drjohn,

Remeber that statement about a year from now when intel is ramping 65 nm in volume, dual core celeron's will probably be equivilant to today's 820.

Just like those 64 bit Celerons that Intel has been selling since last August, according to you...

Joe

Tenchu,

Then AMD will have to drastically reduce their prices on dual-core.

As I said, AMD may want to introduce lower speed grades, such as:

4000 X2 = 2.0 GHz + 2 x 1MB
3800 X2 = 2.0 GHz + 2 x 512K = 2 x Athlon 64 3200
3400 X2 = 1.8 GHz + 2 x 512K = 2 x Athlon 64 3000

You see how far down in performance AMD needs to go, and still not be as low as Pentium 820D

Joe

Keith,

The end of the dual-core in 2005 hype?

But Elmer said...
And wbmw said...
And...

And Ediot

This was supposed to be a silver bullet for Intel. If the story about the unit volume is true (and we already know the pathetic performance story), silver bullet will be a dud.

Too bad we don't have Jerry to say that ... <g>

Joe

Keith,

One Million Dual-Core Pentium Ds To Ship By Year's End

LOL.

For all the noise, Intel is planning to ship less than 1% of its processor shipments as dual core.

I would not be surprised if AMD beat that 1 million target - handily.

Joe

Nas,

Do you have the latest BIOS:

Here is the FTP site for Deluxe:
ftp://asus.com/pub/ASUS/mb/socket939/A8N-SLI%20Deluxe/

Regular:
ftp://asus.com/pub/ASUS/mb/socket939/A8N-SLI/

(I can't seem to get there from the regular site, it seems to be down).

To get answers, here is a good place:
http://www.nforcershq.com/forum/

Just search for your model number.

Here is another:
http://forums.pcper.com/

Joe

mas,

Nah that's got to be worth $299 I would say.

AMD would only ship those "for review", so the price would be irrelevant.

Joe

wbmw,

AMD can always lower the price and sacrifice manufacturing capacity for more dual core sales. It would probably help them to gain in market share, at the expense of earnings. Wasn't that their previous strategy?

I am not sure I understand your point. The market is, and will continue to be mostly single core.

As far as performance and pricing, AMD seems to be limiting the speed grades in a way that the slowest AMD processor roughly matches the fastest Intel processor. To have parity, AMD would need to price either the lowest speed model 4200, or the second slowest - 4400 at $999, and the 4600, 4800 above that.

So AMD is already prividing better value at overlapping performance levels. What's missing is lower speed grades to go down to the level of 2800 x 2. I don't even know what clock speed it would require, probably 1.8 GHz to match that, and price it at similar price.

Joe

Keith,

it is a shame that X2 gets a lot of "negative" press due to its pricing

Maybe AMD should, just for fun, release a 3800 X2, 2x512k, 2GHz, price it at $240 and have it on a permanent backorder. It would still most likely mop the floor with 820, and be cheaper. <g>

For AMD, they will have to watch the market closely and try their best to maximize profit from available capacity. It should be an interesting H2 this year. For now, AMD´s strategy looks right to me.

Agreed, assuming they ammend it with the model above.

Joe

"Our desktop dual-core AMD Athlon 64 X2 products are priced based on performance," said Cathy Abbinanti, an AMD spokeswoman, when asked for comment on the pricing discrepancy. "Based on the performance benchmark information in recent third-party reviews of the competition's dual-core desktop product, we believe our lowest performing AMD Athlon 64 X2 dual-core processor will outperform Intel's highest performing dual-core part."

That's true, according to ExtremeTech tests. However, it may end up being a moot point.

I see. When AMD mops the floor with Intel in performance, as in dual core processors, it is a moot point.

Joe

NaS,

I find it hard to believe that CnQ would work in SP1 and not in SP2.

The standard way to enable CnQ is:
- make sure BIOS supports it
- if BIOS has a setting to turn it on, do so
- download the latest processor driver from AMD web site
- in device manager, locate the processor driver, select Update Driver, and update to the latest processor driver you downloaded
- in Control Panel, Power options, set power scheme to Minimal Power management

Joe

Windsock,

When you are caught telling bending the truth, keep on defending the indefensible.

LOL. Coming from you, it means sooooo much.

Joe

K,

I've heard a claim Celeron-M lacks support for Speedstep. Is that so?

