Sign in with
Sign up | Sign in
Your question

Intel Roadmap

Last response: in CPUs
Share
July 30, 2006 5:48:26 AM

So intel quad-core will be out in December? If so, that will kill conroe. why dump $400 CAN on e6600 when you can get D805 for peanuts, and get quad in 6months? Conroe is king for now but that won't last for long.
for home use and gaming, a D805 oc'ed to 3.6ghz is more than fine with a 7900gt alongside. i'm gonna buy that with a quad-core supporting motherboard, preferably nVidia570 when they come out.
anyone have link to intel roadmap?

More about : intel roadmap

July 30, 2006 12:41:39 PM

The reason you may want conroe now instead of quad core in decemeber is iirc, quad core is at the enthuasit level, so it will be very expensive, you'll most likely not have a use for four cores, and it won't be native quad core, it will basically be two conroes glued together.
July 30, 2006 5:12:30 PM

Excellent analysis Jack, thanks for the thoughts.
Related resources
July 30, 2006 5:20:31 PM

Quote:
The reason you may want conroe now instead of quad core in decemeber is iirc, quad core is at the enthuasit level, so it will be very expensive, you'll most likely not have a use for four cores, and it won't be native quad core, it will basically be two conroes glued together.


This is not unsound reasoning, what Kentsfield will do though is push the remaining Core 2 Duo down the price scale and will have two effects:

X6800 will become the 400-600 dollar part, X6700 the 300, X6600 the 200 dollar part -- as such AMD will need to continue sliding their price down to adjust accordingly. While I have no news, data, or announcements to back this up --- my opinion is this will happen and it will happen sooner rather than later (as they pulled in Kentsfield).

The second point I need to address is the "two die glued together" attitude. In dual core the criticism is somewhat rooted in sound logic, as any two core optimized threaded software will see a latency hit due to the bussing -- this is actually a number of apps overall but not the entire breadth of software. In multi-tasking environments, the only concern is overloading the bandwidth, and the data does not suggest that the FSB is overloading in multitasking environments --- in fact, the multitasking speed up on P4D's is as good and sometimes better than the X2s.

Two dual core in the same package is mitigated lower as multithreaded apps are now not really optimized to slice up between 4 core, but are using two core nicely, some do but most don't. So the difference between a true quad verses a dual dual core package will be minimal; however, we need to wait for the data to start coming in before we make that a firm conclusion.

Also, what is the difference between two die in a package vs two die in two separate sockets.... in fact the dual die/package is more elegant to get to 4 core and two die two sockets (aka 4x4). So while AMD's 4x4 is cool, then Intel's 2x1 (2 die, 1 socket) is not? This is not only incorrect, but hypocritical.

Finally, let AMD go true quad core --- watch their yields tank like the Cell processor, the quickest path to multi core at costs that are not prohibitive is the ability to do what Intel does.... that is, pair up good, functional die and increase the yield. Ultimately, to be competitive AMD will need to lower price but yield will be bad so they will lose (or make much less) money otherwise.


Interesting,

IRT the bussing (forgive my ignorance) is it possible the whole 4x4 or 2x1 is ultimately moot (at this point in time) as with such high capacity processors the bottlnecking is likely (or is it?) to occur off the buss at the slower components, i.e the HDD, CD/DVD PCIe/AGP buss?

Peace
July 30, 2006 5:40:22 PM

Quote:
The reason you may want conroe now instead of quad core in decemeber is iirc, quad core is at the enthuasit level, so it will be very expensive, you'll most likely not have a use for four cores, and it won't be native quad core, it will basically be two conroes glued together.


This is not unsound reasoning, what Kentsfield will do though is push the remaining Core 2 Duo down the price scale and will have two effects:

X6800 will become the 400-600 dollar part, X6700 the 300, X6600 the 200 dollar part -- as such AMD will need to continue sliding their price down to adjust accordingly. While I have no news, data, or announcements to back this up --- my opinion is this will happen and it will happen sooner rather than later (as they pulled in Kentsfield).

The second point I need to address is the "two die glued together" attitude. In dual core the criticism is somewhat rooted in sound logic, as any two core optimized threaded software will see a latency hit due to the bussing -- this is actually a number of apps overall but not the entire breadth of software. In multi-tasking environments, the only concern is overloading the bandwidth, and the data does not suggest that the FSB is overloading in multitasking environments --- in fact, the multitasking speed up on P4D's is as good and sometimes better than the X2s.

Two dual core in the same package is mitigated lower as multithreaded apps are now not really optimized to slice up between 4 core, but are using two core nicely, some do but most don't. So the difference between a true quad verses a dual dual core package will be minimal; however, we need to wait for the data to start coming in before we make that a firm conclusion.

