Correct me as you see fit but..

[darkdemyze]

From the things that I have been reading, it looks like AMD is messing things up by trying to make things simpler.

It's bad enough they released a platform that offered practically no performance boost over their previous offerings, but they wanted to get on the DDR2 boat - understood. It also would help "unify" things a bit by having budget, mainstream, and enthusiest parts all on one socket.

Yet, now they are killing off the 1MBx2 parts:
http://www.dailytech.com/article.aspx?newsid=2858
Which I would see as their way of recoupping their losses they're going to recieve from the massive price cuts, that I'm sure everyone who will be reading this knows about by now, after Conroe.

On top of that, they kill off half of their CPU line-up by december:
http://www.dailytech.com/article.aspx?newsid=2974

Plus Socket F is going to be sporting 1207pin LGA, meaning a larger package, upping the cost-per-proc, even with the 65nm process.

So it seems to me that by getting rid of nearly half their desktop offerings and virtually diminishing 939 support by Q207 ( about the time when Conroe will be much more widely avalible ), when people goto build a new system, why would anyone want to get a lower-performing AM2 proc when they could get a Conroe for what looks to be lower cost and more power efficiant. At which time, Intel will also be gearing up for Penryn.

What I'm trying to say is if AMD cutting support for the platform that most of their chips are selling on ( which can be seen in the hype on any forum about price cuts ); therefore losing profit on the chips that ARE selling, and encouraging people to change platform, you would think in order for them to get people to buy their new platform they would have something that performs at least on par with the competition's right? Which is set back even further by killing off 1mbx2 procs.

And is it just me, or is the new Opteron naming system stupid or what?

Comments?

 

[Mogadon]

Originally posted by: darkdemyze
So it seems to me that by getting rid of nearly half their desktop offerings and virtually diminishing 939 support by Q207 ( about the time when Conroe will be much more widely avalible )

Yeh, but by then we should / may have 65nm AMD chips.

 

[broly8877]

I think AMD deserves a beating from Intel. (just like Intel deserved a beating from the K8)

They got cocky, they should've known Intel would eventually strike, and strike hard. Since 2002 (when the K8 was supposedly ready), they've stagnated. A die shrink there, a few additions here, over all a mere 600MHz increase in frequency without any major changes.

They could've been a step ahead (again) by having an overhauled K8 (K8L) core for the AM2 launch.

 

[Ika]

Originally posted by: broly8877
I think AMD deserves a beating from Intel. (just like Intel deserved a beating from the K8)

They got cocky, they should've known Intel would eventually strike, and strike hard. Since 2002 (when the K8 was supposedly ready), they've stagnated. A die shrink there, a few additions here, over all a mere 600MHz increase in frequency without any major changes.

They could've been a step ahead (again) by having an overhauled K8 (K8L) core for the AM2 launch.

QFT. definitely true. I do wonder what AMD was thinking these past few years...

 

[Griswold]

Originally posted by: darkdemyze
Plus Socket F is going to be sporting 1207pin LGA, meaning a larger package, upping the cost-per-proc, even with the 65nm process.

Probably because they need it? Like, quad cores on same socket? The added cost for bigger package is negligible especially in the server segment its aimed at.

So it seems to me that by getting rid of nearly half their desktop offerings and virtually diminishing 939 support by Q207 ( about the time when Conroe will be much more widely avalible ), when people goto build a new system, why would anyone want to get a lower-performing AM2 proc when they could get a Conroe for what looks to be lower cost and more power efficiant. At which time, Intel will also be gearing up for Penryn.

How much do you know about 65nm AM2 parts? Not much. Like everyone else. However, since AMD claims their 65nm quad core opterons will eat as much (or little) power as a dual core, the performance-watt-price ratio for desktop parts based on AM2 and 65nm could very well be much more competitive than it is now (compared to the still not launched core2). You dont have to have the better performing part to make profit. Intel showed that over a long time in a impressive way.

What I'm trying to say is if AMD cutting support for the platform that most of their chips are selling on ( which can be seen in the hype on any forum about price cuts ); therefore losing profit on the chips that ARE selling, and encouraging people to change platform, you would think in order for them to get people to buy their new platform they would have something that performs at least on par with the competition's right? Which is set back even further by killing off 1mbx2 procs.

