Thursday, April 19, 2007

AMD's 45nm Shanghai quad core running

The 45nm Shanghai is at least 40% faster than Penryn. The low power 65nm K10 can do 1.9GHZ at 45 watts, about 10 watts per core.

AMD's game plan is simple. Barcelona will take the server market, render all Intel Xeons total junk. The 4x4 platform will get Barcelonas, so high end gamers will be happy. On the lower end, high clock K8 dual cores will be competing with Intel's Core 2 in the $100 market.

Patty will be out on the streets by 2Q08.

Intel's collapse will be fast and violent.

33 Comments:

Blogger Ho Ho said...

"The 45nm Shanghai is at least 40% faster than Penryn"

versus

"K10 core will have 40% IPC lead over Core 2 on integer (per core).

K10 will have 200% IPC lead over Core 2 on floating point (per core)."


Are you saying that going from 65nm to 45nm will make it slower?


"The low power 65nm K10 can do 1.9GHZ at 45 watts, about 10 watts per core"

Intel has had 1.86Ghz 50W quadcores for some time, you know. That is a whole 40MHz less than K10.

12:43 AM, April 19, 2007  
Blogger ThomasHOL said...

You seem to be forgetting that gamers don’t need multicore CPU’s as most games don’t even take advantage of a dualcore CPU. It will be a long time until games will need more than 2 cores.

And could you cut out the BK bulshit. Even if Intel shut down everything now they would have money to last till Q208. The reality of the things is that Intel is showing strengths and AMD is having serious cash flow problems and we will have to wait for Q4 for that to change. I personally hope that Barcelona will be a strong chip as I don’t see AMD continuing as today if Barcelona is not a real killer. The thing is that Intel will be able to close much of the gap by using higher clockspeeds. And before you tell us that it is old then please look at the numbers, people don’t care how the performance comes as long as the CPU is fast and don’t use to much power.

1:51 AM, April 19, 2007  
Blogger Tanrack said...

Any news on how AMD is doing with getting 50% market share by en of 2006? From http://sharikou.blogspot.com/2006/03/amd-poised-to-exit-2006-with-55-market.html

5:48 AM, April 19, 2007  
Blogger Tim said...

You quite Fudzilla? You really ARE that stupid....

6:31 AM, April 19, 2007  
Blogger Evil said...

Keep dreaming Ph(ake)d.


So how was AMD's Q this time? Oh wait, not so good.

So how was AMD's market share this Q?

Oh wait, they lost some back to Intel.

You really do live in a dream world don't you?

7:47 AM, April 19, 2007  
Blogger DaSickNinja said...

"The 45nm Shanghai is at least 40% faster than Penryn"

Based of of what? A site with only text, no links, no benches, not even a damn slide show, and a dubious history of AMD bias?

Besides, where are you pulling this 50% better number from?

8:39 AM, April 19, 2007  
Blogger DaSickNinja said...

My bad, that 40% better number. But the same question applies. Where is your evidence. The link to Fudzilla only mentioned TDP, not any performance numbers. Not even on what spec bench.

8:41 AM, April 19, 2007  
Blogger Ho Ho said...

Whats the difference what number he sais? There is exactly no information backing it up anyways.

9:09 AM, April 19, 2007  
Blogger PENIX said...

Tim said...

You quite Fudzilla? You really ARE that stupid....


When you attempt to insult someone's intelligence, you really need to proofread your post.

9:25 AM, April 19, 2007  
Blogger PENIX said...

ho ho said...
Are you saying that going from 65nm to 45nm will make it slower?


I don't understand how you came to this conclusion. He compares one to Penryn and one to the Core 2.

9:27 AM, April 19, 2007  
Blogger PENIX said...

ThomasHOL said...

You seem to be forgetting that gamers don’t need multicore CPU’s as most games don’t even take advantage of a dualcore CPU. It will be a long time until games will need more than 2 cores.


This is quickly changing. Since multi-core is now the standard, games are now adopting it. Many games are developed for pc and consoles, and share most of their core engine code. Both the XBox 360 and the PS3 are multi-core systems. Most of the games coming out for these systems are designed for multi-core in some respect.

9:31 AM, April 19, 2007  
Blogger PENIX said...

Ho Ho said...

Whats the difference what number he sais? There is exactly no information backing it up anyways.


