So if that holds true and they do get the 45nm process running so fast it'll be all the boost they need to get even, or at least so close to Intel that it'll make it much more interesting for us tech geeks :D

[Basic yeilds right now. That means some chips are working, most aren't. Mature yields are required to approach production. That's typically 80% or higher. -Rick]

on theinquirer.net it said that AMD will not make 45nm chips until 2009 and here i am reading something else... whom I would have to believe?

[I'm guessing The Inquirer meant to say 32nm, for that is the timeframe for 32nm products, though possibly into early 2010. Dirk Meyer was very specific that 45nm would have a production ramp in the first half of 2008. -Rick]

Interesting development. Still too early to say how this will pan out. Much to do still for AMD. Yields, leakage testing, ramping to come. Its going to be a while before we see some finished chips. But, still a good sign for AMD. I wonder if their relationship with IBM is starting to pay dividends now? Only time and 45nm process specs will tell.

AMD may yet gain some ground on Intel. I can't help but wonder how much it's costing them to do so....

I think its great that AMD has plans but I highly doubt that they will be able to really produce these with 80% or better yeilds and make a good product in that short amount of time.

Intel had been working on its 45nm since before Conroe was released. They were even working on Nehalem before Conroe was released.

I just hope AMD knows what they are doing. This is exactally what Intel did when the K8s were the top. Just threw somehing out of the Fab as fast as possible and what did you get? Dual cores that could keep a house in Russia warm.

Isn't Shanghai the 45nm lead vehicle product for AMD?

Not a word on Shanghai. Why didn't Dirk metion it by name? Why didn't any analyst at the conference call ask about Shanghai?

Something does not compute

As the author said, he was VERY specific saying they WILL have them in 1h or 2008 and that they ARE making 45nm chips right now.

AMD has been working on 45nm for quite some time too, go back and look at when they had the SRAM chips made using the 45nm process.

If you look back you will see IBM started taping out 45nm in the same quarter that Intel did, which means that AMD benefits from that as they 45nm development agreement with IBM. So it really isn't surprising that they are closing in on pumping them out.

The fact (I don't remember hearing Shanghai either), that they didn't mention Shanghai doesn't really worry me. All that would concern me is making 45nm, the process switch seems to be the most difficult part for AMD as 65nm hasn't been working great for them.

That's what Fabtech says in line with the rest of the "45nm" non-Intel crowd. Look at the delays at getting 65nm out, this looks to be the likely scenario for AMD again. That's 2009. On the other hand, AMD have yield issues on their Barcelona if they can only ship "tens of thousands"? I hardly see any Barcelonas around either up to now, where are the servers from Dell, HP, Sun and IBM? All "Coming soon..."?

It says nothing about HK/MG, one can only assume this is a shrink of the K10 project that still doesn't work at 65nm.

We desperately need some competition on both the cpu and gpu side because if Intel ad nVidia keep that top dog spot for too long, we will end up with high prices and slow or minor chip upgrades each time...

AMD clearly stated that they were not yield problems, but rather they are having to tweak the Uarch (Barcy) to get the clock speed up. They claim the defects for K10 are the same as K8 on 65nm. The conference call sounded very positive. Just because they had issues with 65nm doesn't mean they will have the same issues with 45nm even though it is likely.

I think that they will get this one right.

It just sounds too good to be true. And by the time their 45nm is out, I'll already have a Penryn, and so will many others... No guarantee their 45nm's will even be priced reasonably, or not that much different then intel's offerings anyways.

Now let's pretend for a minute that AMD does somehow get this product out on time...... What about Nahalem? Won't that likely squash their 45nm tech anyways? Intel has had so much lead time to develop a big nasty Nahalem, I just don't see AMD being the performance king for quite some time =(

Is this good news? Absolutely! Is it likely to happen by next summer? Very questionable.......

Don't get your hopes up too high AMD fans.