Speedstep, is obsolete. I think they call the version where processor speed depends on processing demand Enhanced SpeedStep. Banias was the first processor to have that functionality, so Celeron M should be automatically capable of it, unless it has been crippled to disable it.

Joe

chipdesigner,

And yet, AMD CPG revenues improved in Q1 during this "problem", while Intel's fell.

CPG revenue improvements would have been even better without the problems with AMD based retail desktops and notebooks from HP. Isn't that what we want?

Joe

Keith,

wasn't there some quote about Sony possibly using AMD again(may have been Turion)?

Samsung and IBM/Lenovo - it may be possible, but I would not expect anything any time soon.

Toshiba and Dell are lost causes for now, IMO.

Joe

Keith,

I think the likelihood of using AMD notebook chips is good for:
Sony
Samsung
IBM/Lenovo
(probably in that order)

Highly unlikely are
Toshiba
Dell

Toshiba has in the past rather gone out of desktop PC business than use AMD.

Joe

Tenchu,

Count me under virtualization sceptics, except for niches in server market, posibly software developers.

Joe

wbmw,

Everything that's 32-bit x86 today will be 64-bit x86 in the next 2-4 years, no matter if it's high end or low end, mainstream or niche. That's a 50+M market per quarter, so there's no question that it will vastly outpace IPF.

I was thinking roughly 25M for high end of the market, 25M for Celeron and Sempron that don't have 64 bit capability yet (soon I guess).

On the other hand, the market has been porting software to architectures for years with small fractions of this size market, so I don't agree with the argument that software support for IPF will slow down due to x86 supporting 64-bits.

I didn't say that, but chipguy seems to thing there is more software for IA64 than AMD64. I don't know if it is true, but my point is that the ratio will change dramatically to favor AMD64, by a large margin in a year or 2. The ratio of the installed base and current sales will dictate that.

But of course it is not just hardware that counts. It is hardware running 64 bit OS, which will be lagging the hardware installed base for some time, but for the software developers, it is where their future sales will come from...

Joe

wbmw,

So how long do you think it will take before AMD captures 50% of the x86 CPU market? Does "soon" mean in the next 2 years? It might be funny to revisit your prediction then.

I am talking Opteron, Athlon 64, Pentium 6xx, Pentium 8xx, Nocona, Irvingdale(?), Potomac combined.

edit: and some Pentium 5xx apparently.

Joe

wbmw,

Not that it's hard to score a big chunk of the virtualization market (it's a small market), so AMD may end up with just as many design wins eventually.

I am just trying to understand this. What do you mean by design wins? Isn't this a question of software (OS and software like VirtualPC) supporting it?

Joe

chipdesigner,

In 2006, it will probably approach 1000x.

Actually, you are right. I think I lost a zero somewhere along the way.

Itanium may be shipping at rate of 25,000 units per quarter (WAG). AMD64 may soon approach 25,000,000 per quarter, which would be 1,000x, as you pointed out, not 100x.

AMD64 is well over 100x quarterly unit shipment already.

Joe

chipdesigner,

The install base of AMD64-compatible systems is currently on the order of 100x that of IPF systems, and growing rapidly.

Installed base may not be there yet, if you count the frebie Itaniums sitting in the corner gathering dust, that have been given out in years of pilot production, but ongoing sales are moving in that direction.

It is interesting that chipguy wants to talk about software for Itanium,, #1 reason for Itanium failure in the market, and one category where AMD64 will surely prevail (in #s of apps, cost, availability, quality, being up-to-date, etc).

Joe

chipguy,

BTW, at last account there are 3500+ server applications
shipping for IPF. How many native AMD64 apps are there?
Besides video games, I mean.

I don't know you want to go that way... <g>

Joe

Dan,

AMD now has a very rapidly growing share of the notebook segment as well as a rapidly growing share of the corporate desktop, workstation, blade, server, and storage segments.

AMD may be growing in the markets you mentioned (somewhat, not rapidly) except notebook. That one still is in the "potential" category. AMD has a potential to grow market there, when Turion based notebooks finally make it to market.

Joe

chipguy,

The comment you apparently would rather not acknowledge:

"IBM's problems may have to do as much (or more) with the PPC design as with the process technology."

Well, I posted it, so why would I rather not acknowledge it?

