brisbane dual core 65nm

Quote:

---

Originally Posted by ***Deimos***

OMG (*smack self in face and tries to regain composure*)
I'm not sure what exactly we're debating here, but I'll assume you're questioning how the same 65nm processor can be hotter than a 90nm.
1. Typically, in a perfect world, linearly scaled down transistors would run faster and cooler. However, when gate oxides shrank down to around num of atoms you can count on your toes, we ran into trouble.. big BIG trouble.
2. Have you heard of static leakage current. When just recently Intel had such huge problems with Prescots, why would you think AMD would be immune? Static power is growing exponentially with each die shrink!
3. The smaller the die size, the greater the power/heat concentration. We're approaching levels in a nuclear reactor. Those bygone days of dinky aluminum heatsink are long over. And perhaps soon we might reach limits of copper's heat conductivity.
4. This is AMD's first attempt at 65nm. Ofcourse it will take some time to fine tune the technology. Especially if new materials and techniques are being used. Intel wisely had already gone through the gruesome trials nearly a year earlier before Core2Duo.

---

SOI-2 was perfected 90nm on AM2 and got rid of a lot of the leakage but hit a 3ghz wall. SOI-1 for the 1st 939's was pretty good and didn't leak nearly as bad as prescots but was capable of 4ghz speeds in the FX's. I thought SOI-3 at 65nm would have no leakage at all because its so crazy complex and advanced of a process compared to Intels COMS Cu Germinanium, only process. SOI with DSL is much harder, longer to make, and creates more reliable quality dies that can take more and last longer. As I understand it then AMD made a paper launch with something thats not ready. If it was then it would be better then SOI-2... But they are having trouble with creating the SOI gates in 65nm then huh so guess not. As anybody should expect for the 1st time then. Forgot about that prescot for a wile.

Can you tell me what Intel has done different in Core2's FAB processes since a year ago?

Soon we will use different materials like diamonds to go any faster.

Serge Intel went to strained silicon on the Prescots. Since it was their first time using strained silicon they had a learning process to go threw. Plus Prescots had 2x as many transitors.
Intel finally got the right mix of Germinanium in their SS added more . Intel really isn't going to tell all their process secrets .
Now with the Amd process of Strained silicon on SOI . The Germinanium is removed after its stressed. SOI Serge has a pretty poor leakage problem the smaller the processs becomes. I tried to explain this to you a few months back but you wouldn't listen.
I don't think you well see SOI on AMD 45nm process. But who knows.
This may help AMD in the long run . As now they will have to get this figured out.
Because Intel used Stressed silicon on the 90nm process they are ahead of everyone in this area. Ibm is going to Strained silicon on the 45nm process no SOI. Along with the fab club partners of Samsung Infeneon and Chartered. AMD is not in the Fab Club. So its hard telling if IBM will beable to help them in this area as they have 3 other partners to consider. What I find interesting between IBM and Intel is That with the High K metal gates IBM says it will be useless on the 45nm process so they won't use High K until 32nm process. Intel says differant they are going High K on the 45nm process and 3Dgates on the 32nm process. Either Intel is way ahead or IBM knows something Intel doesn't. Its all going to be very interesting thats for sure.

another true SS )
http://www.overclockers.ru/images/ne.../12/pop_02.gif
low voltage EE model
http://www.overclockers.ru/images/ne.../12/pop_01.jpg

Quote:

---

Originally Posted by Serge84

SOI-2 was perfected 90nm on AM2 and got rid of a lot of the leakage but hit a 3ghz wall. SOI-1 for the 1st 939's was pretty good and didn't leak nearly as bad as prescots but was capable of 4ghz speeds in the FX's. I thought SOI-3 at 65nm would have no leakage at all because its so crazy complex and advanced of a process compared to Intels COMS Cu Germinanium, only process. SOI with DSL is much harder, longer to make, and creates more reliable quality dies that can take more and last longer. As I understand it then AMD made a paper launch with something thats not ready. If it was then it would be better then SOI-2... But they are having trouble with creating the SOI gates in 65nm then huh so guess not. As anybody should expect for the 1st time then. Forgot about that prescot for a wile.

Can you tell me what Intel has done different in Core2's FAB processes since a year ago?