Also, what is the difference between two die in a package vs two die in two separate sockets.... in fact the dual die/package is more elegant to get to 4 core and two die two sockets (aka 4x4). So while AMD's 4x4 is cool, then Intel's 2x1 (2 die, 1 socket) is not? This is not only incorrect, but hypocritical.

Finally, let AMD go true quad core --- watch their yields tank like the Cell processor, the quickest path to multi core at costs that are not prohibitive is the ability to do what Intel does.... that is, pair up good, functional die and increase the yield. Ultimately, to be competitive AMD will need to lower price but yield will be bad so they will lose (or make much less) money otherwise.


I have to disagree. AMD didn't HAVE TO lower prices. They did it to let Intel know they were ready for a price war. I think Kentsfield is just as good an idea as 4x4, but it will be a close race based on initial tests. That will mean that "certain" X2s can be used for 4x4 with a slight premium. Intel can't afford to keep lowering their prices. Especially not for Core 2 which will have only been in the market a few months when KF is due.

That would be another shot in the foot. AMDs method worked fine going from single to dual and K8L will benefit from AMDs experience with multicore. K8L just needs to rotate the die and use two XBars. I don't see that being that difficult. They have gotten their 90nm TDP from 89W to 35W.

That is continuous improvement and is a good sign that both 65nm and K8L will be a smooth transition. I owuld bet that alot of the techniques in Rev F will help Rev G.
July 30, 2006 5:51:49 PM

well 4x4 will be faster than 2x1, if they had equal processors. 4x4 is connected by superior hypertransport bus, much like dual opt. systems today. The hyper transport bus, i would say, is the only reason AMD hasn't went belly up in the server market.

Intel does have the advantage though, in some respect that is. The have a clean architecture to play with, so they can add hundreds of thousands of fixes and improvments to the current fastest CPU on the planet.
July 30, 2006 5:52:47 PM

Quote:
The reason you may want conroe now instead of quad core in decemeber is iirc, quad core is at the enthuasit level, so it will be very expensive, you'll most likely not have a use for four cores, and it won't be native quad core, it will basically be two conroes glued together.


This is not unsound reasoning, what Kentsfield will do though is push the remaining Core 2 Duo down the price scale and will have two effects:

X6800 will become the 400-600 dollar part, X6700 the 300, X6600 the 200 dollar part -- as such AMD will need to continue sliding their price down to adjust accordingly. While I have no news, data, or announcements to back this up --- my opinion is this will happen and it will happen sooner rather than later (as they pulled in Kentsfield).

The second point I need to address is the "two die glued together" attitude. In dual core the criticism is somewhat rooted in sound logic, as any two core optimized threaded software will see a latency hit due to the bussing -- this is actually a number of apps overall but not the entire breadth of software. In multi-tasking environments, the only concern is overloading the bandwidth, and the data does not suggest that the FSB is overloading in multitasking environments --- in fact, the multitasking speed up on P4D's is as good and sometimes better than the X2s.

Two dual core in the same package is mitigated lower as multithreaded apps are now not really optimized to slice up between 4 core, but are using two core nicely, some do but most don't. So the difference between a true quad verses a dual dual core package will be minimal; however, we need to wait for the data to start coming in before we make that a firm conclusion.

Also, what is the difference between two die in a package vs two die in two separate sockets.... in fact the dual die/package is more elegant to get to 4 core and two die two sockets (aka 4x4). So while AMD's 4x4 is cool, then Intel's 2x1 (2 die, 1 socket) is not? This is not only incorrect, but hypocritical.

Finally, let AMD go true quad core --- watch their yields tank like the Cell processor, the quickest path to multi core at costs that are not prohibitive is the ability to do what Intel does.... that is, pair up good, functional die and increase the yield. Ultimately, to be competitive AMD will need to lower price but yield will be bad so they will lose (or make much less) money otherwise.


I have to disagree. AMD didn't HAVE TO lower prices. They did it to let Intel know they were ready for a price war. I think Kentsfield is just as good an idea as 4x4, but it will be a close race based on initial tests. That will mean that "certain" X2s can be used for 4x4 with a slight premium. Intel can't afford to keep lowering their prices. Especially not for Core 2 which will have only been in the market a few months when KF is due.

That would be another shot in the foot. AMDs method worked fine going from single to dual and K8L will benefit from AMDs experience with multicore. K8L just needs to rotate the die and use two XBars. I don't see that being that difficult. They have gotten their 90nm TDP from 89W to 35W.

That is continuous improvement and is a good sign that both 65nm and K8L will be a smooth transition. I owuld bet that alot of the techniques in Rev F will help Rev G.


