Sign in with
Sign up | Sign in
Your question

Yet another AMD roadmap

Last response: in CPUs
Share
September 30, 2006 12:14:50 AM

AMD is moving fast to get rid of 939 and 754 as the latest raodmap shows. It seems like they want Brisbane to be the main supply as fast as possible.

Though FAB 30 is still making 90nm at > 100%, they can still get 65nm off the ground for OEMs and then later for retail.

I can't believe that they won't make 65nm 4x4 chips ASAP to keep the cost down for 3GHz chips.

I guess we can only hope that ALL 939 cores will go into AM2 for the high clock 4x4 chips. But then I don't care.

I just wante dto not e that non-AM2 seem to be going by the wayside VERY quickly.

http://www.planet3dnow.de/cgi-bin/newspub/viewnews.cgi?...

More about : amd roadmap

September 30, 2006 12:19:40 AM

BM posting an AMD thread, how unusual.
September 30, 2006 12:35:39 AM

Quote:
Yet another AMD roadmap


Are AMD lost?
Related resources
September 30, 2006 2:49:29 AM

Quote:
Yet another AMD roadmap


Are AMD lost?

So it seems. Apparently the blow delivered by Core2 knocked them silly and in need of directions
September 30, 2006 3:10:43 AM

Quote:
AMD is moving fast to get rid of 939 and 754 as the latest raodmap shows. It seems like they want Brisbane to be the main supply as fast as possible.

Though FAB 30 is still making 90nm at > 100%, they can still get 65nm off the ground for OEMs and then later for retail.

I can't believe that they won't make 65nm 4x4 chips ASAP to keep the cost down for 3GHz chips.

I guess we can only hope that ALL 939 cores will go into AM2 for the high clock 4x4 chips. But then I don't care.

I just wante dto not e that non-AM2 seem to be going by the wayside VERY quickly.

http://www.planet3dnow.de/cgi-bin/newspub/viewnews.cgi?...


They still make 754? I didn't know that. As for 939 they said they'd support it into 2007 last year and i haven't heard differently.
September 30, 2006 4:58:51 AM

Quote:
AMD is moving fast to get rid of 939 and 754 as the latest raodmap shows. It seems like they want Brisbane to be the main supply as fast as possible.

Though FAB 30 is still making 90nm at > 100%, they can still get 65nm off the ground for OEMs and then later for retail.

I can't believe that they won't make 65nm 4x4 chips ASAP to keep the cost down for 3GHz chips.

I guess we can only hope that ALL 939 cores will go into AM2 for the high clock 4x4 chips. But then I don't care.

I just wante dto not e that non-AM2 seem to be going by the wayside VERY quickly.

http://www.planet3dnow.de/cgi-bin/newspub/viewnews.cgi?...
Say it ain't so......They aren't listening to you ? :?
September 30, 2006 5:46:10 AM

I am impressed by the new AMD line up.

Getting entirely rid of the FX series is a nice, honorable and decent thing to do. Any concept of "truth in advertising" demands that AMD aknowledge they have nothing that comes close to the X6800, so dropping completely the FX line simply acknowledges that they can't compete.

Also, I like the 100 mhz increments in speed. we now have a wider range of AMD cpus that perform worse and cost more than their core2 equivalents.

With the 5400+, 5600+, and 6000+ , for example, plus have three choices of chips that cost more than an E6600, but perform less well than an E6600...

What's not to like?
September 30, 2006 5:58:57 AM

Quote:
I am impressed by the new AMD line up.

Getting entirely rid of the FX series is a nice, honorable and decent thing to do. Any concept of "truth in advertising" demands that AMD aknowledge they have nothing that comes close to the X6800, so dropping completely the FX line simply acknowledges that they can't compete.

The FX line doesn't seem like it's going to be discontinued so much as reposistioned.
http://www.neoseeker.com/news/story/6164/
Apparently, the FX-70/72/74 are being released on Socket F for the 4x4 platform. I'm going to wait and see if this is true.

Quote:
With the 5400+, 5600+, and 6000+ , for example, plus have three choices of chips that cost more than an E6600, but perform less well than an E6400...