I belive 4Q is there turnibg point.The folks who holds stock and the fanboys want miracles.396 mil is a positive from the 1st and 2nd Q's.The barcy's that are coming out are high to low midrange,thats where the money is.What folks are missing is that when all this is said and done AMD will be a powerhouse(cpu,m-board,gpu).

That is a pretty pessimistic view imo. I think what you are missing is that even though you will have Penryn and many other will, if AMD produces a competitive chip, they will still sell. CPU sale won't slow down because of Penryn, in all likely hood CPU sales will increase rapidly next year because of emerging markets.

You mention Nehalem, well AMD won't be standing still on the Uarch either. They will be steadily improving Barcy, and will be releasing Shanghai which will be a reworked K10 according to speculation.

It seems like you are only focusing on Intel and not what AMD can do. By the time Nehalem comes out AMD could drop a bomb on Intel who knows.

I tend to try to look at things in a more optimistic manner. You theory revolves around the idea that Intel will not have any setbacks which is doubtful but possible. If anything AMD has proven before that they can make a very competitive cpu.

Meyer chose his sentences wisely.

manufacturing 45nm processors "right now" DOES NOT EQUAL working 45nm processors.

...though Meyer said he would provide additional guidance in December COULD MEAN other factors could surface during this time and delay 45nm production further

So far in Q3, "tens of thousands" of quad-core Opterons have been shipped BUT NO EXACT FIGURES? Intel said 2 million, so why AMD gave a shabby reply?

Phenom in on track and will be launched "in the middle of Q4," - REMINDER that's next month and still no samples or proper demos of Phenom

Before I go....

Seems all 4-core (and 3-core) Phenoms are destined for AM2 socket (most probably to accomodate the dual power planes) means NEW motherboards. And from what I read recently NOT ALL current and even older server boards can be upgraded with Barcelona (not even with BIOS upgrades!) and only sepcific NEW motherboards can be used. Another AMD hype busted again.

So all those out there thinking of buying cheap AM2 (soon to be outdated like Socket 939 previously) expecting to upgrade to quad/tri-core.... just beware!

I wonder if the lack of 65nm enthusiasm on AMD's part has been because of the focus on 45nm. It's unlikely, but a little conjecture when talking about AMD's future products has been the rule of thumb for a little while now.

If they honestly can have 45nm parts on the market by June or so of '08, it would be an incredible move on AMD's part. If they really are STILL tweaking the K10 uarch, they've been having some struggles indeed. If they do manage a ~2.8-3.0Ghz part on 65nm it would bode rather well for AMD's 45nm parts.

Requote... Seems all 4-core (and 3-core) Phenoms are destined for AM2 (AM2-plus) socket most probably to accomodate the dual power planes means NEW motherboards.

There was a rumor regarding AMD which I believed more than I believed release dates for new CPUs and GPUs, that they were going to skip 65nm and go for 45 straight away. The rumor was logically based on AMD's power (or lack of it) to change it's fabs all the way like Intel did. Because intel still produced 90 nm when producing 65, and simply could switch the most out of date production facility to the newest while offering massive quantities for the tech in between. AMD had better listened to that rumor IMO... Competition is good, bleeding of AMD is not.

the BIG problem with amd.... is its SOCKET! THEY KEEP ON CHANGING IT..!!!! WTF...!!!how you build trust if you keep on changing it..! a,939,940,am2,am2 ,am3.... come on...!!!!!

They did only say that they're just making 45nm, they never said anything about working processors. In fact they never said anything about processors. Maybe this would be usable only by around 2009. Intel has been making 45nm for quite a while, if those were working out of the gate then we should have gotten 45's by last year.

sounds like AMD is just trying to catch up by skipping the 65 and heading straight for the 45. If i remember correctly, AMD has less fabs than Intel, whitch would be why they are selling only tens of thousands, instead of millions. Also i have a hunch that they are kinda focused on their Fusion processors.

