Shanghia die shots.

Interesting pics. I guess the area farthest right must be the bigger L3 cache. Shanghai appears most different from the original Barcelona die in that particular area.
 
In comparison to Shanghai, Nehalem seems to have much more logic circuitry. 2x from the looks of it. :eek:

*dum da dum dum*
 
debello_64 said:
Hyper Threading worked out so well for Intel last time they used it, didnt it.
Click to expand...

Actually it worked pretty well, making the systems scale better for various server tasks. Even 3d rendering and video encoding got a small boost out of it (Pixar used a Xeon-powered renderfarm with 2.8 GHz HT processors).
IBM and Sun also implement SMT on their server systems, which does wonders for things like webservers and databases.
Especially Sun's Niagara CPUs are interesting. They don't have a lot of raw performance, but they can hold a lot of threads 'in the air' at the same time, giving really good response times for simple transactions as mostly used with webpages and simple database access.

It may not improve your gaming performance, but I suppose Intel is mainly using this technology for its server products, but doesn't bother to disable it in desktop products because it won't hurt.
 
Mr. Bluntman said:
In comparison to Shanghai, Nehalem seems to have much more logic circuitry. 2x from the looks of it. :eek:
Click to expand...
Intel has much better cache density than AMD. You have to factor that in when you're looking at the area. And the chips have very different I/Os, with Nehalem having more complex QP links and a tri-channel memory controller. Eventually Intel will directly or indirectly release the number of core transistors. When both companies do the same, then you can make a judgement about who uses more logic transistors. ;)

Nehalem has slimmed down vs the transistor count of 2 die Yorkfield CPUs, partially thanks to the lower total cache: 8MB L3 + 2MB L2 in Nehalem vs 12MB L2 in Yorkfield. But there's obviously more to it because Nehalem has a lot more integrated now (NB, QP, ODMC).

It will be interesting if the February AMD roadmap is correct and 2.5GHz Shanghai CPUs are not out until 2009 because that wouldn't speak well of the "savior" CPU.
 
It should be interesting to note that AMD is going with single-pass 45nm immersion litho step, and Intel is still using DP/DE. Overall process takt will be lower for AMD, which is good, but I doubt it will make much of a difference.
 
Scali2 said:
Actually it worked pretty well, making the systems scale better for various server tasks. Even 3d rendering and video encoding got a small boost out of it (Pixar used a Xeon-powered renderfarm with 2.8 GHz HT processors).
IBM and Sun also implement SMT on their server systems, which does wonders for things like webservers and databases.
Especially Sun's Niagara CPUs are interesting. They don't have a lot of raw performance, but they can hold a lot of threads 'in the air' at the same time, giving really good response times for simple transactions as mostly used with webpages and simple database access.

It may not improve your gaming performance, but I suppose Intel is mainly using this technology for its server products, but doesn't bother to disable it in desktop products because it won't hurt.
Click to expand...

It really doesnt have anything to do with this topic... Please dont thread crap this one.
 
Mr. Bluntman said:
In comparison to Shanghai, Nehalem seems to have much more logic circuitry. 2x from the looks of it. :eek:

*dum da dum dum*
Click to expand...

I noticed that too. Taking into consideration the size of a single core minus the L2 cache and comparing that with the the size of one of AMD's cores minus the L2 cache, Intels is much larger. How this will play out is still largely unknown. Generally history has proven that a larger transistor budget equates to a larger performance. There are a few notable exceptions, but the general rule still holds.

I guess at this point we'll have to wait and see.

One of the things that I am interested in seeing is the result of Low-K Dialectrics. AMD claims a 15 percent performance gain. How this wil translate into processing performance is unknown, but but a 15 percent boost in circuit performance is impressive. Another thing that I am interested in keeping my eye on is CTI, and AMD's implementation of High-K Metal Gates. They claim an additional 15% performance boost from this technology. Again how that will translate to processing performance is unknown, but a boost of that magnitude is very impressive.
 
duby229 said:
I noticed that too. Taking into consideration the size of a single core minus the L2 cache and comparing that with the the size of one of AMD's cores minus the L2 cache, Intels is much larger. How this will play out is still largely unknown. Generally history has proven that a larger transistor budget equates to a larger performance. There are a few notable exceptions, but the general rule still holds.
Click to expand...

Like Netbust?

duby229 said:
I guess at this point we'll have to wait and see.
Click to expand...

