LMAO, are you serious?
Higher performance vs. Lower power consumption
- Thread starter Techno+
- Start date
LMAO, are you serious?
Rainbow Man
Veteran
No it isn't. Let's put things into perspective here. 100-400W for high-end graphics cards is NOT a whole lot of electricity. A plasma TV alone or a couple light bulbs can dissipate more than that. Not to mention things like AC and other household appliances like dishwasher washing machines dryers etc.
But what if the electricity doesn't come from a greenhouse gas-generating source?
It could be generated via water/wind/nuclear means for example. Even solar (though I admit that's unlikely).
Peace.
Well, the government doesn't normally give you a choice in such matters, now do they.
If you want low power consumption, buy a laptop or an old DX5 GPU. Higher performance (for irreversible computations) inevitably requires more power (Landauer's principle). Simply plot the performance vs power requirements for either CPUs or GPUs over the last number of years. So if you want faster and faster GPUs, you need to get comfortable with continually higher power requirements.
The push by chip vendors to make chips more power efficient is simply an economic concern from their part. When power consumption was relatively low compared to other equipment or parts of the system, costs were mainly silicon driven. Now that performance has scaled to the level that the costs of acceptably quiet cooling solutions are becoming a significant part of overall delivery costs, then making chips more power efficient becomes more of a priority, with a tradeoff between more complex silicon design and more die area given toward power efficiency.
When chips were silicon constrained, the priority was given to minimizing die area and maximizing freqency to maximize the performance/die area ratio and thus maximize the performance/cost ratio. As chips become power constained, you maximize the performance/cost ratio by maximizing the performance/power ratio instead. This leads to the rather interesting reversal of emphasis, since increasing the power efficency means reducing the frequency and increasing the die area (more transisters running at lower frequencies). This is because power scales up nonlinearly with frequency but performance only scales linearly with frequency. However, both power and performance scale linearly with transistor count. As a result, in the future you will see more and more chips primarily increase performance by dramatically increasing transitor count rather than frequency.
From a buyers point of view, the main differences are total cost of the equipment as well (the cost of the power differences are small). The other considerations (noise, etc.) all have solutions that simply add to the cost.
So if you had two machines available that had the same quietness, the same features, the same reliability and the same performance, etc. but one was 25% less expensive, most would choose the less expensive one. 12 months later, if they took their machine apart, examined it with a volt meter and noticed a sophisticated heat exchanger mounted on the chip, and realized that it had a higher power consumption than the other machine that they didn't buy, would they care much?
The push by chip vendors to make chips more power efficient is simply an economic concern from their part. When power consumption was relatively low compared to other equipment or parts of the system, costs were mainly silicon driven. Now that performance has scaled to the level that the costs of acceptably quiet cooling solutions are becoming a significant part of overall delivery costs, then making chips more power efficient becomes more of a priority, with a tradeoff between more complex silicon design and more die area given toward power efficiency.
When chips were silicon constrained, the priority was given to minimizing die area and maximizing freqency to maximize the performance/die area ratio and thus maximize the performance/cost ratio. As chips become power constained, you maximize the performance/cost ratio by maximizing the performance/power ratio instead. This leads to the rather interesting reversal of emphasis, since increasing the power efficency means reducing the frequency and increasing the die area (more transisters running at lower frequencies). This is because power scales up nonlinearly with frequency but performance only scales linearly with frequency. However, both power and performance scale linearly with transistor count. As a result, in the future you will see more and more chips primarily increase performance by dramatically increasing transitor count rather than frequency.
From a buyers point of view, the main differences are total cost of the equipment as well (the cost of the power differences are small). The other considerations (noise, etc.) all have solutions that simply add to the cost.
So if you had two machines available that had the same quietness, the same features, the same reliability and the same performance, etc. but one was 25% less expensive, most would choose the less expensive one. 12 months later, if they took their machine apart, examined it with a volt meter and noticed a sophisticated heat exchanger mounted on the chip, and realized that it had a higher power consumption than the other machine that they didn't buy, would they care much?
Blazkowicz
Legend
if you buy upper midrange or lower high end, such as 7600GT, 7900GS or 8800GTS : you save money (price and electric bill), don't need to buy a new PSU, are a bit less ripped off and anyway you get the very same exponential progress.
Aerows
Regular
As previously mentioned, a balance is definitely needed. I don't want performance to be so low that upgrades are meaningless, yet IMHO both the GTX and R600* are too large and power hungry. I got an 8800 GTS instead, when I could have afforded a GTX just because I didn't want to deal with the heat + power consumption.
The GTS easily suits my needs with regards to performance, and while it does draw a good amount of power, I don't need to alter my A/C settings to compensate. That is a big deal to me, because I live on the Gulf Coast. We had the A/C on several times in February; in July it is so humid and hot it's like breathing soup outside.
