Wednesday, December 13, 2006

AMD X2 4600+ will replace X2 3800+

AMD introduced the X2 5400+, and X2 5600+, the latter is basically the FX62 at lower power consumption. I am surprised that AMD keeps 90nm processors at the top end. It may have done so deliberately, so FAB30 remains a valuable asset. Now the spectrum looks too colorful, and AMD may eliminate the X2 3800+ to X2 4400+ soon, leaving X2 4600+ as the lowest dual core, which is more than enough to frag the Conroe E6300.

So we would have x2 4600+ at $150, x2 5000+ at $200, X2 5400 at $300 and x2 5600 at $400.

I also expect AMD to pump up frequency with its 65nm parts.

AMD's single core needs some major clock upgrade. Its mobile single core has a top frequency of 2.6GHZ and desktop single core stuck at 2.4GHZ. AMD should definitely introduce some high speed single core desktop chips.

25 Comments:

Anonymous Anonymous said...

I am surprised that AMD keeps 90nm processors at the top end.

Amd have already done that with the 90nm where 130nm where higher clocked initially. No surprise for me.

It may have done so deliberately, so FAB30 remains a valuable asset.

I think it’s deliberated but has nothing to do with the fab, maybe because they are still optimizing 65nm, or don’t want the 90nm get crushed by the 65 nm parts, specially the higher end parts where the difference should be higher.

Now the spectrum looks too colorful, and AMD may eliminated the X2 3800+ to X2 4400+ soon, leaving X2 4600+ as the lowest dual core, which is more than enough to frag the Conroe E6300.

Completely wrong! You are not aware of Intel road map. Intel is going to released new low end conroe based processors. The E4xxx and E2xxx line. These lines are inferior products that will have to compete with X2 3800+. They will also be the first native 2MB L2 cache cores and not half disabled cache like the E6300 and E6400.

I also expect AMD to pump up frequency with its 65nm parts.

Oh really?

AMD's single core needs some major clock upgrade. Its mobile single core has a top frequency of 2.6GHZ and desktop single core stuck at 2.4GHZ.

Let’s wait what Intel will do with the E2xxx line (Core 2 Duo with one core disabled). If its fast AMD will release faster parts, if not it will stay the same. We need AMD fighting with Intel because right now Intel single core P4 suck vs any AMD offering even at just 2.4Ghz.
New Egg is selling FX55 at just $124.10 (Open Box). That’s faster than the 2.4Ghz you where talking about.

2:30 AM, December 13, 2006  
Anonymous Anonymous said...

Your so right.. AMD appears at the moment to be treading water..hopefully the 65nm can overclock.?

2:50 AM, December 13, 2006  
Anonymous Anonymous said...

"I also expect AMD to pump up frequency with its 65nm parts."

Sure, it will bump them but so wil it pump it's 90nm speeds. Latest roadmaps I've seen 90nm will be the top dualcore CPU until K8L arrives and on plain K8 product line 90nm CPU's will be higher clocked than 65nm ones at least until the end of 2007.

3:49 AM, December 13, 2006  
Anonymous Anonymous said...

There's a lot involved with spinning an established or creating a new design in a the next technology. Sore spots like I/O's and Array's are notorious for not scaling with the expected speed up. You remind me of a VP that knows a lot about business but a little technology. What if I told you that the NMOS's gain 2% performance from 90nm to 65nm but the PMOS's gained 15%? Do you still think a frequency bump is a feasible goal? It's not true, of course (not yet :-) ) but do you honestly think AMD hasn't already thought of running their chips at a faster frequency? Do you think Hector is telling his Senior Fellows, "MAN! HOW CAN WE BE SO STUPID? HOW COME WE HAVEN'T THOUGHT OF RUNNING OUR CHIPS AT A FASTER FREQUENCY?!!??!?"

Nothing against you Sharikou... ok... everything against you including my kitchen sink. I wish you could get some background information on these ideas you propose. Granted - they are great commentaries on what companies *could* be doing. But sometimes they're just not grounded in reality.

Still - great blog. Keep the posts coming.

6:55 AM, December 13, 2006  
Anonymous Anonymous said...

Hey! Where have you been Sharikou? We've missed you - with every bullet so far! ;)

7:17 AM, December 13, 2006  
Anonymous Anonymous said...

I am surprised that AMD keeps 90nm processors at the top end

I suspect once AMD mature their 65nm process they will move to the faster chips. For now I suspect they get better yields with the slower chips.

AMD may eliminate the X2 3800+ to X2 4400+ soon, leaving X2 4600+ as the lowest dual core

AMD are releasing a 4000+ and 4400+ on 65nm (not sure about 3800) so the 4600+ will not be the lowest dual speed core especially as the prices look like they are not changing.

leaving X2 4600+ as the lowest dual core, which is more than enough to frag the Conroe E6300

The E6300 is not that far behind 4600+ performance (comparable to 4400+) but will probably overclock higher especially until AMD's 65nm process improves. And as mentioned before the 4600+ will not be the lowest 65nm CPU and will also be more expensive than than the E6300.

