Sandy Bridge Power Consumption Readings?

[pressingonalways]

Hi,

Anyone got the Sandy Bridge i5-2500T CPU yet? I'm curious to know how good the idle power efficiency on these things and if they can possibly do 20 or 25 watt idle load with a picoPSU.

If not, maybe the i5-2390T will be able to reach that level? Anyways, if anyone has these... please post some readings!!

Thanks!

[ces]

Try here:
http://www.legitreviews.com/article/1501/6/

and here:
http://www.silentpcreview.com/article1156-page3.html

and here
http://www.silentpcreview.com/article1148-page5.html

and here
http://www.tomshardware.com/reviews/cor ... ,2838.html

I think if you are talking about just the chip and you extrapolate from these reviews... it appears the likely answer to your question is a YES.

[tim851]

A review of the 2400S on Xbitlabs indicates that there is absolutely NO difference in idle power consumption between it and the regular 2400. Seems logical, as the CPUs are designed to underclock and -volt as much as possible when idle. I don't see why the 2500T would be any different.

If it's low idle power consumption that you're after, I don't think it's worth paying the upmark.

[pressingonalways]

A review of the 2400S on Xbitlabs indicates that there is absolutely NO difference in idle power consumption between it and the regular 2400. Seems logical, as the CPUs are designed to underclock and -volt as much as possible when idle. I don't see why the 2500T would be any different.

If it's low idle power consumption that you're after, I don't think it's worth paying the upmark.

perhaps... but the 2500T has a TDP of 45 watts and the 2390T even lower at 35 watts...

[pressingonalways]

Try here:
http://www.legitreviews.com/article/1501/6/

and here:
http://www.silentpcreview.com/article1156-page3.html

and here
http://www.silentpcreview.com/article1148-page5.html

and here
http://www.tomshardware.com/reviews/cor ... ,2838.html

I think if you are talking about just the chip and you extrapolate from these reviews... it appears the likely answer to your question is a YES.

nice stuff on sandy bridge stuff but not the extreme low energy usage T version I'm looking for. Perhaps I should have been more specific in my topic.

[ces]

nice stuff on sandy bridge stuff but not the extreme low energy usage T version I'm looking for. Perhaps I should have been more specific in my topic.

Read the xbits article:
http://www.xbitlabs.com/articles/cpu/di ... html#sect0

They found not a lot of difference between them at the low end.... and explain why.

http://www.xbitlabs.com/articles/cpu/di ... html#sect0

[tim851]

perhaps... but the 2500T has a TDP of 45 watts and the 2390T even lower at 35 watts...

The TDP is an indicator for the maximum power consumption. Under load conditions, surely the 2500T uses less power than the 2500, and the 2390T even less. But you asked about idle conditions. And under those, there doesn't seem to be much difference. The 2390T is a dual core, it could be a little lower on idle.

They are more expensive than their counterparts though. The 2500T retails for a little more than the 2500, but it's quite a bit less powerful.

In my opinion these low voltage CPUs only make sense if you need to stay within a low thermal envelope. Say you have a small enclosure with a tiny heatsink and don't want the CPU fan to start screaming bloody murder.

[ces]

In my opinion these low voltage CPUs only make sense if you need to stay within a low thermal envelope. Say you have a small enclosure with a tiny heatsink and don't want the CPU fan to start screaming bloody murder.

I would refine what you say a bit more. I would say a small enclosure with a tiny heatsink which you expect to put under heavy load on a consistent basis.

I takes a lot to put any of these Sandy Bridges under a heavy load. Most people will just never do it. You can test your usage to see if you do. Get CPUID TM Monitor and run it for a few weeks. It will record the highest and lowest temp on your current CPU. If your current usage never pushes your current CPU... it really really really will never push a Sandy Bridge.

Apparently, a high end Sandy Bridge that is not pushed, runs no hotter than one of the low TDP Sandy Bridges. Lesson: if you want a low power Sandy Bridge, just get the cheapest one.

[HFat]

The OP asked about idle power consumption but that's not the only reason to buy a low-TDP CPU.
It's unsafe to rely on software to keep your system from overheating. You should make sure your cooling system is able to deal with the full power dissipation. This can sometimes be moderated by underclocking in the BIOS and such but, in general, a low-TDP CPU means a less beefy power supply and cooling system is OK. Perhaps some future motherboards with efficient VRMs (I don't think there are any now) will only take chips rated at 65W or under.
Note that the cheapest CPUs do not have the same features as the more expensive ones. If you want the features but not a lot of computing power, Intel wants you to spend so much you might as well take a low-TDP chip I think. Of course most people do not care for the features but in that case the cheapest CPUs will be slow dual-cores which have a relatively low TDP to begin with and the 35W TDP premium drops to about 10$.

More generally, one also has to wonder about the wisdom of buying a Sandy Bridge system if you're never going to use the processing power. Sandy Bridges comes with unique advantages but current prices are pretty high...

[ces]

The OP asked about idle power consumption but that's not the only reason to buy a low-TDP CPU. It's unsafe to rely on software to keep your system from overheating.