You really think AMD would clear their shelves if they had not instituted a swift and concise motion to slash prices/profits on their CPU's? I tend to think that it prevented an exodus from occurring with AMD customers... one that if AMD does not produce ample stocks of competitive chips in the next year, will surely occur anyway transforming the company once more...
And yes AMD can not withstand a true barrage of long term price cuts in a war with Intel, they just can't, since Intel has a better and larger production process and can actually make them for less than AMD can... So this is not so much a price war, but a case of Intel simply setting a cost on their chips which will make them nice profits for the road ahead, but at the same time being much less than what AMD has been doing in the past... for they have so far drifted from their roots...

And of course AMD will build upon and learn from rev F and apply that to rev G, duh!?! :roll:
July 30, 2006 7:05:00 PM

Quote:
You really think AMD would clear their shelves if they had not instituted a swift and concise motion to slash prices/profits on their CPU's? I tend to think that it prevented an exodus from occurring with AMD customers... one that if AMD does not produce ample stocks of competitive chips in the next year, will surely occur anyway transforming the company once more...
And yes AMD can not withstand a true barrage of long term price cuts in a war with Intel, they just can't, since Intel has a better and larger production process and can actually make them for less than AMD can... So this is not so much a price war, but a case of Intel simply setting a cost on their chips which will make them nice profits for the road ahead, but at the same time being much less than what AMD has been doing in the past... for they have so far drifted from their roots...


Yes, I think so. At this point OEMs and ODMs are behind AMD so their chips will end up on shelves. They are moreso competing with P4 not Core 2 as of yet. SInce everyone knows that X2 is faster than P4, OEMs do too and they realize that they can sell higer quality chips even if it is at a slight premium. People will still spend the extra money for the Volvo. The same thing works for CPUs.

Because of the low peripheral prices and abundance of cheap SLI, AMD has the advantage. Though Intel makes a lot more money they have a lot more bills and right now AMD can survive on opteron sales moreso than Intel can on INCREASED XEON sales.

They dropped their prices to fire sale levels and P4 is still being outdistanced in mainstream retail, as eveidienced by the addtional % or share gained by AMD.

With their closer relationship to a chispet vendor for mobile they can make up even more ground in design wins. The Turion X2 is an excellent performer vs. Core Duo and again now every Intel buyer will want Core 2 desktop and mobile. This will cause differentiation issues and make stocking both very difficult with no real price point differential except mobo cost.


I mean if a person goes into Best Buy or online with $600 and they can get either or, how do you diferentiate enough to have equal sales? Without equal sales you get a glut of one or the other. If non Core 2 chips go out more they can't raise their ASP, if more Core 2 goes out they can't equalize their inventory.

AMD is coasting with equal availabiltiy of AM2 and 939 in retail sooner than 2 months after the intro. Intel can not do that due to their nonCore 2 inventory. They themselves have said they will be 30% Core 2 in Q1 07.

By then 939 will be gone and Brisbane will be rampng while AM2 is the majority >90% total retail. Then the cost of new chips will drop dramatically, and ATi income SHOULD be going to AMD 65nm Turion X2s will start to be ramped again lowering their costs.

AMD has operated like that for more than 4 months before so they can afford it now, ESPECIALLY if they can ramp Dell to volume quickly. It was said 6 months ago(the Inq) that ADM would hold back chips for Dell. The 5000+ is only available to sys builders, so that's a sign that these maybe the first Dell desktops along with the recently sampled X2 3600+.

SInce they have been shipping revenue from Dresden, they can shift around resources and have a much lower gross profit necessity than Intel.
It looks really good for AMD right now in that Intel will have to give up more market share while the Core 2 P4 situation stabilizes.

MS may be interested in 4x4 for their devs, so that is more money in the bank. They were represented in the 4x4 demos and interviews. WIth MS using Opteron and X2, volume is where they need to be. Cancelling 95 chips will allow volume in fewer SKUs even at a 1:1 ratio of new chips vs old. Since they will be concentrating on Turion X2 and 3600+ for the volume market they make get 3 for 2 from 1MB single and dual cores, especially FX57 and 4400+.


Of course AMD has to be smart and invest in new masks that don't waste die space with shrinking the caches but they have tended to be frugal yet "investment-friendly."
July 30, 2006 9:54:03 PM

Quote:


Finally, let AMD go true quad core --- watch their yields tank like the Cell processor, the quickest path to multi core at costs that are not prohibitive is the ability to do what Intel does.... that is, pair up good, functional die and increase the yield. Ultimately, to be competitive AMD will need to lower price but yield will be bad so they will lose (or make much less) money otherwise.