Soon we will use different materials like diamonds to go any faster.

---

the right proportion of pixie dust to fairy spice

;()

Quote:

---

Originally Posted by terrace215

Unfortunately, shrink sucked (.68 or so factor overall), and 65nm Brisbane is ~125mm^2.

---

If that is true, the argument that it won't overclock at all due to running too hot makes no sense. The power density on the 90nm and 65nm versions should be nearly identical. If the CPU really runs that hot, I see two much more probable reasons:

1. The ES in question is not final silicon.

2. The heatspreader is not flat or not attached correctly.

MAS: Who has that 4400+ EE? Can you get any overclocking info?

http://www.pcpop.com/doc/0/168/168366.shtml

http://img5.pcpop.com/ArticleImages/.../000382787.jpg
http://img5.pcpop.com/ArticleImages/.../000382788.jpg
http://img5.pcpop.com/ArticleImages/.../000382789.jpg
http://img5.pcpop.com/ArticleImages/.../000382790.jpg
http://img5.pcpop.com/ArticleImages/.../000382791.jpg
http://img5.pcpop.com/ArticleImages/.../000382792.jpg
http://img5.pcpop.com/ArticleImages/.../000382793.jpg
stock freq.

I don't quit know what your trying to show here. Mas. A 2.3 ghz amd 65nm cpu. Up against a 1.86ghz C2D . If you think that impressive . OK . But don't you think an apples to apples compare would be more suitable.

If your looking at the cost of the cpu. I guesss thats alright. But this is XS and most here are more interested in performance per ghz . Than the cost of a cpu that may or may not O/C very well. AMDprice $214 Intel E6400 $217. You can guy an E6300 even cheaper. power usage is the same. So why the comparison with an E6300. Both are 65watt parts. What I find most interesting about hat short review is no power consumption test. Thats what where all waiting for and O/C test.

if you knew Chinese you would read Brisbane 4400+ based system consumes less power than C2D 6300 system when idle and more power when in burn
( http://www.pcpop.com/doc/0/168/168366.shtml )

Can you read ENGLISH? <<< a little

mentioned brisbane was not OCed -> so i can't tell you its OC-ing limit as well as OC temperatures

Hay guys lets not get this thing closed down. Mas the fact that AMD uses less power than C2D could be a big deal. But 20 watts cost for home use is about 40cents a month. For office use in large company its almost none existant as there rates are lower than ours. Productivity however can easily over come 20 watts of power usage at idle.

Quote:

---

Originally Posted by zir_blazer

Read my last Post. We can make no assumptions solely based on the TDP because that is not what the Processor truely consumes. Besides, I don't recall that any Rev. F K8s Processor only power consumption Benchmarks have been done to do an accurate comparision (The DDR-II Memory Controller should consume more).

---

An AMD processor will never consum more than it's TDP, it may consum alot less than it's TDP, but that's not the point.
X2 5000+ is at the edge of 89w TDP at 90nm and 65w at 65nm, so I think it's good for this kind of comparison.

A X2 consumes less power when idle than a Core2 because Cool 'n Quiet is way better than SpeedStep.
An normal (not Energy Efficient) idle 90nm X2 consumes about 15w, an idle Core2 consumes about 26w:
http://www.xbitlabs.com/articles/cpu...ficient_6.html

BTW, any news on "half" memory dividers?

Quote:

---

Originally Posted by MAS

if you knew Chinese you would read Brisbane 4400+ based system consumes less power than C2D 6300 system when idle and more power when in burn
( http://www.pcpop.com/doc/0/168/168366.shtml )

---

This doesn't help us . Because in burn or load the Intel cpu has the advantage because its less Mgz. This is just a really bad compare.

Brent go to inquirer check out the short story on Intel LOL.

Mas I think your review is = to this one as far as what were discussing here. This is the kind of review were looking for. Temps and O/C ability.

I don't know the price of this cpu but it won't be much.

http://topic.expreview.com/2006-12-1...1666d1617.html

Quote:

---

Originally Posted by Turtle 1

This doesn't help us . Because in burn or load the Intel cpu has the advantage because its less Mgz.