LOL. The 5000+ is overpriced, the 5200+ is MIA, and now AMD is releasing four new X2s. It's like Hector Ruiz let 9-inch into a management posistion in return for promoting AMD:

Quote:
Ruiz: "For your esteemed service to the AMD cause, I am making you the VP of Worldwide Sales/Marketing because face it, Henri Richard was never any good."

9-inch (In tears): "I AM GOD!!!"

The Inquirer: "9-inch was made VP of Worldwide Sales/Marketing. Today, AMD enters a golden age."
September 30, 2006 6:09:29 AM

Quote:
I am impressed by the new AMD line up.

Getting entirely rid of the FX series is a nice, honorable and decent thing to do. Any concept of "truth in advertising" demands that AMD aknowledge they have nothing that comes close to the X6800, so dropping completely the FX line simply acknowledges that they can't compete.

Also, I like the 100 mhz increments in speed. we now have a wider range of AMD cpus that perform worse and cost more than their core2 equivalents.

With the 5400+, 5600+, and 6000+ , for example, plus have three choices of chips that cost more than an E6600, but perform less well than an E6400...

What's not to like?


They are certainly segmenting in fine divisions their product lineup, not sure why. The 3 GHz 6000+ will just about catch the E6600 in performance.

Honestly, I am more impresses with how quickly the adjusted their product portfolio to be able to weather the price war.... just amazing.

I do not see the K8 core catching up to C2D, it will take K8L to get back the performance lead. If a 3 GHz K8 core is barely meeting the E6600, then it would take a 3.6-3.8 GHz K8 core to just nudge out a lead.

Agreed.......

AMD is a fast, smart, adaptable company. Intel is huge, ruthless, and very smart, and AMD is not dead. That says volumes about how smart AMD is.

The REAL issue with the brisbane 65 nano chips is if they have any frequency headroom.

If the 3.0 ghz parts top out at 3.3 ghz AMD is in for a rough ride.

Clock for clock, the Core 2s are about 25% faster than the Athlon x2s.

It's pretty clear Intel could release a 3.5 ghz part tomorrow if they needed to, so to be back in the game AMD needs a 3.5 ghz X 1.25 => 4.3-4.4 ghz part to be fully competative.

Historically, every process shift has give "about" a 50% boost in clock speed on the same design....

Fastest .25 PIII was 1.0 ghz, went to 1.4 ghz on .18... => 40%
Fastest .18 P4 was 2.0 ghz, went to 3.46 ghz on .13=> 70%

Granted 90 nano Prescotts gained little (got to 3.8 ghz) but the subsequent shift to 65 nano gave big jumps in reality..

Intel never officailly released dual core parts faster than 3.73 ghz, but just about any D9xx goes deep into the 4s on good cooling... while the fastest dual core 90 nano part (the D840 at 3.2 ghz) was on the edge of meltdown at that speed, where as 4.xxx ghz D9xxs are pretty routine.

IF (huge if) AMD gets 50% going to 65 nanos and these parts get to 4+ ghz, we have a whole new ball game :) 

if not, then well, not....
September 30, 2006 6:59:22 AM

Quote:
Yet another AMD roadmap


Are AMD lost?

I think we should buy them some sat-nav.
September 30, 2006 9:20:47 AM

I dunno if AMD's marketing personnel clipped coupons to get their MBAs, but this constant leaking of roadmaps that change every couple of weeks is absolutely embarassing. It's vaporware released to cover the previous vaporware. I would strongly suggest that if AMD really wants to recover from the s##tkicking that Intel is giving it, they should shut the hell up until they can demonstrate a Quadcore Brisbane running 4GHz with 8MB cache that makes a Kentsfield look like a Pentium Pro. Then everyone can genuflect and acknowledge that AMD has the crown once again. Until then, all this tossing around of ethereal nonsense is doing nothing but devaluing the brand.

