CPU With Low Idle Power Consumption Recommendation

[Collider]

What would you recommend for a CPU with low idle power consumption that's good enough for a NAS system ?

 

[VirtualLarry]

Brazos / Puma / AM1?

 

[Shivansps]

AM1 is hard to get howdays, at best you can get the soldered mobile variants Like AMD E1-XX00 and AMD E2-XX00... Gigabyte and Biostar have those.

The other option is an J4105 or J4005... that whould be easier to get.

If you are looking something with a bit more punch and more mainstream, the G3930 and I5-7400 are low power as well. Try to avoid AM4 and Intel 8th gen if you are looking for low idle power.

 

[chrisjames61]

AM1 is hard to get howdays, at best you can get the soldered mobile variants Like AMD E1-XX00 and AMD E2-XX00... Gigabyte and Biostar have those.

The other option is an J4105 or J4005... that whould be easier to get.

If you are looking something with a bit more punch and more mainstream, the G3930 and I5-7400 are low power as well. Try to avoid AM4 and Intel 8th gen if you are looking for low idle power.

Isn
t there a 35 watt TDP AM4 variant?

 

[Collider]

Thanks for the suggestions.

Apollo Lake chips have me quite intrigued, 10 TDP sounds pretty nice:
https://www.tomshardware.co.uk/asrock-apollo-lake-motherboard-intel,news-54041.html

Curious if its enough to run UnRaid for storage only (no VM / docker usage).

 

[Collider]

So far thinking on settling on G3930 or G3930T (low power variant) with ASRock Z270 KILLER mobo as it has dual m2 slots I want for unRaid cache.

I also plan on running 10GBe Mellanox ConnectX-2 card - would G3930 be a enough for 10GBe or would it bottleneck ?

 

[coercitiv]

What would you recommend for a CPU with low idle power consumption that's good enough for a NAS system ?

Any modern CPU with integrated GPU will do, you'll also need to make sure the motherboard uses minimal power. As long as you have sleep state support and there's no dGPU in the system to limit deeper sleep states for the package as well, any Intel CPU starting with Haswell will idle at 1-2W. Choose an efficient motherboard and you're all set whether you choose a dual, a quad or a more-core.

To give you an example, last time I measured idle power consumption on my AM1 based system it was close to 11-12W at the main (Athlon 5350, mITX MB, 2x4 GB RAM @ 1.35V, gold rated 350W PSU). The next system I bought added a few watts more up to 15-17W but the specs changed drastically ( i5 6600K, Z370 mITX board with bigger VRM and lots more I/O including integrated WLAN, 2x8GB RAM running at 1.35V, gold rated 450W PSU).

Personally I would go straight for an i3 8100 these days, why worry about speed when the price difference is quite low and power levels are configurable via BIOS anyway.

 

[Shivansps]

Isn
t there a 35 watt TDP AM4 variant?

There is, but the 50W G3930 uses less power.

 

[PingSpike]

If you are looking something with a bit more punch and more mainstream, the G3930 and I5-7400 are low power as well. Try to avoid AM4 and Intel 8th gen if you are looking for low idle power.

I haven't seen anything to suggest that 8th generation should have worse idle power consumption than 7th.

 

[Collider]

I haven't seen anything to suggest that 8th generation should have worse idle power consumption than 7th.

Haven't checked all CPUs, but from what I've seen 8th gen Pentiums idle about ~ +8 watts more. Here are some of the sources I've seen:
https://www.techpowerup.com/reviews/Intel/Pentium_Gold_G5600/16.html
https://www.gamersnexus.net/hwrevie...-g5600-review-benchmarks-vs-amd-ryzen-3-1300x
https://www.anandtech.com/show/13660/amd-athlon-200ge-vs-intel-pentium-gold-g5400-review/20
https://www.techspot.com/review/1698-amd-athlon-200ge/page2.html

 

[IntelUser2000]

Apollo Lake chips have me quite intrigued, 10 TDP sounds pretty nice:
https://www.tomshardware.co.uk/asrock-apollo-lake-motherboard-intel,news-54041.html

The latest is Gemini Lake(Like J5005). Apollo is last gen.

The desktop platforms idle at 30-40W. The Gemini/Apollo can go under 10W if you have the right power supply(60-100W) and frugal components. Same is true for AM1, though that's becoming a bit outdated.