You seem to think that most people only buy processors when they upgrade. In fact, the mass market sells the whole package and that includes the mobo and ram. So, from a business point of view, you're totally off.

And is it just me, or is the new Opteron naming system stupid or what?

Its just you.

 

[darkdemyze]

First of all, I'm not trying to bash AMD. If anything my post was to express my distaught from AMD's recent, and imposing slip-ups. So you can drop the fanboi crap defense right now.

Originally posted by: Griswold
Probably because they need it? Like, quad cores on same socket? The added cost for bigger package is negligible especially in the server segment its aimed at.

I wont disagree, I'm just pointing out another expense they are shelling out which I'm questioning how much they will get back. Or, better said, how much is it going to raise the prices of the processors? Which I understand like you say, it is aimed toward the server space no doubt, but that doesn't mean that if Intel has something better ( lower cost / performance:watt ), that everyone is going to still buy AMD.

How much do you know about 65nm AM2 parts? Not much. Like everyone else. However, since AMD claims their 65nm quad core opterons will eat as much (or little) power as a dual core, the performance-watt-price ratio for desktop parts based on AM2 and 65nm could very well be much more competitive than it is now (compared to the still not launched core2). You dont have to have the better performing part to make profit. Intel showed that over a long time in a impressive way.

"Claims" which are made against what? Un-official estimations. Besides, my point is you are not making a large profit if you are not selling chips. Although I know now AMD is on board with Dell for example, they obviously will not sell no chips. I'm just stating if Intel still has the upper hand in performance ( demand ), and they already obviously have the upper hand in manufacturing ( supply ), that seems to spell better B4YB, which will attract more customers when they decide to change platform, thus raising their market share for enthusiest, mainstream and budget alike.

You seem to think that most people only buy processors when they upgrade. In fact, the mass market sells the whole package and that includes the mobo and ram. So, from a business point of view, you're totally off.

Well no. Hence why I didn't use the word upgrade. I'm actually saying when someone decides to ditch their old system for an entirely new platform ( meaning "the whole package" ), they will look at the numbers and see Conroe is currently doing better and go with an Intel platform instead of AM2.

And is it just me, or is the new Opteron naming system stupid or what?

Its just you.

If you say so..


Honestly I like to see the competition though. When businesses war, the consumer wins. Here's to $170 3800+ OCed to FX-60 speeds :beer:

 

[Praxis1452]

the 2x1mb cache is a huge fricking hit on every piece. They need to cut them. +Performance is very very minimal. The k8 architecture doesn't really need a big cache. The performance is pretty minimal. Up the clock .2ghz is worth it over 512kb cache. So in fact you'll probably get more performance and they'll get more money at stock. THey need to ditch the cache to compete especially with the price cuts. See as how no 2x1mb cache's are getting cuts...

 

[blackllotus]

Originally posted by: broly8877
They got cocky, they should've known Intel would eventually strike, and strike hard. Since 2002 (when the K8 was supposedly ready), they've stagnated. A die shrink there, a few additions here, over all a mere 600MHz increase in frequency without any major changes.

This is an ignorant comment. Its clear that AMD hasn't just sat on their ass for the past few years. AMD and Intel spend a huge amount of resources on research and development, and they can't decide when their technology is useable. The micro-optimizations they made to K8 were to ensure that people would still buy their products. Focusing on K8 is not the way for AMD to make money in the long run.

 

[furballi]

AMD and Intel make the most $ selling CPUs to the average users in the public and private sectors. Power users generate very little income because they account for a very tiny portion of total sale!

The average +3000 class processor is more than adequate for most users. Most large corporations and retailers are still using sub-2GHz class processors because they are fast enough to do the job.

 

[darkdemyze]

Originally posted by: Praxis1452
the 2x1mb cache is a huge fricking hit on every piece. They need to cut them. +Performance is very very minimal. The k8 architecture doesn't really need a big cache. The performance is pretty minimal. Up the clock .2ghz is worth it over 512kb cache. So in fact you'll probably get more performance and they'll get more money at stock. THey need to ditch the cache to compete especially with the price cuts. See as how no 2x1mb cache's are getting cuts...