AMD, hire me as your Mkt. Mgr. I'll straighten you out... for a "healthy" package of salary, perqs and stock options... :D 
September 30, 2006 9:34:39 AM

Quote:
I dunno if AMD's marketing personnel clipped coupons to get their MBAs, but this constant leaking of roadmaps that change every couple of weeks is absolutely embarassing. It's vaporware released to cover the previous vaporware. I would strongly suggest that if AMD really wants to recover from the s##tkicking that Intel is giving it, they should shut the hell up until they can demonstrate a Quadcore Brisbane running 4GHz with 8MB cache that makes a Kentsfield look like a Pentium Pro. Then everyone can genuflect and acknowledge that AMD has the crown once again. Until then, all this tossing around of ethereal nonsense is doing nothing but devaluing the brand.

AMD, hire me as your Mkt. Mgr. I'll straighten you out... for a "healthy" package of salary, perqs and stock options... :D 
You drive a hard bargain. :D 
September 30, 2006 11:34:16 AM

Quote:
You drive a hard bargain. :D 


Dilbert's "Salary Theorem" states that "Engineers and scientists can never earn as much as business executives and sales people."

This theorem can now be supported by a mathematical equation based on the following two postulates:
Postulate 1: Knowledge is Power.
Postulate 2: Time is Money.

As every engineer knows:
Power = Work / Time

And since: Knowledge = Power
And: Time = Money
It is, therefore, true that
Knowledge = Work / Money

Solving this equation for Money, we get:
Money = Work / Knowledge

Thus, as Knowledge approaches zero, Money approaches infinity, regardless of the amount of Work done.

Conclusion: The less you know, the more you make.
September 30, 2006 11:44:24 AM

Quote:
BM posting an AMD thread

No, you are wrong!
BM is posting only BS!
September 30, 2006 1:03:55 PM

Quote:
BM posting an AMD thread, how unusual.


But his included the link.... stand up and cheer!!!

Typical Jumping Jagoff. I'm just glad that I can provide you with a common antagonist to disprove your theory about "strength in numbers"
September 30, 2006 1:06:29 PM

Quote:
I dunno if AMD's marketing personnel clipped coupons to get their MBAs, but this constant leaking of roadmaps that change every couple of weeks is absolutely embarassing. It's vaporware released to cover the previous vaporware. I would strongly suggest that if AMD really wants to recover from the s##tkicking that Intel is giving it, they should shut the hell up until they can demonstrate a Quadcore Brisbane running 4GHz with 8MB cache that makes a Kentsfield look like a Pentium Pro. Then everyone can genuflect and acknowledge that AMD has the crown once again. Until then, all this tossing around of ethereal nonsense is doing nothing but devaluing the brand.

AMD, hire me as your Mkt. Mgr. I'll straighten you out... for a "healthy" package of salary, perqs and stock options... :D 



Intel limped along for 3 years, leaking every new chip they woul release, while providing no competition or even competent processors. I guess now people want AMD news.
September 30, 2006 1:07:18 PM

Quote:
BM posting an AMD thread

No, you are wrong!
BM is posting only BS!

SO says the great follower.
September 30, 2006 1:14:45 PM

Quote:
I am impressed by the new AMD line up.

Getting entirely rid of the FX series is a nice, honorable and decent thing to do. Any concept of "truth in advertising" demands that AMD aknowledge they have nothing that comes close to the X6800, so dropping completely the FX line simply acknowledges that they can't compete.

Also, I like the 100 mhz increments in speed. we now have a wider range of AMD cpus that perform worse and cost more than their core2 equivalents.

With the 5400+, 5600+, and 6000+ , for example, plus have three choices of chips that cost more than an E6600, but perform less well than an E6600...

What's not to like?



You mean like how the bargain basement 9xx series are still appearing and still slower than X2 3800+? Only 1 out of 10 people can get a Core 2, while 10 out of 10 people CAN get an X2.
September 30, 2006 1:21:02 PM

Quote:
AMD is moving fast to get rid of 939 and 754 as the latest raodmap shows. It seems like they want Brisbane to be the main supply as fast as possible.

Though FAB 30 is still making 90nm at > 100%, they can still get 65nm off the ground for OEMs and then later for retail.

I can't believe that they won't make 65nm 4x4 chips ASAP to keep the cost down for 3GHz chips.