I disagree with this. I do not think that Intel benefits much in terms of yield by packaging 2 die together. In order to do this, you would have to COMPLETELY test both die before they are packaged. This is VERY hard to do because it is hard to get full connectivity before the part is packaged and even harder to test at speed. For this reason, the parts can not be fulled tested until it is packaged, and if 1 die is bad, it is too late to change the die. Due to these reasons, do they even pair up good die, or just keep 2 die connected from the wafer?
July 30, 2006 10:09:40 PM

Quote:

Yes, I think so. At this point OEMs and ODMs are behind AMD so their chips will end up on shelves. They are moreso competing with P4 not Core 2 as of yet. SInce everyone knows that X2 is faster than P4, OEMs do too and they realize that they can sell higer quality chips even if it is at a slight premium. People will still spend the extra money for the Volvo. The same thing works for CPUs.

Because of the low peripheral prices and abundance of cheap SLI, AMD has the advantage. Though Intel makes a lot more money they have a lot more bills and right now AMD can survive on opteron sales moreso than Intel can on INCREASED XEON sales.


Man, this is completely ridiculous and could not be further from the truth!!!! I think this thread should be locked for the blatant lies!

As of friday, AMD IS competing with the core2!!!! And people will spend extra money for a volvo, but why buy a volvo when you can pay a little more to get a BMW? Oh wait, i forgot to tell you, you can get the BMW for the SAME PRICE or CHEAPER than the volvo!!! There goes all of your top end volvos.... WOW, your post is ridiculous! And why pay a premium for the BOTTOM end volvo when you can get a honda for the same price, except the bottom end volvo is SO BAD that the Honda just TEARS IT APART?! Yeah, makes PERFECT sense!

And Intel has more bills to pay? Where do you come up with this JUNK? AMD has to pay for ATI, has a fab that they are starting, another fab that they built but haven't started paying for, and they barely made $100 million in profit! AND they just reduced profits 50% by cutting all of their prices!!!! It costs AMD more to make chips, and they sell them for less than 1/2 the price! But yet you seem to think that AMD has a better chance of surviving a price war than Intel. Please tell me where i can find this dream world!
July 30, 2006 10:10:02 PM

well quad-core is gonna be cool n all, but nothing will need 4 cores, i mean games still run better on single-core cpu's, they are just becoming dual-thread, so, i just can say get a dual-core if ur a gamer, if u do some other stupid thing then u might hold off, but seeing how u have a 7900gt or will get one go dual-core, plus, it will be the first quad-core, they may have difficulties n all, u know?, and i u want to overclock get an E6300 and overclock that, because u pretty much need the exact same mobo, good luck me lad
July 30, 2006 10:25:43 PM

multicore is where the market has been headed for many years now, just finally mainstream is acknowledgeing it too. I have had a 2gig memory dual P3 550MHz PC at home now for 8 years or so... there is much software other than games to benefit from multi-core.
for mainstream dual processors made their debut with hyperthreading, then dual-core and soon quad core, shortly thereafter 8 core Cpu and so on, it is a logical answer for more computing power while keeping clock rates within a reasonable thermal range.
Quite a few games over the past few years support multi-threading, and virtually every hot new release today does too.
The hardware industry is pushing the technology and the software industry will no doubt program alongside, otherwise they will be passed by the wayside...
July 30, 2006 10:25:49 PM

Quote:
well quad-core is gonna be cool n all, but nothing will need 4 cores


I think that you are taking an overly extreme position. Currently there are a few work load mixes that can benefit from 4 cores. That will change over time as the S/W vendors become smarter & more experienced with multi-threading their applications. This will not occur in one step. It will continue as a gradual progression over time. However, the transition has begun...
July 30, 2006 10:27:18 PM

Hi,
It is too early to launch desktop quad core CPUs. Neither 4x4. The software is not ready for this yet. Hardly will the situation change within a few months. I am a bit sceptical about intel's tryings to push the market into multi-core. There are many users who don't really need more than one core , for at least a couple of years. Word-editing , internet brownsing - all this has nothing to do with multicores , and a great deal of users don't need more.
Many programs are very crappy written and could be optimised to perform much better. But software developers don't pay much attention to this. It is more difficult to write fast and properly written multi-threaded Apps.

That's my opinion.
July 30, 2006 10:34:01 PM

You do realize that 10 years ago we were cutting edge technology using a original Pentium running at a blazing top line speed of 150/166/200MHz on a 350nm design?
July 30, 2006 10:48:40 PM

Quote:
The reason you may want conroe now instead of quad core in decemeber is iirc, quad core is at the enthuasit level, so it will be very expensive, you'll most likely not have a use for four cores, and it won't be native quad core, it will basically be two conroes glued together.


