Tolliman X3 gets a name

I thought they were going to be called Defecteron?

What the hell is the point of a 3 core CPU if you have to disable one of four cores on the die to make it so????

Makes it so they can sell the chip rather than throw it away if it has a bad core.

-mcg

But that's different than disabling a working core...

Who said that it was a WORKING core? This is done to increase yields of otherwise defective chips.

Don't call me Chris until you post your real name. And even then my name is Christian NOT Chris. Here you go again. So the TACC SuperComputer will be crap because it uses Barcelona? Or Dell was just lying when they said Barcelona creates the most powerful VM Host ever? Or perhaps the near 100% scalign with clock is a figment of everyone's imagination?

No, I know, Anand and David Kanter and everyone else who doesn't share your opinion is smoking CRACK? I can't prove their yields and neither can you. STFU!!

BTW, why are you so rabid?

Christian, since when did I state that K10 sucked? I stated that K10 will be outperformed in single threaded environment, but will do pretty good when scaled. You can do a search, ask people, or do whatever, but I've never said K10 will suck in terms of scaled performance.

On the other hand, AMD's 65nm sucked, because they can't yield Barcelona at an acceptable rate (a little bit more than 30%), and at competitive clockspeed. How do you think AMD's going to compete with 2.0Ghz Barcelona? How do you think AMD's going to save themselves if they can only yield at 30%?

What about Anand and David Kanter? I used both of them as a reference to my argument. I used Anand's figure on memory bandwidth and latency to illustrate that K10's latency is too high. I used David Kanter's comment to argue that K10 will fall short of expected IPC, and fall short of its competitor, who has been out on the market for 6 months!

Christian, as I said, stating your real name doesn't mean jack if you still can't get the most basic things right. You still can't distinguish between latency and bandwidth. You still can't accept the fact that having a tri-core is a yield issue, rather than a binning issue. You simply can't accept the fact that AMD is in the mud, and they're sinking day by day, while Intel is poised to move in for a kill. If AMD goes down like this, do you think FCC will go after Intel for its monopolistic behaviors? No, because AMD basically bled itself to death.

And cut the crap on Intel's "selling under-cost" argument. Just because Q6600 is 70 USD more than E6600, doesn't mean an E6600 die cost 70 bucks. You claimed that you were once a software programmer for Microsoft, yet you still haven't demonstrate that. Simply spreading mis-information, and use "because I was a software programmer from M$", is not an act of heroism, but an act of irresponsibility. As I said, you're embarrassing yourself.

By the way, why are you so rabid?

i am no fanboy. of amd or intel. i am just a hardcore gamer. that is all. i think amd is doing a good at trying to make a full quad-core like intel did. amd just needs to work on it some more. i mean really nor intel or amd have room. to talk they have both made awesome cpus in the past amd had the crown and now intel does. i mean i just got a q6600 and its awesome. i got my buddy a amd x2 and its awesome. i mean really all this comes down to is peoples person options. on this. i mean amd is for the budget builders. and intel is for the more powerful user's. but in the end both are good companies. i think its awesome amd made a cool x3 cpu. i think its awesome because its something differnet in the market. and i dont know how much they will cost. but amd and intel are both good companies. i know you think i am a noob or something. but i am just saying. like i have said before companys have there good days and there bad days just like we do. but amd is slacking right now. and intel is on its game this is just how i see it. i mean look at netburst. it was bad and look at barton core. they where both hot. and run ok. or good. which ever you pref. but in the end both are good companies. i mean as long as your ocmputer does what you want it to do its good. and awesome.


just saying it how i see it.

Brian

I want an X3, screw X4! Cheap asses ftw!

I mean, I should be able to run Xbox360 ports better than anyone! w00t!

Word. I remember seeing somewhere that an AMD exec said that if they had the time over again, they would have gone with 2 duals glued together.

Oh? I need to do more reading, I thought they used candle wax.

It seems fairly obvious to me that the goal of AMD under Hector Ruiz is to "break the monopoly", even at the expense of profitability.

we'll so far i like the quad-core. its awsome. i did a prim95 test and never went over 107F with 4 running full load. which i think is good. unless i am wrong but amd will make its cpu better it just takes some time.

Brian

They don't have time!

They'll break the bank first.

I'll just say you can get more Barcelona dies from 65nm 300mm than Opteron dies at 90nm 200mm.

I wouldn't necessarily say that AMD's manufacturing process is a joke at all. AMD has been able to get 800 more MHz out of their 90 nm process on similar chips with little to no increase in TDP, which is NOT a trivial feat. The first 90 nm dual-cores went up to 2.40 GHz and had 110-watt TDPs. The current 90 nm duals go up to 3.20 GHz with a 125-watt TDP and the 3.00 GHz unit has an 89-watt TDP.

