Sign in with
Sign up | Sign in
Your question

What would happen when AMD goes 65nm

Last response: in CPUs
Share
October 7, 2006 10:27:30 PM

I have a theory but from fear of starting a flame war, the increase Processor count at 65nm would mean that AMD could pump out more GHz per wafer size?

your thoughts?

More about : happen amd 65nm

October 7, 2006 11:08:10 PM

In theory they can, but AMD is going to need a lot more than a MHz increase to beat Intel in the performance war.
October 7, 2006 11:10:41 PM

Quote:
I have a theory but from fear of starting a flame war, the increase Processor count at 65nm would mean that AMD could pump out more GHz per wafer size?

your thoughts?


It'll mean that Intel ho's have to complain about how Intel got there first. Plus it'll mean that AND can save money on their total output as 65nm costs less to produce.
Related resources
October 7, 2006 11:24:23 PM

No flame wars from me. It means AMD will be able to manufacture chips at a lower cost, allowing them to become more competitive with Intel. They also need 65nm for their upcoming quad-cores. Plus they'll be able to gradually lower the power requirements of their current and future parts. So it will be great for AMD and great for consumers. I'm looking forward to it.

The speed improvements aren't automatic but they will come with time. For instance, Intel wasn't able to increase the clock speeds of their P4s when they transitioned to 65nm, although they did lower the power consumption significantly. I'm sure AMD will be able to squeeze a few MHz out of their current lineup, but the real performance improvements will come from their K8L line, which will be available in the second half of next year if things go well.
October 7, 2006 11:27:47 PM

Quote:
I have a theory but from fear of starting a flame war, the increase Processor count at 65nm would mean that AMD could pump out more GHz per wafer size?

your thoughts?


It'll mean that Intel ho's have to complain about how Intel got there first. Plus it'll mean that AND can save money on their total output as 65nm costs less to produce.
BMS Remember that AMD is going to Start Producing or atleast Transition to 45nm Process with in 18 Months of the first 65nm Chip to be Retail sold. And 32nm transtion by ????? any one know if it's by 2010 or Any time later? I know that End of 65nm and Quad Core Might be in a few months close to each othersight< they might be hoping for a 45nm Quadfather and maybe a "OctoMother" or "OctoMom" Processer by the beginning of the 32nm transition
October 8, 2006 1:10:13 AM

The past always bites <insert name> on the rear. 65nm will be good for AMD, and will bring them closer to competing with Intel again.
October 8, 2006 1:19:03 AM

This might seem a bit round about, but with lower cost they can devote more money to development and with better development, they can build more competitive processors, can't they?
October 8, 2006 1:25:05 AM

Quote:
I have a theory but from fear of starting a flame war, the increase Processor count at 65nm would mean that AMD could pump out more GHz per wafer size?

your thoughts?


I'm not sure if you meant "increase processor count" or "increase transistor count", @65nm, per-wafer-size.

Anyway, if >processor count per-wafer-size, then yes: The bigger the wafer, the more dies it would yield at the same node, adding up to total GHz delivered per wafer:

Example 1: 333 dies*4 cores-per-die*3.0GHz per-core = 3 996GHz = 3.9THz (@65nm@200mm wafer);
Example 2: 500 dies*4 cores-per-die*3.0GHz-per-core = 6 000GHz = 6.0THz (@65nm@300mm wafer);

Notes: These are mere examples and might not reflect reality; areal scaling factor between different wafers: 2.25.


If >transistor count (all things being equal), not necessarily, since a mere node transition does not imply a speed grade improvement; and, you could still have bigger dies at a lower node, if you increased enough logic units & cache into it, for instance.

(I fail to see your point, but don't take it as an ironic stance towards your question; maybe it's just me... :wink: ).


And, your theory would be?


Cheers!
October 8, 2006 4:00:55 AM

Quote:
I have a theory but from fear of starting a flame war, the increase Processor count at 65nm would mean that AMD could pump out more GHz per wafer size?

your thoughts?


It'll mean that Intel ho's have to complain about how Intel got there first. Plus it'll mean that AND can save money on their total output as 65nm costs less to produce.

