About seller
Not all my systems have been so successful. In 2022 I measured a pair other systems at 19 watts and 27 watts as part of Curbing the “Gas-Guzzling” tendencies of AMD Radeon with Multi-Monitor . Whereas I did handle to get that 27 watt AMD system down in energy a while later, not every CPU/motherboard combo is destined for the 10 watt ballpark. Earlier than going further, the 7 watt determine for this system was earlier than storage has been added. Be aware that if I don’t enable powertop to disable the keyboard, I get eight watts measured at the wall.Let’s get into detailed specs and part choices. DDR4 and reuse a CPU I have already got. Typically you just should take your greatest educated guesses when it comes to parts, build your rig, tune what you possibly can, and let the chips fall the place they may. Earlier than I begin, here is a fast look at the motherboard layout. The Inexperienced CPU-linked slots and ORANGE chipset-connected slots will develop into relevant all through this write-up. Certainly one of my goals above was the capacity for an eventual 6 NVMe drives. Digging into deeper particulars as to why this generally is a challenge (feel free to skip this section)… Drawback: There are zero shopper motherboards with 6x M.2 slots that may all be used at the same time in PCIe mode.’ll discover that when you try to populate all 6, the last one has to be a SATA variant. The ASRock Z790 PG Sonic also has 6 slots, however you possibly can only use 5 of them (with a respectable excuse: they offer a Gen5 NVMe slot however it comes with an both/or caveat). Why This Downside Exists: There are chipset lane limitations on shopper boards. Assuming I need the power to run all M.2 in Gen4x4 and assuming a producer had been truly keen to devote all of the lanes to M.2 NVMe slots (they’re not), AMD X570 and Intel B760 would max at three M.2 slots, with AMD B650 and Intel H670/Q670/Z690/W680 managing 4.5 M.2 slots is feasible on AMD X670 and Intel H770 boards. Six on a Z790 board. Past that, extraordinary measures like robbing the main PCIE slot of lanes would be required. The solution: PCI-E to M.2 adapters turned mandatory. Now when trying to find a motherboard, it became a matter if including the M.2 slots included to any obtainable PCI-E slots capable of x4 or larger. My options had been now restricted to AMD X570, Intel H770, and Intel Z790 motherboards.Word that whereas using bifurcation is a risk on some motherboards to get greater than 1 NVMe out of the primary PCIe slot, I decided not to depend on it. 7w). AMD chipset energy consumption has concerned me for some time, as earlier X570 chipsets had a TDP of 11w and wanted a fan. Chipset Velocity: Intel H670/Q670/W680/Z690/H770/Z790 chipsets have a DMI 4.Zero x8 hyperlink to the CPU. AMD X570/B650/X670 have a PCIe 4.Zero x4 link to the CPU.1. I already had 64GB of DDR4 that the Intel system may use. AMD 600-sequence chipsets are all DDR5-only. 2. I already had an Intel 12th Gen CPU. 3. I’ve but to see any optimistic dialogue round AM5 power consumption. In any respect. Replace: as I used to be writing this, news truly got here out about AMD 7000-collection CPUs burning/bulging the place the motherboard socket pins meet the CPU. Yeah, sorry AMD, not this time.So Intel it was. After trying out available DDR4 motherboards on the market, I rapidly narrowed options to 2 manufacturers: MSI and ASUS. Don’t care about the board comparisons? Be happy to skip this. The enticing MSI boards had been the Professional Z790-P WIFI DDR4 and Z790-A WIFI DDR4. Almost similar on the floor, except the “A” is a bit of extra premium (audio, rear ports, heatsinks, energy phases, and so forth). Con: My present B660 board that results in increased idle consumption than expected is an MSI. Enticing ASUS options have been the Prime H770-Plus D4 and Prime Z790-P D4 (optionally available WIFI version). Moving into the TUF, Strix, or ProArt was just too costly. The Z790-P D4 is analogous except it has more power phases, higher heatsinking, more USB ports, further fan header, and for our functions…Finally the ASUS Prime H770-Plus D4 was about $a hundred cheaper at the time and is what I chose. One upside I’ve discovered with “cheaper” boards is they are likely to have fewer elements and thus less vampire energy drain at idle, though this isn’t at all times a certainty. I already had this CPU as part of a previous desktop construct. I was getting a new motherboard with 2xDP anyway, and going older-gen didn’t make sense to me.That desktop construct turned out to be a disappointment, and ranks as considered one of my least favorite builds. I had points where sometimes solely 1 of two DP-attached monitors would wake in Linux which meant I needed to either pull/reconnect the opposite DP connector, or manually suspend/resume the system so it may strive once more. One other difficulty was that rebooting between Home windows/Linux sometimes induced odd points which necessitated