Welcome Guest ( Log In | Register )

Outline · [ Standard ] · Linear+

 AMD Puma, a success or a failure?

views
     
cks2k2
post May 24 2007, 04:12 PM

...
******
Senior Member
1,966 posts

Joined: Jan 2003
From: No longer hanging by a NUS

Puma isn't out yet so we can't really compare a future product to existing stuff like Centrino.

Seems like AMD is copying Intel (souped up P3 for P-M) by rehashing an older architecture for the mobile space; which isn't a bad thing since the K8 is a pretty solid architecture.

QUOTE
well centrino was based on P2 n P3 technology...

Pentium-M was based off work done on a heavily souped up P3 derivative called Tinma which was never released. Tinma even had an IMC.

This post has been edited by cks2k2: May 24 2007, 04:15 PM
cks2k2
post May 25 2007, 01:40 PM

...
******
Senior Member
1,966 posts

Joined: Jan 2003
From: No longer hanging by a NUS

QUOTE(charge-n-go @ May 25 2007, 09:54 AM)
Currently Intel has the best Desktop and Mobile lineup. Server is a half-half, depends on the server configurations.

Next year, the story might be a lil different. If Barcelona family has the performance as promised, AMD will dominate desktop and server performance again. Mobile is a half-half though, with Penryn vs Turionx2.

FYI, Core2Duo mobile processor is more energy efficient than Turionx2, esp. the upcoming Penryn series.
*
QUOTE(Kagaya @ May 25 2007, 10:57 AM)
I wonder how TS could start a 'speculating' thread on something JUST paper-launched. Counting the chicks before the eggs? doh.gif

So what with the use of older technology with a lil' upgrades, improvisations? It depends on the feasibility and the potential of the previous technology where it still hasn't reached its headroom yet. This is cost saving and innovation. Using completely obsolete technology is when no improvisation occured but rebranded the LOGO or CALLNAME under a new barebone chassis, that's obsolete.

For the moment it seems like the TurionX2 power management architecture won't be able to counter the 45nm Penryn. I believe the level of miniaturization as well as using High-K material would helps a lot in the power department for Penryn, despite the success shown by Core2 technology.

There's still a long way to go for AMD's 65nm technology to go into a mature state before venturing into 45nm parts, as laid in their recent roadmap, but surprises keeps on popping up for 2008. Especially the decision on whether to finally include DDR3 support for their AM2+ socket. P35 is already out but DDR3 is still way too expensive, perhaps we need the industry to start reassign some DRAM pipelines dedicated to DDR2 reshuffled to DDR3+DDR2 production pipeline to achieve a better economic of scale.

It seems like PUMA is offering a more general, power-management-oriented platform for the industry player. The best part is  the DUAL GRAPHIC implementation for different settings; PCI Graphics when in AC Power Mode and On-Chip (Fusion) or IGP (Northbridge built-in) when in Power Saving Mode. This is quite brilliant, as long as they don't bump into a driver issues laugh.gif

The second thing about the dedicated current plane allows for more sophisticated power controlling scheme. If I'm not mistaken, VIA C7 got 2 time-clock running at different frequency with one at nominal speed the other at a lower speed thus speed-changing is just the immediate change of clock speed rather than voltage level changing or FSB changing. The similar concept would be like turning OFF one of the plane to turn OFF some transistors on chip to achieve better granularity.

I'm quite bored that the mobile platform is just about Centrino and the latest Centrino Pro, perhaps some Puma/Griffin offering will offset the power in mobile computing.
As a matter of fact, IT NEEDED, but because of the power-sapping nature of such core, it's not suitable for mobile computing. If the same could occurs to K8 where it'll be tweaked and improvise for a better mobile computing part, then the same could occur to K10 in the near future as soon as the technology matured on the desktop/server scheme.

On the battery part, I can't agree with you more. Li-On battery these days, though up to 9-cells, aren't really that ENOUGH, it seems like everyone wishes to achieve a double-digit operating hours at full speed, no downclocking and no more KABLOOIE scared laugh.gif thanks to SONY, now everyone is afraid the battery will explode while operating the lappy on the lap, afraid that it'll kill the bird, burn the nest and smash the eggs rolleyes.gif
*
QUOTE(charge-n-go @ May 25 2007, 11:28 AM)
I love this - kill the bird and smash the eggs laugh.gif

anyway, Penryn power management architecture isn't great (identical to merom), but the process tech itself is great. There is no voltage plane control and per core frequency throttle like Puma. If 1 of the core in Penryn is loaded, the whole CPU will be running at the maximum voltage (and probably frequency, depends on the load pattern).

What i meant is, laptop CPU doesn't need the performance ike a server or desktop CPU, because the HDD, RAM and graphic is so much behind the desktop/server parts. Even if it has a powerful CPU like K10, it still can manage the power envelop below 35W if the clock and voltage is low enough (like 1.2GHz/1.1V). However, it is not needed given that HDD space and RAM size is so small for virtualization or huge database / scientific processing. The increased internal BW and higher parallel processing in K10 won't have a great effect when the RAM is so slow and limited, and the HDD is access time is so darn long. Of course, hardcore gamers can't even benefit from K10 due to the pathetic internal graphics laugh.gif This scenario is very well reflected in Core 2 Duo. Conroe smokes K8 in desktop when paired with fast RAM and FSB, while in mobile system Merom leads K8 just a little.

IMHO, maybe one day when the whole laptop system reached a certain level of performance, K10 shall be implemented for lappy CPU design.
*
I'm actually thinking that Griffin using K8 is more of a manufacturing decision rather than architectural decision: AMD is not expected to ramp K10 until next year and even then they'll be going for servers first. In the meantime their K8 desktop products are being slaughtered by Intel and they're pretty much selling them at below production prices to move inventory. The only market left with good margins is the mobility market and AMD can undercut Intel's offerings by pricing. The ATI acquisition really gives them strength in building a Centrino-like platform to challenge Intel in the mobility segment which until now Intel is virtually unchallenged.
With Griffin they can continue churning dual-core 65nm chips at good yields and capture a slice of the mobility market.

Of course this is provided that AMD actually executes flawlessly on all fronts and their recent track record (late and poor 65nm/R600) leaves a lot to desire.

 

Change to:
| Lo-Fi Version
0.0133sec    0.61    6 queries    GZIP Disabled
Time is now: 21st December 2025 - 02:55 PM