---

Yes, that gives it an advantage, but the K8 has the (potentially) big advantage of having much less transistors. The 2W lower power consumption of the C2D E6300 in the pcpop test should also be taken with a fistfull of salt. Not being able to use the same motherboard makes the measurements pretty much useless for a processor power consumption analysis. An inefficient motherboard could account for a 10-15W difference.

Quote:

---

Originally Posted by doompc

X2 5000+ is at the edge of 89w TDP at 90nm and 65w at 65nm, so I think it's good for this kind of comparison.

---

That is the deal, we do not know if a 65nm K8 running at 2.6 GHz is around 65W on Full Load or considerabily less as there are no Processor only power consumption Benchmarks to do a fair comparision.

Why are they not posting all the numbers? Maybe someone needs to get AMD to give them to us ;-)
Surely someone overclocks them from within!

I'm actouly going twords intel on this one. The C2D's are better. Maybe until K8L But if there was only a mobo out there that was matx and was NF-M2 like in features and OCability options then count me in for a upgrade cus I got the cpu all I need is a great mobo for my case. Why in the hell doesn't anybody make a bloody high end intel matx. I really think thats the only thing thats holding me back. I want a matx that can do 500FSB then I'll def switch.

Yeah hell froze over for another AMD fan.

Quote:

---

Originally Posted by Turtle 1

Serge Intel went to strained silicon on the Prescots. Since it was their first time using strained silicon they had a learning process to go threw. Plus Prescots had 2x as many transitors.
Intel finally got the right mix of Germinanium in their SS added more . Intel really isn't going to tell all their process secrets .
Now with the Amd process of Strained silicon on SOI . The Germinanium is removed after its stressed. SOI Serge has a pretty poor leakage problem the smaller the processs becomes. I tried to explain this to you a few months back but you wouldn't listen.
I don't think you well see SOI on AMD 45nm process. But who knows.
This may help AMD in the long run . As now they will have to get this figured out.
Because Intel used Stressed silicon on the 90nm process they are ahead of everyone in this area. Ibm is going to Strained silicon on the 45nm process no SOI. Along with the fab club partners of Samsung Infeneon and Chartered. AMD is not in the Fab Club. So its hard telling if IBM will beable to help them in this area as they have 3 other partners to consider. What I find interesting between IBM and Intel is That with the High K metal gates IBM says it will be useless on the 45nm process so they won't use High K until 32nm process. Intel says differant they are going High K on the 45nm process and 3Dgates on the 32nm process. Either Intel is way ahead or IBM knows something Intel doesn't. Its all going to be very interesting thats for sure.

---

*Now with the Amd process of Strained silicon on SOI . The Germinanium is removed after its stressed.*

And intel leaves it in right? Wouldn't it be better to leave it in to keep the effects of the material forever? As far as I understand it Germinanium is being used for the 1st time in AMD with 65nm's. Germinanium gives 40% better performance in transistors when used. It would help in OCing because of this material and part of the reason Intel can go so high in OCing also because there pipes are longer then AMD's. But removing a fab process after its applied doesn't sound right because it would make applying the process in the 1st place almost useless. Almost like a cheap way of saving cash so you don't have to buy more of the stuff.

Quote:

---

Originally Posted by Serge84

I'm actouly going twords intel on this one.

Yeah hell froze over for another AMD fan.

---

Sorry, I just had to:D
---
http://www.dailytech.com/article.aspx?newsid=5328

Quote:

---

Internal roadmaps showed the Brisbane 65nm desktop processors would launch on December 5th, 2006. Yet almost every contact I had spoken to in the supply chain -- from the engineers to the distributors -- unanimously claimed there was virtually no chance the company would ship a 65nm product this year.

---

Quote:

---

It would help in OCing because of this material and part of the reason Intel can go so high in OCing also because there pipes are longer then AMD's

---

Authenticating...
Verified! - genuine noob.

Why is it every time I come to xtremesystems it seems like its a kindergarten playhouse with jumbo colorful stuff animals, and those toys where you have to fit the square into the square hole, the circle into the... etc..

Look, its really simple... there are no "pipes". They are not being made longer. The terms have been horribly mis-used my internet folks. Processors have buffers where they keep intermediate results. They partition work on each instruction into steps or stages. "Pipeline", "pipe" or any other similar term, is just a METAPHOR. Its like saying that a Ford Mustang has a lot of "muscle".