This is true, as is the nature of exclusive cache. Honestly I almost feel like AMD used 1MB L2 cache banking on the idea people would think "more is better." To tell the truth, I wish I had known that when I bought my 4400+. I was drawn in by the "omg, the 1mb cache is killer" posts all about forum threads accross the net. Not knowing back when I bought it the marginal performance gain was not worth the nominal price premium. But at this point I figure live and learn, then deal with the loss. Besides I've been happy with it anyway and I don't plan on upgrading anytime soon.

 

[broly8877]

Originally posted by: blackllotus

Originally posted by: broly8877
They got cocky, they should've known Intel would eventually strike, and strike hard. Since 2002 (when the K8 was supposedly ready), they've stagnated. A die shrink there, a few additions here, over all a mere 600MHz increase in frequency without any major changes.

This is an ignorant comment. Its clear that AMD hasn't just sat on their ass for the past few years. AMD and Intel spend a huge amount of resources on research and development, and they can't decide when their technology is useable.

Never said they did nothing, just that I believe they slacked off a bit in R&D.

I've heard developing the K8 took a good 6-8 years, and that's:
x86-64
Ondie memory controller
Implementing hyper transport
Improved branch prediction/global history counter/enhanced ICP/new package/ new chipsets needed/ etc

Pretty radical changes, and that's when AMD had less cash to throw around.

I'm sure AMD could've had some sort of K8L-esque CPU this year.

 

[dguy6789]

The jump from 939 to AM2 was around the same performance difference as going from 754 to 939, so I do not see why so many people are complaining. Even if AM2 does not provide a large benefit now, it probably will in the future. Quad core processors may take advantage of the extra memory bandwidth offered via DDR2 and eventually 3. AMD has a significant lead over Intel at the moment in terms of price vs performance in nearly every market except for the ultra low end dual cores. Conroe is probably superior to K8. This is to be expected. AMD will have an answer eventually.

Remember, the highest end does not really make either company much money. The main reason they make high end processors at all is so that the average Joe will say "I remember hearing that AMD has the fastest processors" and pick up a new AMD based system. By far the largest profit the companies make comes out of the value lines of processors. Athlon 64 3000+, Semprons, Celerons, Lower speed Pentium 4s and the lower end Pentium Ds are the most important cpus. All of the high end Conroe and FX processors are small niche markets and will remain that way for a while.

 

[Furen]

- AMD is simplifying its offerings, which is a great idea in my opinion. AMD has an insane amount of SKUs: currently there are 7 AM2 X2s, 4 AM2 A64s, 2 FXs (AM2/939), 5 s939 X2s, 7 s939 A64s, loads of Semprons, like 10 s939 Opterons, like 20 s940 Opterons and a few s754 A64s. So 60+ SKUs... and that's without counting all the mobile parts.

- A normal person doesn't understand the difference between an X2 4200 and a 4400, especially considering that the performance difference is pretty slim, and you'll still be able to buy the 100/1000 series Opterons if you really want 2x1MB L2 so simplification here will avoid OEMs lots of headaches.

- AMD WILL NOT BE TAKING ANY LOSSES from its price cuts, it'll just cut into its margins for X2s but it will also increse the demand for them massively, so they may end up breaking even. Dropping the average die-size for the Dual-core parts will help out a lot in the production side, though.

- Notice that Dailytech's chart doesn't have the Manchester X2s being EOLed, it's just the 1MB L2 parts and the low-end A64s (the 3500+ and 3800+ aren't mentioned either). I'd guess this means we'll still see S939 X2s for a while yet.

- The increased packaging cost for LGA1207 is very likely negligible and will likely not result in any significant price increases.

- I kind of liked AMD's new Opteron naming scheme since the numbers will give more information about the chip itself. For example, a 2.8GHz K8 will be an x220, a 2.8GHz K8-L could end up being x320 or an x420. If AMD plans on staying with Socket F for a while then this new "processor architecture" digit could be very useful.

AMD will likely be slightly behind in power consumption and performance but nothing too dramatic unless you plan on overclocking massively (Conroe does look to be the better overclocker... If I had to venture a guess I'd say that AMD's memory controller is the limiting factor here). Also, Hexus claims that AMD will launch the FX-64 on August 8 so the 10-15% performance advantage the XE version of Conroe has over the FX may shrink slightly.