Baron, the 'Intel ho's' as you want demean them, don't need to tout how Intel got there first. Intel has always achieved the next litho node before AMD since the microprocessor has been in existence.

Though you are correct, 65 nm will allow more cost effective produciton.... oddly you have stated in the past that this is unimportant and that AMD did not need 65 nm.


I never said they didn't need 65nm. I may have said that they seem to have gotten power down without it, but of course they will need 65nm for quad and L3. Even the X4s and X2 Semprons will get L3.

I'm sure that Brisbane is the proof of concept for 65nm and Arcturus, etc will be really working that DDR2 bandwidth.
October 8, 2006 5:07:49 AM

Quote:

.... the initial indicators are that Intel will launch 45 nm products at 3.46 to 3.73 Ghz or around there, this will come around or shortly after K8L debuts. These early numbers are in agreement with Intel's future FUSI process, if this is the one they choose, in terms of the intrinsic transistor performance. So I happen to have some faith in those numbers.

Jack

AMD too is working with nickel silicide! :lol: 
October 8, 2006 5:13:30 AM

Quote:


Baron, the 'Intel ho's' as you want demean them, don't need to tout how Intel got there first. Intel has always achieved the next litho node before AMD since the microprocessor has been in existence.

Though you are correct, 65 nm will allow more cost effective produciton.... oddly you have stated in the past that this is unimportant and that AMD did not need 65 nm.



Awww, look Jack. Baron voted for you. Wasnt that sweet? I guess he doesnt like it when you revive previous statements he's made that contradict his current ramblings. Tsk Tsk.
October 8, 2006 5:20:33 AM

Quote:


I never said they didn't need 65nm. I may have said that they seem to have gotten power down without it, but of course they will need 65nm for quad and L3. Even the X4s and X2 Semprons will get L3.

I'm sure that Brisbane is the proof of concept for 65nm and Arcturus, etc will be really working that DDR2 bandwidth.


Come Baron, you have said and you said it more than once.... I will go dig up the quote.

EDIT: Still looking, it is alot of BS to read through....

Jack

OOHH!!!! BURN :!:

LOL :wink:
October 8, 2006 5:33:13 AM

Just release yourself to your inner nature. Join the darkside of the force Jack


Seriously, I'm looking for the quote and I'm finding BS and a whole lot of it but no quote yet.
October 8, 2006 5:37:19 AM

Quote:


Baron, the 'Intel ho's' as you want demean them, don't need to tout how Intel got there first. Intel has always achieved the next litho node before AMD since the microprocessor has been in existence.

Though you are correct, 65 nm will allow more cost effective produciton.... oddly you have stated in the past that this is unimportant and that AMD did not need 65 nm.



Awww, look Jack. Baron voted for you. Wasnt that sweet? I guess he doesnt like it when you revive previous statements he's made that contradict his current ramblings. Tsk Tsk.

You know what is sad --- there was a 65 nm discussion where Baron stated AMD did not need 65 nm. Where I responded, .."are you nuts. the cost incentive alone is enough to drive them to 65 nm...", damn if I cannot find the darn post.

In searching for it, I have waded through more BaronMatrix posts than I care to and it simply reminds me how jaw dropping poor his understanding is.... we need to start collecting the BaronMatrix classics for future reference --- there was the post he claimed to be an analyst, then later stated he never said that --- lucily someone dug it up, this is the same case.

Oh, I remember that discussion. Barons conclusions were, well let’s say more optimistic opinion than fact.

I remember some of Barons other claims as well:
-Mechanical Engineer working with:
Fuzzy logic
Reverse Kinematics
-Developer
-From another board:
Former member of the US Army 82nd Airborne

Hmmm. He’s made so many statements, you could find yourself writing a book. Or perhaps a thesis. Pick yourself up another PhD while creating a new field of study: Metacognitive Spatial Disorientation Disorder in Speculative Micro Architecture :wink:
October 8, 2006 5:38:32 AM

Quote:

AMD too is working with nickel silicide! :lol: 


Not in terms of FUSI at the moment, ...

FUSI was reported by AMD, but they are not talking implementation until 45 or 32 nm from what I recall.