This is not unsound reasoning, what Kentsfield will do though is push the remaining Core 2 Duo down the price scale and will have two effects:

X6800 will become the 400-600 dollar part, X6700 the 300, X6600 the 200 dollar part -- as such AMD will need to continue sliding their price down to adjust accordingly. While I have no news, data, or announcements to back this up --- my opinion is this will happen and it will happen sooner rather than later (as they pulled in Kentsfield).

The second point I need to address is the "two die glued together" attitude. In dual core the criticism is somewhat rooted in sound logic, as any two core optimized threaded software will see a latency hit due to the bussing -- this is actually a number of apps overall but not the entire breadth of software. In multi-tasking environments, the only concern is overloading the bandwidth, and the data does not suggest that the FSB is overloading in multitasking environments --- in fact, the multitasking speed up on P4D's is as good and sometimes better than the X2s.

Two dual core in the same package is mitigated lower as multithreaded apps are now not really optimized to slice up between 4 core, but are using two core nicely, some do but most don't. So the difference between a true quad verses a dual dual core package will be minimal; however, we need to wait for the data to start coming in before we make that a firm conclusion.

Also, what is the difference between two die in a package vs two die in two separate sockets.... in fact the dual die/package is more elegant to get to 4 core and two die two sockets (aka 4x4). So while AMD's 4x4 is cool, then Intel's 2x1 (2 die, 1 socket) is not? This is not only incorrect, but hypocritical.

Finally, let AMD go true quad core --- watch their yields tank like the Cell processor, the quickest path to multi core at costs that are not prohibitive is the ability to do what Intel does.... that is, pair up good, functional die and increase the yield. Ultimately, to be competitive AMD will need to lower price but yield will be bad so they will lose (or make much less) money otherwise.

I may wait to see what the quads do, but I see a real bottleneck problem
with quads. It's a step in the right direction - But the FSB bandwidth
problem needs to be resolved. Conclusion: buy a conroe now - upgrade
in a 2/3 years when they resolve the bandwidth problem.

Jack - if games start to be multi-threading, does it matter if there are
2 or 4 cores?
July 30, 2006 11:46:40 PM

New technology, same equation.

Given that upgrades are a necessity now every couple years or so, arguments about the current usefulness of quad-core are moot; it's very likely someone can dream up of an awesome way to exploit that much processing power...but not right now.

Immediate results of the release of Intel's and AMD's quad lineups will only help put dual-core into the mainstream, which will benefit average system buyers with even better components and keep enthusiasts happy with more funds in their wallets as prices level off...keep in mind this is the way it always goes with upgrades: following-edge tech becomes the norm and leading-edge will WOW! you before it gets incorporated into the masses when newer stuff comes out. Business can be pretty linear sometimes. :) 

The question isn't "Why would I need quad core," but "I wonder what we could use quad cores for?"

Or something, I dunno.
July 30, 2006 11:46:50 PM

Quote:
well quad-core is gonna be cool n all, but nothing will need 4 cores, i mean games still run better on single-core cpu's, they are just becoming dual-thread




this is completely FALSE. Programs should not be written for DUAL thread, they should be written for MULTI-thread. And pretty much EVERY program will be coded this way and it is almost ALWAYS easier to code for MULTI-threaded rather than DUAL-threaded. This is because it is a lot easier to have a program create a thread for each operation, instead of creating only 2 threads.

Think about writing a game. It would be much easier to generate a thread for every computer generated opponent, rather than having 1 thread try to keep track of half of the opponents and the other thread the other half. You either want 1 thread doing EVERYTHING, or each thread doing their own thing, not somewhere in between.
July 31, 2006 12:23:08 AM

Quote:

Yes, I think so. At this point OEMs and ODMs are behind AMD so their chips will end up on shelves. They are moreso competing with P4 not Core 2 as of yet. SInce everyone knows that X2 is faster than P4, OEMs do too and they realize that they can sell higer quality chips even if it is at a slight premium. People will still spend the extra money for the Volvo. The same thing works for CPUs.

Because of the low peripheral prices and abundance of cheap SLI, AMD has the advantage. Though Intel makes a lot more money they have a lot more bills and right now AMD can survive on opteron sales moreso than Intel can on INCREASED XEON sales.


Man, this is completely ridiculous and could not be further from the truth!!!! I think this thread should be locked for the blatant lies!

As of friday, AMD IS competing with the core2!!!! And people will spend extra money for a volvo, but why buy a volvo when you can pay a little more to get a BMW? Oh wait, i forgot to tell you, you can get the BMW for the SAME PRICE or CHEAPER than the volvo!!! There goes all of your top end volvos.... WOW, your post is ridiculous! And why pay a premium for the BOTTOM end volvo when you can get a honda for the same price, except the bottom end volvo is SO BAD that the Honda just TEARS IT APART?! Yeah, makes PERFECT sense!

