AMD ON TRACK TO RECEIVE FIRST CHIPS FROM CHARTERED

[spud]

The fact of the matter is if you take a 3.6 GHz P4 and underclock it to 2.4 GHz it will run at 20-30 Watts, while the Athlon at 2.4 GHz still dissipates 80-90 watts.

Not quite the Word I was looking for. The prescott 2.4 was only able to get down to ~70 watts because it used a 133 fsb.
Since watts is what generates the heat, and watts is mostly a measure of work, and losses, @ 2.4 ghz that Intel is very inefficient.
I doubt anyone here has a real grasp of Amd's process technology, but you have to admit that it's output has made Intel's best efforts, over the last 4 years look rather sad.
I'm pretty sure Amd has a better idea of what they are doing than we do.

So what makes you grasp the process while the rest of us lesser minds can't grasp?

 

[luminaris]

JumpingJack wrote.. Actually, not a bad point....Chartered is a singapore company, I would not feel comfortable supporting them for various reasons. Why? Well, something similar to why I would not support Nike by buying their products.

trust me... i work there before... chartered really sucks... their yield sucks...
do u ever own the 1st generation colour phone from sony ericsson? the phone
got a lot of problem.. the chip inside was made by chartered.

Xbox 360 also uses chartered chips .... hmmmm heat problems.... keep them well ventilated and you should have not problems.... also recall the initial press release from MS, beside other HW shortages, the CPU yields were lower than had been hoped for.... hmmmmm.

Heat problems???? Oh Geez, so that means we could potentially see 'heat' problems kind of like what Intels experience somewhat now?
No it is a different beast all together, to keep costs down, the PPC in the xBox does not have all the nifty whistles and bells of the process I suspect. I just have not been impressed with chartered quality... the process is not great and their yields are always depressed, or at least from what one can infer from various statements over the years.

Yeah, I've read about it and alot of people have raised an eyebrow over the fact that AMD is partnered with Charter. I guess we won't know till we see em.

 

[9-inch]

Nice job.... here is the short and skinny of it.... 9-inch, MMM, and all others take this info to make their claim:

An Athlon running at 2.4 GHz runs cooler and out performs a 90 nm Intel processor running at 3.6 GHz. Thus, they conclude that AMD must have a better 90 nm process ....

The correct argment is to realize that

Performance (observed) = speed (GHz) x IRC (Instructions retired per clock)

Intel, by using netburst, focused on Speed. AMD, because they could not squeeze more from the process focused on IRC, hence, the reality is AMD has a better architecture with K8 but not necessarily a better process, even with their beloved SOI

The fact of the matter is if you take a 3.6 GHz P4 and underclock it to 2.4 GHz it will run at 20-30 Watts, while the Athlon at 2.4 GHz still dissipates 80-90 watts. You see this in the overclockability too...you also see their process short comings in the "Cold Bug". It is ludicrious the way thay make their arguments.

Did you forgot that Athlons were designed to execute more instructions per clock than intel's shitty architecture??
A 2.4GHz P4 is almost equal to an 1.6GHz Athlon and still the Athlon at that speed consumes less power.

 

[9-inch]

I wonder what Keyboad_Man is doing. He is late with giving the best advice for 9-inch, that 9-inch should accept once and finally before he posts another garbage from the trash.

So then, why are you posting in my thread? :roll:

 

[Legenic]

Nice job.... here is the short and skinny of it.... 9-inch, MMM, and all others take this info to make their claim:

An Athlon running at 2.4 GHz runs cooler and out performs a 90 nm Intel processor running at 3.6 GHz. Thus, they conclude that AMD must have a better 90 nm process ....

The correct argment is to realize that

Performance (observed) = speed (GHz) x IRC (Instructions retired per clock)

Intel, by using netburst, focused on Speed. AMD, because they could not squeeze more from the process focused on IRC, hence, the reality is AMD has a better architecture with K8 but not necessarily a better process, even with their beloved SOI

The fact of the matter is if you take a 3.6 GHz P4 and underclock it to 2.4 GHz it will run at 20-30 Watts, while the Athlon at 2.4 GHz still dissipates 80-90 watts. You see this in the overclockability too...you also see their process short comings in the "Cold Bug". It is ludicrious the way thay make their arguments.

