Submitted by SilentPCReview tries to answer an important question: what is the best power efficiency achievable with currently available AMD and Intel processors that can be used on a desktop PC? The answer: “Our focus on thermals, power and energy efficiency led to mostly predictable results: mobile processors are best, followed by AMD desktop processor in general, and then Intel desktop processors. The power efficiency of AMD Athlon 64 single and dual core processors is excellent, even for their highest performance models. The Intel desktop processors suffer from inefficiency, even on the 65nm die. The Core Duo is a delightful exception in Intel’s camp, with probably the highest performance-per-watt ratio of all the processors in our survey.”
“The Core Duo is a delightful exception in Intel’s camp, with probably the highest performance-per-watt ratio of all the processors in our survey.”
And pretty soon Merom/Conroe/Woodcrest will wipe the floor with other x86 CPU’s, both in terms of performance, but more importantly, in efficiency. Intel’s performance-per-watt Arch is the real deal…and it will only get better and better. 🙂
And most hope Intel won’t be the only one with a high performance-per-watt ratio
Considering how long it has taken for Intel to get to this point, the amount of resources Intel has, and the mistakes they’ve made along the way, I’d say that it’s about time Intel came up with a decent response to AMD’s K8 platform. I am very excited to see Intel finally getting so much right because it will only help spur competition, but I wish that they had achieved something as good before now. This should certainly prompt AMD to invest even more to compete on the mobile side of things.
Sales figures show Laptops / Mobile Workstations are increasing at a impressive rate suprassing desktops except in some areas (gaming, servers etc). For the consumer why buy 3.5″ drives for the desktop when now 2.5″ work at BETTER performance and power! (unless you are into storage systems like NAS boxes). Chips, mobile chips win again they are now dual core with speed steppings. Alot of people put mobile chips in desktops. The only area where desktops win is Graphics cards but who wants a 600 wattage box in their floor? Thats more power demand than my fridge freezer and sounce like a vacume cleaner!
I had no idea processors slurped up that much power! The past months I’ve been reading on OSNews about the next low-power chip consuming less than 1W at full load (ARM and PPC), and these puppies use no less than 23 while idle!
Now I realise there’s quite a gap regarding performance of the best ARM chip and the ‘best’ low-power PPC chip (ignoring the G5s that couldn’t be made usefull for laptops), but I wonder how much general power consumption in the world would have been lower if our general purpose computers were running on MIPS, ARM/XScale and PPC instead of x86 and Itaniums.
People also say the real batteryslurpers in notebooks are the screen and the harddisk, but even if laptop LCDs consume 50 W when used, there’d be significant gains by not going with the energy hogs that are for example Dothan and Yonah.
> consuming less than 1W at full load (ARM and PPC), and these puppies use no less than 23 while idle
Note that FP performance of ARM probably sucks: AFAIK most ARMs don’t have a FPU.
Those <1W chips would have a hard time running Doom3, encoding movies, etc..
Different purpose, different priority/trade-off, you’re comparing apple and oranges.
Now I realise there’s quite a gap regarding performance of the best ARM chip and the ‘best’ low-power PPC chip (ignoring the G5s that couldn’t be made usefull for laptops), but I wonder how much general power consumption in the world would have been lower if our general purpose computers were running on MIPS, ARM/XScale and PPC instead of x86 and Itaniums.
Instruction set is mostly irrelevent in the matter. The types of CPUs you’re talking about (eg: ARM), achieve 1W of power by not having out-of-order execution or “extranous” features like a floating-point unit. You achieve low power by being slow, not by using MIPS or PPC. There are, for example, x86 chips (AMD’s Geode line) that consume as little as 0.9W. Nobody uses them in laptops because they’re too slow.
The complexity of instruction decoding is relevant to power consumption, its significance in desktop/laptop processors is simply reduced by comparison with the memory structures devoted to implementing the remainder of the core. The density of instruction encoding plays an important role in power efficiency, even at the cost of increased decoder complexity. In low-power devices this is a more significant area of concern, because I-caches can be responsible for a significant percentage of power dissipation.
I think that people too often take for granted that the x86 somehow retards the industry. The idea that relying on a MIPS or PPC ISA (certainly not the most compact) would make a significant difference in the power utilization landscape of the desktop/laptop is certainly a waste of mental energy. If people had valued power consumption over cost and performance, that would have changed the landscape.