PPC is still lagging Intel and AMD in performance, even after resorting to exotic cooling solutions. I don't know enough to know what the bottleneck is, what stops it from scaling higher in clock speeds, or why IBM has not been more aggressive in squeezing higher IPCs, if the higher clock speeds are a problem.

K8 and Dothan are achieving higher IPCs running in the same clock speed ballpark (without water cooling). And, BTW, wasn't the RISCy nature of Power supposed to lead to easier clock speed scaling? This would point to process problem.

On the other hand, they are targetting a variant of PPC for Cell to run at 3.2 GHz, apparently using the same 90nm process technology with much lower power comsumption. So who knows.

What's your take anyway?

Joe

Phil,

IBM at the same time sells semiconductors and buys them. What they sell doesn't seem to be high enough volume and high enough ASP for their microelectronics division to be consistently profitable.

At the same time, IBM is paying top dollar to Intel for a large volume of microprocessors, and is even promoting them. As a result, IBMs semiconductor division is losing money, their PC division lost money, their server division Power is a loss leader to sell a bunch of hardware accessories, software, suppor and consulting.

So IBM makes nothing on hardware, gives all the profits from hardware to Intel, and IBM is even promoting lines where all the profit goes to Intel. And Intel, after all is a competitor. Interesting way to run a business, where the objective seems to be to give most if not all of the profits to a competitor.

Joe

chipguy,

Care to explain this convenient theory?

Which part, if IBM is still leading edge in process technology (I know the answer, and I don't know if there is enough data, I guess IBM still is) or if there is anyone better for leading edge technology for logic devices besides IBM for AMD to partner with?

Joe

Phil,

re: IBM

I am not so sure about this !
Judging from the negative press in recent months - I have the impression that IBM is not the leading-edge in chip-production.
( which makes me wonder what AMD is doing with them - apart from paying large bills )

Doing your own process development, having a development fab is a lot more expensive than what AMD is paying right now to IBM.

As far as IBM not being leading edge, I think they still are, and besides, there is no one else to partner with. IBM's problems may have to do as much (or more) with the PPC design as with the process technology.

Their other problem may be that a leading edge fab needs to be making high ASP products, and be nearly fully utilized. I don't know how much of the high ASP stuff IBM is making in high volume to pay for all the overhead associated with a leading edge manufacturer / foundry.

In their infinite wisdom, IBM executives give most of their high ASP semiconductor money to Intel in form of Xeon and Dothan...

Joe

Phil,

Is there a link to this ?

Firstly, Spansion lost a big chunk of market share, then, there is a move to smaller geometries with Mirrorbit (which is still a slow portion of overall total, and is lagging regular flash in terms of moving to 130, 110 and 90nm.

Then there is JV3, which is only partially equipped.

My WAG is that AMD could double the bit capacity several times over if it were to use latest generation of the process technology (110, 90nm) with Mirrorbit, and if JV3 was fully equipped.

But then, maybe Spansion does not want to put any money into JV3, and concentrate capital expenditure for conversion of JV3 + JV4 into combined SP1 fab with 300 mm wafers.

Joe

T64,

Is it possible that AMD converted Spansion space for manufacturing CPU's like maybe Socket 754 Semprons at 110 nm...can this be done?

No. K8 needs SOI and copper interconnects. I don't think flash is using copper yet, and definitely not SOI. Converting Austin or any part of JV/SP fabs into processor making fabs is out of the question.

First of all, AMD is barely selling out current capacity.
Second of all, Fab36 is less than a year away from commercial shipments.
Third, Chartered is closer
Forth, IBM could produce AMD CPUs tomorrow (if the decision was made sufficiently early) and AMD had plentiful demand at good ASPs.

Joe

chipguy,

The 970FX is about as far behind current x86 chips in
performance as the historical norm for the Mac vs PCs.
The big difference is the 970 isn't as far behind x86 in
clock rate.

Now when the world is finally convinced that the clock speed does not equal performance...

But anyway, I thought that IBM went from 2 GHz to 2.7 GHz, while in the PC world, the gain was smaller percentage-wise, while there were no great performance enhancing breakthroughs aiding IPC on the PC side (ignoring dual core for now).

Joe

Keith,

Taiwan's UMC seen winning Spansion order for NOR flash memory chips - report

It doesn't seem to make any sense. I thought Spansion had unused capacity.

At some point Spansion may bring out Ornand, but it is going to take a while before they get more orders for it than they can fill using existing capacity.

