone would let them be able to control time to market, etc. ATI has always had good enginering and not much else, IE poor marketing, manufacturing issues. On the otherhand AMD has shown that they can put products out there and get them sold as they have proved by taking 25% more of Intel's market share in desktop CPUs sales over the last two years.

AMD will be able to make the most out of ATI's technologies by providing them better markets, managagement, and access to more advanced IC processes. AMD also is able to produce 65nm technology as well as having 90nm technology well in place. The pure-play foundries the ATI was using in the past (TSMC) had no tools for the 65nm process and limited ability for the 90nm process. This alone will help develop ATI's GPU and chipset ICs.

There is going to be a move to a CPU/GPU chip as mentioned in this thread. With both companies technology and the recent technology to better manage on chip transistor power use, AMD will now be in a much better position for this new arcitucture. This may be the leading cause for the merger along with inhouse chipset design which only Intel enjoyed in the past.

As for a Nvidia-Intel merger.... don't kid your self. With the move toward making CPU's to be more general purpose ie a CPU that handles 3d rendering well and with the disagreement over chip set technologies the two have been having, the chances of a merger or even a sharing of technology looks like slim/none

Nvidia will get stronger is the market with the AMD/ATI merger. ATI and Nvidia would have less direct competion and would be able to share graphics and chipset technologies.

The only one I really see not benefiting from this merger is Intel who now has a serious competitor in all markets. While they have just realesed the the conroe and a soon to come quad core chip. AMD is keeping very quite about its new transitor power management technology. While Intel turns off the sections not being used, AMD will use six stages of power consumtion. This mean that transistors or sections can use say a 1/3 of the power for low priorety requests while high priortey requests will recieve the wattage they need.

Starting to ramble now, signing off

Welll amd is going to drop the ATI moniker. Wonder what will happen to radeon will it become AMD RADEON?

If they start integrating the chips into one, then possible, although it will probably be some new name. As of now they will probably stay with ATI.

I disagree, AMD having the fabs to produce at the 90nm and 65nm level will have no effect on ATi.

Why?

AMD can't keep up with demand as it is. They've off-loaded some of their own production needs on third parties jsut to keep up with CPU's. ATi doesn't have much to gain fab-wise, they're still going to rely on others to do it. AMD is going to build another giant fab in New York State, but that won't start until at least 2007 and with that kind of construction and the technology and cleanspace to be set up, it won't be ready for a long, long time.

The other part of it is that AMD probably isn't going to be using its latest-and-greatest fabs on chipsets and GPU's. AMD primarily will still be a CPU company, and to be competetive they need to keep up with Intel in fab process technology. Historically AMD is always behind anyway, look at right now for an example, Intel is a good way into the 65nm cycle and AMD is still converting fabs to be able to start. This means ATi (AMD chipsets and GPUs) are probably going to be pushed off a generation, exactlely how Intel does it - they move CPU's to a new process, and the old fab setups then produce chipsets and integrated graphics. That way they got a longer return on the heavy investment of that fabrication process.

AMD hasn't gotten to 65nm yet. They're still on 90nm, and the GPU market is already moving to 90nm and talking about the half-step to 80nm as well before, presumably, a drop to 65nm. There's no way ATi is going to benefit from this merger in fabrication if they're getting AMD CPU fabs this long after nVidia already introduced GPUs on this transistor size.

Sure, they'll build more fabs to help alleviate this, but those are extremely expensive and the combined companies have enough debt as it is. It'll be a while I think before the two get to that point and in the mean time I see ATi graphics suffering; they won't get any help from AMD's fab and processes, or production capability, and they'll be hurting from the company reorganization and general turmoil surrounding this. What happens after that, depends on how AMD leads the new GPU wing. I hope they keep the same business going, with the insane competition in the GPU market, because I think it'll be hard to come back once falling behind and then nVidia has a monopoly on performance graphics. On the other hand, if thing do start getting combines more that advantage for nVidia may be fairly short-lived, but I'm still not sold on the CPU-GPU integration.



We'll see... I'm still skeptical, as for fast graphics as we see today you need very fast RAM. I suppose the combination of the two will be quite some time down the road, so maybe by that time it won't be as big a deal as system RAM improves, but historically it seems GPU RAM is developed much more quickly. Graphics performance may be held back by this. Unless with future process shrinks and smaller core size AMD decides to add a ZRAM bank as a kind of large cache, but then that technology is supposed to die as doping techniques change on the fabrication process anyway, I think.

On chipsets, I absolutely agree. It's more income for AMD and having chipsets and CPUs under one roof has to help both along. It could be death for some of the lower-end chipsets though, a la VIA.



Absolutely agree, Intel gains nothing.



I'm not sure what you mean.

ATi and nVidia have less direct competition? They're still the only two that make performance graphics, and they still both make chipsets for the same platforms. They're still competing, unless AMD decides to take ATi's graphics and chipsets to the budget level and let nVidia keep the performance/enthusiast market.