 

[Collider]

One thing that makes comparing power consumption quite difficult is that reviewers don't disclose if they're on 110v or 220v

220v consumes half the wattage of 110v.

Plus different motherboards used in tests.

 

[Markfw]

One thing that makes comparing power consumption quite difficult is that reviewers don't disclose if they're on 110v or 220v

220v consumes half the wattage of 110v.

Plus different motherboards used in tests.

Half the amperage, not half the wattage, thats the same.

Yes, different motherboards make a small difference, but different platforms make way more difference. Example would AMD Ryzen platform vs socket 1151 platform. Not even sure which takes more.

The motherboard is supplied with 12 volts, 3.3 volts and 5 volts, no matter if the PSU is 110 or 220. Its needs are a constant (assuming the same CPU). So if you are running on 220v, then the amps to the PSU are half that of 110, but the motherboard still takes the same wattage.

 

[chrisjames61]

There is, but the 50W G3930 uses less power.

The AM4 chip comes with Vega graphics so I would think it would be a much better choice?

 

[Shivansps]

The AM4 chip comes with Vega graphics so I would think it would be a much better choice?

It is a much better choice if you are looking for cpu and gpu perf... Not su much when looking for low power, the Celeron is better for that.

 

[PingSpike]

Haven't checked all CPUs, but from what I've seen 8th gen Pentiums idle about ~ +8 watts more. Here are some of the sources I've seen:
https://www.techpowerup.com/reviews/Intel/Pentium_Gold_G5600/16.html
https://www.gamersnexus.net/hwrevie...-g5600-review-benchmarks-vs-amd-ryzen-3-1300x
https://www.anandtech.com/show/13660/amd-athlon-200ge-vs-intel-pentium-gold-g5400-review/20
https://www.techspot.com/review/1698-amd-athlon-200ge/page2.html

Thanks for that. Those results are surprising to me. Particularly examples like the i3-8300 vs 7700K. I don't see any reason the idle should be higher on the new chip with the same number of cores and lower clock speeds. Maybe the platform uses more power?

Its hard to compare since people seem to measure idle different ways, if they measure it at all and the rest of the platform can make a big difference sometimes.

 

[Collider]

Thanks for that. Those results are surprising to me. Particularly examples like the i3-8300 vs 7700K. I don't see any reason the idle should be higher on the new chip with the same number of cores and lower clock speeds. Maybe the platform uses more power?

Its hard to compare since people seem to measure idle different ways, if they measure it at all and the rest of the platform can make a big difference sometimes.

Yeah, now I'm not so sure. I think they might be from different motherboards. Other benchmarks I saw were very close with gen 7 Pentium idling ~2w lower but being ~2-3w higher under full load. So gen 8 Pentium idles slightly higher but is slightly more efficient under load. Anandtech turns out has a pretty good benchmark database and I think they measure at the rail so that takes motherboard out of the equation completely.

 

[IntelUser2000]

Anandtech turns out has a pretty good benchmark database and I think they measure at the rail so that takes motherboard out of the equation completely.

It's not easy to measure, but motherboard can contribute quite a bit. The Techspot reviews doesn't show higher power for 8th gen chips, so it has to be the motherboard.

My i3 7100 idles at 1.4W but the system idles at 35W.

If you want low idle power, forget the regular chips. You have to go for the mobile versions or Gemini Lake/AM1. Then you can get under 10W assuming you get everything right.

One thing that makes comparing power consumption quite difficult is that reviewers don't disclose if they're on 110v or 220v

Power is work done, and what you want to care about. Volts are electric potential. They are not the same.

Just to make things clear, 220V appliances are slightly more efficient, but we're talking 1-5% difference.

Other things matter way more. Like platform choices. And whether you choose the right power supply. If your system uses under 50W, don't get a power supply greater than 250W, because they are less efficient under load. You want to look for something under 100W. For Gemini Lake you'd want PicoPSUs at 60W.

 

[coercitiv]

If you want low idle power, forget the regular chips. You have to go for the mobile versions or Gemini Lake/AM1. Then you can get under 10W assuming you get everything right

The OP wants dual m2 slots for unRaid cache - meaning bigger motherboards with proper chipsets - but even Z chipset boards can be configured for low power usage. As I already mentioned, my initial testing for a mITX Z170 board + 6600K had the system running around 15-17W idle for the entire system measured at the wall. (mechanical disks were disabled though, only the SSD was online).