Don't the two also go hand in hand somewhat? An efficient chip won't draw as much power as one that isn't engineered as efficiently. Then you also get into needing better cooling, a better case and a bigger power supply. I think both got a bit out of hand this generation.
* Standard disclaimer of "Based upon what we have heard"
The GTS easily suits my needs with regards to performance, and while it does draw a good amount of power, I don't need to alter my A/C settings to compensate. That is a big deal to me, because I live on the Gulf Coast. We had the A/C on several times in February; in July it is so humid and hot it's like breathing soup outside.
Don't the two also go hand in hand somewhat? An efficient chip won't draw as much power as one that isn't engineered as efficiently. Then you also get into needing better cooling, a better case and a bigger power supply. I think both got a bit out of hand this generation.
* Standard disclaimer of "Based upon what we have heard"
- Static power consumption scales linearly with area and accounts for a large amount of overall power consumption, but it stays constant with frequency.
- Area can't keep on going up forever the way it's going now, so frequency has to go up to increase performance per area.
- Power scales linearly with frequency... all other things equal. Going beyond a certain point, additional logic is needed to allow higher frequencies, but that amount depends big time with the frequency increase you're aiming for. And there are regions along this curve where increasing frequency is a better deal.
In other words, I don't really see a reversal of emphasis at all.
It's all a big multi variable optimization. Which is why it is fun.
- Static power consumption scales linearly with area and accounts for a large amount of overall power consumption, but it stays constant with frequency.
- Area can't keep on going up forever the way it's going now, so frequency has to go up to increase performance per area.
- Power scales linearly with frequency... all other things equal. Going beyond a certain point, additional logic is needed to allow higher frequencies, but that amount depends big time with the frequency increase you're aiming for. And there are regions along this curve where increasing frequency is a better deal.
In other words, I don't really see a reversal of emphasis at all.
Increasing parallelism (#transistors) allows you to increase performance for a given frequency, or reduce frequency for a given performance (and reduce voltage as well, which reduces leakage exponentially, and/or use more slower/low leakage transistors). Increasing frequency usually requires an increase in voltage too, and that is what kills the power scaling. Also to increase the frequency you may have to use more low Vt transistors, which leak more. I think additional logic is always needed to increase frequency, you need to make your critical path units faster, and to make them faster you generally have to spend more transistors even assuming pipeline depth is kept the same. Area can be kept manageable via process shrinks and more hand tuning of the design, which also can help with speed.
Not saying frequency is ebil. It will continue to increase, no doubt about that. All depends on your design targets i guess.
Well I tend to severely doubt that. Otherwise noone could use any heaters, boilers, AC, ovens etc. since they're all waaay over 1kW.
EDIT: wiki says...
NEMA 1-15 (North American 15 A/125 V ungrounded) - so it's max. 1875W
NEMA 5-15 (North American 15 A/125 V grounded) - same
And here we have:
CEE 7/7 (French/German 16 A/250 V earthed) - 4 kW
Duh, I thought the US way used to be "bigger is better"...
Last edited by a moderator:
Since the main problem is to keep current flow low, lower voltage means lower max power. That's why many air conditioners use 220V sockets.
Regarding the power/performance issue, I think, in general, if you can accept lower performance, you can generally get much lower in power requirement. It's generally not linear. For example, if you turned down the frequency, you can generally turned down the voltage, which can provide super linear reduction. On the other hand, if you double the number of (logic) transistors, generally you won't get double performance while you may need more than double power.
So, in theory, the problem is mostly how to get acceptable performance with least power requirement. Of course, in reality cost is also a very important issue.
Regarding the power/performance issue, I think, in general, if you can accept lower performance, you can generally get much lower in power requirement. It's generally not linear. For example, if you turned down the frequency, you can generally turned down the voltage, which can provide super linear reduction. On the other hand, if you double the number of (logic) transistors, generally you won't get double performance while you may need more than double power.
So, in theory, the problem is mostly how to get acceptable performance with least power requirement. Of course, in reality cost is also a very important issue.
But desktop GPUs tend to be manufactured with processes that don't optimize for static leakage. My understanding is that TSMC's low-power process (which is used for handhelds and generally laptops, iirc) has much lower static leakage.
Because the low-power process also has the advantage of being available before any other process variant nowadays, I've been wondering if more and more desktop GPUs are going to use it in the future. Well, eventually everyone will be using high-k anyway...
But that's apparently not happening at TSMC before 32nm, sadly.Anarchist4000
Veteran
_xxx_ said:
Nobody plugs those into a standard outlet though. I can't think of any major appliance that plugs into a normal wall outlet. We've got big heavy duty outlets for whenever those show up. Also keep in mind that a typical household circuit is 20A and likely used to wire an entire room.
Wow, that's heavy. So 60's. We only have heavy duty outlets for the stove and maybe some huge-arse heaters, but otherwise it's all just normal wall plugs.