Did you forgot that Athlons were designed to execute more instructions per clock than intel's shitty architecture??
A 2.4GHz P4 is almost equal to an 1.6GHz Athlon and still the Athlon at that speed consumes less power.

he's not talking about performance, he's talking about clock speed and pipeline length vs. heat. you don't even understand your own thread.

 

[9-inch]

he's not talking about performance, he's talking about clock speed and pipeline length vs. heat. you don't even understand your own thread.

Sorry to tell you this but please read again the post before saying such non sense.

 

[sonoran63]

For those less physics inclined the Ioff (sometimes called Idsat) the higher the number the faster the transistor can switch on and off, those numbers above in a nut shell explain why Intel can go to > 3 GHz and can overclock to > 5 GHz in some cases, where as AMD's 90 nm process has barely been able to break 3 GHz (with a single core Opty).

I wonder how much of that is chip design vs how much is due to SOI? Are you familiar with any stats on fastest obtained switching speeds for transistors on SOI vs non-SOI strained silicon? With Conroe moving back to a high-IPC design more like A64, one wonders if the process technologies chosen will have a larger effect on the highest obtainable clock speeds? A major difference might be enough to give one technology or the other a sizeable advantage.

 

[Legenic]

he's not talking about performance, he's talking about clock speed and pipeline length vs. heat. you don't even understand your own thread.

Sorry to tell you this but please read again the post before saying such non sense.

wow. you're just disregarding anything that makes AMD non-godlike.

jumpingjack just backed up what I said with data. read his post.

we're talking about process tech and how it reflects against different architectures regarding heat, etc. we're not talking about performance at certain clock speeds.

 

[dvdpiddy]

cool thanks for the link 9-inch

 

[9-inch]

My point: If you simply ignore just performance and compare power as related to the capability of the process, then you need to clock both processors the same, set Vcore the same, and then compare TDP. Then, and only then, can you say which process is better in delivering lower dissipated power metrics (normalized to total transistor count of course).

I wouldn't in life use your metric for evaluating X process vs Y process just the way you can't compare grapes to watermelons.

In my case, I'd rather use performance/watt to see which process is the best. Since intel's ngma architecture is somewhat comparable to K8 (both focus on more IPC), then I'd rather compare both of them in term of process efficiency at the same clock speed. I can bet you a million dollars and a kidney that intel's process lack AMD's efficiency.

I said it before and I would say it again:
intel is the best one to deliver at process shrinking (I admit this one with lots of pain in my @ss), but they suck at process efficiency. Once AMD fully utilizes ssgoi in 65nm, intel better have ready their 45nm process or they'll suffer the consequences just the way they did in the 90nm era.

 

[9-inch]

[/quote]

However, take the architecture out of the equation and compare processing tech to each other and Intel is superior.

Do you live in a delusional world or you just CAN'T accept facts?
Intel's process stink and they'll continue so because they won't admit that DSL-SOI (AMD's current process) is way better than theirs. Even intel admited that they won't use SOI since long time ago because that would make them look like a follower instead of an innovator.
If intel did ever wanted to lead AMD in all forms, they would have used SOI in their chips lon time ago and incorporated HTT on their ngma architecture instead of creating a useless technology like CSI. Well, that's another story I won't talk about but I hope you get the message.

 

[9-inch]

On this point.... there is data on the net that proves Intel's leadership position on process technology, I will leave ito you to research, I have posted the data in various places thoughout the forum

Just he way there's data on the net that proves that AMD's and IBM process is better than Intel's. I've posted links about SOI and DSL-SOI process on this forum. I'll leave it to you to search on your own.

 

[brainysmurf]

I think Chartered is supposed to concentrate on lower-speed, entry-level parts, like Sempron and the cheaper Athlon 64s. The FX series and Opterons will be produced by AMD.
Don't hold me to that, I seem to recall having read that somewhere a few weeks ago but I can't recall the source.

 

[KingLoftusXII]

My H2

An H2? I get the feeling someone has a small pee-pee.

 

[spud]

However, take the architecture out of the equation and compare processing tech to each other and Intel is superior.

Do you live in a delusional world or you just CAN'T accept facts?
Intel's process stink and they'll continue so because they won't admit that DSL-SOI (AMD's current process) is way better than theirs. Even intel admited that they won't use SOI since long time ago because that would make them look like a follower instead of an innovator.
If intel did ever wanted to lead AMD in all forms, they would have used SOI in their chips lon time ago and incorporated HTT on their ngma architecture instead of creating a useless technology like CSI. Well, that's another story I won't talk about but I hope you get the message. [/quote]

Actually they said it was a cost issue.