What AMD lacks in manufacturing is the ability to do process shrinks and ramping as fast as Intel does. This is not all that surprising since AMD has two main fabs with one at 90 nm and one at 65 nm. Intel has roughly a dozen big fabs with several at 65 nm and a couple at 45 nm. They have the fab capacity to shut one down, install new tooling, and then ramp while AMD has to be very careful with transitioning as they need the fab to remain somewhat operational during the retooling period.

And as far as stubbornness goes, I am pretty sure you're referring to AMD not making dual-die MCM multi-core chips. This is also not surprising as AMD's IMC platform is not nearly as amenable to it as Intel's FSB platform is. It's as easy to put two dies in one socket as it is two dies in two sockets with an FSB. An MCM with an IMC is a little trickier to accomplish. The best way to accomplish that is by having one IMC per die (two per CPU) and a wide backside bus between the two dies (such as HT 3.0). This necessitated a redesign of the die to accomplish- specifically the independent 64-bit IMCs per die and HT 3.0- and AMD waited to incorporate this into the 10h rather than port it to the K8. They had to redesign the core anyway, so why not make a quad-core version while they were at it?

I think neither camp is more stubborn than the other. Intel is finally getting rid of the reincarnation-of-RDRAM FB-DIMMs and the FSBs in their servers, both of which are overdue. And AMD has 8-core dual-die MCMs in their roadmap as well.

What will be interesting to see is how Intel handles its first IMC chips...

MU, I'm glad you still post here. Thank you.

Amd made a huge mistake buying ati, the concept is solid but the price was too high. AMD is only going to get back in the game with dirt low prices which they can't do or very impressive new technology which they don't have.

The future is not so bright that they have to wear shades.

TC

You're right, I don't. But I can guess, based on the data available, right?

Based on AMD's own slide, current 65nm Barcelona are yielding at 30%, maybe a little more. Then AMD came out and said, "look, we really can't make 2.5Ghz Barcelona now, but 2.0Ghz will do". After a while, AMD said, "hey, let's launch a new processor that was never on the roadmap."

Don't you think 30% yield has a little merit now?

My point was that I wouldn't be happy with 30% yields and AMD says they're happy with the yields. If they really wanted to harp on yields they would have rather than defect density. I still don't think the yields are anywhere near that low. The higher bins could be lower for steppings prior to B2F or B3, but you are saying that ONLY 3 out of 10 chips EVEN WORK.

That's nearly impossible. They would never have enough chips for Newegg, much less Dell HP, and Sun. Get yields out of your head. When AMD or a representative that has been to Fab 36 says that, I'll believe it. Until then their yields in my estimation, with the near maturity of 65nm, are in the high 60-70%.

The B2F stepping will be the release of Phenom. B3 will probably be 3GHz+ Kuma and Tolliman.

By stating 30%, it means AMD can only yield 30% of the functional Barcelona, with four cores intact. Adding Tolliman, I'm not sure about the yield, but if the defects are not clustered, then I'm sure the yield can go up to 50% or more. Then, adding Kuma, that adds another 20%.

So for AMD to yield all processors lineup with K10 core, I would suspect about 70~80%, which is reasonable. But for Barcelona and Phenom, I'm suspecting no more than 40% as of now.

Let me see, after Barcelona went on sale in channel, the chips were gone within several days, and most vendors do not get another shipment until November.

On the other hand, several OEMs complained about their availability.
Dell still has no Barcelona available for their servers. As of now, you cannot buy a Dell server with Barcelona. Same with HP and Sun.

Wow!....but so far AMD only plan to release them at 2.5Ghz maximum for Tolliman....

Guess AMD isn't too confident about B3 either.

And the reason why OEM still don't have enough processors several weeks after launch because?

Intact means, a die with four working cores. The clockspeed is hampered by AMD's 65nm process.

They should! This is the first time I've heard that several weeks after launch, there were still no availability from OEMs.

Oh wait.. quad fx?

I know that Chartered is not receiving orders from AMD as of now. I'm not too familiar with the fabrication process. I'll dig a little more to answer this.

The problem with using a foundry is that you have to place orders in advance, no matter if they sell or not. And usually those orders are not small, but pretty large. Once the order is placed, it cannot be canceled easily.
Once AMD does have it's K10 line running at a level acceptable to fulfill orders, I'm sure they will hit up Chartered to run more of their other products.

We shall fight now, to the death.