This is true. No information has been given to Sharikou's true sources. I suspect that Sharikou has inside connections which he does not mention to protect them. How else is it that Sharikou manages to make so many predictions that seem unlikely, but turn out to be true?

9:35 AM, April 19, 2007  
Blogger Giant said...

This is true. No information has been given to Sharikou's true sources. I suspect that Sharikou has inside connections which he does not mention to protect them. How else is it that Sharikou manages to make so many predictions that seem unlikely, but turn out to be true?

Like how AMD would exit 2006 with 50% market share? Or how Intel would post "massive operational losses" in Q3'06 and beyond? (For the record, in each quarter that Intel was supposed to post a loss they've posted profit of well over 1bn each time!)

9:52 AM, April 19, 2007  
Blogger Ho Ho said...

penix
"He compares one to Penryn and one to the Core 2."

So K10 is 200% faster than Conroe and Shanghai is 40% faster than Penryn then wouldn't that mean that Penryn is roughly 160% faster than K10?


"How else is it that Sharikou manages to make so many predictions that seem unlikely, but turn out to be true?"

What has come true? Definitely nothing about BK or marketshare.
I can make a truckload of predictions also, perhaps couple of them will come true too.

9:53 AM, April 19, 2007  
Blogger abinstein said...

"The 45nm Shanghai is at least 40% faster than Penryn. The low power 65nm K10 can do 1.9GHZ at 45 watts, about 10 watts per core."

Sharikou... just how credit is the FUDzilla news (the correct link)? FWIK, AMD's 45nm requires immersion while its 65nm doesn't. How can these two be made without changing the fab?

10:00 AM, April 19, 2007  
Blogger DaSickNinja said...

Penix:
This is true. No information has been given to Sharikou's true sources. I suspect that Sharikou has inside connections which he does not mention to protect them. How else is it that Sharikou manages to make so many predictions that seem unlikely, but turn out to be true?

You got any definite proof of this?

10:24 AM, April 19, 2007  
Blogger DaSickNinja said...

Ho Ho
Whats the difference what number he sais? There is exactly no information backing it up anyways.

Considering the validity of the arguments of those who support the good "doctor", Any slight mistake has to be corrected less they focus on that and not the actual issue.

10:25 AM, April 19, 2007  
Blogger Bubba said...

Congrats AMD!

In just a couple of hours you will announce that you have managed to wipe 40 years of effort off the books in a mear six months.

Nice accomplishment.

10:27 AM, April 19, 2007  
Blogger PENIX said...

Ho Ho said...
So K10 is 200% faster than Conroe and Shanghai is 40% faster than Penryn then wouldn't that mean that Penryn is roughly 160% faster than K10?


I do not know how you came to your conclusion based on your description. Please spell out each step to your conclusion. Also, please keep in mind that it was unspecified if the 40% was int, fp, or over all.

10:29 AM, April 19, 2007  
Blogger Ho Ho said...

I just used the numbers from Sharikous latest posts. You can do the math yourself if you want to and point out where is the mistake.

10:43 AM, April 19, 2007  
Blogger DaSickNinja said...

Penix, that doesn't take much learning to figure out. Basic arithmetic and word problem solving. You know, the kind you learn in 4th grade.

10:55 AM, April 19, 2007  
Blogger Dahaka said...

Penryn is not much better than kentsfield only %10-15.

Intel is thinking penryn will beat k10 but they are wrong.:)

K10 is faster than kentsfield at %40-50 .

K10 will faster than penryn at %30-35.

Intel will be doomed.

Intel will defeat.

11:18 AM, April 19, 2007  
Blogger Bubba said...

"Intel will defeat AMD".

Fixed that for you.

11:40 AM, April 19, 2007  
Blogger Dahaka said...

@Bubba you will see!

Intel will end.

AMD will defeat intellllll!

Wait and see!

IDIOT!

12:11 PM, April 19, 2007  
Blogger enumae said...

This comment has been removed by the author.

12:29 PM, April 19, 2007  
Blogger Ho Ho said...

"@Bubba you will see!"

So why not sit quietly and wait? Assuming AMD won't push it back any more there isn't that long waiting. Who knows, perhaps they even have the courage to release some real benchmarking results!

12:30 PM, April 19, 2007  
Blogger PENIX said...

Ho Ho said...

I just used the numbers from Sharikous latest posts. You can do the math yourself if you want to and point out where is the mistake.

DaSickNinja said...