And Intel has more bills to pay? Where do you come up with this JUNK? AMD has to pay for ATI, has a fab that they are starting, another fab that they built but haven't started paying for, and they barely made $100 million in profit! AND they just reduced profits 50% by cutting all of their prices!!!! It costs AMD more to make chips, and they sell them for less than 1/2 the price! But yet you seem to think that AMD has a better chance of surviving a price war than Intel. Please tell me where i can find this dream world!


Intel has 100,000 - excuse me 99,000 employees to feed, 12 Fabs to power and a much larger channel to fill with mobos.

Are you on a hallucinogen? Lock the post for blatant lies???

It's obvious it's useless talking to you so I won't.
July 31, 2006 12:27:29 AM

Quote:
well quad-core is gonna be cool n all, but nothing will need 4 cores, i mean games still run better on single-core cpu's, they are just becoming dual-thread




this is completely FALSE. Programs should not be written for DUAL thread, they should be written for MULTI-thread. And pretty much EVERY program will be coded this way and it is almost ALWAYS easier to code for MULTI-threaded rather than DUAL-threaded. This is because it is a lot easier to have a program create a thread for each operation, instead of creating only 2 threads.

Think about writing a game. It would be much easier to generate a thread for every computer generated opponent, rather than having 1 thread try to keep track of half of the opponents and the other thread the other half. You either want 1 thread doing EVERYTHING, or each thread doing their own thing, not somewhere in between.


Please stop it. Joh Carmack doesn't even really have an extended multithreaded game engine so it's not that easy. CrySis will be the first after a long dev cycle. Patches for SMP DO NOT enable pure multithreading it just slices off the easiest routines.

Sync'ing video and audio streams running on separate threads is not trivial.
July 31, 2006 12:27:49 AM

Quote:

Yes, I think so. At this point OEMs and ODMs are behind AMD so their chips will end up on shelves. They are moreso competing with P4 not Core 2 as of yet. SInce everyone knows that X2 is faster than P4, OEMs do too and they realize that they can sell higer quality chips even if it is at a slight premium. People will still spend the extra money for the Volvo. The same thing works for CPUs.

Because of the low peripheral prices and abundance of cheap SLI, AMD has the advantage. Though Intel makes a lot more money they have a lot more bills and right now AMD can survive on opteron sales moreso than Intel can on INCREASED XEON sales.


Man, this is completely ridiculous and could not be further from the truth!!!! I think this thread should be locked for the blatant lies!

As of friday, AMD IS competing with the core2!!!! And people will spend extra money for a volvo, but why buy a volvo when you can pay a little more to get a BMW? Oh wait, i forgot to tell you, you can get the BMW for the SAME PRICE or CHEAPER than the volvo!!! There goes all of your top end volvos.... WOW, your post is ridiculous! And why pay a premium for the BOTTOM end volvo when you can get a honda for the same price, except the bottom end volvo is SO BAD that the Honda just TEARS IT APART?! Yeah, makes PERFECT sense!

And Intel has more bills to pay? Where do you come up with this JUNK? AMD has to pay for ATI, has a fab that they are starting, another fab that they built but haven't started paying for, and they barely made $100 million in profit! AND they just reduced profits 50% by cutting all of their prices!!!! It costs AMD more to make chips, and they sell them for less than 1/2 the price! But yet you seem to think that AMD has a better chance of surviving a price war than Intel. Please tell me where i can find this dream world!


Intel has 100,000 - excuse me 99,000 employees to feed, 12 Fabs to power and a much larger channel to fill with mobos.

Are you on a hallucinogen? Lock the post for blatant lies???

It's obvious it's useless talking to you so I won't.

Correct, but Intel made 8 BILLION dollar profit in the last year!!!! What are you smoking?! Obviously you dont know simple accounting. When you make a LARGE profit, youc an afford to decrease your ASP. Intel could build 2 fabs today, and pay for it today, without going into the red. Amd can not even come close to this.

Hopefully other people can see through your BS.
July 31, 2006 12:30:32 AM

Quote:
well quad-core is gonna be cool n all, but nothing will need 4 cores, i mean games still run better on single-core cpu's, they are just becoming dual-thread




this is completely FALSE. Programs should not be written for DUAL thread, they should be written for MULTI-thread. And pretty much EVERY program will be coded this way and it is almost ALWAYS easier to code for MULTI-threaded rather than DUAL-threaded. This is because it is a lot easier to have a program create a thread for each operation, instead of creating only 2 threads.

Think about writing a game. It would be much easier to generate a thread for every computer generated opponent, rather than having 1 thread try to keep track of half of the opponents and the other thread the other half. You either want 1 thread doing EVERYTHING, or each thread doing their own thing, not somewhere in between.