As far as sharing technologies... chipsets yes, graphics I don't think so. AMD already shares chipset info with nVidia and ATi, they have to so that each can continue producing high-quality chipsets. I'm not sure what you mean by technology in this case, maybe that's not what you mean at all. But on the graphics front, if they're still competing on the enthusiast front, I don't see why they would share tech any more than they do now.



Meh... Nothing ATi does with AMD is going to affect Intel much except for thing perhaps far down the road (integration of CPU/GPU for example). ATi may bring onboard graphics in-house to AMD, but they've been available for a long time on ATi and other chipsets anyway. ATi already produced chipsets for AMD and Intel, being bought by AMD doesn't change what they do for AMD that much and if anything Intel stands a chance to ship a few more units of their own motherboards as ATi is going to completely fall out of the Intel chipset market it seems. They're actually sans a competitor there, with no percievable strength gained by AMD in the short term.

Conroe looks to kick AMD around for a while to come, this is true. Their 'quad-core' is crap though, it's the same thing they did with Pentium D, just squashed a couple processors together on the same package... hopefully this time they will at least by some on-package interconnect instead of communicating over an outside bus. Here AMD still stands a chance, they'll be moving straight to 'proper' quad-core and some changes to the core that seem to make it more similar to Conroe as well, such as added ability to crunch SSE instructions which AMD currently receives little benefit from.




That's very cool, if they can get it to work and scale okay with clockspeed while not harming performance. But what application does it have? Portables, it'd be great, but desktops? Servers? When those are cranking away all they'll be running at full wattage, when they're not heat isn't a problem. They'd have to demonstrate some heavy power (and money) savings.



PS - Glad to see you around again MuTTAR!

ATI previously had no access to 65nm process, Nvida has none. The 80nm process you refered to is a revision of the 90nm process, just like the x850xl was 110nm which was a revision of the 130nm process. They are made on the same tools. When you have to actually switch tools is 130,90,65,..

ATI now having access to 65nm technology will let them begin to design chips the couldn't have thought they would be able to send to fab in the near future. Nvida in the meanwhile also does not have their own fab either so they will be stuck with 90nm tech until the pureplay foundry they use (TSMC also) gets tools in to produce 65nm or better. But that is a million to billion dollar investment, when 90nm still only makes up less than 10% of their current production.



Right! It wont be ready until ~2010 or so, but we are not talking about next month anyhow we are talking about the furture of the big 4 in the enthusist PC market. Having a new plant in 2010 is an important step. Them having already laid down prelimnary plans before the merger speaks volumns for the kind of managment that will take ATI to the next level, as stated

So ATI can set a plan to move to 65nm GPU by 2010. Can Nvidia say that or can Intel say they will make a high prefomance GPU based on the 65nm process by 2010? I think not.

So for a short time there will be a reorganization but in RL the dirty word reorganization often shakes out the rug so to speak. Also, they are not cutting there production ties with TSMC. At the most there will be a slow down for a month or so, then they will come back stronger than ever.




Nuf said dap dap ty Anand and Rahul... peace out.

ATI for this first time will be able to make in house chip sets like Intel, onboard graphics like intel. They will be able to compete in every market now with intel.(strong period) This is like Sony publishing a good windowed OS and also making PC apps for every day use. Intel has enjoyed a bit of a monolpoly for several years and now AMD is "just like mike"
if not better.

If it is avalible in stores before AMD releases its next flagship for the sktAM2

The on/off is the technology already used in the Conroe. and in the newest Xeon chip. While the 1-6 stage will be in the next AMD chip(which will also be 65nm based). If we have learned anything for AMD in the last 2-3 years clockspeed does not = fast *looks at archnaid*. As for performance I have never seen a chip even in beta benchmarks but differnet levels of on or off seem better than the black and white approach that Intel is using as of today.

I realize that, that's why I called 80nm a half-step.



But they don't have access to it. That's the problem. AMD does, but none of that 65nm capability is going to be used for GPUs and chipsets.

In the future they will... in the future they would have anyway via other routes. AMD is just a different partner to fab chips for them, except with AMD their stuff will be on the backburner. The point I'm trying to make is that it's possible they could have gotten 65nm parts from TSMC or other sources before AMD opened up the 65nm fabs as the CPU division moved on to something smaller.

ATi is basically going to get the CPU division's leftovers. It means that ATi parts are only going to be able to advance as far as the CPUs can lead, and say we hit a hiccup in that requires some time to iron out - 90nm for example presented problems - that prevents ATi from going to the bleeding-edge process, the limit of what can be done at the time, because CPUs will be fabbed at that process with GPUs and chipsets in the background.

10% of the third parties' production is 90nm? I'm surprised at the low number, but that doesn't necessarily help your point - they may not hit every step along the way, but nothing is preventing them from jumping from 130nm to 65nm.