For the would-be reader, the power consumption of the Geode LX requires a little elucidation. The Geode LX at any given speed will use more power than reported by PR@Watt figures. The Geode family integrate the memory controller and video processing. To obtain the number provided the processor is given a fixed workload and then the power utilized by the structures on the processor not strictly utilized for execution is removed. I think the max power of a Geode LX will be somewhere over the 3W mark, but I’m too lazy to look for the exact figure. It does however serve nicely to demonstrate that the x86 ISA doesn’t require a processor to dissipate 100W.
“(ignoring the G5s that couldn’t be made usefull for laptops),”
Oh my god is that sad.
Unless you actually meant “IBM decided not to bothering with a mobile 970 design for Apple since they comprised less than 4 percent of their chip sales and after the nightmare of five years of dealing with Apple as a customer IBM decided to dump Apple as a chip customer and focus on the 150 million console machines that will be sold over the next five years.”
Let me guess…you also believe Microsoft’s equally bogus damage control meme about their problems with security being ‘due to just how darn popular Windows is’
Unless you actually meant “IBM decided not to bothering with a mobile 970 design for Apple since they comprised less than 4 percent of their chip sales and after the nightmare of five years of dealing with Apple as a customer IBM decided to dump Apple as a chip customer and focus on the 150 million console machines that will be sold over the next five years.”
I’m convinced that if it would’ve been easy to do for IBM, it would have been done.
The rest of your post was blatant trolling.
My PSU is 40 watts for my laptop. I doubt the LCD uses that much, 2.5″ drives now are perpindicular and 5400 drives use the power of 4200. I have 7200 RPM in mine and 5400 perpindicular as my external drives. Big boxes are only for egomainiacs
What I would like is a PCI express slot dock for a nice High power GFX card though for my Acer Ferrari Laptop or some way of upgrading the graphics.
To solve power issues, they should implement software adjustable clocks, like they do with certain microcontrollers.
They do. Turions, FI, start at either 0.8 or 1GHz.
>And most hope Intel won’t be the only one with a high performance-per-watt ratio
they won’t – sun and IBM have some nice proc’s for sale… and not just them.
are great for DL machines where bandwidth and storage space is the only requirement along with NAS boxes. I would love a NanoPC for this, fanless silent headless practically powerless.
I’m more interested in knowing which CPU manufacturer FPU is more optimized for UNIX/linux OS?if it makes a difference at all.
FPU optimised for a platform? Shouldnt the platform be optimised for an FPU?
Shouldnt the platform be optimised for an FPU?
after giving it a second thought, yes. i was wondering which CPU company like intel and amd, had the better FPU.
I believe AMD has a better general FPU, but for best performance you should go Intel with SSE code.
This study was annoying, almost a marketing gift for Intel.
So they ended up measuring the power of the AMD products which include an integrated memory controller and comparing the numbers directly to Intel parts which don’t have an integrated memory controller. And then they even fail to mention this fundamental difference.
The only real numbers that have any bearing was the total system power dissapated, not the CPU power.
“This study was annoying, almost a marketing gift for Intel.”
Load:
Intel PD 930 Presler (3.0 GHz), 1.22V: 93.6W DC for CPU, 146W AC for system.
AMD A64 X2 3800+ Toledo (2.0 GHz), 1.39V: 58.0W DC for CPU, 109W AC for system.
Idle:
AMD A64 X2 3800+ Toledo (1.0GHz), 1.18V: 5.6W DC for CPU, 51W AC for system.
Intel PD 930 Presler (2.4~3.0 GHz), 1.22V: 32W DC for CPU, 75W AC for system.
Intel looks fantastic, there, doesn’t it?
“The only real numbers that have any bearing was the total system power dissapated, not the CPU power.” (emphassis added)
…bearing on what?
I know I’m more interested in the DC power numbers than the AC ones: they show the heat dissipated by the CPU’s heatsink, and show how far they could undervolt the CPU, and how much that affected the heat it created. That means more or less room to undervolt fans, and maybe even remove fans from the system (I’ve got 4, after all!).
“So they ended up measuring the power of the AMD products which include an integrated memory controller and comparing the numbers directly to Intel parts which don’t have an integrated memory controller. And then they even fail to mention this fundamental difference.”
It’s not a fundamental difference, in reference to silencing the computer. The CPU heatsink has to cool X watts. Whether that X watts includes a memory controller or not does not matter. The chipset can be passively cooled.
I believe Intel has a better general FPU, but for best performenace you should go AMD with their code.