But at this point I really have lost all hope for AMD. They were saying that it would be 10% faster clock-for-clock than Kentsfield. In what? What we got was something that was maybe 5-10% faster than Windsor, not Kentsfield. It would be great if they came back and wowed me, but I'm convinced that's NOT going to happen, and Shanghai is going to get it's arse handed to it on a platinum platter by Nehalem. AMD is going to be behind even farther unless they can pull a rabbit out of their asses...

duby229 said:
One of the things that I am interested in seeing is the result of Low-K Dialectrics. AMD claims a 15 percent performance gain. How this wil translate into processing performance is unknown, but but a 15 percent boost in circuit performance is impressive. Another thing that I am interested in keeping my eye on is CTI, and AMD's implementation of High-K Metal Gates. They claim an additional 15% performance boost from this technology. Again how that will translate to processing performance is unknown, but a boost of that magnitude is very impressive.
Click to expand...

I thought AMD's implementation of High-K metal gates was called "gate-first?" :confused:

At any rate, AMD needs all the help that they can get with a lower IPC. But I fail to be convinced otherwise than Shanghai is going to be nothing but a massive FAIL when stacked up to the competition at the end of this year/beginning of next.
 
Mr. Bluntman said:
Like Netbust?

But at this point I really have lost all hope for AMD. They were saying that it would be 10% faster clock-for-clock than Kentsfield. In what? What we got was something that was maybe 5-10% faster than Windsor, not Kentsfield. It would be great if they came back and wowed me, but I'm convinced that's NOT going to happen, and Shanghai is going to get it's arse handed to it on a platinum platter by Nehalem. AMD is going to be behind even farther unless they can pull a rabbit out of their asses...

I thought AMD's implementation of High-K metal gates was called "gate-first?" :confused:

At any rate, AMD needs all the help that they can get with a lower IPC. But I fail to be convinced otherwise than Shanghai is going to be nothing but a massive FAIL when stacked up to the competition at the end of this year/beginning of next.
Click to expand...

It is entirely possible that you'll be right, but I'm a glass half full kinda guy, and prefer to give it the benefit of a doubt. AMD recent track record hasnt been that great, some may even say that it has been aweful, but I tend to beleive that history repeats itself, and at the end of the day, whether it be this generation or the next, AMD will regain the crown. I had hoped it would be Barcelona, but It didnt pan out that way. Just like I am hoping it'll be Shanghia. It may not pan out that way this time either, but one has to keep hope alive right?

Anyhow in the end it all depends on how well transistor and circuit performance translates into processing performance. It may well come out the gat kickin ass and taking names, though at this point based on the die shots, it doesnt look like IPC will be much different from Barcelona.

Anyhow either way it will be a little bit cheaper then Barcelona and should clock higher. It may not take the crown, but I'll be willing to bet that it closes the gap.
 
If you want to debate technology gaps between the two major companies why not make a thread and do so there instead.

Also, if you see a post that violates the rules, please Report it. No need to pile on with more thread crapping.
 
Also I dont know why, but I just noticed this for the first time , but it appears that Shanghia has 4 HT controllers? One on top, one on the bottom, and two on the side. Anybody else notice that?

Maybe we'll see this die on socket g3?
 
Rich tate comin down on yez like the hamma! dats da law!

duby229 said:
Maybe we'll see this die on socket g3?
Click to expand...

I thought it was confirmed as AM3 and AM2+ compatable. Do we have any shots of the AM3 socket? is it another 940 pinner or does it have something new?
 
duby229 said:
It is entirely possible that you'll be right, but I'm a glass half full kinda guy, and prefer to give it the benefit of a doubt. AMD recent track record hasnt been that great, some may even say that it has been aweful, but I tend to beleive that history repeats itself, and at the end of the day, whether it be this generation or the next, AMD will regain the crown. I had hoped it would be Barcelona, but It didnt pan out that way. Just like I am hoping it'll be Shanghia. It may not pan out that way this time either, but one has to keep hope alive right?

Anyhow in the end it all depends on how well transistor and circuit performance translates into processing performance. It may well come out the gat kickin ass and taking names, though at this point based on the die shots, it doesnt look like IPC will be much different from Barcelona.

Anyhow either way it will be a little bit cheaper then Barcelona and should clock higher. It may not take the crown, but I'll be willing to bet that it closes the gap.
Click to expand...

Based on the die shots, it seems that Shanghai is nothing more than a die shrink with an added 4MB of L3 cache and possibly some IMC tweaks. The processing elements themselves look un-changed. Hence the ass-handing will commence...