 

[darkdemyze]

Maybe I should have clearified a bit on what I meant with the naming system, I think the first 2 numbers are great. Okay, insted of 1.8 dual proc being 2210, why not 2218. 2.6 8-way be 8226? It can just be confusing to someone who isn't in the know I guess. Personally to me and many others, this wont be a problem. But for others, the numbers may seem awkward as they don't exactly reflect performance other than "higher number means faster" which may change if they decide to raise those last 2 numbers according features or die / cache size. Seems like a problem AMD has had for quite some time, just my opinion I suppose..

 

[Regs]

You're exactly right dguy6789. Except on this forum, everybody wants performance now, now, and now. Including me. We're that 2% of the market that pays for 4000 dollar water-cooled SLI crap and those bleeding edge media freaks who encode everything from family vacations to 3 hour long feature films in high definition. Or mostly...to brag while we wait for the next best selling game to hit.

 

[Markbnj]

QFT. definitely true. I do wonder what AMD was thinking these past few years...

Well they did introduce the consumer world to dual core processors a year ago, and pretty much single-handedly changed the market for desktop and laptop systems.

"All right, but apart from the sanitation, medicine, education, wine, public order, irrigation, roads, the fresh water system and public health, what have the Romans ever done for us?"
"Brought peace?"
"Oh, peace - shut up! There is not one of us who would not gladly suffer death to rid this country of the Romans once and for all."
"Uh, well, one."
"Oh, yeah, yeah, there's one. But otherwise, we're solid."

 

[imported_ITguy]

Originally posted by: Aflac

Originally posted by: broly8877
I think AMD deserves a beating from Intel. (just like Intel deserved a beating from the K8)

They got cocky, they should've known Intel would eventually strike, and strike hard. Since 2002 (when the K8 was supposedly ready), they've stagnated. A die shrink there, a few additions here, over all a mere 600MHz increase in frequency without any major changes.

They could've been a step ahead (again) by having an overhauled K8 (K8L) core for the AM2 launch.

QFT. definitely true. I do wonder what AMD was thinking these past few years...

Maybe they were desperately trying to beat a massive monopoly that prevented their sales from growing eventhough they had unargueably the superior product for the industry for atleast 3 years and argueably much longer. In order to create new designs you need profits and revenues to pay expensive designers and AMD only just started making a profit last year eventhough they've had the superior product for many years and the only 64bit product as well. That's why it may seem like they were sitting on their ass but it is next to impossible to survive when one company is soo large they control every aspect of the industry, from what new standards are set, to what products venders will but and not buy.

Just a thought.

Every sale that goes to a monopolist, ensures that we will get slower performing processors for higher prices in the future. That's why we should give the benefit of the doubt to the little guy and support them. That will serve to ensure that competition will give us, the consumer the best products at the best prices in the future.

 

[zsdersw]

and the only 64bit product as well.

Wrong. Itanium is a 64bit product. And so too are the EM64T P4's and Celerons.

That's why we should give the benefit of the doubt to the little guy and support them

Ridiculous. Just because AMD is "the little guy" doesn't mean their purposes are any different from Intel's.. or that they're altruistic and Intel is greedy.. or that they're worthy of the pedestal that some on here (apparently including you) put them on.

Both AMD and Intel are out to make money. Both will do what they can to do so.

Neither are worthy of any "benefit of the doubt".

 

[SexyK]

Originally posted by: Markbnj

QFT. definitely true. I do wonder what AMD was thinking these past few years...

Well they did introduce the consumer world to dual core processors a year ago, and pretty much single-handedly changed the market for desktop and laptop systems.

"All right, but apart from the sanitation, medicine, education, wine, public order, irrigation, roads, the fresh water system and public health, what have the Romans ever done for us?"
"Brought peace?"
"Oh, peace - shut up! There is not one of us who would not gladly suffer death to rid this country of the Romans once and for all."
"Uh, well, one."
"Oh, yeah, yeah, there's one. But otherwise, we're solid."

Huh? Intel released the first consumer dual core CPU, so I'm not sure what you're getting at with that comment...

 

[Markbnj]

Huh? Intel released the first consumer dual core CPU, so I'm not sure what you're getting at with that comment...

That's true, but I think it was AMD that really suceeded with them in the market. 'Single-handedly" probably wasn't the best way to put it.

