Printable View
Had some people bothered to display the slightest bit of logic and not take my comment ad literam we wouldn't have steered off course.Quote:
Originally Posted by PetNorth
I don't see why we should take for granted a company's claims when based on their track record , meeting deadlines is the exception and not the rule.
All the nitpicks that surface show that AMD will hardly be able to meet its deadline.What's wrong in having a conversation on this subject ? :confused:
For the desktop, that's simply not true!Quote:
Originally Posted by PetNorth
Honetly I'm not optimistic about server part :(
Do you know more than AMD?
PR my dear Watson , PR.;)Quote:
Originally Posted by PetNorth
PR though it may be that doesn't automatically mean that its a lie.Quote:
Originally Posted by savantu
Where the heck did you get that slide BTW? AFAIK AMD never released any info. at all about the K10 to the public or to closed door meetings, I don't think it ever even existed outside of VHDL....
I can't believe that AMD's .65 is going to ramp that slowly either or that they'll make so few clock speed improvements to the K8 with a die shrink like that.
ah yeah... I forgot we have here to Hector's son...Quote:
Originally Posted by savantu
figure it...
Hector Ruiz said a week later that they plan to introduce a new microarch in late 2007 and ramp it in 2008.Quote:
Originally Posted by mesyn191
The slide is from a a technical PDF when AMD released it Q3 results in 2003.Quote:
Where the heck did you get that slide BTW? AFAIK AMD never released any info. at all about the K10 to the public or to closed door meetings, I don't think it ever even existed outside of VHDL....
I can't believe that AMD's .65 is going to ramp that slowly either or that they'll make so few clock speed improvements to the K8 with a die shrink like that.
K10 was very real back then with K9 close to having completed its design phase when it was killed.
K10 work started in mid 2002 and the architecture was finished by Q2 2004.The design phase continued at a slow pace until late 2005 when it was presumably killed/mothballed.
As for AMD's 65nm it is obvious the ramp will be slow since FAB36 was build for 90nm tech.
A die shrink doesn't automatically mean you'll get a frequency bonus.For that to happen you need extensive work on the chip's speedpaths since shrinking ussually makes a mess in that area.
The question is : can AMD make a good shrink while working hard on K8L ?
So? You can't know they're lying... For all you know they didn't know until a week later what they could/couldn't do.Quote:
Originally Posted by savantu
Looking at that slide if they were serious its easy to see why they had to can it, even if they were able to reach those clockspeeds the chip would probably be to hot, and the market is moving away from monolithic single core chips anyways.Quote:
Originally Posted by savantu
That doesn't mean anything. FAB30 was origionally built with 180nm quipment and they transitioned it to a 130nm process with out too much trouble. IIRC it was when they had to switch from 130nm to 90nm that they ran into serious issues because they had to replace most of their equipment. AFAIK FAB36 was always built to start out on 90nm and transition to 65nm, the next jump will probably be more difficult.Quote:
Originally Posted by savantu
Its a whole new process, of course they did significant work transitioning it to .65, AFAIK its impossible to do it any other way...Quote:
Originally Posted by savantu
They've got like 3 design teams now right?Quote:
Originally Posted by savantu
Quote:
Originally Posted by mesyn191
Remember IBMs 500GHz chip thing? It ran at 350GHz on air cooling
...watching this thread, and unfortunately not because of the K8L news....
Keep it clean and on topic....official warnings will go out to anybody that tries to stray/bash/flame/bait/etc.
I wouldn't use the term lying , I prefer "distorting reality".They say everything is fine until you see the thing being pushed further and further down the roadmap.Quote:
Originally Posted by mesyn191
First K8L roadmaps had it in early Q1 and many thought Brisbane was K8L based.Then it was moved to Q2 .Now you can't find it in Q3 , so it is Q4.
You don't have to be a genius to see that K8L is sliding further and further back...
