INTEL MATH: 2+2=QUAD-CORE PROCESSOR

[9-inch]

Although Intel executives at IDF did not confirm plans to eventually introduce a monolithic quad-core processor, Brookwood says he believes Intel will make the move when it transitions from its current 65 nanometer manufacturing process technology to the 45 nanometer technology. The smaller geometry manufacturing process will let Intel put the additional transistors of the quad-core structure in a smaller physical size than would be possible with a 65 nanometer process.

This totally proves what I've said about Intel needing 45nm for their quad core processors. AMD will deliver their quad core solutions on 65nm and still run cooler than Intels sh!tty process.
As I said before, Intel is good ar shrinking die sizes but they suck at process technology. 8)

http://www.informationweek.com/news/showArticle.jhtml?articleID=181502198

 

[dvdpiddy]

dude you lost me there for a second but yeah amd could deliver quad core on 90 nm if they want but it would have alot less cache

 

[ltcommander_data]

Yeah Intel either is unable or unwilling (cost issues) to make a quad core on 65nm.

However, did you notice the critical sentence from that article?

Brent Kerby, product marketing manager for AMD, says the company will again use a monolithic design for its first quad-core offerings, scheduled for the second half of 2007.

Hmm, first we hear beginning of 2007, then that their aiming even higher for late 2006, and now amazingly they admit that they won't have anything until H2 2007. You can complain that Intel's may not be the best quad core solution, but at least they have one that looks to have a 6 month lead on AMD.

It's interesting that you mention that Intel is good at shrinking die sizes, because AMD's ability to do so is very vague at this point.

http://www.theinquirer.net/?article=30183

Serious questions have to be asked about whether AMD really has got problems with its move to 65 nanometres – perhaps relying on IBM knowhow has its limitations because the whisper is that the delays are meaningful. We would have asked AMD about this, but our meeting with Chimpzilla got canned, and that happened to other individuals here too.

AMD had originally aimed for late 2006, but more recently early 2007 has seemed more likely. Sadly, The Inquirer's meeting with AMD was cancelled before they could elaborate on the situation. Problems with 65nm may explain why AMD's quad cores are delayed to H2 2007. I seriously doubt AMD could make a quad core solution on 90nm, especially if they want to be competitive power-wise.

And then there’s AMD. The firm acquired a reputation over the years for having an unerring aim when it came to firing at its own foot. For the last three to four years AMD has done well. But if it doesn’t start making some meaningful noises over its own 65 nanometre technology, it could well find itself in some trouble.

Between the continuted memory controller problems with AM2, their silence on their supposedly superior SOI 65nm process, and Intel's new architecture gaining momentum, AMD's position seems increasingly precarious. I'm really hoping that they are going to make some major announcements at CeBit to clear things up.

 

[Action_Man]

Caps lock button still jammed in troll?

 

[luminaris]

LOL

Can't help but laugh man. We're here watching ltcommander_data with his incredible knowledge and yet, your here to throw in the ole, what's up with the caps lock thing. Your a trip man! :lol:

 

[Action_Man]

I'm over arguing with the trolls and fanboys for the moment. Its easier just to ridicule them.

 

[dvdpiddy]

exactly like ycon being born from sewage

 

[Action_Man]

Woot!

 

[dvdpiddy]

excuse me but the only one thats gay here is ycon when he gets a boner for intel propaganda

 

[Action_Man]

Ooooooooh, I thought you meant I was born from sewage like ycon, sorry my mistake!

 

[dvdpiddy]

good then sorry for insulting you on my previus post i'll erase it now

 

[Heyyou27]

dude you lost me there for a second but yeah amd could deliver quad core on 90 nm if they want but it would have alot less cache

In your dreams.

 

[dvdpiddy]

its possible if intel did dual core on 90 nano then i dont see why amd cant do quad core on 90 nano

 

[Action_Man]

If they did quad core on 90nm it'd be huge, like 300mm2 with 512k L2 cache per core.

 

[qurious69ss]

Let's wait and see what the next generation does am2/conroe, before we go into the whole quad core thing. Things change alot in a year, especially since there is all this talk about a chip recession, oh well that's neither here nor their.

 

[dvdpiddy]

If they did quad core on 90nm it'd be huge, like 300mm2 with 512k L2 cache per core.

exactly thats what i said and action if i was born from sewage like you i would'nt be posting here now would i?