I guess we can only hope that ALL 939 cores will go into AM2 for the high clock 4x4 chips. But then I don't care.

I just wante dto not e that non-AM2 seem to be going by the wayside VERY quickly.

http://www.planet3dnow.de/cgi-bin/newspub/viewnews.cgi?...
Say it ain't so......They aren't listening to you ? :?

They're not making 65 nm 4x4 chips because they can't, 65 nm is not binning high enough..... :) 

@ Baron, they aren't listening to you because they already know what they need to do to keep costs down and if they could they would!!!


They are just producing the first 65nm chips. Intel took a long time to get their 65nm parts below SuperNova. AMD will do it by the middle of next year. I'm sure the EE chips are practice for 65nm. AMD has yet to say they are having any yield issues and the taiwan mobo community would have leaked the news by now.

I would say that they will get them out on schedule with the lower voltages. They have gotten 90nm down to 54W and still have the perf\watt title.

But then I'm probably making it up.
September 30, 2006 1:24:34 PM

Quote:
I am impressed by the new AMD line up.

Getting entirely rid of the FX series is a nice, honorable and decent thing to do. Any concept of "truth in advertising" demands that AMD aknowledge they have nothing that comes close to the X6800, so dropping completely the FX line simply acknowledges that they can't compete.

Also, I like the 100 mhz increments in speed. we now have a wider range of AMD cpus that perform worse and cost more than their core2 equivalents.

With the 5400+, 5600+, and 6000+ , for example, plus have three choices of chips that cost more than an E6600, but perform less well than an E6400...

What's not to like?


They are certainly segmenting in fine divisions their product lineup, not sure why (EDIT: I posted a likely reason below). The 3 GHz 6000+ will just about catch the E6600 in performance.

Honestly, I am more impressed with how quickly the adjusted their product portfolio to be able to weather the price war.... just amazing.

I do not see the K8 core catching up to C2D, it will take K8L to get back the performance lead. If a 3 GHz K8 core is barely meeting the E6600, then it would take a 3.6-3.8 GHz K8 core to just nudge out a lead.

(EDIT: Here is a link showing the compare -- http://www.hardwarezone.com/articles/view.php?cid=2&id=... )


AMD is good at what they do. If the rumors are true that ALL 65nm chips will get L3, a 3GHz chip will overtake Core 2 and perhaps improve clock for clock.
September 30, 2006 1:36:05 PM

Hey, Baron, I've been tryin' to stay outta this, but I think with your latest sextuple post, I can hereby, with the power vested in me, bestow upon you the hallowed title of:

OFFICIAL THG POST HO.

Wear this emblem of distinction with pride!

:lol: 
September 30, 2006 2:15:41 PM

AMD roadmaps are always so depressing.
September 30, 2006 2:23:12 PM

Enjoy the current intel lead as long as you can boys....

As soon as the AMD quad core comes out intel will be behind another 3 years....

You all know glueing conroes together is very inefficient and can't compete.... Unless they come up with their own HT technology they are toast....

It will probably take 16 core of intels(8 glued conroes) to equal 8 cores of AMD(true octo)....
September 30, 2006 2:52:09 PM

Quote:
BM posting an AMD thread

No, you are wrong!
BM is posting only BS!

SO says the great follower.
Yeah, right....almost everyone here is a follower of AM. why do you think we all hate you so much?
September 30, 2006 2:56:51 PM

Quote:
Enjoy the current intel lead as long as you can boys....

As soon as the AMD quad core comes out intel will be behind another 3 years....

You all know glueing conroes together is very inefficient and can't compete.... Unless they come up with their own HT technology they are toast....

It will probably take 16 core of intels(8 glued conroes) to equal 8 cores of AMD(true octo)....

You have no idea what are you talking about. Glueing 2 dualcores is as much efficeint as 4 cores on the same corssbar, without a latency penalty. Why don't you look at the Kentsfield/Clovertown benchmarks before you make another extremly stupid radical fanboy-istic conclusion without any knowledge about the subject?
September 30, 2006 4:01:33 PM

Quote:
Enjoy the current intel lead as long as you can boys....

