AMD praying ‘Barcelona’ makes up for four-core mistake

[m25]

They also say that Shanghai and the 8-core Montreal will still be aimed on the same sockets of Barcelona.
http://www.theregister.co.uk/2007/03/17/amd_rivas_barcelona/

 

[m25]

Actually, it allows you to double core count without reworking the arch or doubling sockets. Don't know if AMD can do it on K8/K10 arch but that is definitely a point Intel got better than AMD.

 

[m25]

Up until now they have boasted their exclusively NATIVE chips and now they tell a different story of that. They're also planning the 8-core Montreal to fit on existing sockets so I don't realy see where the crap lays here.

 

[qcmadness]

They also say that Shanghai and the 8-core Montreal will still be aimed on the same sockets of Barcelona.
http://www.theregister.co.uk/2007/03/17/amd_rivas_barcelona/

After reading the article, I think your title is misleading. :?

 

[m25]

That was just one aspect of the article, however, you're the second who thinks this, so I just put the article's title.

 

[m25]

:?:

 

[BaronMatrix]

bit of a misleading title, it sounded as if they were saying it was better performance or other technical areas. instead they were basically saying it was more mpressive to the sheep of the world who will blindly believe 4 is better than 2.

it is a shame companies have to worry about marketing and not actual technical merit if you will.

this post is something i would expect form lordpope or BM, not you.

I don't think you know me so don't assume I would do something similar to ANYONE here, you follower.

 

[BaronMatrix]

They also say that Shanghai and the 8-core Montreal will still be aimed on the same sockets of Barcelona.
http://www.theregister.co.uk/2007/03/17/amd_rivas_barcelona/

From the article:

Sources tell us that the chip is cranking through software at an unreal clip in the labs, beating out AMD’s oft-cited 40 per cent performance gain figure.

 

[jamiepotter]

I'm confused. Has the title been changed?

I got the feeling in any case that they specifically _weren't_ saying it was a better idea to just stick two dual-cores together and call it 'quad', and that they were almost accusing Intel of cheating. (A bit like when you mix malt whiskies from different breweries and call it a 'pure malt': i.e. false advertising, basically).

 

[jamiepotter]

I guess they're making a wild inductive leap from your past forum behaviour. People are so unfair, aren't they, Baron?

 

[wickedmonster]

What Intel did by slapping two dual cores together is a good thing for a lot of IT companies. Remember, enterprise software charge per socket so if you can fit 4 cores in one socket instead of two, you effectively reduced your software cost in half. And enterprise software is alot more expensive than any piece of hardware. Intel should be applauded for bringing quad core so fast to market.

 

[BaronMatrix]

I guess they're making a wild inductive leap from your past forum behaviour. People are so unfair, aren't they, Baron?

You mean posting AMD news? Or slapping you for having Borat as your avatar?

 

[BaronMatrix]

bit of a misleading title, it sounded as if they were saying it was better performance or other technical areas. instead they were basically saying it was more mpressive to the sheep of the world who will blindly believe 4 is better than 2.

it is a shame companies have to worry about marketing and not actual technical merit if you will.

this post is something i would expect form lordpope or BM, not you.

I don't think you know me so don't assume I would do something similar to ANYONE here, you follower.

Baron, you do this routinely. It isn't that he knows you personally, he knows your posting habits, patterns, and methods. This is the reputation you have built for yourself, it has nothing to do about 'knowing you' but has everything to about knowing you.

Sources tell us that the chip is cranking through software at an unreal clip in the labs, beating out AMD’s oft-cited 40 per cent performance gain figure.

Reply to this Brood Queen.

 

[jamiepotter]

I didn't know you had. Does it upset you terribly?

 

[BaronMatrix]

are you saying you have not created threads with a title that is based on a sentence of an article that you have twisted for your own purposes?

you cannot seriously say this is not your style of thread.

That's exactly what I'm saying. There was a Larrabee article that was straddling, but I take usually the exact title and then my commentary may be speculative or theoretical.

 

[BaronMatrix]

I didn't know you had. Does it upset you terribly?

From the article:

Sources tell us that the chip is cranking through software at an unreal clip in the labs, beating out AMD’s oft-cited 40 per cent performance gain figure.

That statement shou dbe the entire discussion fo the rest of this thread. It was said that Henri's numbers were "very conservative."

I guess a sideline could be how AMD could have done an MCM, but they have to cognizant of resource limitations.

Continue. I'm going to the gym.

 

[jamiepotter]

Continue. I'm going to the gym.

It's good to know that at least your body is getting a workout.