Jack
I wouldn't put any money on that!! :wink:
October 8, 2006 5:39:32 AM

Quote:


I never said they didn't need 65nm. I may have said that they seem to have gotten power down without it, but of course they will need 65nm for quad and L3. Even the X4s and X2 Semprons will get L3.

I'm sure that Brisbane is the proof of concept for 65nm and Arcturus, etc will be really working that DDR2 bandwidth.


Come Baron, you have said and you said it more than once.... I will go dig up the quote.

EDIT: Still looking, it is alot of BS to read through....

Jack

OOHH!!!! BURN :!:

LOL :wink:

You know I am working to stop giving Baron such a hard time, it is not working :)  ...

I don't even know why i thought it was so funny, it just laughed so much. It was just such an off the cuff casual insult, like you'd rehursed it in the mirror that it made the idea that everything BM posts is BS seem like a fact. :lol: 
October 8, 2006 5:45:35 AM

Quote:
Just release yourself to your inner nature. Join the darkside of the force Jack


Seriously, I'm looking for the quote and I'm finding BS and a whole lot of it but no quote yet.



How dare you post a picture of Paul Otellini with out the express written permission of Intel!! No horde lightsaber for you!
October 8, 2006 5:51:46 AM

In response to your previous post I give a certain amount of dignity to such a worthless excitement of electrons by simply saying, Meh.
October 8, 2006 5:58:01 AM

Can I start a "Place your Bets" thread?
So we could post what future techs we think would get implimented and then go back, see what passed what failed, and who gets owned and losses some form of currency or whatever they put up.
Or would that get me banned faster than <insert name>?
October 8, 2006 6:00:50 AM

This might make me look stupid, but what is FUSI?
October 8, 2006 6:06:39 AM

Quote:
This might make me look stupid, but what is FUSI?


Obviously florineuranium plated silicon. GOD!!!!


LOL, i have no idea. 8)
October 8, 2006 6:14:08 AM

Quote:
In response to your previous post I give a certain amount of dignity to such a worthless excitement of electrons by simply saying, Meh.


Im sorry. The joke was in reference to one of our more delusional, (but actually sometimes smart) horde members: MadModMike. You know him by his puppet logins, MrsD and MrsBytch. He was into the whole "Evil Empire" of Intel, and coined one of the most ridiculous phrases Ive ever seen written here: "the darkside of Conroe". Im sure you are familiar with the type, "I am the light, the truth shall be known, yada yada yada".
October 8, 2006 6:20:24 AM

Yeah I know, it was meant to be a joke though. We always say "Meh" in response to things we don't have a sufficiently witty response to. A joking tone doesn't carry well over this medium.
MadModMike, he really was like that?
@ Jack,
Thanks....
October 8, 2006 6:33:07 AM

Quote:
Yeah I know, it was meant to be a joke though. We always say "Meh" in response to things we don't have a sufficiently witty response to. A joking tone doesn't carry well over this medium.
MadModMike, he really was like that?
@ Jack,
Thanks....



In a word: Yes
October 8, 2006 6:34:55 AM

8O I say no more... I need coffee now.
Maybe we should institute some kind of sanity test before you can become a member... but then it wouldn't be so much fun either.
October 8, 2006 6:52:40 AM

so where does ATi come into this ? ,

do you guys think that ATi will combine resources and start Researching and Helping AMD develop better CPU's ?

im hoping they do , inovation , inovation , inovation , more competition , better technology = good hardware for consumers.
October 8, 2006 6:55:01 AM

They kinda have to because:
AMD < ATI
AMD = ATI
ATI = AMD
ATI > AMD

They're one in the same.
October 8, 2006 7:29:39 AM

Quote:
so where does ATi come into this ? ,

do you guys think that ATi will combine resources and start Researching and Helping AMD develop better CPU's ?

im hoping they do , inovation , inovation , inovation , more competition , better technology = good hardware for consumers.


I am curious, could you list the innovations that AMD has brought to date?Ummmm.....The term PIB(processor in a box). :p 
October 8, 2006 7:40:45 AM

Quote:
In theory they can, but AMD is going to need a lot more than a MHz increase to beat Intel in the performance war.


