Lower power usage for Barton

[herbocharged]

http://www.amdzone.com/articleview.cfm?articleid=1240&page=2

How did AMD manage to reduce the power while adding 17million transistors. Barton has the same voltage as T-bred but a lower Icc typ. The frequecy is 67Mhz lower, but that shouldn't make that much of a difference.

Athlon XP model number: 3000+
Cache Size: L1 - 128KB and L2 - 512KB = 640KB Total Cache
Frequency / Bus / Multiplier: 2.167GHz / 333FSB / 13.0x multiplier
Infrastructure Support: Socket A motherboards
Fab location: AMD's Fab 30 wafer fabrication facility in Dresden, Germany
Process Technology: 0.13 micron copper process technology
Die Size: 101mm2
Approximate Transistor count: 54.3 million
Nominal Voltage: 1.65v
Max Die Temp: 85 degrees Celsius
Typical Thermal Power: 58.4 W
Max Thermal Power: 74.3 W
Icc Typical (low power state): 7.2 A
Icc Typical (working state): 35.4 A
Icc (processor current) Max: 45.0 A

Athlon XP model number: 2800+
Operating Frequency: 2.25GHz/333FSB (2.17GHz/333FSB)
Manufactured: AMD's Fab 30 wafer fabrication facility in Dresden, Germany.
Process Technology: 0.13 micron copper process technology / Thoroughbred rev-B
Cache Size: L1 - 128KB and L2 - 256KB = 384KB Total Cache
Nominal Voltage: 1.65v
Die Size 84mm2
Transistor count: Approx: 37.6 million
Infrastructure: Socket A
Max Die Temp: 85 degrees Celsius
Typical Thermal Power: 64.0 W (62.0 W)
Max Thermal Power: 74.3 W (68.3 W)
Icc Typical (low power state): 5.4 A (5.9 A)
Icc Typical (working state): 38.8 A (37.6 A)
Icc (processor current) Max: 44.9 A (41.4 A)

 

[AbsolutDealage]

Power consumption is not just based on the number of transistors and the process size. The design of a complex chip is a delicate balance of area, power, speed. The engineers at AMD obviously have opted to design the latest core with a mind for less power consumption (probably due to the ever increasing power demands we have seen recently). In short, they have done this by manipulating the sizes and ratios of the individual transistors to make them more power friendly.

 

[herbocharged]

All the reviews I read made it seem like the only change was the addition of 256kb of cache. I suspected they altered transistors, critical paths, etc. Just don't know enough about cpu layout to discern this.

 

[AbsolutDealage]

Originally posted by: herbocharged
All the reviews I read made it seem like the only change was the addition of 256kb of cache. I suspected they altered transistors, critical paths, etc. Just don't know enough about cpu layout to discern this.

Yes, I would assume that the cache accounts for a good deal of those extra transistors in Barton. The remaining transistors added are probably due to the re-design of logic for timing optimization and area optimization.

 

[foxkm]

I bet cause Barton has a larger core, it can dissapate the heat a little bit better. Maybe I'm wrong.

 

[Peter]

Let's note that the TYPICAL power went down, not the MAX power. Now typical power is measured while running a certain set of applications. With L2 cache access consuming less power than having to drive the front side bus to access the main RAM array through the chipset, it's quite obvious why typical power goes down when L2 cache size goes up. Maximum power, measured with a certain set of instructions that generates the worst possible scenario, does not benefit from this effect at all. The slightly higher max power is also expected, those extra transistors need to be fed from somewhere.