I also expect AMD to pump up frequency with its 65nm parts.

Obviously, only question is how far up and how fast.

7:33 AM, December 13, 2006  
Anonymous Anonymous said...

Sharikou said...

"Now the spectrum looks too colorful, and AMD may eliminate the X2 3800+ to X2 4400+ soon, leaving X2 4600+ as the lowest dual core, which is more than enough to frag the Conroe E6300."

The X2 4600+ ($220) is not up against the E6300 ($180), it is up against the E6400 ($220).

* all prices from Newegg *

----------------------------------

PS:The only reason that is remotely possible is thanks to Intel, and the price war, here is an example...

The week prior to the release of E6300/E6400 the price of the X2 4600+ was $547, the week after the launch the price was $243, and now its about $220.

AMD fans and Intel fans should be thanking Intel.

Like them or not, they made processors cheaper.

9:09 AM, December 13, 2006  
Anonymous Anonymous said...

Bumping single core frequencies might be useful for emerging markets, but not for the US market and other industrial countries.

It is about time AMD used Z-ram.

Also instead of removing X2 3800 and X2 4200+, AMD should lower their prices for these chips once they hit 65nm. X2 3800+ at 99$ and the X2 4200+ at 135$.

At the 99$ price for a true dual core hopefully single cores will go extinct in 18-24months.

I think the 65nm Turions X2 will really shine and provide amazing battery life. I hope AMD introduces them soon.

9:09 AM, December 13, 2006  
Anonymous Anonymous said...

Actually, keeping high end processors at 90nm have a lot of sense. My guess is that AMD will not introduce high end parts on 65 nm till the crossover.

10:57 AM, December 13, 2006  
Anonymous Anonymous said...

I'm betting that AMD is keeping it's high end CPUs at 90 nm for now because that process is much more mature, so their yields are better at those higher frequencies.

65 nm makes plenty of sense for the mainstream, as that is where volume is the most important, and volume is AMD's greatest weakness.

fxyefx

11:43 AM, December 13, 2006  
Anonymous Anonymous said...

Does any of this single core stuff really matter?

These low end processors are destined to run pirated versions of Windows in third world countries.

What will turn out to be a bust is 2P quad-core. There is no mainstream OS that handles threads well enough to put 8 threads on the road and keep the machine going full speed.

Some will say that 'virtualization' will be the salvation of 4+ core processors, but again no OS has good virtualization features either.

So the short term winner will be Core 2 Duo and its Xeon equivalent that give you superior performance. We can only hope that quadcoritis pushes down the cost of dual core.

12:11 PM, December 13, 2006  
Anonymous Anonymous said...

"AMD introduced the X2 5400+, and X2 5600+, the latter is basically the FX62 at lower power consumption."

So why is the 5600+ cost >$200 LESS than the FX62? AMD's pricing has become ridiculous - they are gouging an additional $200 for a same spec process (FX62) which uses more power?

They are also gouging the AM2 customers as the price of a single AM2 chip is less than the price/chip of a 4x4 (processor) - which one would think would cost more due to the additional HT links.

And if you new anything about AMD's Si processing you would not be surprised that the high end processors are on 90nm - the transistor right now is basically a dumb shrink of the 90nm and given the process is far less mature, bin splits are probably no where near as healthy as 90nm. Over time this will change but as you continue to laud the "continuous improvement" process of AMD - what this means is they start the technology ramp not much better than 90nm, but will over time get to their target improvement (probably take at least another year or so). Intel puts the bulk of the improvements in before ramp - a different approach - each has its pros and cons...

Or do you somehow think AMD produces these chips that have bigger die area on 90nm just to show that they can? Why would you want to make more of them on 65nm if you can make fewer on 90nm and remain capacity constrained - makes perfect sense to me! It should be rather obvious that there are some limitations (either 65nm yield, binsplits, or issues with the quick shrink) - these will likely be overcome over time.

2:56 PM, December 13, 2006  
Anonymous Anonymous said...

when are you going to retract this story

5:18 PM, December 13, 2006  
Anonymous Anonymous said...

Umm, are you kidding about 4600+ "fragging" the E6300? It simply doesn't.

It's marginally faster in some areas, marginally slower in others.
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2802&p=9 (note that even at stock speed the E6300 wins several benchmarks).

And, of course, it trounces the 4600+ if you throw overclocking into the mix - it's not even close.

Please try to keep the facts straight.

9:26 PM, December 13, 2006  
Anonymous Anonymous said...

"So we would have x2 4600+ at $150, x2 5000+ at $200, X2 5400 at $300 and x2 5600 at $400"

Source?

Your imagination I guess.

10:55 PM, December 13, 2006  
Anonymous Anonymous said...

The week prior to the release of E6300/E6400 the price of the X2 4600+ was $547, the week after the launch the price was $243, and now its about $220.

Does the fact that the proce dropped $327 mean that AMD was milking customers for the most it could ? surely not...