 

[Action_Man]

*confused*

 

[dvdpiddy]

*more confused*

 

[9-inch]

Problems with 65nm may explain why AMD's quad cores are delayed to H2 2007. I seriously doubt AMD could make a quad core solution on 90nm, especially if they want to be competitive power-wise.

Are you sure about this?? 8)

Here's a link I want you to read:
News.com

...And here's a quote:

"Right now, we use a 90-nanometer manufacturing process for the chip. We could lower that to 60-nanometer, 45-nanometer or 30-nanometer," said Bacher, adding that IBM could also increase the amount of memory within each blade.

And they're using the same process AMD will use for their upcoming CPUs. This means that AMD is still on schedule with their 65nm process and Socket F will be the first one to use it, the rest of AMD's CPUs will be manufacured on 90nm but with the new FD-SOI.

And if you still don't believe about the benefits of AMD's new process, here's a link that will convince you even more:
http://nanotechwire.com/news.asp?nid=2659

The companies announced that they have successfully combined embedded Silicon Germanium (e-SiGe) with Dual Stress Liner (DSL) and Stress Memorization technology (SMT) on Silicon-On-Insulator (SOI) wafers, resulting in a 40 percent increase in transistor performance compared to similar chips produced without stress technology, while controlling power consumption and heat dissipation. The new process technologies reduce interconnect delay through the use of lower dielectric constant (lower-K) insulators, which can improve overall product performance and lower power consumption. In addition, the new technologies have shown ability to be manufactured at the 65nm generation and scaleable for use in future generations.

Maybe, AM2 could beat Conroe with this 40% improvement just from the process itself.

 

[ltcommander_data]

The thing with the news.com article is that it is about the cell processor. Personally, I don't no much about the Cell so I have no idea how representative the fact that the 90nm process works on the cell is to having the 90nm process making quad core general purpose CPUs.

The article itself saids:

It's also not clear exactly what the performance increase will be compared to existing blade servers.

While the Cell may be called 9-cored in the article, my understanding of the cell architecture is that only 1 of those is a full CPU, while the others are just specialized SIMD modules. Sounds kind of like a bunch of glorified vector units, but they are no doubt great in multimedia applications. Somehow I doubt IBM could release their cell processors on 30nm right now at the drop of a hat.

It's not that I doubt AMD can create a quad core using a 90nm process, it's just whether the end result is worth it. Die size and yields are one issue, but heat is another. You may point to improvements from FD-SOI, but the dual core FX-62 already needed the TDP increased to 125W. It doesn't seem realistic that AMD could double the cores of the 2.8GHz FX-62 and still remain within that TDP.

A 40% increase in transistor performance always appears good on paper, but it's always hard to maintain in the real world when many other design factors and interactions come into play. The funny thing is, while AMD may be playing up their new addition of e-SiGe to SOI, their real process transition hasn't even caught up with their previous process refinements.

When Strained Silicon on Insulator (SSOI) was announced it was supposed to pave the way for higher speed processors. However, in truth the introduction of SSOI chips has been a lot slower than what you might think.

http://www.theinquirer.net/?article=30154

However, not all of the processors are manufactured using this new improvement. There is still plenty of classic 90nm SOI being manufactured, with a conservative introduction of new parts.

It's hard to tell what AMD's focus is. You'd think they would use the SSOI process on their flagship product the FX-60 to allow better overclocking to justify the higher price they are selling or at least ensure higher yields of this higher clocked processor, but in fact the FX-60 is still using the old SOI process. One way of looking at it would be that that means the FX-60 probably has more room to grow if SSOI was ever used, but then the question is why wasn't it used on their flagship product in the first place?

You can also argue that the FX-60 wasn't manufactured in SSOI process, since overclockability of the CPU wasn't as stellar as its predecessors.

It also begs into question whether the FX-62 is also using the older SOI process if it's TDP had to increase to 125W. If AMD hasn't even transitioned their products to the SSOI process, it just makes me wonder how long the transition to FD-SOI will take.

 

[Action_Man]

Maybe, AM2 could beat Conroe with this 40% improvement just from the process itself.

Intel can't improve their process at all?

 

[Sarcastic]

Yeah Intel either is unable or unwilling (cost issues) to make a quad core on 65nm.

Don't forget that both AMD and Intel are in business to make money. There is still alot of money in duel core they will not make it obsolete before making a chunk of change.

