AMD Resurrects K8 Architecture for 2008 Roadmap

[Zephyn]

Hey all check this out,

I pulled it from: http://www.dailytech.com/article.aspx?newsid=9899

AMD plans to keep "Brisbane" around, releases new chips based on it

Things at AMD may have gone from bad to worse with the lackluster Phenom launch in late November. Not only did Phenom fail to appeal to professional reviewers, but the company ended up removing one third of its CPU lineup just after the big day.

Last week AMD CEO Hector Ruiz vowed that the company would stop hemorrhaging cash and return to profitability soon. "That is our number one goal right now," Ruiz said in a conference in Bangalore.

Making a profit at AMD apparently means refocusing on its older K8 architecture. The company will introduce eleven 65nm K8 processors over the next two quarters. By comparison: AMD launched two quad-core K10 Phenom processors in November with three more scheduled over the next two quarters. Two tri-core Phenom processors will follow in March 2008.

Essentially, AMD will move any remaining Athlon 64 processors from the 90nm node to the 65nm node, with a few new frequency and TDP variations.

The AMD Athlon 64 X2 5600+ will be the first to jump on the new 65nm K8 bandwagon with a 65W TDP. The previous Windsor-based chip of the same featured an 89-Watt TDP. AMD will also add 100 MHz to the core frequency of the Athlon 64 X2 5600+, now rated at 2.9 GHz. Total L2 cache will be halved in the move to the Brisbane core, and the updated Athlon 64 X2 5600+ chips will feature only 1MB of L2 cache. Availability of these processors is scheduled for Q1 2008.

AMD's higher-end Athlon 64 X2 6400+ and Athlon 64 X2 6000+ will both be discontinued.

AMD will also update its "Energy Efficient" series and will release three new chips, the AMD Athlon 4850e, Athlon 4450e, and Athlon 4050e in Q2 2008. All of the new offerings will be based on AMD's Brisbane core and will feature a 45-Watt thermal envelope. AMD's current energy efficient "BE-2xxx" series will be phased out at that time. Respectively, the new chips will run at 2.5GHz, 2.3GHz and 2.1GHz.

All new Brisbane chips will be based on the Socket AM2 interface. These processors are compatible with AMD's AM2+ socket designated for Phenom processors.

 

[cfvh600]

Interesting

 

[pogsnet]

I think they should improve the Athlon X2 to better heights like putting Cool n Quite 2.0 and increasing the HT plus shrinking the die. Overall this is bugless idea. Hey "no L3 cache" please. L2 is more than enough. Just focus to improve the heat dissipation and power usage. And how about removing the memory controller on your chip... maybe thats the big bug why Intel dont do that and has no plans righht now. They say they will try only on 2009 but I guess they are testing that in labs right now but to no success, makes things slow.

 

[bwdsmart]

I think they should improve the Athlon X2 to better heights like putting Cool n Quite 2.0 and increasing the HT plus shrinking the die. Overall this is bugless idea. Hey "no L3 cache" please. L2 is more than enough. Just focus to improve the heat dissipation and power usage. And how about removing the memory controller on your chip... maybe thats the big bug why Intel dont do that and has no plans righht now. They say they will try only on 2009 but I guess they are testing that in labs right now but to no success, makes things slow.

well actually intel is gona use an imc this yeah on nehalm, and it greatly increases memory bandwidth. L3 cache is a great idea for amd because of the IMC, but the errata bug is whats making the l3 preatty much useless right now. if im wrong someone feel free to point that out to me but IMC is whats going standard for intel for and after nehalm.

...i think i just butcherd the spelling but oh well.

 

[Turas]

I thought nehalem coming in 2008 will have the integrated memory controller.

 

[bwdsmart]

it is, Q4 i believe is the planned release of consumer level products.

 

[spaztic7]

This is the first smart thing AMD has said in a long time.... Its funny how Hector sounds all deprest now.

 

[ethel]

This is great news - maybe they will go back to s939 so I can upgrade my Opteron

 

[BaronMatrix]

This is rather smart and something I expected. By cancelling the 90nm chips, they automatically get 2X the chips going from 100% 200mm wafers to 100% 300mm wafers.
That also means that IF the TLB bug is causing lower perf, then Kuma will sit perhaps 20% faster than X2 (it's already 10-15% faster).