a full poweroff/restart. Hardware decode on Ubuntu using Wayland continues to be problematic and when applications tried to make use of it to play video, issues would ensue. Lastly, unlike my earlier Intel systems which could all be introduced down close to the 10 watt mark, this one was idling at 19 watts, although I suspected the MSI motherboard I was utilizing may have been an element.Many of the headaches I skilled have been associated to the GPU and show. Since I used to be about to build something server-oriented, that was no longer an element. This was memory I already had. I ran the four sticks of memory on the XMP profile of the dual-rank equipment which was 16-18-18-36. Every part else was basically left to the defaults besides that I ran the RAM at 1.25 volts (higher than stock 1.20, but lower than the XMP 1.35v setting).This element wasn’t deliberately chosen. When i needed a fresh Ubuntu Server set up for testing, this happened to be the only SSD I had kicking around that wasn’t at present getting used. I used to be going to be doing lots of A/B testing on PCIE and NVMe devices, so putting in Ubuntu 23.04 to a SATA SSD made sense to maintain PCIE slots free. Word that after testing, the primary OS was to be run on a Samsung SSD 970 EVO Plus 500GB NVMe. Not much to say besides that Samsung stuff tends to reliably go into low energy modes. Having used both drives, I can’t measure any power difference between them in my testing.Tom’s Hardware tested the Samsung idle at 0.072 watts (through ASPM/APST), and Anandtech tested the Sandisk Ultra 3D idle to be 0.056 watts (by way of ALPM). Each are well below the 1W decision of my Kill-A-Watt meter. As a lot as this 750W PSU could look like overkill for a system meant to sit round 10 watts, when 12 drive motors spin up at the identical time, the instantaneous load is more likely to be fairly high.Seagate states 2A/3A DC/AC peak currents on the 12V rail for considered one of their 10TB 3.5″ drives. This bursty energy demand has the potential to be problematic if the PSU isn’t up to the task. If an array of 6 drives collectively pull 150-200 watts at the same second the CPU spikes to drag a peak 120W, that’s a leap from around 10 watts idle to around 400 watts. This could simply cause an instantaneous voltage dip - if it dips enough to cause an instantaneous crash/reboot it’s probably not an enormous deal, but when it dips simply sufficient that knowledge is corrupted throughout a reminiscence refresh or when another drive is mid-write…’s a extra painful downside. Oversizing the PSU to a point (or including some in-line capacitors to the power rails) is smart. Fortuitously, regardless of operating exterior of the peak effectivity vary, a lot of the Corsair RM sequence is fairly environment friendly across a variety. 9-10 watts was the consumption when the show output was on. 600 kernel boot parameter for a 600s timer), which is the place this system sits many of the week. 8 watts was the consumption if the USB keyboard power management was disabled. If you happen to don’t SSH into the server from elsewhere, spending the additional watt for keyboard use may be crucial.As talked about in the beginning, I began with 12 laborious drives. Half have been 2.5″ and the opposite half were 3.5″. As a result of the motherboard only has 4 SATA ports, a SATA controller and a port multiplier were used to handle the remaining drives. Moreover, 4 NVMe drives had been used early on: certainly one of them, a Western Digital SN770 had a tendency to get fairly sizzling even at idle which signifies it in all probability wasn’t going right into a low energy mode.With all the gear related, at idle, with display off, and with the 12 drives spun down to standby, I used to be shocked to see that my idle energy consumption had gone from 7 watts all the way in which up to a whopping 24-25 watts. Far a lot! Something was amiss. I disconnected the exhausting drives and began testing parts one at a time. ee88 club had been fairly crude tests meant to get a tough idea as to the culprit, so numbers here aren’t exact. I rapidly discovered that the JMB585 SATA controller I was utilizing induced energy consumption to increase by one thing within the 6-10 watt vary (precise measurements in a later part).The controller itself is just purported to take a pair watts, and the tiny heatsink stayed cool, so there was obviously extra happening. Where was the ability going? I decided to watch the CPU package deal C-states. Without the JMB585 SATA controller, the system hit C6. When the JMB585 was reconnected, the best the system hit was C3. Ah ha! However why? Seems that if a PCIE-connected device won’t go into ASPM L1, the CPU won’t go into as deep a sleep.The JMB585 controller playing cards don’t seem to have ASPM help. A bit further experimentation revealed one thing else that I hadn’t identified, and it has to do with C6 vs C8. The system will solely hit C8 if there’s nothing hooked