 

[Caboose-1]

I still can't believe that the first chips will be based on 90nm and not 65nm. AMD is still far behind in the technology race against Intel.

What are you talking about??
AMD's process is A LOT better than Intels. Even you are witness that 90nm A64s perform better and consume less power than intel's 65nm offerings.

Let's just give them time for them to mature the process. I believe once we see 65nm offerings from AMD they will perform 40% better than current 90nm processors. I do believe intel is scared to sh!t about AMD's new and improved ssgoi process, that's why they're shouting out loud about their 45nm experiments.

Interesting, care to back that up with some data ?

Word.
Adverb. Hydroxypropyltrimonium Chloride.

 

[dvdpiddy]

I still can't believe that the first chips will be based on 90nm and not 65nm. AMD is still far behind in the technology race against Intel.

What are you talking about??
AMD's process is A LOT better than Intels. Even you are witness that 90nm A64s perform better and consume less power than intel's 65nm offerings.

Let's just give them time for them to mature the process. I believe once we see 65nm offerings from AMD they will perform 40% better than current 90nm processors. I do believe intel is scared to sh!t about AMD's new and improved ssgoi process, that's why they're shouting out loud about their 45nm experiments.

Interesting, care to back that up with some data ?

Word.
Adverb. Hydroxypropyltrimonium Chloride.

ethylene diamine tetraacetic acid

Since you are in to shampoo's, here is a metallic chelating agent for ya' pronoun

 

[joefriday]

EDTA is the coolest! gotta love 1:1 chelating agents!

 

[stOrmy]

I still can't believe that the first chips will be based on 90nm and not 65nm. AMD is still far behind in the technology race against Intel.

What are you talking about??
AMD's process is A LOT better than Intels. Even you are witness that 90nm A64s perform better and consume less power than intel's 65nm offerings.

Let's just give them time for them to mature the process. I believe once we see 65nm offerings from AMD they will perform 40% better than current 90nm processors. I do believe intel is scared to sh!t about AMD's new and improved ssgoi process, that's why they're shouting out loud about their 45nm experiments.

intel comes out with data, you say it's bs and disregard it, "until you see real proof". AMD comes out with similar information, and you praise it as being fact despite there being no hard data to back it up.

40% better transistor performance won't mean 40% increase in performance. it means that with this process AMD will be able to reach higher clock speeds, upwards of (from what I've read, if I recall correctly) 3.4ghz or similar.

AMD's process performance isn't far and away better than intels. they may be able to create similar numbers in terms of power usage and temeratures, but that's very much because of the efficient architecture they're using. you can't compare say a p4 on 90nm to an A64 on 90nm and say AMD is kicking intels anus in process performance. the two architectures are vastly different. long pipelines and high clockspeed means high temperatures.

a plus side for intel on manufacturing is that they can produce chips at a much lower cost than AMD. this means that in any sort of pricing war, intel will have an advantage.

another point: there are TDP numbers floating around for the turion X2 processors that show them as 35w on AMD's 90nm process, which is great. but the numbers also list future 65nm turions as being in the same power envelope; 35w.

and people will see your thread if it's not all in capitals you know.

Nice job.... here is the short and skinny of it.... 9-inch, MMM, and all others take this info to make their claim:

An Athlon running at 2.4 GHz runs cooler and out performs a 90 nm Intel processor running at 3.6 GHz. Thus, they conclude that AMD must have a better 90 nm process ....

The correct argment is to realize that

Performance (observed) = speed (GHz) x IRC (Instructions retired per clock)

Intel, by using netburst, focused on Speed. AMD, because they could not squeeze more from the process focused on IRC, hence, the reality is AMD has a better architecture with K8 but not necessarily a better process, even with their beloved SOI

The fact of the matter is if you take a 3.6 GHz P4 and underclock it to 2.4 GHz it will run at 20-30 Watts, while the Athlon at 2.4 GHz still dissipates 80-90 watts. You see this in the overclockability too...you also see their process short comings in the "Cold Bug". It is ludicrious the way thay make their arguments.