That means that 2.9GHz K8s can be entry level while 2.9GHz Kuma will be midrange, 2.9GHz Tolliman will be entry level high end and 2.9GHz Phenom will be high end.

And there will be NO MORE 90nm Opterons, which means a cost savings from having the larger wafers, though the die is 25% larger for Barcelona.

And soon you will ll see why the timing for buying ATi was perfect. nVidia has not released a Phenom chipset so there may not have been one.

Since a lot of debt was incurred, it helps them to hit bottom faster so next year at this time, it's all gravy.

K8 has NOTHING to do with profitability except in terms of cost savings over 90nm. Remember that FabTech reported that Fab 30 is CLOSED to fully convert to 300mm @ 65nm. They may even put more 45nm there and leave Fab 36 as the 65nm fab. That would actually be the most efficient thing to do as Fab 36 just went to 100% 65nm.

Why rip it out now?

 

[spaztic7]

Yes, AMD stock is below $9. The market is affraid of what will happen next. People are jumping the ship before they are drownd with it.

It is nice that you still like amd and yes i think it is smart to bring out the K8 even though it is older (but faster then K10). But good lord Baron. You are the black knight from monty phthon holy grail. You have no more limbs. I seriously think this is AMD's final blow.

 

[xrider]

This is rather smart and something I expected. By cancelling the 90nm chips, they automatically get 2X the chips going from 100% 200mm wafers to 100% 300mm wafers.
That also means that IF the TLB bug is causing lower perf, then Kuma will sit perhaps 20% faster than X2 (it's already 10-15% faster).

That means that 2.9GHz K8s can be entry level while 2.9GHz Kuma will be midrange, 2.9GHz Tolliman will be entry level high end and 2.9GHz Phenom will be high end.

And there will be NO MORE 90nm Opterons, which means a cost savings from having the larger wafers, though the die is 25% larger for Barcelona.

And soon you will ll see why the timing for buying ATi was perfect. nVidia has not released a Phenom chipset so there may not have been one.

Since a lot of debt was incurred, it helps them to hit bottom faster so next year at this time, it's all gravy.

K8 has NOTHING to do with profitability except in terms of cost savings over 90nm. Remember that FabTech reported that Fab 30 is CLOSED to fully convert to 300mm @ 65nm. They may even put more 45nm there and leave Fab 36 as the 65nm fab. That would actually be the most efficient thing to do as Fab 36 just went to 100% 65nm.

Why rip it out now?

I am new to this forum but Baron your unbridled enthusiasm For AMD is laughable If I did'nt know any better I'd say you are a comedian. If AMD came out and said all of its fab where burning you'd say great AMD has found a way to solve all of the worlds energy problems. Bad news is bad news you can put lipstick on a pig............ I'll give you one thing AMD should seek you out and hire you for a PR position talk about spin.

 

[nightscope]

Sounds like AMD is trying to get the highest clock they can get from their k8 architecture to buy them some time... Sound familiar?

Smart thing to do nonetheless.

 

[zenmaster]

This is a smart move by AMD, but it speaks VOLUMES about the problems with the K10.

The 5000+ BE shows that AMD is getting faster cooler chips.
If they can keep improving, they should be able to make some money.

 

[weskurtz81]

Is anyone aware of the fact that AMD did a "stop ship" on all K10 processors? They will still sell a few, but only to screened customers. They will not have the bug figured out until the B3 revision, but it is very bad news.

 

[speedemon]

Why not a Quad k8?

 

[cnumartyr]

This is great.. I might pick up a 5000+ BE to replace my P4 machine.

 

[3Ball]

I am new to this forum but Baron your unbridled enthusiasm For AMD is laughable If I did'nt know any better I'd say you are a comedian. If AMD came out and said all of its fab where burning you'd say great AMD has found a way to solve all of the worlds energy problems. Bad news is bad news you can put lipstick on a pig............ I'll give you one thing AMD should seek you out and hire you for a PR position talk about spin.

Yea man get used to this...all of his posts will be in very similar fashion.

Best,

3Ball

 

[3Ball]

Why not a Quad k8?

Because of the way the cache and imc work on the K8's it wont allow for an efficient "glued" together quad such as the FSB on the Intel's allows and would require a whole revamp of the arch...which is what K10 is, only its having a few problems...to say the least! lol

P.S. This is just a general/simply put way...it is slightly more technical than this.