as much as the CPU-hooked up PCIE lanes. In other phrases, if something is plugged in to the top PCIE slot or the highest NVMe slot, C6 is the maximum. So whereas C8 would be a luxurious, hitting C6 was a should.C3 makes use of a lot power. If SATA controllers have been going to forestall the CPU from hitting the very best energy saving states, I began to wonder whether or not I ought to have been in search of a motherboard with 6-8 SATA ports so that I wouldn’t need to rely on add-on controllers… A little trying to find SATA HBAs confirmed that while there aren’t many options right here, the ASM1166 SATA controller should support ASPM L1, although the firmware must be flashed for it to work correctly (and to work in any respect on newer Intel boards).This was one thing I’d have to order: I've Marvel and JMicron spares, however they don’t assist ASPM. I’d actually been avoiding ASMedia for years, but out of necessity they were now getting one other likelihood: I ordered a pair ASM1166 6 port SATA controllers. Worth a mention… Throughout initial testing with a WD Black SN770 (Gen4 NVMe), I found an issue when the primary (top CPU-hooked up) PCIE and NVMe ports had been used. ’t cease the errors, but diminished them substantially. Some analysis showed the root trigger will be any number of issues. Because swapping the motherboard or CPU wasn’t a pleasing thought, my best hope was swapping to a distinct brand of NVMe.I ordered some Crucial P3 NVMe drives. This turned out to be a profitable endeavor: with the WD drives replaced by the Crucial drives, I used to be now not getting any errors, though keep in mind these are Gen3 drives. When i had the 2 Essential P3 NVMe drives installed in the CPU-linked PCIEx16 slot and the highest M2 slot, I noticed greater idle temps than expected. While the NAND sat at about 27-29C, the controllers have been reporting 49-50C - a lot higher than I anticipated for these explicit drives. I moved the one from the PCIEx16 slot to a chipset-related PCIEx4 slot. L1 sub-states solely seem to get enabled on the chipset-linked slots.Unlucky, but it surely does seem to coincide with the available BIOS settings in the screenshot above. I put each NVMe drives on chipset-related PCIE slots. Sadly when trying numerous A/B exams utilizing these 2 Crucial NVMe drives with different slot configurations and varied BIOS settings, I saw very inconsistent conduct when it comes to temperature, although it’s value noting the JMB585 and an NVMe boot drive have been also connected during these tests. For example, both drives may idle at around 40C until a soft reboot at which point 1 (or each) might now idle on the 50C vary. Sometimes it seemed doable to maintain 1 drive on the CPU-related M.2 and retain 40C temperatures on each drives as lengthy because the x16 slot wasn’t populated.Presumably I was hitting some sort of bug. The Samsung boot NVMe seemed to maintain a constant idle temperature regardless of what was taking place with the Crucial NVMe drives, so I suspected the Essential drives themselves are not less than partly to blame. Curiously, generally one (or each) controller temps would drop all the way down to the 29C range when on the chipset-related slots. Late Update: I sadly didn’t keep track of temperatures throughout the remainder of the testing, and heatsinks/airflow between drives have all been jumbled round. Power Consumption Puzzles - Swapping from the JMB585 to the ASM1166.After a couple weeks the ASM1166 arrived. First a pair bits regarding the card which you would possibly discover useful if you’re contemplating it… I began with a firmware flash - ASM1166 playing cards often have old firmware which doesn’t work with Intel 600-sequence motherboards and from what I understand can have points with energy administration. For anybody planning to purchase one of those ASMedia playing cards, I ought to observe that like most SATA controllers and HBAs out there, the standard actually varies. One among my cards had a heatsink that was on a bit crooked: the thermal pad was thick enough to stop it from shorting close by parts, but be aware that these merchandise can be really hit-and-miss.This is among the situations the place paying somewhat more to buy from someplace with an excellent return policy can be prudent. I did fairly a little bit of A/B testing, so right here is a fast “JMicron JMB585 vs ASMedia ASM1166” by way of total system energy consumption, although it may solely be applicable to this platform (or even perhaps this specific motherboard). First, power consumption without any drives related to the cards (the SATA SSD boot drive is related to the motherboard) to get a baseline. 1 watt). Measurements after the display output went to sleep. 