Penix, that doesn't take much learning to figure out. Basic arithmetic and word problem solving. You know, the kind you learn in 4th grade.


Sharikou said...

The 45nm Shanghai is at least 40% faster than Penryn.

K10 core will have 40% IPC lead over Core 2 on integer (per core).

K10 will have 200% IPC lead over Core 2 on floating point (per core).

Anandtech said...

Penryn vs. Conroe improvements to be in the 5 - 10% range at minimum depending on the application.

Sharikou said...

Anand showed almost %0 performance boost in modern 3D games.


Once again, the logic of Intelers has been hindered by low IQ. I tried to offer the oportunity to correct the obvious flawed logic, but my offer was rejected. Once again, I will educate the Intelers on the basics of logic and reasoning. Please pay attention.

From the information given we can gather the following:

Core 2: 100% (our base for comparisson)
Penryn: 110% (0% to 10% increase, but we will assume 10%)
K10: 140% (40% to 200% depending on test, we will assume 40%)
Shanghai: 154% (40% gain over 110% penryn)

A die shrink from 65nm to 45nm will not result in a huge gain. For both Intel and AMD, the results are similar; 10% gain for penryn and a 14% gain for Shanghai.

12:36 PM, April 19, 2007  
Blogger BONER said...

"You seem to be forgetting that gamers don’t need multicore CPU’s as most games don’t even take advantage of a dualcore CPU. It will be a long time until games will need more than 2 cores."

There's already games that support 2 and more cores such as Supreme Commander and in the coming months many more games which can utilize four cores for the best gameplay will be out like Alan Wake.

3:06 PM, April 19, 2007  
Blogger PENIX said...

BONER said...
There's already games that support 2 and more cores such as Supreme Commander and in the coming months many more games which can utilize four cores for the best gameplay will be out like Alan Wake.


The X360 is 3 core, the PS3 is 9 core. Multi-core is already here.

3:16 PM, April 19, 2007  
Blogger abinstein said...

"Penix, that doesn't take much learning to figure out. Basic arithmetic and word problem solving. You know, the kind you learn in 4th grade."

which apparently and unfortunately you haven't graduated it from?

now again, where does that 200% come from?

9:26 PM, April 19, 2007  
Blogger Ho Ho said...

penix
"Core 2: 100% (our base for comparisson)
Penryn: 110% (0% to 10% increase, but we will assume 10%)
K10: 140% (40% to 200% depending on test, we will assume 40%)
Shanghai: 154% (40% gain over 110% penryn)"


From Sharikou posts:

"K10 core will have 40% IPC lead over Core 2 on integer (per core)."
"K10 will have 200% IPC lead over Core 2 on floating point (per core)."
"Anand showed almost %0 performance boost in modern 3D games"
"The 45nm Shanghai is at least 40% faster than Penryn"


One thing that is clear is that "Fundamentally, Shanghai is just a dumb shrink coupled with some more cache" as it won't give performance increase over K10. Had I used the 200% FP IPC number conclusion would have been that Shanghai is considerably slower than K10.

Don't blame me, I just use the numbers our beloved host and entertainer gave us. As we all know, he always knows what he sais and you must not doubt in his words!


"The X360 is 3 core, the PS3 is 9 core."

Actually, 360 has three cores that can run two threads in parallel. Cell in PS3 has one PPC core that can run two threads in parallel and 7 SPU's that run single thread.

Still I agree that multicore is here. It has been for quite some time, actually. HT gave up to 30% performance increase to P4's and that was years before real x86 dualcores.

5:19 AM, April 20, 2007  
Blogger PENIX said...

ho ho said...
Actually, 360 has three cores that can run two threads in parallel. Cell in PS3 has one PPC core that can run two threads in parallel and 7 SPU's that run single thread.


Technically, the PS3 has 1 PPE core and 8 SPE cores, but 1 SPE core is "disabled to improve manufacturing yield". I have not been able to find out what exactly that means, or if the last SPE is even functional.

8:20 AM, April 20, 2007  
Blogger Ho Ho said...

That last SPE is not functioning. You have only seven working SPUs in each PS3. It is basically the same as how GPU manufacturers disable some quads or CPU manufacturers disable parts of cache.

In the blades you buy from IBM or MC all Cells have 8 working SPUs.

8:55 AM, April 20, 2007  

Post a Comment

Links to this post:

Create a Link

<< Home