As soon as the AMD quad core comes out intel will be behind another 3 years....

You all know glueing conroes together is very inefficient and can't compete.... Unless they come up with their own HT technology they are toast....

It will probably take 16 core of intels(8 glued conroes) to equal 8 cores of AMD(true octo)....

You have no idea what are you talking about. Glueing 2 dualcores is as much efficeint as 4 cores on the same corssbar, without a latency penalty. Why don't you look at the Kentsfield/Clovertown benchmarks before you make another extremly stupid radical fanboy-istic conclusion without any knowledge about the subject?

If you think kentsfiled is going to be faster than the new gen AMD quad cores - you are dreaming....

Also - the speed at which cores are connected makes a big difference.... Opteron proved that in the server space....
September 30, 2006 4:10:00 PM

Quote:
If you think kentsfiled is going to be faster than the new gen AMD quad cores - you are dreaming....

I don't think Core2 Quad will be faster than quadcore K8L clock for clock. Maybe equal, maybe slower....but I think that Kentsfield will be clocked higher. So, maybe you are dreaming....

Quote:
Also - the speed at which cores are connected makes a big difference....
Oh.....are you blind or stupid?
Quote:
Opteron proved that in the server space....

We were talking about 1P desktop/workstation. For 4P & 8P servers Opteron will remain to be the best until Intel comes with some decend connecting solution between the CPUs.
September 30, 2006 4:19:00 PM

Quote:
If you think kentsfiled is going to be faster than the new gen AMD quad cores - you are dreaming....

I don't think Core2 Quad will be faster than quadcore K8L clock for clock. Maybe equal, maybe slower....but I think that Kentsfield will be clocked higher. So, maybe you are dreaming....

Quote:
Also - the speed at which cores are connected makes a big difference....
Oh.....are you blind or stupid?
Quote:
Opteron proved that in the server space....

We were talking about 1P desktop/workstation. For 4P & 8P servers Opteron will remain to be the best until Intel comes with some decend connecting solution between the CPUs.

SO all you really disagree with me on is the speed they are connected?

I know you were talking about 1p.... I used that as an example becasue AMD connects it chips together the same way it connects its cores - I could have said:

athlon X2 proves the speed at which cores are connected makes a big difference(as compared to pentium dual cores(glued))....

intel has to get an HT of their own, and put the memory controllers in the cpu(1 per).... I heard rumors on here that they are working on it.... I hope for competitions sake - intel come up with something....
September 30, 2006 4:36:30 PM

Quote:
I know you were talking about 1p.... I used that as an example becasue AMD connects it chips together the same way it connects its cores(HT) -

No!

The cores on the K8 chips are connected via the SRI. The SRI, HT link and the ODMC are connected on the crossbar. The HT link on the K8 DT/WS is used for connecting to the northbridge of the mainboard.

Quote:
I could have said:

athlon X2 proves the speed at which cores are connected makes a big difference(as compared to pentium dual cores(glued))....

You don't understand.... The cores on the K8 are communicating via SRI and L2 cache, the same is with the glued cores on same package.

Quote:
intel has to get an HT of their own,
yes for MP>2P server comuters, but no for DT/WS 1P & 2P.
Quote:
and put the memory controllers in the cpu(1 per core)....
No! Learn what is memory disambigutation and how it works.
September 30, 2006 4:38:58 PM

Stupid AMDroids...

Quote:
Enjoy the current intel lead as long as you can boys....

As soon as the AMD quad core comes out intel will be behind another 3 years....

Classic fanboi statement.

Quote:
You all know glueing conroes together is very inefficient and can't compete.... Unless they come up with their own HT technology they are toast....

Proof? Kentsfield sure is "inefficient" with 99.78% scaling in Cinebench... fscking AMDroid idiots.

Quote:
It will probably take 16 core of intels(8 glued conroes) to equal 8 cores of AMD(true octo)....

Another dumbass fanboi statement.
Stop with teh FUD already!!1oneone

SO all you really disagree with me on is the speed they are connected?

I know you were talking about 1p.... I used that as an example becasue AMD connects it chips together the same way it connects its cores - I could have said:

Quote:
athlon X2 proves the speed at which cores are connected makes a big difference(as compared to pentium dual cores(glued))....

Here, I will refute you on your claim. The Athlon X2 was faster than the Pentium D simply because of architecture, not core interconnect. :roll: :roll: :roll:

Quote:
intel has to get an HT of their own, and put the memory controllers in the cpu(1 per).... I heard rumors on here that they are working on it.... I hope for competitions sake - they intel come up with something....

I don't think Intel will benefit from an IMC on Conroe as much as AMD did with K8. Memory disambiguation does a very good job already.
September 30, 2006 4:44:22 PM

So you really don't think there is any speed difference between glueing cores together and fabricating them all in one(other then the way AMD can share L3 cache now)?
September 30, 2006 4:46:44 PM

Quote:
So you really don't think there is any speed difference between glueing cores together and fabricating them all in one(other then the way AMD can share L3 cache now)?


No.
September 30, 2006 4:51:16 PM

EDIT: Just wanted to let you know, this is NOT the proper way to respond to multiple posts. I wanted to let BM have a post to himself.

Stop spreading the FUD already, stupid forum shill.
Quote:
You mean like how the bargain basement 9xx series are still appearing and still slower than X2 3800+? Only 1 out of 10 people can get a Core 2, while 10 out of 10 people CAN get an X2.

Okay, fanboi: Let me link you to where you can buy a Core 2. Let me also link you to where you can buy an Athlon 5000+.
Core 2:( [78 Reviews])
http://www.newegg.com/Product/Product.asp?Item=N82E1681...
Looks like a lot of people have Core 2's.

5000+: ([3 Reviews])
http://www.newegg.com/Product/Product.asp?Item=N82E1681...
Not the same for the Athlon
(Note: The "flawed logic" above is appropriate for BaronMatrix)

Quote:
You mean like how the bargain basement 9xx series are still appearing

Unlike AMD, Intel has a lot of shit inventory to burn through. "Still appearing" translates into "releasing new products". They're not "releasing" any new Netburst processors. :roll:

Quote:
They are just producing the first 65nm chips. Intel took a long time to get their 65nm parts below SuperNova. AMD will do it by the middle of next year.

You're "sure" about a lot of things in this paragraph. Where is your proof? These just sound like fanboi statements.
Quote:
I'm sure the EE chips are practice for 65nm.

No, they're practice for milking money out of consumers. That, and they run on the 90nm process now
Quote:
AMD has yet to say they are having any yield issues and the taiwan mobo community would have leaked the news by now.

How do you know? The taiwan mobo community would know everything, of course. :roll:

Quote:
I would say that they will get them out on schedule with the lower voltages. They have gotten 90nm down to 54W and still have the perf\watt title.

Refute Perf/Watt:
You quoted a 54W figure for X2 3800+load power consumption. The E6700 is SPEC'ED to 65W(*1.1 for MCH), so we'll just use that for comparison.
Perf(3800+)/54W < Perf(E6700)/73W.

Quote:
But then I'm probably making it up.

You sure as hell are, fanboi.

Quote:
AMD is good at what they do. If the rumors are true that ALL 65nm chips will get L3, a 3GHz chip will overtake Core 2 and perhaps improve clock for clock.

You can't claim that until you have proof. But I hope AMD does overtake Core 2 with their 65nm parts.
September 30, 2006 4:53:03 PM

Quote:
So you really don't think there is any speed difference between glueing cores together and fabricating them all in one(other then the way AMD can share L3 cache now)?

No
September 30, 2006 4:56:06 PM

Quote:
So you really don't think there is any speed difference between glueing cores together and fabricating them all in one(other then the way AMD can share L3 cache now)?


No.

intel(and amd for that matter) should just make 1 kick but tiny Single core then - and 'glue away'.... More efficient....
September 30, 2006 5:05:25 PM

Quote:
* gOJDO = Spud if anone was wondering I replied to him as spud before he deleted the message and reposted as gOJDO....