We'll see. AMD can't keep up with their CPUs yet, and as we go to 65nm that, for a relatively short time, will get worse, not better. If they really deliver on the K8L (and more importantly on the time scale we're talking about, future iterations thereof) demand will continue to rise for AMD chips and they will just be catching up to supply constraints right now. AMD is using all they have to keep up with CPU's, and they have contracted third parties to help production. Even with a new fab, given current constraints and forseeable increase in demand, what chunk of the pie does that leave ATi?



ATi planning on 65nm GPUs by 2010? Maybe. See how fast CPUs shrink, if we're not down to the next big step ATi can forget it, because AMD is going to want all those 65nm fabs turning out CPUs.

By that time some of the third parties could be transitioning to 65nm fabs as well. They may be kind of slow to do it, but nVidia should have quite a bit of leverage there; whoever gets 65nm fabs up and running gets a near-exclusive deal with one of the largest GPU makers in the world. It takes time to transition, this is true, it takes years... but 2010 is ~3.5 years away.

Intel and performance GPUs? Meh. I think they're not too eager to enter that market.



No, they couldn't afford to sever ties with TSMC, that's the only place they're going to get chips from. So how they come back stronger than ever I don't know - same GPUs, same fabs, same process.





Two points.

First, I was thinking closer to the present. Right now it just wouldn't work IMO, the latencies are just too high for system RAM. We'll see how things look when this actually comes to fruition, if it does, because I hope they're not just going to glue a GPU and CPU together on one package a la Intel; and integrated CPU/GPU is going to be a ways down the pipe. Whole new design and architecture, probably.

The second is that we'll half to see how integration plays out. If it does become two cores on a single package, will the market accept it? Cheaper it may be, but if in the short term it's just different packaging and presumably a performance hit, will it be accepted? Supposedly Intel has had this idea in mind for quite some time, but we haven't seen it yet. Perhaps there's a reason for that. Now, full-scale integration... That could be different. It would be an interesting cycle, re-integrating what was separated out. If it can be pulled off perhaps it will be fine, but the market would be far different because of it... no more upgrading CPU or GPU as fits your need, you have to upgrade the whole package (and pay off the company's R&D and everything on each). Graphics and general performance become linked, it's no longer be possible for us to run high-end graphics on an otherwise relatively mediocre system to make a budget gaming platform, for example. Then again, maybe that's not such a big deal as there's not a whole daughter-card and RAM to upgrade, just a single CPU-like package.




In0house chipsets and onboard graphics, I agree. And it will bring them some more revenue if all goes to plan. But that's not really a new market to compete in with Intel. AMD has had good third-party chipsets through both nVidia and ATi, in-housing one of them just redirects cash flow a bit. A chipset is mostly secondary to CPU choice anyway, and that's what will still drive AMD - someone looking at a Dell isn't going to notice (if Dell even tells consumers) what chipset is used, it's all about the CPU. Intel pushing units of motherboards is directly tied to the CPU recognition by consumers, so it's not like a new market they will have to compete with AMD in, AMD will now just be getting a little extra money on some AMD computers.

It's not like Sony making a good windowed-OS because that implies creating an alternative where previously there was no big competitor. The competition exists for Intel already, look at how AMD has eaten into their market share. Whether or not the chipsets come from AMD doesn't matter to Intel, they're still fighting the AMD CPU and brand recognition. Chipsets and onboard graphics don't compete. Consolidating the CPU and chipset won't be a much bigger threat than what was there previously, except that it means more cash flow for AMD.



I still disagree. Conroe kicks the pants of the current AMD sktAM2 flagship, and I don't mean the Extreme Edition or whatever, the FX-62 or w/e it is now has competition in chips half its price or less. Another clock bump, up to IIRC 3GHz on the FX, isn't going to save AMD the performance crown until K8L sometimes next year.

Availability may not be all that bad. Already Conroe or Woodcrest are showing up on the 'net for notification or pre-order. And that's for PIB sales, I'd imagine Intel is more concerned with supplying manufacturers like Dell and Apple with the good before DIY PIB parts are put on sale. If it really is bad, it's a matter of how well the K8L does. Native quad-core and some other goodies, but then, with AMD's limited fab capability and even more limited (as of yet non-existant) 65nm fab capability it may be a while before the K8L is available either.



I hadn't heard that... I need to read some more up on it, wonder if Jon Stokes has his in-depth study on Conroe up yet...



I'm well aware that comparing clockspeed across architectures is a poor way of guaging relative performance, but we're still stuck with the fact that coreclock bumps is one of the few ways we are able to increase performance on a given architecture. We can do some revisions and tweaks and play with cache and the like, but all I was saying that if implementing this tech hurts the ability to scale frequency in any way it's going to die a quick death, because that would basically be selling your future. Inability to scale clocks on the current architecture means you're stagnant until the next big revision.



I agree that on paper it looks nice, I just wasn't aware this was not something on the infamous '3 to 5 years' schedule and that it was already entering mainstream on the Intel side of things.




I hope that makes sense, all that. I need sleep.