 

[Bateluer]

Originally posted by: SexyK



Huh? Intel released the first consumer dual core CPU, so I'm not sure what you're getting at with that comment...

Intel paper launched the first dual core chips. AMD's dual core chips were available on launch day and delivered the performance, whereas when Intel's finally showed up, they didn't.

Did any mention that there is almost zero difference in performance between the 512KB and the 1MB cache A64s? Buying a 1MB cache A62(x2) would be foolish as you are wasting the extra couple hundred dollars.

 

[Viditor]

Originally posted by: Aflac

Originally posted by: broly8877
I think AMD deserves a beating from Intel. (just like Intel deserved a beating from the K8)

They got cocky, they should've known Intel would eventually strike, and strike hard. Since 2002 (when the K8 was supposedly ready), they've stagnated. A die shrink there, a few additions here, over all a mere 600MHz increase in frequency without any major changes.

They could've been a step ahead (again) by having an overhauled K8 (K8L) core for the AM2 launch.

QFT. definitely true. I do wonder what AMD was thinking these past few years...

Guys, you really have to take a step back and understand just how looooooong it takes to get a chip to market (if it didn't take so long, do you really think Intel would have allowed AMD to bash their brains in for the last 3 years???).
From the beginning of design to finished product takes an average of ~5 years.
In the case of the Conroe core, it took even less (closer to 4 years)...but it cost them dearly. Intel cancelled or put on hold almost all of their other projects (like CSI) to get Conroe out early (Conroe was originally to be out next year).
I'm sure that Intel's accelerated program is what caught AMD by surprise, as K8L dual core is due out near the time when Conroe was originally expected to launch.
In the end, while Intel will have a very good 6 months (at least) as the performance king, they won't have huge volumes till next year...the price drop from AMD is a stop-gap measure to keep sales up against the PD/P4 firesale (which was the second whammy Intel dealt).
After all, whether you realise it or not we all look for price/performance...and since AMD's profit margin is now better than Intel's, they have some room to work with there.

Next year, we'll all be amazed at K8L...and after that I'm sure Intel will be touting their 45nm chips (which no doubt will be most excellent as well).

The good news for us all is that Intel can no longer supress AMD with their marketing, so we should see some really good innovation happening from here on out!

 

[darkdemyze]

Originally posted by: Bateluer

Did any mention that there is almost zero difference in performance between the 512KB and the 1MB cache A64s? Buying a 1MB cache A62(x2) would be foolish as you are wasting the extra couple hundred dollars.

Yes. And I know.

Originally posted by: darkdemyze

Originally posted by: Praxis1452
the 2x1mb cache is a huge fricking hit on every piece. They need to cut them. +Performance is very very minimal. The k8 architecture doesn't really need a big cache. The performance is pretty minimal. Up the clock .2ghz is worth it over 512kb cache. So in fact you'll probably get more performance and they'll get more money at stock. THey need to ditch the cache to compete especially with the price cuts. See as how no 2x1mb cache's are getting cuts...

This is true, as is the nature of exclusive cache. Honestly I almost feel like AMD used 1MB L2 cache banking on the idea people would think "more is better." To tell the truth, I wish I had known that when I bought my 4400+. I was drawn in by the "omg, the 1mb cache is killer" posts all about forum threads accross the net. Not knowing back when I bought it the marginal performance gain was not worth the nominal price premium. But at this point I figure live and learn, then deal with the loss. Besides I've been happy with it anyway and I don't plan on upgrading anytime soon.

 

[darkdemyze]

Originally posted by: Viditor

In the case of the Conroe core, it took even less (closer to 4 years)...but it cost them dearly. Intel cancelled or put on hold almost all of their other projects (like CSI) to get Conroe out early (Conroe was originally to be out next year).
I'm sure that Intel's accelerated program is what caught AMD by surprise, as K8L dual core is due out near the time when Conroe was originally expected to launch.

I did not know this. Ha well said.

Anyone interested in a nice read about the actual "5 year process" that goes into building an Intel chip ( the book covers the P6 ), the book is called "Pentium Chronicals: The People, Passion, and Politics Behind Intels Landmark Chips" by Robert Colwell. Good book too. Offers an entertaining insight on what those engineers really have to go through in order to get these chips that we critic so heavily to the mainstream market. :thumbsup: