K8L uarchitecture to arrive mid 07

[Hard Ball]

The official word is, and has been for a while (if you are not aware), that the plan for K8L launch is mid-07; which is also what I have been saying for a while. It's not Q1 07, or Q1 08, but middle of next year.

http://www.informationweek.com/hardware...icleID=190900525&subSection=Processors

Coming from Dirk Meyer himself:

The battle for quad-core market position has already begun for processor rivals Advanced Micro Devices and Intel.

On a second quarter earnings conference call Thursday, AMD executives said the company's planned quad-core processor?based on what executives characterized as a new microarchitecture?will be demonstrated before the end of the year.

"We have a new microarchitecture under development and the first substantiation of that will be the quad-core to be launched in mid 2007," said Dirk Meyer, AMD's president and chief operating officer.

While AMD promises better financial results for Q3; I don't see how that can really happen. Given that they just slashed their entire desktop lineup prices by ~%50; they either would have to sell more than double the amount of Q2 for the desktop segment; or they would have to increase their sales in their other segments (Server/Workstation, Budget, Mobile); just don't see that happening.

Edit: sorry, forgot link;

 

[dexvx]

Link

http://www.tgdaily.com/2006/07/20/amd_to_accelerate_roadmap_to_45nm/

On other news, 45nm from Intel to debut Late 2007/Early 2008

 

[imported_inspire]

Thanks for the news - I wish we could get some news on when they plan to launch 4x4, I've been all over google and can only find some pretty shaky rumors.

 

[coldpower27]

Don't we know this already K8L debuing in Mid 2007? Though it's interesting to read this information again.

 

[twjr]

And i also thought it was only the server versions that were being released then and the desktop versions being released later, much like intel did with woodcrest then conroe.

 

[videoclone]

AMD is going to be dead in the water until they get out K10

 

[twjr]

Originally posted by: videoclone
AMD is going to be dead in the water until they get out K10

..............They always have 4x4.

 

[AnandThenMan]

Originally posted by: videoclone
AMD is going to be dead in the water until they get out K10

By that reasoning, Intel should have been dead years ago, considering they had inferior products for years, not just the predicted 2-3 quarters.

 

[twjr]

Originally posted by: AnandThenMan

Originally posted by: videoclone
AMD is going to be dead in the water until they get out K10

By that reasoning, Intel should have been dead years ago, considering they had inferior products for years, not just the predicted 2-3 quarters.

You have to remember how much smaller AMD is as a company compared to intel. also intel was always doing well in the mobile market and they make chipsets too. their entire earnings do not come from cpus.

Also the degree to which p4 was beaten was not so decisevly as the conroe has beaten a64. Their third tier cpu outperforms amd's flagship in nearly every benchmark.

 

[AnandThenMan]

Originally posted by: twjr
Also the degree to which p4 was beaten was not so decisevly as the conroe has beaten a64. Their third tier cpu outperforms amd's flagship in nearly every benchmark.

I'm not saying you are wrong here, but I bet you just made that up. I would take an educated guess and say the performance/watt/price of netbust vs. A64 is worse compared to A64 vs. Conroe.

edit: ESPECIALLY in the server space.

 

[ZOXXO]

Originally posted by: AnandThenMan

Originally posted by: twjr
Also the degree to which p4 was beaten was not so decisevly as the conroe has beaten a64. Their third tier cpu outperforms amd's flagship in nearly every benchmark.

I'm not saying you are wrong here, but I bet you just made that up. I would take an educated guess and say the performance/watt/price of netbust vs. A64 is worse compared to A64 vs. Conroe.

edit: ESPECIALLY in the server space.

What prices would you use in such an analysis?

 

[twjr]

Originally posted by: AnandThenMan

Originally posted by: twjr
Also the degree to which p4 was beaten was not so decisevly as the conroe has beaten a64. Their third tier cpu outperforms amd's flagship in nearly every benchmark.

I'm not saying you are wrong here, but I bet you just made that up. I would take an educated guess and say the performance/watt/price of netbust vs. A64 is worse compared to A64 vs. Conroe.

edit: ESPECIALLY in the server space.

What i mean is that when AMD released their A64's there were still some things that the P4 could do faster, whereas now there is nothing that an Fx62 can do faster than an x6800 and maybe even an e6700.

 

[Hard Ball]

Originally posted by: videoclone
AMD is going to be dead in the water until they get out K10

Typical fanboy talk.
IT's like saying that Intel will be dead in the water until they have Nehalem.

K8L will likely be around the same IPC as CMA in scalar int and Vector performance as CMA; while will have significantly better performance at FP. But Intel will still be ahead in terms of process technology next year. So what this basically means is that Intel will have an edge in Desktop markets, while AMD will be in servers/workstaions by the beginning of 08 (when Intel transitions to 45nm). Notebook will be more or less a crapshoot to predict; since AMD will have a dedicated mobile uarchitecture; and no one really knows any details or performance estimates yet.

 

[dexvx]

Originally posted by: Hard Ball
K8L will likely be around the same IPC as CMA in scalar int and Vector performance as CMA; while will have significantly better performance at FP.

Thats basically an unfounded extrapolation.

 

[Kromis]

Alright, in order to express how my opinion, I'm gonna quote a Magic: The Gathering card named "Sengir Vampire"...

"Empires rise and fall, but evil is eternal."

Lol...everybody has their ups and downs. I'm tired of this bullsh!t. It's just the way life is. Either you damn fanboys have something smart to say or else GTFO. Lol, sorry for my outburst there but it just had to come out...

 

[Hard Ball]

Originally posted by: dexvx

Originally posted by: Hard Ball
K8L will likely be around the same IPC as CMA in scalar int and Vector performance as CMA; while will have significantly better performance at FP.

Thats basically an unfounded extrapolation.

Extrapolation through these improvement on core throughput:
--increased width of the datapath for FP and Vector ops
--32Byte fetch
--larger size of BTB
--out of order and 128 bit ld into register file
--extension to the ISA
--extensible L3 to handle cache coherency traffic
--additional microcode flash ROM
--independent scaling of voltage for individual die components
--etc, etc.

and these on interface bandwidth and efficiency:
--48bit addressing and 1GB page
--memory mirroring
--HTX device support
--HTT3.0 w/higher bandwidth and unganging (splitting 16bit to 8bit links to maintain point to point configuration)
--FBDIMM support (may not be in initial version)
--etc, etc

All in all, on 65nm, it should about pull even with CMA at A/L ops, match the throughput of new Intel wide SIMD units; and definitely will be superior in FP, and will offer significant improvments at BP (not sure about compared to CMA's). And at the same time, this should allow for lower latency in memory access, much greater overall system bandwidth, one-hop NUMA supports, as well as much less problems with cache snoop than a Kentsfield or Clovertown. We will see what Tigerton brings.

 

[R3MF]

Originally posted by: twjr

Originally posted by: videoclone
AMD is going to be dead in the water until they get out K10

..............They always have 4x4.

means nothing if it's only usable with £700 FX chips.