Best,

3Ball

 

[chookman]

I'll give you one thing AMD should seek you out and hire you for a PR position talk about spin.

LOL

I think everyone is rooting for AMD to pull something out of the bag to make things competitive again, hopefully this move will close the gap a little and keep losses to a mininum while they fix the issues with Phenom and crew. The only problem with this is by the time AMD get it right we could be seeing Nehalem's release and if it does anything close to whats been touted so far, AMD might have to start over yet again.

 

[Mathos]

Kinda both bad and good news indeed. I don't like the fact that they're movings everything to Brisbane is going to lower the L2 cache size though. As was seen in the 5000+ black edition reviews, even at 3.2 it still performs slower than the 6400+, due to having half the L2 cache. If they were gonna phase out the 90nm core, but put 1+1mb caches on the brisbane variants I'd be all for it. Just means I'll be buying an AM2+ board, and probably a 6400+ black edition here in the next couple weeks. Though, I'd prefer to wait for more manufacturers to get AM2+ boards out other than gigabyte.

This is probably the best move for AMD right now, as they may be able to push k8 for a bit more performance, on a revised brisbane core. In the end with the k10 though, we all know even with b3 stepping they may not be able to fix the L3 cache problems. I personally believe that the only thing that will fix this chip and it's performance problems is switching to high K hafnium or AMD's/IBM's Germanium Version of that.

*For those who haven't taken electrical engineering classes in college. Germanium is a material commonly used in high performance diodes and resistors.

 

[speedemon]

Because of the way the cache and imc work on the K8's it wont allow for an efficient "glued" together quad such as the FSB on the Intel's allows and would require a whole revamp of the arch...which is what K10 is, only its having a few problems...to say the least! lol

P.S. This is just a general/simply put way...it is slightly more technical than this.

Best,

3Ball

Thanks , I was hopin it was as easy as a revamped mem controller

 

[BaronMatrix]

I am new to this forum but Baron your unbridled enthusiasm For AMD is laughable If I did'nt know any better I'd say you are a comedian. If AMD came out and said all of its fab where burning you'd say great AMD has found a way to solve all of the worlds energy problems. Bad news is bad news you can put lipstick on a pig............ I'll give you one thing AMD should seek you out and hire you for a PR position talk about spin.

What are you talking about? I laid out what they need to do to keep selling CPUs. That's not spin. Since Ars Technica is saying that they are installing Fab 38 as 45nm, I guess I know what I'm talking about.

 

[BaronMatrix]

This is a smart move by AMD, but it speaks VOLUMES about the problems with the K10.

The 5000+ BE shows that AMD is getting faster cooler chips.
If they can keep improving, they should be able to make some money.

How does it do that? It shows that AMD wants to have multiple levels of perf for different power envelopes and use scenarios. They have to get rid of 90nm, so rather than only having K10 variants, they can have K8 for the low end and ULV mobile. I mean mobile doesn't really need L3 but it could benefit from HT3 and split power planes.

ISuppli shows that AMD doubled Intel's share increase for Q3, so I think they will still be selling enough CPUs. They just need to gain more cost savings with larger wafers and smaller processes.

Oh yeah, I forgot. They just suck right now and I shouldn't excuse them, but rather sing them a lullaby.

 

[TechnologyCoordinator]

What are you talking about? I laid out what they need to do to keep selling CPUs. That's not spin. Since Ars Technica is saying that they are installing Fab 38 as 45nm, I guess I know what I'm talking about.

What the hell is a 2.9 ghz Phenom? I haven't heard of such a thing?


If putting K10 on hold and concentrating on last generation technology will return them to profitability, then great. However, I just think this is desperation. Perhaps they have better yeilds with K8 and will be able to make money, who knows.

 

[BaronMatrix]

Because of the way the cache and imc work on the K8's it wont allow for an efficient "glued" together quad such as the FSB on the Intel's allows and would require a whole revamp of the arch...which is what K10 is, only its having a few problems...to say the least! lol

P.S. This is just a general/simply put way...it is slightly more technical than this.

Best,

3Ball

That's not correct. Direct Connect actually makes it easy as you can have a coherent connection between the packages. That's why they can make a tri-core. Montreal is set to be an 8 core MCM with two Shanghai chips.