That period was a GHz frenzy mood , with Intel working on the 5Ghz Prescott , 7 Ghz Tejas and 10Ghz Nehalem.We all know how miserably it pan out...Quote:
Looking at that slide if they were serious its easy to see why they had to can it, even if they were able to reach those clockspeeds the chip would probably be to hot, and the market is moving away from monolithic single core chips anyways.
True , but look here http://www.dailytech.com/article.aspx?newsid=418Quote:
That doesn't mean anything. FAB30 was origionally built with 180nm quipment and they transitioned it to a 130nm process with out too much trouble. IIRC it was when they had to switch from 130nm to 90nm that they ran into serious issues because they had to replace most of their equipment. AFAIK FAB36 was always built to start out on 90nm and transition to 65nm, the next jump will probably be more difficult.
At best they started to install 65nm tech in Q1.9 months later and you are in 2007.Until you reach good yeilds and start mass manufacturing , add another 3 months at least.
Exactly.K8 is the best example of a highly tuned chip.AMD spent more time on it than any other CPU or Intel.The huge number of revisions and steppings is incredible.How else can you get a 2.8 DC in 125w ?Quote:
Its a whole new process, of course they did significant work transitioning it to .65, AFAIK its impossible to do it any other way...
IIRC they said they have 3x more men.How that translates into more teams I don't know.Quote:
They've got like 3 design teams now right?
What I do know is this :
-K8 is further refined on 90nm , new transistor is scheduled for late 2006 to allow FX64/66 be created in a 125w TDP
-K8L should tape out this quarter or the next one , Validation takes 1 year , lots of bug fixes , new steppings => lot of work
-a mobile optimized microarch is at work scheduled for late 2k7/early 2k8
With all the above , who's left to shrink K8 to 65nm ? Not many IMO.
Add to this the fact that they are busy working probably on K8L replacement , since Nehalem looks to be a formidable threat , more so than Conroe.
al this guessing..omg i'm getting paranoia!
yet good points from many members in here :)
although...:stick:
AFAIK it wasn't really a whole chip, they just did a few transistors at that speed and the news got reported wrong.Quote:
Originally Posted by Lead Head
http://arstechnica.com/news.ars/post/20060622-7117.html
FWIW Intel was able to get a single transistor to 20Ghz or so back in the PIII days.
Yea, it seems everyone pulls that crap from time to time in the tech. biz. Hopefully this isn't one of those times, but either which way none of us can know for sure at this point of time, just premature to dismiss what they're saying out of hand to me.Quote:
Originally Posted by savantu
AFAIK all the reports saying it was a sched. Q1 release were wrong. David Kanter had the original news on K8L and from what he has said it was always supposed to be out in Q2.Quote:
Originally Posted by savantu
http://www.realworldtech.com/page.cf...0206035626&p=3
Dailytech is the one who keeps saying Q4 2007 or Q1 2008 for K8L, AFAIK AMD is still saying it'll be Q2 2007.
The article is from Jan. 2006, if they started then they'd have all the equip. in place by Sept. 2006. Just enough time to have some chips out before 2007. Certainly low volume production at first but AMD never said they'd have volume .65 in late 2006. Sounds consistent with what AMD have said so far and during thier Tech Analyst Day in June of this year. They've got some good .pdf slides on thier website.Quote:
Originally Posted by savantu
I don't think anyone expects them to be shipping 4Ghz+ A64's on their .65 process in Q1 2007, but you'd think they'd be able to do at least a little better than this... Don't they have some sort of new SOI tech. for .65?Quote:
Originally Posted by savantu
Well according to this article we're both wrong, apparently they've got 5 or 6 design teams and this was back in 2004:Quote:
Originally Posted by savantu
http://news.com.com/Critical+Mass.+f...3-5371721.html
Things might've changed since then, but point is we don't know, and they've done alot with what they've got so far so don't count them out just yet.