You imply they need to beat Intel?

They don't, the highest performance CPUs barely sell at all, just too expensive. They merely need to market, market, market, and beat Intel in the mid to low end. They won't of course, Celerons on Intel Integrated video will still dominate but it's still possible for AMD to take some marketshare if they could get prices low enough and do something interesting with ATI chipsets.
October 8, 2006 8:05:15 AM

Quote:
I am curious, could you list the innovations that AMD has brought to date?

I'm guessing that A64 and dual core dont count.
October 8, 2006 8:24:37 AM

Quote:
This might make me look stupid, but what is FUSI?


It's a Female WUSY ! :) 
October 8, 2006 9:39:21 AM

Quote:
Just release yourself to your inner nature. Join the darkside of the force Jack


Seriously, I'm looking for the quote and I'm finding BS and a whole lot of it but no quote yet.



How dare you post a picture of Paul Otellini with out the express written permission of Intel!! No horde lightsaber for you!

Can i have one????
October 8, 2006 9:40:26 AM

Quote:
I am curious, could you list the innovations that AMD has brought to date?

I'm guessing that A64 and dual core dont count.

Intel got 64 bit CPUs out first (Itanium) and they brought out the first Dual Cores (D8** series).

In other words, YOU LOOOOSE!
October 8, 2006 12:34:11 PM

Quote:
so where does ATi come into this ? ,

do you guys think that ATi will combine resources and start Researching and Helping AMD develop better CPU's ?

im hoping they do , inovation , inovation , inovation , more competition , better technology = good hardware for consumers.


I am curious, could you list the innovations that AMD has brought to date?
I think he meant ATI's innovations, not AMD's.
Anyway, some AMD innovations which come right off my mind (in x86 CPUs, not in the whole CPU industry, be it clear from now):

* SIMD FP operations (3dNow! came out earlier than SSE)
* OOO superscalar FPU (K7)
* Integrated memory controller (Intel's Timna was cancelled) & Hypertransport; conceptually nothing earth shattering, but these 2 combined make a highly scalable architecture
* 64 bit addressing and extensions
* NX bit :) 
* "Inclusive Cache"
* native dual core (for what it's worth..)
* support for DDR memory (while Intel was pushing RDRAM)
* i'm not 100% sure about this, but i think i recall PowerNow! being launched (K6+) a bit earlier than Intel's Speedstep (Pentium 3?)

That said, to mention all Intel's innovations, a whole 20 pages thread wouldn't be enough.. ;) 
October 8, 2006 12:37:20 PM

Quote:
I am curious, could you list the innovations that AMD has brought to date?

I'm guessing that A64 and dual core dont count.

Intel got 64 bit CPUs out first (Itanium) and they brought out the first Dual Cores (D8** series).

In other words, YOU LOOOOSE!
Nay, IA64 doesn't count cause it's:
* neither the first 64 bit CPU (MIPS, SPARC, Alpha, HP-PA... and many others preceded it)
* nor the first 64 bit x86 CPU (it's not x86 at all ;)  )
October 8, 2006 12:59:20 PM

Quote:
I am curious, could you list the innovations that AMD has brought to date?

I'm guessing that A64 and dual core dont count.

Intel got 64 bit CPUs out first (Itanium) and they brought out the first Dual Cores (D8** series).

In other words, YOU LOOOOSE!
Nay, IA64 doesn't count cause it's:
* neither the first 64 bit CPU (MIPS, SPARC, Alpha, HP-PA... and many others preceded it)
* nor the first 64 bit x86 CPU (it's not x86 at all ;)  )

True points there...Guess it's 1 - 1.
October 8, 2006 2:09:06 PM

Quote:
so where does ATi come into this ? ,

do you guys think that ATi will combine resources and start Researching and Helping AMD develop better CPU's ?

im hoping they do , inovation , inovation , inovation , more competition , better technology = good hardware for consumers.


I am curious, could you list the innovations that AMD has brought to date?

better read my post again , until you understand it , i didnt say anything bout AMD or ATi bringing any inovations at all to the market , i just said with both of them joining forces , they can inovate together , and bring out some great products.
October 8, 2006 3:06:04 PM

Quote:

AMD too is working with nickel silicide! :lol: 


Not in terms of FUSI at the moment, ...