EDIT: Endyen challenged me on the numbers above, and so I need to disclaim 20-30 watts is a rough guess, I do not have data to support that, my point was to emphasize that it would be very low. EDIT2: Wow, I was way off...I just underclocked my 3.6 GHz prescott (idles at 47 deg C, load is 56 de C), to 2.8 GHz and tweaked down the Vcore, based on Power=CV^2*F, I would estimate roughly 65 watts of TDP, and a temperature (heat dissipation) would put my temperature at ambient at about 38 deg C, I am measuring 41 Deg C. So a 2.4 GHz would do something along the order of 50-60 Watts. Thanks Endyen for ensuring accuracy.


Exactly! It is good to see an excellent post that brings good information to the forums instead of half truths.

 

[Jackalope73]

Not to get into the middle of a pissing match, just get hosed from every direction, but I believe the major advantage to moving to a more compact architecture is more beneficial in terms of processor yields regardless of performance it’s price per unit. Intel is making more money per CPU currently. That is were the superiority of process size matters to Intel and AMD's stock holders. Maybe one day everyone will let IBM just build there CPUs. Kilocore cell anyone, Nanotubes, Quantum, or Photonic CPUs? At Crazy IBM’s basement we’ve got out of this world alien technology at prices that are INSANE! What will IBM do when the Roswell Aliens in their basement die? Hey Joefriday I own a 1987.5 Mustang gt and a 1989 Bronco with a 351, all with fuel injection I guess that makes me an expert, an expert on gas prices! :lol: Chill guys, we don’t need anymore heart attacks in IT! :tongue:

 

[joefriday]

Glad to see that my post did not fall entirely on deaf ears. I own an 87 Nebraska State Patrol Mustang and an 88 Thunderbird Turbo Coupe clone with an 88 Mustang GT 5.0/5 speed swap. The contents of that post are actually 100% factual. Good stuff to know if you ever plan to hop up an 86-88 V8 T-bird/Cougar or 86-89 Towncar/Grand Marquis/Crown Vic.

to JumpingJack: Unfortunately I was forced to Google the Tris-bipyridine ruthium(II) chloride complex. My BS double major in Chemistry can only get me so far! :lol: My wife is a QA Chemist at Cargill, while I'm finishing my first year of med school. I can't imagine going double PhDs, the amount lab work would drive me insane! I'd never want to be a research chemist...not much of a lab rat. From what I've found, that complex looks like it has the ability to be either an electron donor or electron acceptor. Do I win cookie?

 

[tamalsmith]

They are behind and still can outperform. 70% of the market proves that.


Looks like a good start from AMD. Let's hope they can ramp 65nm faster with the help of Chartered.

http://www.infoworld.com/article/06/04/14/77428_HNamdchartered_1.html

I still can't believe that the first chips will be based on 90nm and not 65nm. AMD is still far behind in the technology race against Intel.

The first AMD chips that come out of Chartered's factories will be made using a 90-nanometer manufacturing process, not a 65nm process as some reports have claimed, he said.

Over time, AMD and Chartered eventually plan to shift production to a 65nm process, he said.

LOLS! intel and microsoft were a duo-poly, did you know?

microsoft put more useless crap in their OS, and intel sold exclusive cpus for microsoft's products!

thats why they got that much %.

egads, these guys with no brains...
and yes, this treat is funny " my penis is bigger than yours.." kind of shit is hilarius.

*edit*
btw there are ppl spatting shit about cars in a post about AMD receiving chips from chartered company?

 

[spud]

However, take the architecture out of the equation and compare processing tech to each other and Intel is superior.

Do you live in a delusional world or you just CAN'T accept facts?
Intel's process stink and they'll continue so because they won't admit that DSL-SOI (AMD's current process) is way better than theirs. Even intel admited that they won't use SOI since long time ago because that would make them look like a follower instead of an innovator.
If intel did ever wanted to lead AMD in all forms, they would have used SOI in their chips lon time ago and incorporated HTT on their ngma architecture instead of creating a useless technology like CSI. Well, that's another story I won't talk about but I hope you get the message.
Actually they said it was a cost issue.

Dude, I fixed your quote to make it look "correct" some funny bug's in the formz.

Just for the record: Here is a PDF link on page 49, figure 22 showing a TEM of an Intel constructed SOI transistor, along with IV curves showing the improvement over bulk. Intel publically stated that the technology was feasible, but declined to implement for cost reasons. The paper I reference is actually being used in an Electrical Engineering course at Standord.... the mecca of churning out engineering talent for the semiconductor industry.