I could've _sworn_ I read somewhere they'd already taped out on K8L on Aces in late June hence the die shots, that'd line up nicely with a late Q2 2007 launch...
From reading this thread, I have to agree a bit on what Vapor had said.
Lets keep this to a fair and solid argument guys.
Time to put in my $.02 here.
This is what I think. AMD's main strategy at this point is to "revolutionize" the approach again. They are probably doing this by focusing on different parts of the CPU, such as implement Hypertransport 3.0 and many other things.
Instead of focusing on "lower watts and high performance", it seems to me that AMD is revolutionizing their design yet again. I guess in 2007, this is when we will find out.
AMD can take the hit from Intel... we will just have to wait for their response.
I think the only thing really revolutionary that they're trying to do is make FPGA's or at least DSP-like processor support available for everyone via the HTX slot. Application and OS support is a must of course but being able to use something similar to Cell's general purpose computing power on any old PC would be impressive IMO and could spur development as it'll offer developers access to ~10x or so performance of the CPUs available at that time (on a single core basis of course).
I think they plan on putting something like that on die or at least on package at some point via MCM to but that is all so far off into the future we can only make WAG's at this point in time.
Everything else they do could be considered generic really, which is thier strength IMO.
HT 3 will really only matter for servers with 4 or more sockets, no one in the desktop or even enthusiast market uses systems like that in any volume worth mentioning.
K8L seems to have alot of power saving features to me (independant vCore, adjustable power for the on die memory controller too) and looking at the latest roadmaps they seem to be trying to get more 65w and 89w TDP chips out there too. Only the FX line and perhaps some of the higher clocked Opterons will be in the 125w TDP range. I'd say they're doing the best they can to compete with Intel's power vs. performance approach.
Somethings make me laugh. Duo2 isn't out YET... that means it doesn't matter if it can reach 5ghz. No normal person can buy it and if you do the jokes on you. Yet I wouldn't buy a flawed cpu just yet. And thats a real swift kick in the duo's for ya. lol That will be solved with a new stepping Ah?
How many ppl will get the flawed ones needing recall? Woodchest has even worse flaws. Duo 2 still has alot to prove in the real world as a real production product. Not as a proto-type. Prove this cpu is so great when you can buy it at a store when its really released.
Wow that really hurts... Top goverment of the world rejects the Duo2... "Is speachless" (Thinks: isn't surprising!)
http://www.theinquirer.net/default.aspx?article=32842
http://www.theinquirer.net/default.aspx?article=32818
http://sharikou.blogspot.com/2006/07...stability.html
That sure is a beautiful cpu filled with so many flaws... would seriously hurt my performance. Oh and its more then just a raid problem Duo2 is having here. AM2 all the way for me.
The only ones out ATM AFAIK are all ES CPUs so the bug is a non-issue...
I would love to see those roadmaps. I would love to see the roadmaps with K8L in Q1 and with K8L in Q2.Quote:
Originally Posted by savantu
This is the latest AMD roadmap, and no K8L till Q4 2K7!
http://images.dailytech.com/nimage/2...e_amd_q307.jpg
Google is your friend...:stick:Quote:
Originally Posted by DoubleZero
http://www.hkepc.com/hwdb/am2-4800-1.htm
http://www.digitimes.com/mobos/a20060327A5019.html
You know, rumors, speculations, bla bla bla yada yada yada from tabloids are more reliable than AMD direct statements...Quote:
Originally Posted by DoubleZero
K8L isn't coming before H2 2007.Quote:
Originally Posted by PetNorth
Already the most recent roadmaps show AMD will miss the December launch schedule for Brisbane 65nm CPUs. All this because converting a fab from 90nm to 65nm takes time and resources.
Perkam
Of course. Like I've said before several times, AMD has stated mid 2007. This is the only official timeframe. Mid 2007 is an aprox timeframe, it isn't an exact month, so it may be june, july... who cares really a month earlier or later.
and money :DQuote:
Originally Posted by perkam