Do you think that I am that much lame.....Let me quote my self then:
Quote:
No.
September 30, 2006 5:12:15 PM

Quote:
* gOJDO = Spud if anone was wondering I replied to him as spud before he deleted the message and reposted as gOJDO....

Do you think that I am that much lame.....Let me quote my self then:
Quote:
No.


MY MISTAKE: I thought something happened that did not!!!!

BACK ON TOPIC:
intel(and amd for that matter) should just make 1 kick but tiny Single core then - and 'glue away'.... More efficient....
September 30, 2006 5:31:46 PM

Quote:
Enjoy the current intel lead as long as you can boys....

As soon as the AMD quad core comes out intel will be behind another 3 years....

You all know glueing conroes together is very inefficient and can't compete.... Unless they come up with their own HT technology they are toast....

It will probably take 16 core of intels(8 glued conroes) to equal 8 cores of AMD(true octo)....
Well, each pair of cores can communicate through the L2 cache, so latency and bottlenecking won't really be a problem.
Quote:
So you really don't think there is any speed difference between glueing cores together and fabricating them all in one(other then the way AMD can share L3 cache now)?
No current AMD chip has L3 cache, and K8L will be the first to feature it. By the time this is available, Intel's CSI will be on it's way along with 45nm (if it's not already out before K8L hits the scene).
September 30, 2006 5:34:41 PM

Making singlecores and gluing them together has significant disadvantages:
1. You need separate caches for each one.
2. Cache coherency must be through the bus. This will hurt performance when scaling to more cores.

When properly implemented (imo, a shared cache), a multicore solution on one die will not suffer from cache coherency problems.
September 30, 2006 5:43:31 PM

Quote:
]No current AMD chip has L3 cache, and K8L will be the first to feature it. By the time this is available, Intel's CSI will be on it's way along with 45nm (if it's not already out before K8L hits the scene).


good point as long as the 45nm process is a speed-up and not just a die shrink like AM2 ....
September 30, 2006 5:48:10 PM

Quote:
Making singlecores and gluing them together has significant disadvantages:
1. You need separate caches for each one.
2. Cache coherency must be through the bus. This will hurt performance when scaling to more cores.

When properly implemented (imo, a shared cache), a multicore solution on one die will not suffer from cache coherency problems.


What you are saying is what I was saying originally, but everyone on this thread insists I(we) are wrong....
September 30, 2006 5:58:48 PM

Huh? Both the Athlon X2 and Pentium D are not properly implemented dual core solutions. They have separate caches and must communicate over their buses to ensure cache coherency. The Athlon X2's bus is a lot faster and lower latency, but the FSB is not far behind.
September 30, 2006 6:17:33 PM

While communicating through both L2 cache and a Hyper Transport bus would be ideal, the way Athlon X2s communicate is still better than the ol' Pentium Ds.
September 30, 2006 6:22:40 PM

Quote:
I am impressed by the new AMD line up.

Getting entirely rid of the FX series is a nice, honorable and decent thing to do. Any concept of "truth in advertising" demands that AMD aknowledge they have nothing that comes close to the X6800, so dropping completely the FX line simply acknowledges that they can't compete.

Also, I like the 100 mhz increments in speed. we now have a wider range of AMD cpus that perform worse and cost more than their core2 equivalents.

With the 5400+, 5600+, and 6000+ , for example, plus have three choices of chips that cost more than an E6600, but perform less well than an E6600...

What's not to like?



You mean like how the bargain basement 9xx series are still appearing and still slower than X2 3800+? Only 1 out of 10 people can get a Core 2, while 10 out of 10 people CAN get an X2.


WTF are you talking about?

Please tell me your not talking about the availability issue again.
Anyone can find any core 2 they want, any time, while supplires of X2 5000 are still perilously thin and x2 5200 is still little more than a rumour.

Do you mean x2 3800s? Oh yeah, they're all over the place---big whoop de friggin do.

Are you talking about prices? You beter go review the retailers prices.

As usual you post BS without a single thin thread of evidence to back it up
September 30, 2006 6:24:49 PM

Quote:
While communicating through both L2 cache and a Hyper Transport bus would be ideal, the way Athlon X2s communicate is still better than the ol' Pentium Ds.

