AMD says new 'Shanghai' chip is ready to go

[Idontcare]

"We're in full production right now in the factory," he said. "People will start getting first silicon from the final production very shortly."

Patla asserted that Shanghai is a "very power efficient product" and will perform much better than Barcelona because the smaller 45-nanometer process yields "a lot more (clock) frequency."

At the same frequency (speed), Shanghai will outperform Barcelona by about 20 percent, Patla said.

http://news.cnet.com/8301-13924_3-10054038-64.html

20% :shocked: Man oh man I sure hope so!

And they've got an AMD roadmap in there, 6-core Istanbul coming Q4/09.

Magny-Cours and Sao Paulo Q2/10. (32nm?)

 

[extra]

Well, I like competition and I hope this is true, it'd be nice to see them deliver again like the 4xxx series video cards, but honestly I have my doubts I'm betting 10% in most situations.

 

[blckgrffn]

That's good news. Let's hope it substantially outscales barcelona as far as clock speeds go. I mean, they need this thing to do 4+ Ghz...


Nat

 

[Lonyo]

Man, Q4/09 is a long time away for "only" 6 cores. It wouldn't surprise me to see 8 core/16 thread Nehalem coming by then.
Intel already managed 6 cores with Dunnington, and since Nehalem is supposed to be fairly modular, they should be able to get 6 or 8 cores before then.

Good news that they are finally ready to go though. This improves the AMD platform, which is arguably better in all ways except CPU (mobo/graphics - integrated or not).
As long as the pricing is right it could potentially steal away from Intel in the consumer market, especially at the lower end where Nehalem won't really compete too much initially.

Potentially the most interesting aspect is the long continued support for DDR2 over DDR3.
Looks like DDR3 for AMD won't happen until 2010, which is the sort of timeframe I think where DDR2 and DDR3 will reach parity (if I remember some chart correctly).
Again, with continued CDDR2 support that gives AMD the short term price advantage over Intel for end users.

 

[bradley]

AM3 motherboards are being released in 09Q2, around March timeframe, and will support DDR2 and DDR3 using AM3 chips.

 

[jones377]

where is bulldozer.....

 

[Idontcare]

Originally posted by: jones377
where is bulldozer.....

Everyone would like to know, including AMD :laugh:

 

[jones377]

Originally posted by: Idontcare

Originally posted by: jones377
where is bulldozer.....

Everyone would like to know, including AMD :laugh:

Well this roadmap runs all the way through 2010 at least, so if it shows up in 2011 it will run straight into Sandy Bridge. SSE5 vs AVX, won't even be a contest. It will be like 3dnow vs SSE all over again.

 

[Phynaz]

At the same frequency (speed), Shanghai will outperform Barcelona by about 20 percent, Patla said.

See sig.

 

[coolpurplefan]

Ack! Did you say more GHz? And 20% faster clock-per-clock? Yum! CPU-hungry games will love it!

 

[Idontcare]

Originally posted by: Phynaz

At the same frequency (speed), Shanghai will outperform Barcelona by about 20 percent, Patla said.

See sig.

I suppose this time they didn't say what specific metric Shanghai is going to outperform Barcelona.

He could simply mean power consumption is reduced 20% at the same clockspeed.

You sound like you got fooled once and can't get fooled again :laugh:

 

[Foxery]

Amazing news. AMD can afford Christmas this year after all!

 

[Martimus]

Thanks for the article. That was a very nice read. Hopefully we will get an Anandtech review soon on the server chip, and that will help us understand everything that was improved. I still hold out hope that the IMC speed will be greatly increased (I would like to see a 25% increase to 2.5GHz.)

 

[aldamon]

Originally posted by: Foxery
Amazing news. AMD can afford Christmas this year after all!

Don't count on that.

 

[geoffry]

While I do like this and think it is closer to the truth this time around I have one point.

Didn't AMD float some benches of a 40% clock for clock Barcelona improvement over Kentsfield before its release? They cherry picked the 1 thing that they did well and "forgot" to mention that its slower clock for clock in pretty much everything else, lol.

 

[Martimus]

Originally posted by: Idontcare

Originally posted by: jones377
where is bulldozer.....

Everyone would like to know, including AMD :laugh:

Is Sao Paulo the server version of Bulldozer, or is it just the DDR3 enabled version of Istanbul?

 

[Cookie Monster]

I dont think it would be such a wise idea for AMD to falsely state about shanghai's performance increases at this point in time..

 

[sonoran]

Originally posted by: geoffry
While I do like this and think it is closer to the truth this time around I have one point.

Didn't AMD float some benches of a 40% clock for clock Barcelona improvement over Kentsfield before its release? They cherry picked the 1 thing that they did well and "forgot" to mention that its slower clock for clock in pretty much everything else, lol.

To be honest, Intel does almost the same thing - except that Intel will state "up to" X% improvement, where X% is the best improvement you see in any test. I guess you can't blame a company for advertising what they do best - as long as it's not an outright lie.

Anyone who's watched this game for a while should know to wait on results from a range of tests, regardless of whose chip is under discussion, if they want the real story on overall performance.

 

[Martimus]

Originally posted by: Idontcare

Originally posted by: Phynaz

At the same frequency (speed), Shanghai will outperform Barcelona by about 20 percent, Patla said.

See sig.

I suppose this time they didn't say what specific metric Shanghai is going to outperform Barcelona.

He could simply mean power consumption is reduced 20% at the same clockspeed.