At 20 % , selling 400,000 or so in Q4 will mean theyre back, with marketshare

Do you know how many sockets intel make in last few years ?
for AMD its just 939->AM2->AM3 for desktops (being AM2 and AM3 compatible)
or 940->F for servers
as why to make AM3, its targeted for boards with lots of new stuff and features, DDR3, PCI-e 2.0 HT3.0 etc.
And gues what, intel is making again new socket for new CPU coming in 2008.

Jettto: I fully agree with creating a new socket every time. It's only decent. Because thinking you can run a superior processor properly on an inferior mobo is just stupid.

[How can you say that? There are electrical protocols which must be adhered to. If they are within spec, there are no inferior motherboards. CPUs contain large caches to overcome the limitations of the memory architecture. The FSB, or in AMD's case, multi-point HyperTransport bus, adheres to a protocol which allows whatever equipment is on board to function properly. Upgrading the processor should affect nothing on your system, other than additional performance, provided it adheres to the protocols--which it will, by the way. Have you ever heard of an Overdrive processor? These existed back in the early 1990s, and allowed a 80386 core to be plugged in to an 80286 motherboard, for example. Or a 486 in a 386. Or a Pentium in a 486. These allowed upgrade paths to more advanced software potential, but without having to replace something that's often not as important as the CPU itself. Especially back then when the CPU frequency could make a huge difference in performance, not changing motherboards. The ability to upgrade a CPU across process revisions today is limited more by power requirements than design considerations. It would be an easy thing to provide overdrive processors, but they wouldn't make as much money because they wouldn't be selling chipsets to support the new sockets. -Rick]

What good is it if you can upgrade your processor, but your mobo does not support half the specs? AMD could force processor support through bios updates just as well as Intel. But what would the point be.

[What specs are you talking about? Not having the lastest PCIe? Not having SATA? FireWire 800? On-board audio? Faster networking? What? The last time I checked, Windows and Linux ran just fine on even old Pentiums. They don't need the high-end equipment to function, though if it's present they will. Software runs primarily on the processor, with the OS handling the interfacing between CPU and on-board hardware through drivers. Provided those drivers function with the new CPU, then it would only make the system go faster having a newer processor with greater processing capabilities. -Rick]

And I loathe hardware dealers who make abuse of Intel's one socket fits all policy. Because your average Joe just doesn't know that it's practically useless to have a core 2 duo running on a machine with a 400mhz bus. And then they slap in 1066mhz RAM in there just to make the guy think he has the best of the best. While he'd probably be better off getting a decent P4. It's shameful really. Intel creates an avenue for exploitation. But hey, I guess that's how you get rid of obsolete stock...

Backwards compatibility is a shameful practice in a performance driven market. Sure if everybody just wanted to play minesweeper then there would be no problem.

[Not everyone is gamer. In fact, most people today have computers that are 10x beyond what they need. They simply have them because that's what's available today. People in the U.S. and other high-tech countries are so spoiled. We don't have any idea what it means to have a high-speed computer because typically the OS will be so bloated it will bring the machine down to its knees so that it appears a modern computer is only two or three times faster than a Pentium-133 based machine of 10 years ago. It's sad. Very sad. -Rick]

And of course Jettto... AMD does not require you or anybody to upgrade every single time they bring out a socket. So just STFU. Reading you dribble makes my eyes hurt.

[... -Rick]

Amd was having big issues with 65nm and instead of investing huge money and time to solve it, they rather invest it into 45nm whitch is making more sense. I dont know how far is AMD with 45nm but with IBM using basicaly same technology and already having experience with it and posible problems it should be easier then 65nm process.
90,65,32nm etc processes are so diferent that problems with one of them doesnt mean same problem may be in next one and i believe AMD is already making they research into 32nm too.

AMD can't keep this pace much longer. They've already made deals with Chartered and IBM for fabrication so now they're dependent on them. If AMD has to go it alone, they're already finished. If their partners are contributing as they should it is likely that they'll catch back up to Intel.
This needs to be a two horse race or we all lose.