I put great faith in Barcelona and Agena, so much so that I believed AMD's word to a "T" that it was going to be 10% faster than Kentsfield. That turned out to be complete bullshit, and more often than not, was barely faster than the existing K8 core. AMD needs a new architecture to compete, and if the roadmaps are correct, we won't see that until 2010. I've pretty much written off AMD until they prove otherwise; my X2 5400 being my last AMD purchase for a very, very long time.
 
Mr. Bluntman said:
Based on the die shots, it seems that Shanghai is nothing more than a die shrink with an added 4MB of L3 cache and possibly some IMC tweaks. The processing elements themselves look un-changed. Hence the ass-handing will commence...

I put great faith in Barcelona and Agena, so much so that I believed AMD's word to a "T" that it was going to be 10% faster than Kentsfield. That turned out to be complete bullshit, and more often than not, was barely faster than the existing K8 core. AMD needs a new architecture to compete, and if the roadmaps are correct, we won't see that until 2010. I've pretty much written off AMD until they prove otherwise; my X2 5400 being my last AMD purchase for a very, very long time.
Click to expand...

The main problem K10 is facing at the moment is the IMC/L3 frequency. A 1.8GHz, it limits performance scaling from higher clocked SKUs. I'm of the belief that if the IMC/L3 was to run in sync with the CPU clock, it would come close to Kentsfield levels.

If Shanghai can get the IMC/L3 speeds up, along with the larger L3 cache and perhaps some architectural enhancements - it may very well stand a chance, at least clock for clock, against Penryn.

Of course, if Intel feels threatened they could bump up Penryn clockspeeds to levels AMD cannot match, and I have yet to mention Nehalem... ;)
 
harpoon said:
The main problem K10 is facing at the moment is the IMC/L3 frequency. A 1.8GHz, it limits performance scaling from higher clocked SKUs. I'm of the belief that if the IMC/L3 was to run in sync with the CPU clock, it would come close to Kentsfield levels.

If Shanghai can get the IMC/L3 speeds up, along with the larger L3 cache and perhaps some architectural enhancements - it may very well stand a chance, at least clock for clock, against Penryn.

Of course, if Intel feels threatened they could bump up Penryn clockspeeds to levels AMD cannot match, and I have yet to mention Nehalem... ;)
Click to expand...

Agreed. There's much untapped potential even in Conroe and Kentsfield. Let alone Wolfdale and Yorkfield. AMD has seriously faceplanted this round. Can they catch up? Only the Shadow knows...
 
Mr. Bluntman said:
Based on the die shots, it seems that Shanghai is nothing more than a die shrink with an added 4MB of L3 cache and possibly some IMC tweaks. The processing elements themselves look un-changed. Hence the ass-handing will commence...

I put great faith in Barcelona and Agena, so much so that I believed AMD's word to a "T" that it was going to be 10% faster than Kentsfield. That turned out to be complete bullshit, and more often than not, was barely faster than the existing K8 core. AMD needs a new architecture to compete, and if the roadmaps are correct, we won't see that until 2010. I've pretty much written off AMD until they prove otherwise; my X2 5400 being my last AMD purchase for a very, very long time.
Click to expand...

It's an interesting take on things I suppose, but is far too emotional and not enough factual.

I agree that that Shanghia doesnt look much different from Barcelona from a visual perspective and may not see much of an IPC improvement. However from a fabrication perspective it is significantly different, and without seeing the hardware in action nobody here can say that without many grains of salt. We know that transistor performance will be greatly improved, and that circuit performnace will be greatly improved. We know that the Memory controller is overhauled, and that they added an additional HT3 controller. The point being that any IC company in there right mind would die shrink before making major modifications to the architecture.

These are far too many compounded unknown elements to make a solid conclusion. This is the reason I said we'll have to wait and see.
 
Mr. Bluntman said:
Agreed. There's much untapped potential even in Conroe and Kentsfield. Let alone Wolfdale and Yorkfield. AMD has seriously faceplanted this round. Can they catch up? Only the Shadow knows...
Click to expand...

Yeah, I'm convinced Intel is sandbagging at the moment. A look at QX9650 power consumption levels shows that Intel could potentially release it up to 3.8GHz and still remain under the 130W TDP limit.