You sound like you got fooled once and can't get fooled again :laugh:

To be honest, I don't even know what that quote means, since it isn't a direct quote and who knows what context Patla was talking to the interviewer about. I know that I have been misquoted almost every time I have been interviewed by a newspaper/magazine, so it is hard to know what this particular sentence really means. I think IDC's statement of it having a 20% improvement in power consumption at the same clock speed it the most likely.

 

[Idontcare]

Originally posted by: Martimus

Originally posted by: Idontcare

Originally posted by: jones377
where is bulldozer.....

Everyone would like to know, including AMD :laugh:

Is Sao Paulo the server version of Bulldozer, or is it just the DDR3 enabled version of Istanbul?

http://www.dailytech.com/Hello...t+G34/article12400.htm

All the rumors to-date have Sao Paulo being an MCM'ed Shanghai (2x4) and Magny-Cours being an MCM'ed Istanbul (2x6). All will be K10.5 per the speculation to date.

My expectation though for Sao Paulo and Magny-Cours is even if they are K10 derivatives they will likely need to be 32nm (so maybe another 10-20% higher IPC over 45nm Shanghai K10.5 architecture?) just to make the TDP per socket make any sense unless the clockspeeds are going to be sub-2GHz.

 

[Martimus]

Originally posted by: Idontcare

Originally posted by: Martimus

Originally posted by: Idontcare

Originally posted by: jones377
where is bulldozer.....

Everyone would like to know, including AMD :laugh:

Is Sao Paulo the server version of Bulldozer, or is it just the DDR3 enabled version of Istanbul?

http://www.dailytech.com/Hello...t+G34/article12400.htm

All the rumors to-date have Sao Paulo being an MCM'ed Shanghai (2x4) and Magny-Cours being an MCM'ed Istanbul (2x6). All will be K10.5 per the speculation to date.

My expectation though for Sao Paulo and Magny-Cours is even if they are K10 derivatives they will likely need to be 32nm (so maybe another 10-20% higher IPC over 45nm Shanghai K10.5 architecture?) just to make the TDP per socket make any sense unless the clockspeeds are going to be sub-2GHz.

Thanks. Although this article contradicts the Sao-Paulo being a MCM 2X4 core processor, saying that the it will be a six core processor and Magny-Cours will be a MCM Sao Paolo. To be honest, I don't know which one to believe. Daily Tech is wrong about the minor details like that all the time, but I don't know much about Softpedia, so they could also be full of donkey-doo.

 

[thilanliyan]

Thanks for the link...hopefully Deneb will come soon too and then my switch to AMD right now doesn't seem like such a bad idea. Honestly, the only thing I DON'T like about the Phenom chips is the power consumption...other than that...I have my 9850 clocked to 3GHz on the stock cooler and it performs very well...don't regret it one bit (except the damned Gigabyte motherboard that has cold boot issues with ACC on...hopefully a BIOS fix is coming :-( ).

 

[Idontcare]

Originally posted by: Martimus

Originally posted by: Idontcare

Originally posted by: Martimus

Originally posted by: Idontcare

Originally posted by: jones377
where is bulldozer.....

Everyone would like to know, including AMD :laugh:

Is Sao Paulo the server version of Bulldozer, or is it just the DDR3 enabled version of Istanbul?

http://www.dailytech.com/Hello...t+G34/article12400.htm

All the rumors to-date have Sao Paulo being an MCM'ed Shanghai (2x4) and Magny-Cours being an MCM'ed Istanbul (2x6). All will be K10.5 per the speculation to date.

My expectation though for Sao Paulo and Magny-Cours is even if they are K10 derivatives they will likely need to be 32nm (so maybe another 10-20% higher IPC over 45nm Shanghai K10.5 architecture?) just to make the TDP per socket make any sense unless the clockspeeds are going to be sub-2GHz.

Thanks. Although this article contradicts the Sao-Paulo being a MCM 2X4 core processor, saying that the it will be a six core processor and Magny-Cours will be a MCM Sao Paolo. To be honest, I don't know which one to believe. Daily Tech is wrong about the minor details like that all the time, but I don't know much about Softpedia, so they could also be full of donkey-doo.

The only issue I have with that logic is that it makes no room for, or use of, the 6-core Istanbul, which is clearly on AMD's roadmap as shown in the OP link.

Why would they have two 6-core chips listed as if they were distinctly different products?

...Unless...Sao Paulo is a native 6-core 32nm chip from which Magny-Cours is MCM'ed. Now that would make a whole lot more sense from a TDP/socket consideration.

 

[Martimus]

Originally posted by: Idontcare
The only issue I have with that logic is that it makes no room for, or use of, the 6-core Istanbul, which is clearly on AMD's roadmap as shown in the OP link.

Why would they have two 6-core chips listed as if they were distinctly different products?

...Unless...Sao Paulo is a native 6-core 32nm chip from which Magny-Cours is MCM'ed. Now that would make a whole lot more sense from a TDP/socket consideration.

I really don't know. It is posible that Sao Paolo would be the 32nm shrink of Istanbul (It does appear to follow it by ~ 6 months, so that is feasible - for a plan anyway). But it could be exactly as you said, and be a 2x4 core MCM, and just be a huge die with low clock speeds, which seems equally feasible (or maybe even moreso given the timeframe).

 

[taltamir]

ok, so shanghai is shipping... so what is it called? phenom X4 something?