 

[Collider]

The OP wants dual m2 slots for unRaid cache - meaning bigger motherboards with proper chipsets - but even Z chipset boards can be configured for low power usage. As I already mentioned, my initial testing for a mITX Z170 board + 6600K had the system running around 15-17W idle for the entire system measured at the wall. (mechanical disks were disabled though, only the SSD was online).

Interesting. Could you expand on "properly configured", curious what settings you've applied.

Sent from my Nokia 7.1 using Tapatalk

 

[coercitiv]

Interesting. Could you expand on "properly configured", curious what settings you've applied.

I danced around the notion in my previous posts, it matters a lot whether unRaid supports C (sleep) states or not. My bet would be it does, superficial searches yielded no results though. Other than that, no special settings except disabling integrated sound chipset. Further disabling other features may yield marginal savings.

I cannot stress this enough: most reviews use dGPUs in their idle power consumption tests. This helps them be consistent with most of their past and future reviews, but completely skews results from a minimalist power footprint perspective (adds dGPu power consumption and prevents CPU from entering a deep package C state sleep). A modern Intel CPU without a dGPU in the system will properly enter deep package and core C state sleep and therefore consumer under 2W at idle. I do not own an AMD APU right now, but it should behave similarly (AM1 did that with excellent results). This means that any modern system with minimal I/O installed should consume under 20W at idle, most likely closer to 15W if PSU and MB were chosen correctly.

 

[IntelUser2000]

The OP wants dual m2 slots for unRaid cache - meaning bigger motherboards with proper chipsets - but even Z chipset boards can be configured for low power usage.

Oh yea right.

I guess 10-15W difference isn't worth fretting about.

 

[Collider]

Yeah I figured ITX boards would be lower in terms of power consumption but those would be pretty limited in terms of options, extra slots always come in handy later on.

 

[PingSpike]

I danced around the notion in my previous posts, it matters a lot whether unRaid supports C (sleep) states or not. My bet would be it does, superficial searches yielded no results though. Other than that, no special settings except disabling integrated sound chipset. Further disabling other features may yield marginal savings.

I cannot stress this enough: most reviews use dGPUs in their idle power consumption tests. This helps them be consistent with most of their past and future reviews, but completely skews results from a minimalist power footprint perspective (adds dGPu power consumption and prevents CPU from entering a deep package C state sleep). A modern Intel CPU without a dGPU in the system will properly enter deep package and core C state sleep and therefore consumer under 2W at idle. I do not own an AMD APU right now, but it should behave similarly (AM1 did that with excellent results). This means that any modern system with minimal I/O installed should consume under 20W at idle, most likely closer to 15W if PSU and MB were chosen correctly.

Some interesting info. I've browsed a lot of sites looking for information on idle power and have found it frustrating to compare, probably about half of benchmarks don't even include idle power at all.

Like you said though, there's a lot of variables to it. Even from tests I ran on my own. Different motherboards seem to idle differently, different PSUs, ram sticks use power, different disk drives and dedicated video cards all can make large differences in minimum idle.

And then there's software. Video cards often idle a lot lower once their driver is loaded. I seem to recall on my ivybridge system with unraid if I used an ancient PCI video card instead of the igpu I dropped a surprising amount of idle watts. The old card only used 1watt (it seems) and (probably) the igpu wouldn't downclock using the simple driver used by that version of unraid. Booting a VM that used a passed through dedicated video cards would drop idle watts since the driver would load.

It makes the numbers on review sites sort of useless. Some sites do tests to separate out the cpu, but of course the cpu requires the platform to run so even that isn't the whole story.

 

[Collider]

Update:

Decided to go with most recent 9th Gen CPU & MOBO and ended up getting Gigabyte Z370XP SLI & Pentium Gold G5400

Happy to report idle power consumption at the wall in Windows & UnRaid is 21.4W, without any expansion cards or drives, just mobo, cpu, 16gb ram & stock cpu cooler.

This is on a EVGA 500 BQ power supply which is Bronze rated, would have liked to have Gold in there but I already had this PSU laying around.

Glad I went with a full size / featured mobo instead of mini ATX/ITX, perhaps power consumption could have been a few watts lower, but 21W is pretty good IMO.