I'm sure AMD will get some improvement from Shanghai, and it would be a fine achievement if they can muster a 3GHz chip at Penryn IPC levels (as has been rumoured, although by Fudzilla *cough* ;) ), but it must be pretty demoralising to know in the back of your mind that the competition can easily clock their chips nearly 1GHz faster and maintain the same performance advantage as before.
 
duby229 said:
It's an interesting take on things I suppose, but is far too emotional and not enough factual.

I agree that that Shanghia doesnt look much different from Barcelona from a visual perspective and may not see much of an IPC improvement. However from a fabrication perspective it is significantly different, and without seeing the hardware in action nobody here can say that without many grains of salt. We know that transistor performance will be greatly improved, and that circuit performnace will be greatly improved. We know that the Memory controller is overhauled, and that they added an additional HT3 controller. The point being that any IC company in there right mind would die shrink before making major modifications to the architecture.

These are far too many compounded unknown elements to make a solid conclusion. This is the reason I said we'll have to wait and see.
Click to expand...

harpoon said:
Yeah, I'm convinced Intel is sandbagging at the moment. A look at QX9650 power consumption levels shows that Intel could potentially release it up to 3.8GHz and still remain under the 130W TDP limit.

I'm sure AMD will get some improvement from Shanghai, and it would be a fine achievement if they can muster a 3GHz chip at Penryn IPC levels (as has been rumoured, although by Fudzilla *cough* ;) ), but it must be pretty demoralising to know in the back of your mind that the competition can easily clock their chips nearly 1GHz faster and maintain the same performance advantage as before.
Click to expand...

If AMD wants to keep me as a customer, they must prove that they are worthy of my spending dollar. It would be great to see them competitive again, but I'm not getting my hopes up anymore. The burden of proof is on them.
 
Mr. Bluntman said:
If AMD wants to keep me as a customer, they must prove that they are worthy of my spending dollar. It would be great to see them competitive again, but I'm not getting my hopes up anymore. The burden of proof is on them.
Click to expand...

Well said, so lets leave it at that and wait to see what happens. In the mean time we can still discuss the technology, like why did they add a forth HT controller? Will it be available on G3, and if so then does that mean it has a DDR3 memory controller?

What the heck is going on here? Why waste die space with hardware that cant be used on current sockets?
 
duby229 said:
Well said, so lets leave it at that and wait to see what happens. In the mean time we can still discuss the technology, like why did they add a forth HT controller? Will it be available on G3, and if so then does that mean it has a DDR3 memory controller?

What the heck is going on here? Why waste die space with hardware that cant be used on current sockets?
Click to expand...

Agreed. I feel that I have vented enough about AMD's failure to compete. For the longest time, I have defended them, and after seeing a friend's C2D system in action, I have realized what I have been missing. I'm all about getting AMD if the purchase is warranted, but right now, why waste my cash on something that doesn't even really stack up, especially since I can sell my existing motherboard and processor to buy an nForce 750i SLI or 650i SLI based Intel motherboard, leaving me to just pay an additional $20 for an E6750 over an X2 6400, or an additional $50 for an E8400 when they become available for so much more added performance in additional exceptional overclocking ability should I so choose to do so?

I will admit that I'm very curious to see how Kuma and it's 45nm variant will stack up, not to mention Shanghai and Deneb. Unless they have more tweaks that they are keeping under the NDA kimono, I don't see a reason to get excited.

And the so-called "AM2 backwards compatiblilty" has been a joke. Mostly with nVidia chipset motherboards, although there are exceptions to the rule (ASUS, Gigabyte, and a couple of mainstream ECS and Biostar motherboards). I also am curious to see how the added 4MB of L3 will affect performance.
 
duby229 said:
It is entirely possible that you'll be right, but I'm a glass half full kinda guy, and prefer to give it the benefit of a doubt. AMD recent track record hasnt been that great, some may even say that it has been aweful, but I tend to beleive that history repeats itself.
Click to expand...


Oh he is very much correct,and if history does repeat itself,then the Shanghai will be fast turned into a budget chip,as AMD is forced to slash prices,and sell them for nothing,as
they continue to watch ASP's slide farther north.


I predict a sub 3 dollar share price by the fourth quarter,with ASP's even further decimated.I do hope I am wrong of course,but I wasnt last time,or the time before
that...

harpoon said:
The main problem K10 is facing at the moment is the IMC/L3 frequency. A 1.8GHz, it limits performance scaling from higher clocked SKUs. I'm of the belief that if the IMC/L3 was to run in sync with the CPU clock, it would come close to Kentsfield levels.
Click to expand...