11:50 PM, December 13, 2006  
Anonymous Anonymous said...

"It may have done so deliberately, so FAB30 remains a valuable asset."

Yeah this makes a lot of f%^&#& sense - why hurry the transition to 300mm AND 65nm (F30 --> F38), when you can make high end parts on larger dies, on 200mm wafers, with higher power consumption! Really don't want to migrate those to 300mm/65nm...

In fact it is such a valuable asset they probably should NEVER convert it... (in this bizarro world the fab will become even more and more of a valuable asset...)

"So we would have x2 4600+ at $150, x2 5000+ at $200, X2 5400 at $300 and x2 5600 at $400."

Did I misread the official AMD prices? Are you now sales/marketing for AMD?

"AMD's single core needs some major clock upgrade. Its mobile single core has a top frequency of 2.6GHZ and desktop single core stuck at 2.4GHZ. AMD should definitely introduce some high speed single core desktop chips."

Major speed upgrade, not on K8 - that architecture has run out of clockspeed headroom... unless AMD's planning a single core version of K8L I wouldn't expect to see any MAJOR speed upgrades on single core Plus the people (in general) buying single core chips do not need the speed (the market that needs/wants the clockspeed will buy the dual core chips)

It is abundantly clear you have even less an understanding of the market (as a whole) than your technology understanding.

1:22 AM, December 14, 2006  
Blogger Roborat, Ph.D said...

Is "barely keeping up" the code word for "fragging" in this site?

I'm confused because its been consistently used to describe AMD's processors to it's Intel counterpart. i.e., "4x4 will frag Kenstfield", etc.

2:44 AM, December 14, 2006  
Anonymous Anonymous said...

Does any of this single core stuff really matter?

These low end processors are destined to run pirated versions of Windows in third world countries.


You can keep that piece of American trash right there with you in the US.

We will use AMD (which does not have the American attitude) processors to run Linux, BSD or Solaris operating systems thank you very much.

8:54 PM, December 14, 2006  
Anonymous Anonymous said...

"We will use AMD (which does not have the American attitude) processors to run Linux, BSD or Solaris operating systems thank you very much."

Aren't both BSD and Solaris (originally) American, right? That leaves you only linux to use... ;-)

1:17 AM, December 15, 2006  
Anonymous Anonymous said...

"The week prior to the release of E6300/E6400 the price of the X2 4600+ was $547."

Apparently you don't understand the computer/cpu market in general.

Before Conroe, Intel's P-D 960 was close to $600; after Conroe, it's just over $300. Intel is the one who decides to massively cut its price margin, and AMD is the one who has to respond likewise. Customers (for now) are the ones who benefit.

But Why does Intel do so? Because AMD was not winning market share? Because it can make more Conroe than demand? Because it does not mind AMD disrupting its monopolistic tactics?

11:16 AM, December 16, 2006  
Anonymous Anonymous said...

"But Why does Intel do so? "

Maybe to clear inventory of an older chip? (with side benefit of forcing AMD to cut prices in turn to)

It's not just simply about market share fanboy. If it was, why was it done at the same time as Conroe? Just a coincidence?!?! Why not a quarter earlier? Or later?

What was Intel going to do? Price the P4 and Core2 chips at the same price or introduce Core2 at >$1500.

AMD didn't have to respond - if their product was superior they could charge more for it. Obviously there is no justification for charging more for the X2's than Core 2...that is why they responded. The only way they could get away with charging more (since you seem to have a fundamental understanding of economics) is if there brand was worth something over Intel - which at this point in time is obviously not the case in desktop space.

2:40 PM, December 17, 2006  
Anonymous Anonymous said...

"AMD didn't have to respond - if their product was superior they could charge more for it. "

Bullshit. The reason of AMD having to respond is precisely that monopoly is evil. For 2.5 years AMD's chips are clearly superior yet they could not charge more for them.

12:18 PM, December 18, 2006  
Anonymous Anonymous said...

"The only way they could get away with charging more (since you seem to have a fundamental understanding of economics) is if there brand was worth something over Intel - which at this point in time is obviously not the case in desktop space."

Their "brand" worth something? You talk about computers like fashion products, and people use computers like they wear perfumes. Buying too much Apple lately, uh? (I have friends keep saying Apple has decent notebook while he keeps sending his Macintosh to repair.)

And if you think AMD charge more on the server side because of the "Opteron" brand, you're wrong. The big-server and supercomputer guys don't care brands; they care about scalability, reliability, balanced tradeoff, and efficiency.

12:23 PM, December 18, 2006  
Anonymous Anonymous said...

Bullshit. The reason of AMD having to respond is precisely that monopoly is evil. For 2.5 years AMD's chips are clearly superior yet they could not charge more for them.

AMD was quite happy to charge $300 US for their cheapest dual core, and upwards of $600 US for a part like the 4800+. Now you complain that Intel releases faster parts at lower price points? What's wrong with that? Cheaper and faster processors are always good for consumers.

7:38 PM, December 19, 2006  

Post a Comment

<< Home