And the socket thing only matters to enthusiasts anyway (and maybe server engineers). Your parents and your employers don't buy CPU upgrades, they buy new PCs.

xrodney,

AMD has to go through 65 nm route because of the cost.
45 nm process equipment is said to be 4 times more expensive than 65 nm one.
Also if the tech is on the market for some time they can use proven tools.
Moreover, it is impossible to switch architecture and the process at the same time.

I cheering for you AMD, but I only buy based on price vs performance vs quality.

how many socket did intel had? compared to amd...

they are still on 775!!! and its all working fine...

i understand the updates... blah blah blah...
but were not talking smart here... were talking MARKETING!!!
of course you need to update the mobo with the processor... but hey... you could also buy a good mobo with a cheap cpu then later buy a good one when the best one(when it prices go down)... because its not everybody that play games or convert dvd.... 24h/day...

the more complicated it is for the average customer... they less they tend to buy it....
dont wonder why intel has always been #1 even with a worst cpu...!

I know quite a number of them. Many were dismayed and majority went Intel rather than AM2 after waiting so long. AMD lost customers there. Phenom won't change that situation as AMD's negative attitude towards enthusiasts continues. AMD has released little information or nothing of interest about Phenom except low clocks. Since Barcelona is the closest thing out there and benchmarks look disappointing, the migration continues.

The average customer buys from Dell and other OEMs. They don't build the PC themselves. If they were to do that, they would generally know what they are doing.

As far as the socket changes, that is no different from Intel changing the VRM. The end result? You have to get a new motherboard, same thing.

per jetto's comments about AMD making upgrades in CPU more difficult...

Do you remember when Intel went w/ the P2, the one w/ the cpu module. It was a plastic catridge that had the cpu & cooler built into a single unit, that was basically a daughter card that you had to plug into a slot onto the motherboard.

The slot you plug'd it into was sorta like a PCI slot, it wasn't PCI, but just looked simular to one.

I don't remember all the tech names for it, but when AMD came out with one, it was the 1st generation Athlon.

Anyway, the point being, before Intel 1st came out w/ that garbarge, you used to be able to use Intel, AMD & Cyrix CPUs on the same motherboard, as long as BIOS supported it.

Intel went with the cartridge CPU to FORCE people to use a motherboard that would ONLY work w/ Intel CPU.

That FORCED all end-users to choose a platform, you either bought a motherboard for an Intel or AMD CPU.

So, you think AMD make a lot of socket changes, think back a bit, maybe you'll remember when all that started.

I'm not going to support Intel just because thier CPUs cost more & are a few nano-seconds faster. I'm certainly not going to support a company that tries [w/proof w/more than one occasion] to railroad other companies out of buisess.

You think Vista has hefty system requirements, Intel & Microsoft have been working together for years to help force & push upgrades. [[ hence the term WinTel]]

If AMD did that, I'd have to buy a small system w/VIA chips, for now, they don't, so I support AMD 100%.

The whole point is not about who actually wins, but I guess Intel without competition in the future will
price the products their own way , so if this is true I guess its good news.

Look carefully at the quotes:
1. "We are manufacturing 45nm parts right now" This doesn't mean these parts would be sellable and Intel could make an identical claim that they are manufacturing 32nm parts right now.

2. "He said they're on track for production ramp in the first half of 2008"
This could mean anything - it could simply mean they are putting the tools in for 45nm in H1'08

3. "Basic yields are being attained right now at 45nm using existing materials."

Basic yields = ? (probably means poor?). 'Existing materials' likely means this is a rev -1 (more or less a 65nm process with 45nm features - it definitely is not near the final performance targets.

What is most important with all of this "we're ramping", "basic yields", etc... os the one thing that is omitted - when will AMD's first 45nm parts be available for purchase (not 'launch' 'shipping', being validate' etc). The press needs to start pushing a little harder for some actual (and meaningful) data and dates - AMD is like a US politician who speaks a lot, but it is impossible to tell what they really mean. They parse the words carefully and speak in generalities so after the fact they can say they excuted to plan (see K10 launch, 40% better? mid-year? clcokspeeds?)