I'm not,I think it would surpass it by a hair fairly consistantly.
 
Well let's hope that B3 fixes the lack of scaling, and gets those NB clocks up. The TLB issue is fixed this time around for sure though, so that's no longer an issue.

I'm sincerely worried about AMD's future. Intel is getting set to steamroll AMD, and unless they can pull off a miracle (I really don't see it at this point in time - anything they do will be too little too late...) I doubt that we will have two major players in the x86 market anymore. IBM isn't interested AFAIK in competing with Intel, since we are seeing a trend that they are moving away from hardware (selling off their PC division to Lenovo) and gearing up for becoming just services, software, and support. As countless others have pointed out, IBM is staying out of this.

NVIDIA isn't able to buy AMD either for both financial and political reasons.

And Samsung, the last prospector is located in Asia, and something like that is NOT going to sit well with the FTC. Not to mention the total culture shock of the two different philosophies (three if you count ATi) that the companies embrace.

If things don't improve, AMD may well close its doors by the end of 2009, with the vultures circling to get their fill of the carcass.

I certainly hope I am wrong, but my magic eight ball has spoken.
 
Manny Calavera said:
I'm not,I think it would surpass it by a hair fairly consistantly.
Click to expand...

May I ask how you came to such a conclusion?

According to this Techreport article increasing the NB from 1.8GHz to 2GHz yields ~2% improved performance on a Phenom 9600, academically significant but not noticeably better. I fail to see how increasing it just 300MHz further to 2.3GHz will bring it 10% higher performance it needs to match, let alone exceed, a Kentsfield clock for clock.

Right now, we can conclude that a Phenom 9600 is ~5% slower than if it had the NB/L3 running in sync. The performance hit with future higher clocked Phenoms would increase if AMD doesn't address this issue. [H]'s Phenom review showed this clearly, with the Phenom @ 3GHz scaling far worse than the C2Q @ 3GHz.
 
harpoon said:
Right now, we can conclude that a Phenom 9600 is ~5% slower than if it had the NB/L3 running in sync. The performance hit with future higher clocked Phenoms would increase if AMD doesn't address this issue. [H]'s Phenom review showed this clearly, with the Phenom @ 3GHz scaling far worse than the C2Q @ 3GHz.
Click to expand...

I would suspect that the larger/faster caches of the C2Q are also partly the reason why it scales better?
In general, the faster your CPU gets, the higher the (absolute) difference between memory and CPU speed. However, the cache clocks along with the CPU. Since C2Q has a bigger and faster cache, it is less dependent on the memory, and therefore scales more linearly.

If you look at the dualcores, and take the 2 MB models compared to the 4 MB models, you'll see a similar pattern.

Which also explains why clocking the NB/L3 higher is still not enough to beat Kentsfield at higher clockspeeds. Phenom is still more dependent on the RAM than Kentsfield is, which hampers scaling.
We'll probably (hopefully?) see the 45 nm models scaling better because of the increased cache size aswell. Unless ofcourse AMD drops the clockspeeds of the NB/L3 even lower.
 
what happens when Intel is primed to steamroll AMD? What do we do? suddently the X86 market, the X86 R&D field, possibly humanities greatest achievement, is left alone to one player? Thats not good for users at all. Progress grinds to a hault. But I'm betting sanctions legal or otherwise, will be taken against Intel before their allowed their cornered market.

But I dont thinkt hats going to happen. Perhalps AMD will be forced into the low end market but they will keep a small fire lit at Intels feet, and will one day again in one, two, maybe three decades leapfrog Intel when Intel decides to go with their "Core5burst" architecture. Sooner or later Intel will slip up.

anyways, 1am ramblings of a tired student.

I hope Shanghai is able to push a consistant 3% lead over core 2, clock fore clock, and is able to scale to 3.5Ghz, forcing Intel to release a 3.6Ghz model at a reasonable price (with an unlocked multiplier).
 
MrWizard6600 said:
what happens when Intel is primed to steamroll AMD? What do we do? suddently the X86 market, the X86 R&D field, possibly humanities greatest achievement, is left alone to one player? Thats not good for users at all. Progress grinds to a hault. But I'm betting sanctions legal or otherwise, will be taken against Intel before their allowed their cornered market.
Click to expand...