Look, we've heard all this speculative performance shite before. It's not hard to pluck a percentile out of the air. It's not discussion-worthy. Let's wait for the benchies.

What is discussion-worthy is an AMD exec thinking that they dropped the ball in not going for an MCM chip.

 

[MU_Engineer]

AMD could have done an MCM like Intel did, but there are several reasons whey they did not:

1. AMD didn't get to 65nm until recently, whereas Intel had a while ago. A 90nm quad-core would have been very impractical and ill-advised but a 65nm unit is doable.

2. AMD would have had to put two chips in a package and only one would have had an IMC if it were to be compatible with current sockets and motherboards. This would require the second die to only get data over NUMA, which would lead to poor performance.

3. AMD could have made an MCM that did not cut performance by putting two dies, each with an IMC and connecting the two by on-chip NUMA, making an Opteron-22xx-on-a-chip setup, so to speak. This would be fast but would require a new socket or at least a new pinout to handle the pins for four channels of RAM.

4. The one IMC, two dies approach could be done easily enough with using two Opteron 22xx/82xx dies and current hardware, but the two IMCs, two dies would require a new socket and the effort to get it all right with the channel. This would result in quite a bit of effort for AMD at a time when they are working on getting their new K10 chips made. This is why the 65nm chips sold to date are straight die shrinks- no extra effort beyond making new masks was needed.

I personally think that a two-IMC, two-die MCM would do well for AMD as long as they have good planning. Yes, it will take a new socket and new motherboards (with 2 RAM banks) but once the MCMs become monolithic dies, the sockets and boards can still be used. The second IMC's pins and RAM banks can just be left unused until future MCMs come along and need them. This kind of a setup can also be used to feed a ton of RAM bandwidth into a monolithic die by having two 128-bit or four 64-bit IMCs instead of one 128-bit or two 64-bit IMCs. That would allow many more cores to be placed in one socket and push the RAM bandwidth wall back a bit. One disadvantage to this setup would be the fact that you'd have to buy 4 RAM modules at a time and that it would not be ideal for servers as there would be a *ton* of RAM slots on the board.

 

[sonoran63]

Customers don't care if there's one die, two dies, or a herd of hamsters inside the chip - they only care about how much it costs and how it performs. And since this is business, not some teenage bragging contest, sales are all that count.

The FAR more interesting comment in this article is the part about them expecting to use the integrated graphics circuitry as an adjunct to the CPU's math processing capabilities.

 

[Major_Spittle]

I must have missed something. The Threads name came from the article and I saw nothing wrong with his post.

“If I could do something different, I wish we would have immediately done a MCM - two dual cores and call it a quad-core,” said Mario Rivas, an EVP at AMD, during a recent interview in Austin, “because, I guess, the market sucks it up.”

All that matters is if it is in demand and if it will sell better than the competition, thus the title of the article. Obviously 4 is better than two for servers and for creating a upgrade path for people like me who couldn't afford a Quad, but will this summer and can just plug it in.

I don't know why people are getting on M25???

 

[Major_Spittle]

Yeah, the title was from the article and it follows the quote in the article I posted.

You sound a little defensive about native being better when it is not in more ways than just marketing.

It is better for yeilds, less product lines to ramp, higher bin yeilds, quicker to ramp to dual/Quad/Octa..., provides cheaper upgrade path to consumers, less of a low speed ram penalty, low cost production, low cost performance increase in product line.....

I hear you talking, but I disagree with your motives to flame. Especially envoking BM factor in a flame. :lol:

 

[Major_Spittle]

gay or straight?

 

[Major_Spittle]

gay or straight?

:lol: :lol: make sure the knot will slip easilly enough. :lol:

I can only assume you are referring to one of your homosexual practices.

Maybe you should start a CPU thread about that, it would make as much sense as you last two.

 

[Major_Spittle]

you take that in college or work on it your whole pathetic life?

Actually I was getting a college education while you were " getting punched in your head by your vietnam vet. father " .

I guess we can see which is the best path to enlightenment.

 

[m25]

his original title was "AMD admits Intel's non-native Quad is a better idea"
it is misleading as they were only really referring to it in marketing terms, not as was implied by his title a better solution than a native one.
people get on to others like lord pope for using misleading titles and what not so why not get on to other people for the same.

Holy sh!t; are you still arguing about this?! Do you think I put there a 'misleading' title because I wanted people to get on me for that?! I simply put that title because I wanted to highlight that particular paragraph in the article and after people read the article, they understand it was from a marketing standpoint. I did it because for the first time AMD was accepting the real advantage of a non-native multi core and that's all. I am sorry for someone who might have misunderstood but I didn't make any comment on the article so I didn't distort anything.