14 watts) where the CPU bundle power state plummets to C2.Shockingly, the JMB585 behaves in an reverse manner the place it’s consumption is decrease on the CPU-connected slot (and it didn’t cause C2) - nonetheless, you’ll soon see that things change when drives are literally connected… I did further testing with the controllers, including taking part in “musical chairs” with a couple NVMe drives to see if a number of units would throw a wrench into things, but nothing unexpected came about so I’ll skip those particulars. With baseline measurements full, subsequent it was time to truly put some drives on these controllers. The SATA SSD boot drive stayed on the motherboard, 2 NVMe drives were added to the mix (chipset-connected except in any other case famous), and four of the 2.5″ SATA arduous drives were positioned on the controller.I’ll listing the “spun down” consumption after the exhausting drives went into standby - “spun up” was precisely 2 watts larger in every test whereas the drives were idle. 8 watts. No contest. A further profit is that I was ready to make use of both of the ASM1166 playing cards within the system, whereas trying to use each of my JMB575 cards at the same time resulted within the system refusing besides, although that may very well be a platform or motherboard-specific difficulty.There's a commerce-off although - I at all times found the JMB585 to be rock-stable reliable, including when paired with a JMB575 port multiplier. My past experience with ASMedia SATA controllers has been lower than stellar: reliability with the ASM1166 stays to be seen, however on the very least it’s a foul candidate for a port multiplier because it doesn’t assist FBS (only CBS). 1. When removing/reinserting the NVMe boot drive, a BIOS message appeared claiming that it couldn’t boot attributable to GPT corruption. The ASM1166 cards had to be briefly eliminated for the BIOS to “find” the NVMe boot drive again (after which they may very well be reinstalled).Linux has to iterate by means of all of them. One of many comments talked about an older Samsung 840 Pro SSD limiting to C3 whereas a crucial Force GT SSD allowed C8. While these are older drives, I nonetheless discovered this a bit shocking. It was worth investigating. I used the H770 as a testbed with a Samsung 850 EVO SATA SSD boot drive along with an important P3 NVMe and built a customized kernel to allow the Realtek network adapter to achieve L1.2. No ASM1166, just utilizing the Intel onboard SATA.I reached C10 after working powertop with auto-tune and permitting the display to sleep. I tried varied drives I have on hand, powering off the system every time to swap drives and repeat the method. Here were the outcomes. I recommend being cautious when choosing SATA SSD manufacturers and fashions. I’ll try to replace this list over time with drives I’ve examined, but keep in mind certain manufacturers in the storage space have proven a propensity in the direction of silently swapping to inferior elements in a few of of their mainline products, so you should always confirm the claimed efficiency metrics of any storage devices you buy whereas within your return window.Be happy to depart a remark with good/bad drives you come throughout. 1) Motherboard help and BIOS configuration are vital - I’ve had motherboards with very inflexible BIOS’s. On this one, “Native ASPM” and the appropriate L1 states have to be enabled (to permit OS-controlled as an alternative of BIOS-controlled) for low power consumption to work. 2) Units all must support ASPM L1. In any other case you’re really rolling the dice. The hardest half right here as you may need guessed is finding SATA controllers that support it - if doable, get a motherboard sufficient enough Intel chipset-related SATA ports to keep away from needing a separate card.I ought to observe that finding NVMe drives that have working low-energy APST power states beneath ASPM isn’t always a given and you’ll want to do a little analysis there too. 3) If you possibly can hit the C8 energy state, avoid utilizing CPU-connected PCIe lanes (top PCIe and M2 slot). On this particular motherboard, my advice can be to avoiding utilizing them altogether if you can, unless you either want the low-latency full-bandwidth path to the CPU or your devices are so active they by no means sleep anyway.Recall that Each my JMicron and ASMedia SATA playing cards brought about the CPU Package deal C-State to plummet to C2 if plugged into the x16 PCI-E slot. 14USD transport on the time via eBay). At the moment I found our not often-used fax machine which was at all times powered on used 7 watts… 10 years greater than paid for the Kill-A-Watt. Seagate rates standby consumption of those 3.5″ drives at about 0.8w every, and the 2.5″ drives at about 0.18w each. This lines up with what I’m seeing above.My active-idle numbers really match up fairly well to Seagate specs too. The plain commentary: in comparison with the remainder of the system components, the 3.5″ drives are energy-hungry monsters. The HDDs will ultimately be replaced with SSDs. With idle consumption as low as it is during HDD standby, there isn’t a major rush and this process will regularly happen