I hope the end of AMD would mean that alternative architectures get another shot at stardom.
Mainly because of Windows, a lot of software is locked into the x86 architecture.
Now that .NET is starting to take shape, we are not that dependent on x86 anymore. Just as with opensource software, it will become easier to adopt other architectures. In the server world you see Sun and IBM being pretty successful with their alternative architectures, and even Intel's own Itanium still does quite well in some areas.

It would be nice if the end of competition on x86 also means the start of competition between x86 and other architectures.

I'm really glad that things didn't end up this way in the GPU-world. The shader language used in both Direct3D and OpenGL is universal, and is translated by the driver to the specific hardware architecture. This means that the designers have a lot more freedom in their designs, and that is partly why development in the GPU-world is much quicker than in the CPU-world.
Just look at all the different DX9-capable GPUs out there. They differ greatly in their architectures, but still they can all run DX9 applications.
I'd hate to think what GPUs would look like today if we got locked into 3dfx' Glide, and every year we'd just see faster iterations of the same Voodoo architecture, with the same low quality shading.

So I think we need something like the shader language, but for CPUs. Java or .NET are the closest we have to that. I hope that Microsoft's next Office is written for .NET, so it could be the 'killer app' that .NET needs in order to succeed.
 
MrWizard6600 said:
I hope Shanghai is able to push a consistant 3% lead over core 2, clock fore clock, and is able to scale to 3.5Ghz, forcing Intel to release a 3.6Ghz model at a reasonable price (with an unlocked multiplier).
Click to expand...
Shanghai catching up to Kentsfield clock for clock will be a miracle. I wouldn't hold my breath over that lowered bar either.
 
This thread has absolutely nothing to do with interpreted languages. The day that interpreted languages becomes the norm is already here in most OS's. And it aint Java or .NET. Instead we have an array of specialized languages like Python and PHP, Perl and CSS, XMl, and a whole bunch of others, including Java, and .NET, but they will never be the sole survivors. They are both totally inadequate In my opinion you'd be better of in moth cases using Python, or Perl, and in the case of Javascript, then you'd be better of using CSS or PHP..

As was said before however this has nothing to do with this thread, so lets try to keep this tread on topic, and it is about Shanghai.
 
I WANT...

MOAR.jpg


IPC DAMMIT! :p
 
duby229 said:
This thread has absolutely nothing to do with interpreted languages. The day that interpreted languages becomes the norm is already here in most OS's. And it aint Java or .NET. Instead we have an array of specialized languages like Python and PHP, Perl and CSS, XMl, and a whole bunch of others, including Java, and .NET, but they will never be the sole survivors. They are both totally inadequate In my opinion you'd be better of in moth cases using Python, or Perl, and in the case of Javascript, then you'd be better of using CSS or PHP..
Click to expand...

Java and .NET aren't interpreted, nor are they scripting languages, like Python, PHP, Perl etc. They try to scratch a different itch.
CSS and XML aren't even programming languages at all.

duby229 said:
As was said before however this has nothing to do with this thread, so lets try to keep this tread on topic, and it is about Shanghai.
Click to expand...

Everything about Shanghai has already been said anyway.
What's wrong with some technical discussion of what may be in the future? It does affect AMD, and perhaps Shanghai. We're on a tech forum right? We're enthousiasts on technology right?
 
Then create an appropriate thread in the appropriate section instead of derailing this one.

And scratching different itches is the entire point, and is the very reason I said Java and .NET are totally inadequate. Python, Perl, and to a lesser extent Mono, can fill the same needs that Java and .NET fill, while being far more flexible, and other languages like XML, CSS, and others fill other needs that Java and .NET arent even capable of.

Making uneducated blanket statements like that are silly at best.
 
Mr. Bluntman said:
Based on the die shots, it seems that Shanghai is nothing more than a die shrink with an added 4MB of L3 cache and possibly some IMC tweaks. The processing elements themselves look un-changed. Hence the ass-handing will commence...

I put great faith in Barcelona and Agena, so much so that I believed AMD's word to a "T" that it was going to be 10% faster than Kentsfield. That turned out to be complete bullshit, and more often than not, was barely faster than the existing K8 core. AMD needs a new architecture to compete, and if the roadmaps are correct, we won't see that until 2010. I've pretty much written off AMD until they prove otherwise; my X2 5400 being my last AMD purchase for a very, very long time.
Click to expand...

because the k9 failed and AMD had to try to tweak the K8 and tweak some more....AMD will not be back in the game for the high end until the totally new core k11 hits...

and that wont be for some time
 
You must log in or register to reply here.
Back
Top