Also, the Spansion process is optimized for the flash they are making while UMC is not. So while capacity is not sold out, it makes no sense using unoptimized process rather than optimized one.

Joe

Tenchu,

By the way, I thought x86-64 increases the number of SSE registers to 16, no? Maybe Apple is looking into EM64T processors from Intel. Or AMD64 processors? Time to rekindle those Apple-AMD alliance hopes ...

It seems to me that right now, Apple is in less trouble with their processors than in the past. They now have a 64 bit processor, somewhat 64 bit OS, and on the desktop, the processor is not as far behind as in the past (if you ignore the upcoming AMD X2 processors for the moment).

Joe

Maybe they came make them for:
1st year: $500
2nd year: $425
3rd year: $375
4th year: $325
5th year: $275

Joe

chipguy,

Well it can be (i.e. fixed frequency voltage mode) but
what I actually meant was the Foxton *frequency boost*
might be disabled in the performance comprison. I.e. a
2 GHz Montecito wouldn't speed up to ~2.2 GHz when
running TpmC and SPECint2k or speed tp to ~2.1 GHz
when running SPECfp2k to take advantage of the thermal
headroom available for those workloads.

You are assuming there is thermal headroom when running those apps. I think the chances are better that there is one when running TpmC, but it may very well be that computationally intensive, sustained tasks will result in clock speed dropping below 2 GHz.

BTW, I think it is a great technology, and I hope AMD will implement it in the future (as an enhancement to PowerNow / Cool'n'Quiet.

BTW (#2), is this implemented inside the CPU alone, or does this need to work together with the OS?

Joe

chipguy,

4) Foxton enabled or disabled?

Can Foxton really be disabled? I think it can only be partially limited by placing the cap at the nominal clock speed of the chip (disabling the overclocking). But I doubt the clock speed reduction in order to enforce the TDP can be disabled.

Joe

imho,

With the exception of the ocasional inside info that leaks from an employee

Well, how do you beat, if the info is under NDA? Other publications / sites are under NDA, which prevents them from reporting on the info (or have no info). Or it is a nugget from somre industry research paper that othewise costs 10s of thousands of dollars to purchase.

the rags you talk about do nothing more than regurgitate what internet chat sites and newsgroups spout as "fact".

I don't think they do say it as a fact. They generally say where they read it and link to the source. You be the judge.

re: http://www.amazon.com/exec/obidos/ASIN/B00001ZWV7/indrasnet/104-9708445-3340719

So you are admitting you have nothing. Like Paul, you fail the challenge. You have not been able to provide a better source, while in the very same post you knock The Register / Inquirer some more.

I am looking forward to other Intel acolytes taking the challenge: http://www.investorshub.com/boards/read_msg.asp?message_id=6406740

Joe

Paul,

I think the 36,288 posts by Niceguy767

While opinions of other investors have some value (which is why this thread exists), I was thinking more in terms of news / rumor / gossip.

By the way - it's been a few years since you have been trumpeting the imminent success of Cyrix - care to explain why?

And Intel before that.

Can't win them all, but it helps when a bull market bails you out from a poor investment decision.

Joe

BTW, in your younger age, you seemed to have been able to follow the thread of conversation with ease. What happened?

chipguy,

Yeah I guess to *you* thereg is a reliable source. OTOH
the June 22, 1998 Microprocessor Report article about
Merced mentions expectations of 800 MHz for the device.

I have a source that is more credible than either one. It has never been wrong. It is a rock. The downside is, it never told me anything, but it has the least number of instances of being inaccurate.

Then, there are newspapers that tell me what happened yesterday. Good for checking yesterday stock prices, but a little stale.

Then there is The Register and The Inquirer. It tells me what is happenening and what is likely to happen that has not been revealed by official sources.

I challenge you to find me a place or a source that (today) is better for this type of information. I would welcome links to the sources that happen to not be silent as a rock on topics and issues that The Inquirer and The Register cover.

Joe

PS: This is a challenge to every Intel acolyte: When knocking The Register or the Inquierer, please provide the links of alternate sources that were more accurate, and just as timely.

Buggi,

Thanks

Joe

Keith,

a giveaway is not a sale, so it is unlikely to be included in the "ASP". It would however be included in the shipment numbers.

And it would be included in Itanium system revenue numbers, the only "hard" numbers there are, AFAIK.

Joe