as my HDD drives/spares fail and SSD costs fall. nha cai ee88 for “end game” is for an all-SSD build.Initially the plan was for 1 boot drive, 6xNVMe (possible Crucial P3 4TB) for a RAIDZ2 array, and 6xSATA (probably Samsung 870 QVO 8TB) for the 2nd RAIDZ2 array. Since using the CPU-connected M2/PCIe slots not only brings unpredictability but in addition comes at a slight C-state/energy/temperature cost, I'd alter that plan and surrender a couple NVMe in the first array and use SATA instead in order that I don’t have to contact CPU-linked lanes. This half is simply value studying if you’re taken with meticulous particulars concerning the storage.Be at liberty to skip to the ultimate part otherwise. As alluded to earlier, this is a Samsung 970 EVO Plus. Presently lower than 4GB of the 500GB house is used (a 64GB swap partition exists however at all times sits at zero used). It was initially chosen because Samsung had developed a popularity for reliability (which has been falling by the wayside recently), and Samsung also scored nicely in opinions every time it got here to idle energy consumption. This drive is sort of all the time idle and both Controller and NAND temps stayed low all through all testing (20-24C). It might eventually be swapped to a SATA SSD to free up an NVMe port.These drives are used for the first 6-drive ZFS RAIDZ2 array - the one which will get probably the most use. Someday per week it’s busy with a activity that entails studying just a few TB over the course of 24 hours. Usage by way of the remainder of the week is sporadic, and the drives spend many of the week spun down. Power consumption of the 2.5″ Seagate drives is honestly pretty spectacular. Spun down they’re each rated at 0.18w, in low energy idle they’re rated at 0.85w, and the learn/write averages are rated at about 2w. There are many SSDs out there with worse power consumption numbers than this spinning rust.5TB capability provides a number of storage-per-watt. Not nice performers. 80-120MB/s peak learn/write. To be truthful although, many TLC/QLC SSDs fall to these write ranges when their SLC cache is exhausted. SMR (Shingled Magnetic recording). Reads are nice, but write performance absolutely plummets when random writes happen - it acts like a QLC SSD without an SLC cache that also doesn’t have TRIM. Low rated workload (55TB/year vs 550TB/12 months for 3.5″ Exos drives). No configurable error recovery time (SCT ERC), and these drives can grasp for minutes if they hit an error while they relentlessly attempt to re-read the problematic sector. Ubuntu must be configured to attend instead of trying to reset the drive after 30 seconds.Greater error rates in the event that they heat up (I’ve had to change just a few and have discovered they don’t like running hot). Typical HDD pain factors (slow to spin up, and many others). To be completely truthful to Seagate, these are sold as external USB backup drives. Pulling these 15mm tall drives out of the enclosures and using them as RAID members in a NAS isn’t exactly utilizing them as supposed. The extremely low energy consumption is great, but there are obvious trade-offs. Long run, these 2.5″ 4/5TB drives will slowly be replaced by 4TB SSD drives (possibly all NVMe).SSDs in 4TB capability began to develop into accessible on the consumer finish in 2021/2022 at about 4-5x the price of the spinners. Lower than 2 years later they’ve dropped to about 2x the fee, and i count on respectable brands to last greater than 2x as lengthy because the Seagate spinners. If availability of the Crucial P3 (Gen3) mannequin remains, I’ll possible keep with this model regardless of being limited to Gen3 speeds.I strongly thought of the Crucial P3 Plus (Gen4), but power consumption in opinions was larger regardless of only a few conditions where performance was notably increased as nicely. My greatest concern with the P3 Plus (Gen4) was that if I had points with ASPM/APST, Tom’s Hardware showed it with a 0.3w idle energy premium over the P3 (Gen3) for the 2TB model. I desire “worst-case scenario” power to be as low as possible. Used in the secondary 6-drive RAIDZ2 array - a backup array that’s spun up for about 2 hours per week where it receives constant heavy writes.Energy consumption of the 3.5″ Seagate drives is about what you’d expect. These 10TB drives are rated at about 0.8w every in standby, 2-5w idle, and 6-9w studying and writing. These are rated to collectively pull about 45-50 watts when writing. That’s a little bit of additional UPS load I don’t really need if a lengthy energy outage takes place in the course of the backups (I keep on with consumer 1500 watt UPS’s). These are rated to collectively pull about 4.8 watts when in standby. Again, some UPS load I wouldn’t thoughts shaving off. Long-time period these drives will possible be changed by Samsung 870 QVO 8TB SATA drives. The 870 QVO sports 0.041w/0.046w idle with ALPM, 0.224w/0.229w idle with out, and 2.0-2.7w throughout a