 

[BaronMatrix]

The thing with the news.com article is that it is about the cell processor. Personally, I don't no much about the Cell so I have no idea how representative the fact that the 90nm process works on the cell is to having the 90nm process making quad core general purpose CPUs.

The article itself saids:

It's also not clear exactly what the performance increase will be compared to existing blade servers.

While the Cell may be called 9-cored in the article, my understanding of the cell architecture is that only 1 of those is a full CPU, while the others are just specialized SIMD modules. Sounds kind of like a bunch of glorified vector units, but they are no doubt great in multimedia applications. Somehow I doubt IBM could release their cell processors on 30nm right now at the drop of a hat.

It's not that I doubt AMD can create a quad core using a 90nm process, it's just whether the end result is worth it. Die size and yields are one issue, but heat is another. You may point to improvements from FD-SOI, but the dual core FX-62 already needed the TDP increased to 125W. It doesn't seem realistic that AMD could double the cores of the 2.8GHz FX-62 and still remain within that TDP.

A 40% increase in transistor performance always appears good on paper, but it's always hard to maintain in the real world when many other design factors and interactions come into play. The funny thing is, while AMD may be playing up their new addition of e-SiGe to SOI, their real process transition hasn't even caught up with their previous process refinements.

When Strained Silicon on Insulator (SSOI) was announced it was supposed to pave the way for higher speed processors. However, in truth the introduction of SSOI chips has been a lot slower than what you might think.

http://www.theinquirer.net/?article=30154

However, not all of the processors are manufactured using this new improvement. There is still plenty of classic 90nm SOI being manufactured, with a conservative introduction of new parts.

It's hard to tell what AMD's focus is. You'd think they would use the SSOI process on their flagship product the FX-60 to allow better overclocking to justify the higher price they are selling or at least ensure higher yields of this higher clocked processor, but in fact the FX-60 is still using the old SOI process. One way of looking at it would be that that means the FX-60 probably has more room to grow if SSOI was ever used, but then the question is why wasn't it used on their flagship product in the first place?

You can also argue that the FX-60 wasn't manufactured in SSOI process, since overclockability of the CPU wasn't as stellar as its predecessors.

It also begs into question whether the FX-62 is also using the older SOI process if it's TDP had to increase to 125W. If AMD hasn't even transitioned their products to the SSOI process, it just makes me wonder how long the transition to FD-SOI will take.



I could agree except that the FX62 doesn't exist. I am not going to search the net for stories regarding it but the inquirer stated that the process for AM2FX62 would include improved leakage current and gates, which implies SSOI.

And there were also stories that AMD could go 65nm this year but don't think they will need to. Conroe may change that but I still see 20-30% improvement when all is said and done. If I'm right about AMDs plans they will get 3.2GHz out of the AM2 @ 90nm. That should pretty much even the playing field and may give them an advantage, especially in the server space where Socket F is supposed to double the FP registers. Does anyone remember what happened when Intel added 8MB L3 to Xeon?

I think AMD is going to add L3 for all FX and Opteron chips using the ZRAM tech and that should enable them to beat WoodCrestConroe to death. I mean if you look at the benchmarks with Apache, Intel will definitely need 100% improvement to catch (just to catch) a 4Way Opteron. They OWN TPC-C up to 1TB databases.

 

[apache_lives]

Although Intel executives at IDF did not confirm plans to eventually introduce a monolithic quad-core processor, Brookwood says he believes Intel will make the move when it transitions from its current 65 nanometer manufacturing process technology to the 45 nanometer technology. The smaller geometry manufacturing process will let Intel put the additional transistors of the quad-core structure in a smaller physical size than would be possible with a 65 nanometer process.

This totally proves what I've said about Intel needing 45nm for their quad core processors. AMD will deliver their quad core solutions on 65nm and still run cooler than Intels sh!tty process.
As I said before, Intel is good ar shrinking die sizes but they suck at process technology. 8)

http://www.informationweek.com/news/showArticle.jhtml?articleID=181502198

conroe @ 45nm quad core = conroe now, twice the cores, equal heat/power, cheaper production.

Intel does seperate dies so they can pick 2 high speed dies and put them together rather then in one package having one fast core and the other(s) slow, and quad core - if it was one die thats 4 times the "risk" of making a slower cpu.

When AMD gets 65nm it will get colder, but then Intels 45nm will get even colder, faster and cheaper, works both ways.