Quote:

---

Originally Posted by Turtle 1

I don't quit know what your trying to show here. Mas. A 2.3 ghz amd 65nm cpu. Up against a 1.86ghz C2D . If you think that impressive . OK . But don't you think an apples to apples compare would be more suitable.

If your looking at the cost of the cpu. I guesss thats alright. But this is XS and most here are more interested in performance per ghz . Than the cost of a cpu that may or may not O/C very well. AMDprice $214 Intel E6400 $217. You can guy an E6300 even cheaper. power usage is the same. So why the comparison with an E6300. Both are 65watt parts. What I find most interesting about hat short review is no power consumption test. Thats what where all waiting for and O/C test.

---

Uhh huh.. sure, whatever. Where was the almighty performance per Ghz, when throughout 2001-2006, people abandoned their P3's and jumped on the P4 bandwagon?

XS has never been about anything per something. Nobody cares if you can get a little extra IPC, little less noise, performance/W, performance/$ or whatever. XS is mainly about overclocking.. by any creative means imaginable. And over the years, the fountain of creativity here has shown no signs of abating.

Quote:

---

Originally Posted by brentpresley

Your 40% better performance number is a VAST oversimplification. If it were that simple, Ge would have been used in wafers for decades now.

The fact is that too much Ge in the wafer has the OPPOSITE effect and reduces transistor switching efficiency. Add to that the fact that Ge is BLOODY hard to work with. When you add it to wafers your dust level has to be in the PICO particles as the wafers are incredibly sensitive to contamination compared to regular Si-only wafers.

All of this raises costs, which is why it is just NOW being implemented, when we are approaching the limits of Si-only technology.

---

I'm talking about transistor performance not the performance of the cpu lets just get that clear with e-SiGe added.

Quote:

---

Originally Posted by Serge84

I'm actouly going twords intel on this one. The C2D's are better. Maybe until K8L But if there was only a mobo out there that was matx and was NF-M2 like in features and OCability options then count me in for a upgrade cus I got the cpu all I need is a great mobo for my case. Why in the hell doesn't anybody make a bloody high end intel matx. I really think thats the only thing thats holding me back. I want a matx that can do 500FSB then I'll def switch.

Yeah hell froze over for another AMD fan.

---

Congrats Serge you have become an enthusist. I wish I could make that step. But news occurred today that makes it impossiable for myself.

As for your question yes intel leaves Ge in its SS. IBM/AMD remove it. I believe the main reason for this is because dual stressed strained si ge on SOI would interfer with the transiters and gates operating effiently. Not sure.

Quote:

---

Originally Posted by ***Deimos***

Authenticating...
Verified! - genuine noob.

Why is it every time I come to xtremesystems it seems like its a kindergarten playhouse with jumbo colorful stuff animals, and those toys where you have to fit the square into the square hole, the circle into the... etc..

Look, its really simple... there are no "pipes". They are not being made longer. The terms have been horribly mis-used my internet folks. Processors have buffers where they keep intermediate results. They partition work on each instruction into steps or stages. "Pipeline", "pipe" or any other similar term, is just a METAPHOR. Its like saying that a Ford Mustang has a lot of "muscle".

---

When you have a pipelined processor, each cycle is a number of stages. In each stage, you do a few different things, and, depending on the processor design, each stage is one or more oscillator cycles. For example the P4 has 31 stages witch is long, the K8 far less, and the core2 far more then the K8. Now that will tell us how well the K8, Core2, and K8L will overclock depending on stages of the pipeline.

5 stage pipelines are easiest conceptually, and from them, you tend to extrapolate our current pipelines (we use more than 5 stages in modern processors). The five stages are Fetch (get the next instruction), Decode (figure out what it is), Execute (do math), Memory (do memory acess), and Write-back (finish things up write data back to memory). Now when I say "memory", it's really more like cache, not memory. But the processors of today use 30+ stages.

How wide will also determen the cycles per clock in all out performance how it will calculate 64bit or 128bit data chunks per cycle. P4's are very slow, K8's have pretty much a 1000/1500mhz+ clock speed advantage over P4 at a much lower clock speed. And Core2 is 20% faster then K8's because of there cycles per clock and stages. So what am I again?