duplicate (according to Toms/Anandtech).Price-clever, the Samsung 8TB SATA SSD is at the moment a good bit dearer than 8TB spinners (closer to 3x the associated fee) so except these drives begin to see extra frequent use for some purpose, substitute with the SSDs will almost actually wait till I’ve run out of spares. Replacing my spinning rust with SSDs is a process that can likely take some time. “Special” Allocation Class - lets you create a vdev particularly for metadata and for “small” blocks if desired. A cache drive, recognized generally as an L2ARC.Should you create the “special” vdev at pool creation, all of your metadata (and optionally, small blocks of a measurement you choose) will go on the “special” vdev instead of your spinning rust. Very fast file listings and listing traversal while protecting the spinning rust for the files themselves. Yes, you possibly can “ls” a bunch of directories without waking your HDDs from sleep. Biggest draw back is that because all of your metadata is on this vdev, if it ever dies, access to all of your information is essentially gone. So it actually ought to be at the very least mirrored.Possibly even a 3-approach mirror. Say goodbye to some ports. The L2ARC is a bit completely different. It’s essentially a stage 2 cache. When the cache in RAM will get full, ZFS will copy some of the blocks to the L2ARC before it evicts that content material from RAM. The next time that data must be accessed, it’ll be read from the L2ARC instead of the disk. One profit compared to the “special” vdev is that you’re advantageous with solely 1 SSD - if there’s an issue with the information in the L2ARC (bad checksum, drive dies, and so forth), ZFS will just read the content material from the original disk.Additionally, as soon as the L2ARC is full, ZFS will just begin once more firstly of the L2ARC SSD and overwrite stuff it wrote earlier than which has some pros (old data never accessed anymore) and cons (data that was steadily accessed and will need to get written to the L2ARC again). You can even add/take away L2ARC devices from the pool at your leisure - need so as to add a 64GB SSD, 500GB SSD, and 2TB SSD? Go right forward - ZFS will distribute blocks among them.Have to take away the 500GB SSD from the pool a few days later and use it elsewhere? Go proper ahead. The largest downside to the L2ARC is that in case you forget to specify “cache” when including the device, you most likely mucked up your pool. It’s also imperfect: even with cautious tuning it’s laborious to get ZFS to jot down Every part you need to the L2ARC before it will get evicted from memory. At the identical time, relying on your knowledge, the L2ARC may see a variety of writes, and it's possible you'll must rigorously watch the well being of your SSD. Up to now I’ve used the “special”, used L2ARC, and have used both at the identical time (you can even inform the L2ARC to not cache issues already contained in the “special” vdev).NVMe drives to stay in low power mode more…). Conclusion - Regrets? Second-guessing? What could have gone differently? In contrast to the ASRock J4005 build where I realized part approach by means of that I’d kneecapped myself in a number of ways, I don’t get the identical sense right here. This time I ended up with low idle energy AND a reasonably succesful system that should be versatile even if repurposed in the future. I’m fairly proud of my component decisions, though I’d be curious to know how the MSI Pro Z790-P DDR4 (one of the opposite motherboards I thought of) would do in comparison. Performance-smart the MSI has the advantage of 6xSATA ports, however it comes with the plain downside of the notorious Intel 2.5G networking chip.The MSI also has a PS/2 port and I’ve never really checked to see if PS/2 keyboard power consumption is decrease than USB (recall that I save 1 watt if I allow powertop to shut down the USB keyboard port). And of course it could be interesting to check the ASPM and ALPM settings, and to see if the snags I hit with CPU-connected PCIe/M.2 slots exist in the same means. Whereas this system currently sits within the 15-sixteen watt vary when idle with drives in standby, as soon as all HDDs are changed with SSDs, I’d count on idle consumption of round 10-11 watts which isn’t dangerous for 72TB value of drives, 64GB of RAM, and a fairly first rate processor.Replace: Recent Linux kernels disable the L1 energy saving modes of most Realtek NICs which prevents the CPU from entering respectable C-states, thus increasing energy consumption by rather a lot. Whereas there are workarounds, transferring forward I’ll probably restrict myself to motherboards containing Intel 1 Gigabit network adapters (maybe moving to Intel 2.5 Gigabit when it becomes clear they’ve labored out all the kinks). You can find additional details concerning the Realtek NIC scenario within the comments below. The script I mention deep in the feedback will be found at: RTL8125-ASPM.sh.txt .