Please stop it. Joh Carmack doesn't even really have an extended multithreaded game engine so it's not that easy. CrySis will be the first after a long dev cycle. Patches for SMP DO NOT enable pure multithreading it just slices off the easiest routines.

Sync'ing video and audio streams running on separate threads is not trivial.

Man, you are RETARDED!!! Hey, you probably think 1+1=3. No, John Carmack doesn't even have an extended multithreaded game engine. But thats because he has not done it yet. I didn't say it is currently available. I said it is EASIER to program MULTI-threads over ONLY 2 threads! Get it right! Dont comment on my posts unless u actually know what you're talking about!
July 31, 2006 12:35:00 AM

Quote:
Correct, but Intel made 8 BILLION dollar profit in the last year!!!! What are you smoking?! Obviously you dont know simple accounting. When you make a LARGE profit, youc an afford to decrease your ASP. Intel could build 2 fabs today, and pay for it today, without going into the red. Amd can not even come close to this.



8 billion in profit can be spent very quickly on new initiatives like price drops. I have always said I was skpetical about the merger in terms of cost but no payments are due before at least Q4. At the same time the next month's ATi income would be in AMDs hands. A lot fo the money was brrowed so it's not even profit-based.
A company liek AMD can float a loan much easier than a company that doesn't generate 7 billion a year.

So you're saying that even with the price drops, Intel could drop $7B tomorrow in cash?

8O
July 31, 2006 12:41:16 AM

But on topic, Intel's roadmap looks pretty good. They have Kentsfield and Clovertown coming.I hope they start to cancel some of their older Xeon line though. They have gotten power down somewhat but it will affect Intel's design wins as much as AMDs, maybe even moreso.

They don't have a core refresh scheduled until 2008, so they will have to go with clockspeed to maintain an edge. Of course that's dependent upon AMDs success wih power/frequency under 65nm. 35-55W could be the norm for Brisbane.

Intel is also working on their "Sun-Killer" which may make waves at the extreme high end but I don't see it hapenng before 2008-2009. They can rush at it and undermine their market more but hopefully they will just let PD/XeonP4 go away quickly.
July 31, 2006 1:01:15 AM

Quote:
Correct, but Intel made 8 BILLION dollar profit in the last year!!!! What are you smoking?! Obviously you dont know simple accounting. When you make a LARGE profit, youc an afford to decrease your ASP. Intel could build 2 fabs today, and pay for it today, without going into the red. Amd can not even come close to this.



8 billion in profit can be spent very quickly on new initiatives like price drops. I have always said I was skpetical about the merger in terms of cost but no payments are due before at least Q4. At the same time the next month's ATi income would be in AMDs hands. A lot fo the money was brrowed so it's not even profit-based.
A company liek AMD can float a loan much easier than a company that doesn't generate 7 billion a year.

So you're saying that even with the price drops, Intel could drop $7B tomorrow in cash?

8O

Yes, it can be spent very quickly!!!!! You have still not addressed the FACT that it is MUCH easier to spend $100 million than it is to spend $8 billion!!! Sorry that you brain can not comprehend this SIMPLE thought! It shows just how thick your head is!

1, dont put words in my mouth. I didn't say there were going to drop $7 billion in cash tomorrow. I said they made enough profit that they COULD have spent $7 billion in cash and still not be in the red! And YES, as long as Intel makes a profit next quarter, they COULD spend $7 billion in cash tomorrow! And it is very likely that they will make a profit next quarter as they have not been in the red in a LONG time.
July 31, 2006 1:41:09 AM

Quote:

This is not unsound reasoning, what Kentsfield will do though is push the remaining Core 2 Duo down the price scale and will have two effects:

X6800 will become the 400-600 dollar part, X6700 the 300, X6600 the 200 dollar part -- as such AMD will need to continue sliding their price down to adjust accordingly. While I have no news, data, or announcements to back this up --- my opinion is this will happen and it will happen sooner rather than later (as they pulled in Kentsfield).

The second point I need to address is the "two die glued together" attitude. In dual core the criticism is somewhat rooted in sound logic, as any two core optimized threaded software will see a latency hit due to the bussing -- this is actually a number of apps overall but not the entire breadth of software. In multi-tasking environments, the only concern is overloading the bandwidth, and the data does not suggest that the FSB is overloading in multitasking environments --- in fact, the multitasking speed up on P4D's is as good and sometimes better than the X2s.

Two dual core in the same package is mitigated lower as multithreaded apps are now not really optimized to slice up between 4 core, but are using two core nicely, some do but most don't. So the difference between a true quad verses a dual dual core package will be minimal; however, we need to wait for the data to start coming in before we make that a firm conclusion.