FUSI was reported by AMD, but they are not talking implementation until 45 or 32 nm from what I recall.


Jack
I wouldn't put any money on that!! :wink:

I would.

How much?

Quote:

BTW, would you put money on RHT?

The correct term is Speculative Multithreading. Intel and AMD have both been working on it. :lol: 
October 8, 2006 3:40:09 PM

Quote:


Baron, the 'Intel ho's' as you want demean them, don't need to tout how Intel got there first. Intel has always achieved the next litho node before AMD since the microprocessor has been in existence.

Though you are correct, 65 nm will allow more cost effective produciton.... oddly you have stated in the past that this is unimportant and that AMD did not need 65 nm.



Awww, look Jack. Baron voted for you. Wasnt that sweet? I guess he doesnt like it when you revive previous statements he's made that contradict his current ramblings. Tsk Tsk.

You know what is sad --- there was a 65 nm discussion where Baron stated AMD did not need 65 nm. Where I responded, .."are you nuts. the cost incentive alone is enough to drive them to 65 nm...", damn if I cannot find the darn post.

In searching for it, I have waded through more BaronMatrix posts than I care to and it simply reminds me how jaw dropping poor his understanding is.... we need to start collecting the BaronMatrix classics for future reference --- there was the post he claimed to be an analyst, then later stated he never said that --- lucily someone dug it up, this is the same case.


I expect to see this quote. I have a very good memory and nothing to hide. If I don't see it EXACTLY AS YOU QUOTED IT, you will never live it down, cause this isn't the first time you've made up something and attributed it to me.
October 8, 2006 3:44:46 PM

Quote:
I expect to see this quote. I have a very good memory and nothing to hide. If I don't see it EXACTLY AS YOU QUOTED IT, you will never live it down, cause this isn't the first time you've made up something and attributed it to me.

Looks like you have selective amnesia, you remember only about the things you like. We don't need quotes of your BS, we allready know that you are very stupid and ignorant moron without knowledge.
October 8, 2006 3:45:00 PM

yeah, that is why your profile is so complete...

but you are a legend in your own mind.
October 8, 2006 3:54:42 PM

Quote:
I have a theory but from fear of starting a flame war, the increase Processor count at 65nm would mean that AMD could pump out more GHz per wafer size?

your thoughts?


Nope. Release speeds are going to max out at 2.6 GHz according to the most recent released roadmaps, they will benefit from better power characteristics though and the smaller die, as BM and others point out -- 65 nm will increase output at lower costs.

However, AMDs 65 nm transistor metrics are turning out roughly 10% slower than Intels. In Dec. of this year at the IEDM we will be able to see if AMD made any progress on their 65 nm development, if they are going forward with the process they presented at 2005's IEDM (you know that 40% metric they tooted), then it will be unlikely they will push 65 nm much beyond 3.2-3.4 GHz. This will not be enough to overtake Intel on the K8 core.

It does not take much work to research the net to see that AMD's top 65 nm bin will be slower (significantly slower) than the top 90 nm bin. And the picture is not to rosy, AMD has not put on any roadmap (official or leaked) any 65 nm parts showing > 3.0 GHz.

K8L though is a different story, and it will be fun to watch the fireworks fly when that architecture is released. So long as IPC can significantly out reach C2D then AMD has a chance of getting the crown back, but it will need to be almost double.... the initial indicators are that Intel will launch 45 nm products at 3.46 to 3.73 Ghz or around there, this will come around or shortly after K8L debuts. These early numbers are in agreement with Intel's future FUSI process, if this is the one they choose, in terms of the intrinsic transistor performance. So I happen to have some faith in those numbers.

Jack

As I remember the transition to 90nm, AMD tried to push the lower clock CPUs first into the market since AMD cannot bear the risk of the failure of new process. They tried Athlon64 and Sempron and then Opteron at last. So I still think AMD should push the lower bin first.