Nice reply, modded up.

Quote:

So, I have to state that I can’t find any indication of direct data transfers from one execution core to another in the Athlon 64 X2 processor. According to my tests, the most recent copy of data is always read from system RAM. This must be a limitation of the MOESI protocol implementation. The following seems to happen when data are accessed: on receiving a read request probe read that the second core puts on the system bus, the first core performs a write-back of the modified cache line into memory. After this write or at the same time with it, the requested line is transferred to the second core. If the data in the first core’s cache haven’t been modified, they are read from system RAM. Why is there no direct transfer between the cores via the crossbar switch? Ask AMD’s engineers about that! :) 

http://www.xbitlabs.com/articles/cpu/display/dualcore-d...
Don't think it's much better, though there is bypassing of HT in some situations:

Quote:
When randomly reading the modified data (Picture 4), there’s a small growth of data transfer latency for data blocks smaller than 512MB which may be due to the necessity to copy the modified cache lines into system RAM. The growth is very small, though, and there is no such latency growth when the data are accessed sequentially. This probably means that the memory controller doesn’t access the data after having just written them to memory, but returns them to the processor from the internal buffers, which is in fact right.
(See pic 3)

Compare with
http://www.xbitlabs.com/articles/cpu/display/dualcore-d...
pic 14 and 15.
September 30, 2006 6:25:01 PM

Quote:



AMD is good at what they do. If the rumors are true that ALL 65nm chips will get L3, a 3GHz chip will overtake Core 2 and perhaps improve clock for clock.


Rumours true? Proof?
September 30, 2006 6:26:36 PM

the s-939 processor is still a very good piece,but i can understand the necessity of changing over as quickly as possible as the new core structure is better and cooler than the 939.(runs less wattage).anyways from a business stand point it makes perfect sense,just not so cool for the DIY community.ah well,such is life.goodluck AMD.

Dahak

EVGA NF4 SLI MB
X2 4400+@2.4 S-939
2 7800GT'S IN SLI MODE
2X1GIG DDR400 MEMORY IN DC MODE
WD300GIG HD
520WATT PSU
EXTREME 19IN.MONITOR
September 30, 2006 7:31:15 PM

Quote:
Enjoy the current intel lead as long as you can boys....

As soon as the AMD quad core comes out intel will be behind another 3 years....

You all know glueing conroes together is very inefficient and can't compete.... Unless they come up with their own HT technology they are toast....

It will probably take 16 core of intels(8 glued conroes) to equal 8 cores of AMD(true octo)....

You have no idea what are you talking about. Glueing 2 dualcores is as much efficeint as 4 cores on the same corssbar, without a latency penalty. Why don't you look at the Kentsfield/Clovertown benchmarks before you make another extremly stupid radical fanboy-istic conclusion without any knowledge about the subject?


are u sure? have u seen the benchmark of native quad-core processor? no. coz it havent come out yet, and u definitely cant make that statement.
September 30, 2006 7:45:32 PM

Quote:
Enjoy the current intel lead as long as you can boys....

As soon as the AMD quad core comes out intel will be behind another 3 years....

You all know glueing conroes together is very inefficient and can't compete.... Unless they come up with their own HT technology they are toast....

It will probably take 16 core of intels(8 glued conroes) to equal 8 cores of AMD(true octo)....

You have no idea what are you talking about. Glueing 2 dualcores is as much efficeint as 4 cores on the same corssbar, without a latency penalty. Why don't you look at the Kentsfield/Clovertown benchmarks before you make another extremly stupid radical fanboy-istic conclusion without any knowledge about the subject?


are u sure? have u seen the benchmark of native quad-core processor? no. coz it havent come out yet, and u definitely cant make that statement.

gJODO is correct, you don't know what you are talking about. You are correct that no benchmarks are out, but to assume that 'glued' on cores are sub-par performance wise, as you impily, shows you do not understand how it 'fits' together.

ya gJODO is correct...theoritically, real life bencmark says it all.
!