Intel Stress/SOI technical paper

I should also point out that the curve on page 52, figure 30 (the Ion/Ioff cuvre) labeled "DST with raised S/D" is exactly the embedded SiGe in the IBM/AMDs announced "40%" breakthrough. Note that this paper was dated May 16 2002. In otherwords, Intel MADE what IBM/AMD hopes to make back in 2002 Bwhaaahhhaaaaahaaaa

Man, am I gonna have fun with 9-Inch..... I will spread this out over 6-8 weeks and make it educational to boot ....

Word.

 

[joset]

(...) I have been digging around the past few months, collecting references as soon as I can logically order them into something coherent I will post what I have found. (...)

Glad I'm not too late!


Cheers!

 

[rettihSlluB]

Quite frankly, the transistor parametrics for the bulk CMOS process from Intel exceeds the transistor parametrics for the SOI process from IBM/AMD at the same device node (i.e. 90 nm, I have linked this data but it is evident that there is a rudimentary lack of understanding). This is going to be difficult to explain in detail, but I will soon -- trying to make it as educational as I can.

Bullshit. :roll:

Please, read the following:

AMD and IBM also said they were implementing lower dielectric constant (low-k) insulators into the 65nm process to "reduce interconnect delay" and "improve overall product performance and lower power consumption".

http://www.ibmtechnology.com/modules/news/article.php?storyid=473

In the end of 2004 AMD and IBM announced another technological breakthrough in the transistor performance field. The technology jointly developed by the engineers from both companies called Dual Stress Liner allowed them to improve the response time of semiconductor transistors by 24%. This report provides decision makers with the information to make a solid business assesment about AMD's competitive position in conjunction with the Chipworks report SAR-0410-004, AMD 2700BKX4LB Athlon Microprocessor Structural Analysis and SAR-0507-802, AMD ADA3500DAA4BP & 2700BKX4LB Transistor Characterization Report.

http://www.chipworks.com/WebReports/ShowOverview.asp?ReportCode=SAR-0507-801

At the IEEE Solid-State Circuits Conference (ISSCC) in February 2004, IBM announced the migration of their 130nm PowerPC 970 to 90nm, reduced die size by almost 50% , and significant reductions in power consumption. The paper described the 90nm process as strained CMOS with silicon-on-insulator (SOI), with a minimum gate length of 46nm, three (3) gate dielectric thicknesses between 1-2nm, and an SRAM cell size of 1.06 µm². It also detailed 10 levels of metal with fluorine-doped TEOS (FTEOS) dielectric, a form of fluoro-silicate glass (FSG), in the back-end process.

"These process and architecture innovations enable significant cost and power reductions", remarked Chipworks' president, Julia Elvidge. "The PowerTune scaling, together with the SOI, allows operation as low as 15 W."

http://www.chipworks.com/news/2004_IBM_90nm.asp

The transistors on a chip are produced by processing the surface of a silicon wafer as described above. This creates a positive electrical capacity between the microstructures - sort of like in a capacitator - that provides a channel for the charge carriers. This capacitator effect automatically slows down the transistor's switching speed, which simultaneously raises the power consumption of the chip as a whole.
In order to counteract this effect, a thin substrate of insulating oxide is embedded between the individual layers. A silicon-on-insulator structure, or SOI structure for short, can boost transistor circuit speeds by as much as 30%

http://www.tgdaily.com/2005/04/19/behind_the_closed_doors_of_amd/index.html

This one says it all:

Intel has been shipping Pentium 4 processors based on its 90nm Prescott core for well over a year now. Only their high-end Xeon and Pentium 4 Extreme Edition processors are currently still manufactured in 130nm, which partly explains why they carry such a hefty pricetag. Ever since Intel introduced these Prescott processors they have been criticized for their high power drain and excessive heat production. This is especially true in comparison to the previous 130nm generation based on the Northwood core. Nevertheless Intel has been shipping them by the millions and although they announced that there would be no 4GHz Pentium 4 a few months ago they quietly introduced the 3.8GHz just recently. Clearly that means that Intel’s 90nm process works, or at least works well enough to abandon the 130nm process and transition all product lines to 90nm. Regardless of that the end user seems to think that Intel’s 90nm process is broken and that AMD’s 90nm process is working like a charm, but is it?

Intel is fast at introducing smaller process, but they don't perform that well against AMD's. 8)

There are enough links that proves that AMD's process is BETTER than Intel's. Hope you enjoy your self reading.