But this time is quit a bit different from the last one as AMD now is trying to push x2 first.
October 8, 2006 4:37:09 PM

AMD would notice that they are still no match for any Core2
October 8, 2006 5:00:49 PM

Is it just me, or did AMD have nowhere near as many problems transferring from 130nm to 90nm as from 90nm to 65nm?

I may not know too much about the complexitys of producing CPUs but surely there can't be that much difference from shrinking from 130 - 90 and 90 - 65. After all it's the same production methods just smaller as far as i'm aware.

I only thought we'd be seeing problems as large as AMD are having with the transition from 45 - 32 and smaller as different production methods would have to be used.

I'm now prepared to be flamed for being a complete and utter noob :D 

Corrections are also welcome.
October 8, 2006 5:14:02 PM

Quote:
Is it just me, or did AMD have nowhere near as many problems transferring from 130nm to 90nm as from 90nm to 65nm?

I may not know too much about the complexitys of producing CPUs but surely there can't be that much difference from shrinking from 130 - 90 and 90 - 65. After all it's the same production methods just smaller as far as i'm aware.

I only thought we'd be seeing problems as large as AMD are having with the transition from 45 - 32 and smaller as different production methods would have to be used.

I'm now prepared to be flamed for being a complete and utter noob :D 

Corrections are also welcome.


Generally the shrinking process isn't hard; it's the fine tuning to get decent yields that is the problem. That’s where the cost of transition comes into play as well.

I personally don't see AMD even getting close to their .45nm target maybe early 2009 but that’s my personal opinion, as seen with the slowing trend of AMD in node transition. Also I don't see them making the .32nm node at all, well not without help from their partners.

This is where being Intel is a good thing deep pockets ensure that they will be able to continue to smaller nodes while smaller companies will not be able to make that transition. Intel, Samsung, and Texas Instruments will likely be the only .32nm node manufactures for quite some time.
October 8, 2006 5:29:10 PM

The innovative Pentuim Rating...
October 8, 2006 5:29:30 PM

Still looking, and I've found many debunked claims and assertions, but not this one... Damn, I'm out of coffee again.
October 8, 2006 5:47:36 PM

Quote:
Is it just me, or did AMD have nowhere near as many problems transferring from 130nm to 90nm as from 90nm to 65nm?

I may not know too much about the complexitys of producing CPUs but surely there can't be that much difference from shrinking from 130 - 90 and 90 - 65. After all it's the same production methods just smaller as far as i'm aware.

I only thought we'd be seeing problems as large as AMD are having with the transition from 45 - 32 and smaller as different production methods would have to be used.

I'm now prepared to be flamed for being a complete and utter noob :D 

Corrections are also welcome.


Generally the shrinking process isn't hard; it's the fine tuning to get decent yields that is the problem. That’s where the cost of transition comes into play as well.

I personally don't see AMD even getting close to their .45nm target maybe early 2009 but that’s my personal opinion, as seen with the slowing trend of AMD in node transition. Also I don't see them making the .32nm node at all, well not without help from their partners.

This is where being Intel is a good thing deep pockets ensure that they will be able to continue to smaller nodes while smaller companies will not be able to make that transition. Intel, Samsung, and Texas Instruments will likely be the only .32nm node manufactures for quite some time.

Wow...quite a few of my assertions were actually correct, much to my suprise.
October 8, 2006 5:54:26 PM

Quote:
I am curious, could you list the innovations that AMD has brought to date?

I'm guessing that A64 and dual core dont count.

They don't count nor count really --- what AMD has done over time is take peices and put them together to make a great product, but each individual 'innovation' was not necessarily innovation as it had existed in the industry before. However, putting it all together into one package and making it good is innovative, people often though state AMD is the great innovator which is a bit of a stretch....

My opinion --- IBM is the better innovator beyond Intel or AMD. IBM does better research too.

Jack

Word.
October 8, 2006 6:02:21 PM

Quote:
- Inclusive cache -- not sure about this one, need to research who actually brought it to market first.... AMD did not invent it though, this I am certain.

So far from what Google is telling me, and also from Wikipedia, AMD is the first to successfully market the inclusive cache. that doesn't mean they are the first overall. I'll continue looking for this.
Ninja
!