That is a valid concern, but look at this:

65W TDP i5-2400S $204.99 at newegg
95W TDP i5-2300 $184.99 at newegg
73W TDP i3-560 Clarkdale $149.99 newegg
73W TDP i3-540 Clarkdale $119.99 newegg

I didn't do any thorough research but these test results would indicate that the 4 core Sandybridge not unly uses less pwoer at a full burn than does a two core Clarkdale, but also idles lower... perhaps as much as 10 watts lower.
http://www.xbitlabs.com/articles/cpu/di ... html#sect0

It is hard to imagine what computer case wouldn't plausibly be able to handle a 2400S or a 2300? What ITX case would come to your mind that couldn't handle this?

[HFat]

I didn't do any thorough research but these test results would indicate that the 4 core Sandybridge not unly uses less pwoer at a full burn than does a two core Clarkdale, but also idles lower... perhaps as much as 10 watts lower.

The 73W TDP figure is evidently for the whole Clarkdale line, not for individual models. Most of them are never measured as consuming anywhere as much.
Clarkdales can idle very low. Idling 10W lower is simply impossible (the power consumption would be negative). As with Sandy Bridge, it's not the CPU which consumes the most at idle. People have pushed power consumption for the a whole Clarkdale-based system under 15W. I don't think you can do that with Sandy Bridge yet for lack of efficient motherboards.

It's the whole cooling system which matters for overheating, not just the case. There are any number of possible variations including fully fanless cooling or fanless heatsinks (even if MikeC argues against the latter). If you're using a large case with a fanned heatsink plus case fans and so on, you should of course be alright with the regular models.

[pressingonalways]

Thanks all for your input... Understanding that the idle power consumption is fairly the same among all sandy bridge CPUs does help me better understand what to purchase when I upgrade my computer...

I noticed that the xbits lab article ( http://www.xbitlabs.com/articles/cpu/di ... html#sect0 ) said that they got power consumption down to 16 watts on idle not factoring the PSU efficiency.... Does this mean that with a PicoPSU you can see total energy readings below 25 watts measured from the wall socket? That's insane - it beats even my Atom!

[ces]

That's insane - it beats even my Atom!!

Yeah isn't it?

[mdchaser]

I was going to wait for the T version to come out as well but ended up going with a standard 2400 and I'm glad I did (actually I should have picked up a 2500). Idle is identical on all of them no matter the model (the asus software reads at 2.5W at idle, not sure how accurate) and power consumption scales completely linearly with clock rate. That means that all you are giving up is speed by getting a lower clocked variant as efficiency is the same at any clock speed up to somewhere in the mid 4Ghz range!

[ame]

I was going to wait for the T version to come out as well but ended up going with a standard 2400 and I'm glad I did (actually I should have picked up a 2500). Idle is identical on all of them no matter the model (the asus software reads at 2.5W at idle, not sure how accurate) and power consumption scales completely linearly with clock rate. That means that all you are giving up is speed by getting a lower clocked variant as efficiency is the same at any clock speed up to somewhere in the mid 4Ghz range!

I think your right.

I'v seen similar thing when playing around with OC settings. I noticed I can get away with 4.1-4.2 with just about any standard voltage/current rating and with default TDP limits. Its only when going over 4.2 that these limits needed to be raised. This obviously would lead to higher power consumption.

Even when OCed, using dynamic votage means it idles at a very low ~0.9V.

[ces]

That's insane - it beats even my Atom!!

Apparently any and all of the Sandy Bridge chips use under 4 watts at idle. That means despite all the transistors, Intel has learned how to turn off and turn down large swaths of its silicon when not needed.

The justification for the Atom was that it had less silicon and therefore was more efficient than larger chips. The larger chips do more work per clock cycle and it appears you no longer have to pay an energy penalty for keeping that extra silicon available for when you need it.

If you keep your computer on 24/7, most of the energy usage is at idle. At 4 watts, how much closer to zero can you get? Not much. My recollection is that an LCD screen uses 100-150 watts and even a green one uses 50 watts.

Other than the approximate $150 difference between the cost between an atom and a more powerful general purpose Sandy Bridge computer, can anyone make a case for using at atom CPU for any plug in computer? Especially given that a Sandy Bridge compute will likely run Windows 8 and Windows 9 while the same can not be said of an atom computer.

In fact when you factor in the cost of legal software, the only case I can make for an Atom computer is if you are using it to run Linux.

It seems to me that small Atom and Atom like chips are an evolutionary dead end and the future is with CPUs with more silicon that adjust their energy usage to the work at hand.... using that extra silicon to do more work per clock cycle and shutting it down when unneeded.

So.... can anyone propose a compelling case for buying a new Atom CPU for purposes of running Windows?

[ces]

Wow, Sandy Bridge 50% faster on average energy consumption by 50% lower than the Phenom X4 and X6 II AMD really needs to give the BD, which offers performance that is close, but at a lower price.

Unfortunately, by the time Bulldozer shows up, it may be 2 generations behind.

I don't even consider AMD in my purchasing decisions, but their continued falling behind hurts everyone, not just those who buy from them.

They are the only ones that keep Intel honest and keep Intel innovating. Intel technology (and pricing) is no more competitive than AMD forces them to be.