Also, what is the difference between two die in a package vs two die in two separate sockets.... in fact the dual die/package is more elegant to get to 4 core and two die two sockets (aka 4x4). So while AMD's 4x4 is cool, then Intel's 2x1 (2 die, 1 socket) is not? This is not only incorrect, but hypocritical.

Finally, let AMD go true quad core --- watch their yields tank like the Cell processor, the quickest path to multi core at costs that are not prohibitive is the ability to do what Intel does.... that is, pair up good, functional die and increase the yield. Ultimately, to be competitive AMD will need to lower price but yield will be bad so they will lose (or make much less) money otherwise.


It would be intersting if AMD can afford a price cut with or without 65nm fabrication process. :lol: 
July 31, 2006 2:36:08 AM

Quote:
well quad-core is gonna be cool n all, but nothing will need 4 cores, i mean games still run better on single-core cpu's, they are just becoming dual-thread




this is completely FALSE. Programs should not be written for DUAL thread, they should be written for MULTI-thread. And pretty much EVERY program will be coded this way and it is almost ALWAYS easier to code for MULTI-threaded rather than DUAL-threaded. This is because it is a lot easier to have a program create a thread for each operation, instead of creating only 2 threads.

Think about writing a game. It would be much easier to generate a thread for every computer generated opponent, rather than having 1 thread try to keep track of half of the opponents and the other thread the other half. You either want 1 thread doing EVERYTHING, or each thread doing their own thing, not somewhere in between.


Please stop it. Joh Carmack doesn't even really have an extended multithreaded game engine so it's not that easy. CrySis will be the first after a long dev cycle. Patches for SMP DO NOT enable pure multithreading it just slices off the easiest routines.

Sync'ing video and audio streams running on separate threads is not trivial.

Quake 4 gets a 86% or so performance boost from being run on a dual core environment. 86% isn't sliceing off the easiest routines.

But on the point of multithreaded engine I personally think looking at the Unreal 3 engine is more promiseing. All eyes should be on it and not this CrySis engine.
July 31, 2006 2:52:17 AM

Quote:

So you're saying that even with the price drops, Intel could drop $7B tomorrow in cash?


Yes. They have over $7B in cash/STI.
July 31, 2006 3:05:00 AM

Quote:
well quad-core is gonna be cool n all, but nothing will need 4 cores, i mean games still run better on single-core cpu's, they are just becoming dual-thread




this is completely FALSE. Programs should not be written for DUAL thread, they should be written for MULTI-thread. And pretty much EVERY program will be coded this way and it is almost ALWAYS easier to code for MULTI-threaded rather than DUAL-threaded. This is because it is a lot easier to have a program create a thread for each operation, instead of creating only 2 threads.

Think about writing a game. It would be much easier to generate a thread for every computer generated opponent, rather than having 1 thread try to keep track of half of the opponents and the other thread the other half. You either want 1 thread doing EVERYTHING, or each thread doing their own thing, not somewhere in between.


Please stop it. Joh Carmack doesn't even really have an extended multithreaded game engine so it's not that easy. CrySis will be the first after a long dev cycle. Patches for SMP DO NOT enable pure multithreading it just slices off the easiest routines.

Sync'ing video and audio streams running on separate threads is not trivial.

Quake 4 gets a 86% or so performance boost from being run on a dual core environment. 86% isn't sliceing off the easiest routines.

But on the point of multithreaded engine I personally think looking at the Unreal 3 engine is more promiseing. All eyes should be on it and not this CrySis engine.


At LCD res (17 and 21" LCDs run at 1280), that is VERY FAR from the truth.

AMD

Intel


Real perf at low res is one thing, but low res doesn't an experience make. UT is yes also designed for AMD64 and multi-core. There are others but they both have DX9 demos and are designed for DX10.

Kentsfield will come in handy for it with Vista. As will 4x4. Vista DX10 will punish GPUs and CPUs.
July 31, 2006 4:29:26 AM

Hey Baron,

Does your mouth overload your ass in real life too?

I mean seriously, if we could knock all the BS out of you, we could bury you in a matchbox.
July 31, 2006 2:06:05 PM

Quote:
As of friday, AMD IS competing with the core2!!!! And people will spend extra money for a volvo, but why buy a volvo when you can pay a little more to get a BMW? Oh wait, i forgot to tell you, you can get the BMW for the SAME PRICE or CHEAPER than the volvo!!! There goes all of your top end volvos.... WOW, your post is ridiculous! And why pay a premium for the BOTTOM end volvo when you can get a honda for the same price, except the bottom end volvo is SO BAD that the Honda just TEARS IT APART?! Yeah, makes PERFECT sense!


Poo like this keeps many an economist employed. :trophy:
!