You are not logged in.
Hi all,
I have directly measured the Intel/AMD CPU power consumption in a PC machine.
This is easy because power supply offers two +12v power wires for motherboard. Therefore, the CPU power consumption can be get by just measuring the two +12v wires.
And I have a crazy idea to measure cell broadband engine in Playstation 3.
However, things become complicated because there are no dedicated power lines for cell processor, and thus I can't use the same method as I did in PC.
For couple of days surveying on the internet, I found that "IOR ip2003ap" chips might forms power supply system for cell processor, but I can't find which line on the PS3 motherboard is the input voltage for cell processor. There is no circuit layout of ps3 on the internet ...
Could someone here gives me a suggestion?
Elson
Offline
CPU: Cell Processor
* PowerPC-base Core @3.2GHz (Prototype ran at 4.66Ghz)
> 64 bit, "Power Architecture" processor
> Dual issue, dual threaded, in-order processor.
> 235 square mm
> 235 million transistors
> Rambus XDR and FlexIO technology allow up to 100 gigabyte/s memory transfer rates.
> 90nm Process CMOS SOI
> 65nm CMOS SOI Process Started March 07
- 6GHz at 1.3V
- Dual power supply; enhances SRAM stability and performance using an elevated array-specific power supply, while reducing the logic power consumption.
> Power consumption has been estimated at 60 - 80 Watts at 4 GHz
* 9 Core CPU
> 1 Power Processor Element (PPE) - Acts as Controller (PowerPC Core)
> The PPE is dual threaded
> 8 Synergistic Processor Elements (SPEs) with 256KB "Local Stores" per Core
> Each SPE capable of 32 GigaFlops (32 bit)
* 1 VMX vector unit per core
* 512KB L2 cache
* 7 x SPE @3.2GHz
> 1 of 8 SPEs reserved for redundancy
> 6 SPE used for game applications
* 7 x 128b 128 SIMD GPRs
* 7 x 256KB SRAM for SPE
* Element Interconnect Bus (EIB)
* Direct Memory Access Controller (DMAC).
* 2 Rambus XDR memory controllers
* Rambus FlexIO (Input / Output) interface
* Test and Debug Logic
* Total floating point performance: 218 GFLOPS
* Capable of running at speeds beyond 4 GHz
An SPE is a self contained vector processor which acts as an independent processor. They each contain 128 x 128 bit registers, there are also 4 (single precision) floating point units capable of 32 GigaFLOPS* and 4 Integer units capable of 32 GOPS (Billions of integer Operations per Second) at 4GHz. The SPEs also include a small 256 Kilobyte local store instead of a cache. According to IBM a single SPE (which is just 15 square millimetres and consumes less than 5 Watts at 4GHz) can perform as well as a top end (single core) desktop CPU given the right task.
We have all this information here, LOOK...
Also, if you think that any modern 20th century CPU runs at 12V, then you need to go back to school...
Offline
^^^^
Were in the 21st century now, by the way.
Offline
Powerslave wrote:
CPU: Cell Processor
* PowerPC-base Core @3.2GHz (Prototype ran at 4.66Ghz)
> 64 bit, "Power Architecture" processor
> Dual issue, dual threaded, in-order processor.
> 235 square mm
> 235 million transistors
> Rambus XDR and FlexIO technology allow up to 100 gigabyte/s memory transfer rates.
> 90nm Process CMOS SOI
> 65nm CMOS SOI Process Started March 07
- 6GHz at 1.3V
- Dual power supply; enhances SRAM stability and performance using an elevated array-specific power supply, while reducing the logic power consumption.
> Power consumption has been estimated at 60 - 80 Watts at 4 GHz
* 9 Core CPU
> 1 Power Processor Element (PPE) - Acts as Controller (PowerPC Core)
> The PPE is dual threaded
> 8 Synergistic Processor Elements (SPEs) with 256KB "Local Stores" per Core
> Each SPE capable of 32 GigaFlops (32 bit)
* 1 VMX vector unit per core
* 512KB L2 cache
* 7 x SPE @3.2GHz
> 1 of 8 SPEs reserved for redundancy
> 6 SPE used for game applications
* 7 x 128b 128 SIMD GPRs
* 7 x 256KB SRAM for SPE
* Element Interconnect Bus (EIB)
* Direct Memory Access Controller (DMAC).
* 2 Rambus XDR memory controllers
* Rambus FlexIO (Input / Output) interface
* Test and Debug Logic
* Total floating point performance: 218 GFLOPS
* Capable of running at speeds beyond 4 GHz
An SPE is a self contained vector processor which acts as an independent processor. They each contain 128 x 128 bit registers, there are also 4 (single precision) floating point units capable of 32 GigaFLOPS* and 4 Integer units capable of 32 GOPS (Billions of integer Operations per Second) at 4GHz. The SPEs also include a small 256 Kilobyte local store instead of a cache. According to IBM a single SPE (which is just 15 square millimetres and consumes less than 5 Watts at 4GHz) can perform as well as a top end (single core) desktop CPU given the right task.
We have all this information here, LOOK...
Also, if you think that any modern 20th century CPU runs at 12V, then you need to go back to school...
a little off topic, have you seen the price of rambus ram, its worse than ddr3 ram
Offline
nickb827 wrote:
^^^^
Were in the 21st century now, by the way.
Yes, but even in the 20th century when they were using Z80 and 65XX CPUs, they were 5V, none were ever 12Volts. There were no TTL, PIA, CIA, ROM, PROM, KERNEL ROM, chips more than 5V at a any time. I think the Commodore 64s "VIC" chip ran a good wattage though, thus it had to made out of ceramic.
What's makes it funny, is; even back in those days, the GPU was a problem. The VIC chip was a great source of heat, and would not last long without a heat-sink if not in the ceramic package... They did have the plastic VIC, but you were lucky if you had the ceramic one, which I did. It amazes me that even today, we have GPUs that still have the same issues as they did 23 years ago. It only ran at 8Mhz too, where as the 6510 was 4.77Mhz... It could not share ram with the system, so it had to be almost twice as fast to operate in tolerance. I seem to actually remember the VIC may have required almost 12V, I am not sure. I mean, this thing would cause 2nd degree burns if you touched it long enough, yeah, that's 20+ watts for one chip man... That's a HUGE number compared to what they are today. No CPU that I am aware of, ran 12V...
Anyhow, when you measure a 12V source that is stepped down, measuring the wattage is going to be incorrect, because you are not getting the most accurate reading due to the step-down. You'll have a few watts of overhead -vs- measuring the wattage AFTER the step-down, in this guys case, 1.3V to the CELL CPU.
Offline