Dual Intel LAN NUC!
-
@gonzopancho:
Do you have any idea what the idle power consumption might be for the Netgate RCC-VE 2440? Also, any idea when it can be purchased? Webpage says "Available Q1 2015" which is now, but no "add to cart" button.
Yes I do. "Add to cart" will work when they're in-stock.
"Yes I do" answers the idle power consumption question, or the when they'll be in stock question? Or both? So, since you know the answer to one or both questions, I'll ask the direct questions: what is the idle power consumption? And when will they be in stock?
@gonzopancho:
Anyway, TDP on the i5-4250U is 15W, base frequency is 1.3GHz, Turbo is 2.6GHz (http://ark.intel.com/products/75028/Intel-Core-i5-4250U-Processor-3M-Cache-up-to-2_60-GHz).
TDP on a C2358 (2 core C2000) is 7W, base frequency is 1.7GHz, Turbo is 2.0GHz (http://ark.intel.com/products/77978/Intel-Atom-Processor-C2358-1M-Cache-1_70-GHz)TDP is interesting, but not a useful metric for actual system-level power consumption. The CPU/SoC is but one component on the board. There are memory, disks, maybe a BMC/IPMI unit, NICs, graphics chips in some cases, etc. Then you have power supply efficiency (which is a curve, i.e. load-dependent, not constant).
Case in point, despite the 15W TDP of the i5-4250U, my NUC at idle, takes about five watts total from the wall. That's for the CPU, single 4GB DIMM, SSD, NIC, etc. My Desktop i5-2500k has a 95 Watt TDP, big SSD, 16 GB RAM (4x4 DIMMs), and idles at just under 30 Watts from the wall. Of course, either of these systems, when fully cranked up, pull way more than the CPU's TDP.
So that's why I'm interested in actual at-the-wall power consumption, as measured by something like a Kill-a-Watt. It's that, not the TDP, that drives my electric bill.
-
Darek here from Logic Supply, I just wanted to mention that we've recently announced a Dual NIC NUC system - http://www.logicsupply.com/blog/2015/03/12/introducing-next-generation-industrial-nuc-ml100/
I'd be happy to answer any questions you have.
-
Sure - Install the new release of pfsense on it, test it and get back to us with results.
-
@gonzopancho:
Do you have any idea what the idle power consumption might be for the Netgate RCC-VE 2440? Also, any idea when it can be purchased? Webpage says "Available Q1 2015" which is now, but no "add to cart" button.
Yes I do. "Add to cart" will work when they're in-stock.
"Yes I do" answers the idle power consumption question, or the when they'll be in stock question? Or both? So, since you know the answer to one or both questions, I'll ask the direct questions: what is the idle power consumption? And when will they be in stock?
@gonzopancho:
Anyway, TDP on the i5-4250U is 15W, base frequency is 1.3GHz, Turbo is 2.6GHz (http://ark.intel.com/products/75028/Intel-Core-i5-4250U-Processor-3M-Cache-up-to-2_60-GHz).
TDP on a C2358 (2 core C2000) is 7W, base frequency is 1.7GHz, Turbo is 2.0GHz (http://ark.intel.com/products/77978/Intel-Atom-Processor-C2358-1M-Cache-1_70-GHz)TDP is interesting, but not a useful metric for actual system-level power consumption. The CPU/SoC is but one component on the board. There are memory, disks, maybe a BMC/IPMI unit, NICs, graphics chips in some cases, etc. Then you have power supply efficiency (which is a curve, i.e. load-dependent, not constant).
Case in point, despite the 15W TDP of the i5-4250U, my NUC at idle, takes about five watts total from the wall. That's for the CPU, single 4GB DIMM, SSD, NIC, etc. My Desktop i5-2500k has a 95 Watt TDP, big SSD, 16 GB RAM (4x4 DIMMs), and idles at just under 30 Watts from the wall. Of course, either of these systems, when fully cranked up, pull way more than the CPU's TDP.
So that's why I'm interested in actual at-the-wall power consumption, as measured by something like a Kill-a-Watt. It's that, not the TDP, that drives my electric bill.
TDP is useful, all other things being equal. SSDs take more power than eMMC. Powering USB takes more power than you might think. Yes more RAM -> more power.
But the CPU (assuming you're using it) will still be the largest consumer.
Your NUC "at idle" isn't really using the CPU much. It's mostly in C states. If you're using your NUC that little for pfSense, I wonder why you're using it.
So the 8 core variant of the RCC-VE, at full tilt, with the USB, miniPCIe, and m-sata all running, will run off a 36W power supply.
This figure includes headroom (because you don't want to actually pull 36W from a 36W power supply for too long.)They're in-stock now.
At idle, the 2 core doesn't reliably register on a Kill-a-Watt. The Kill-a-Watt is hardly an example of state-of-the-art measurement technology, especially at low loads.
-
Hi,
at the CeBit 2015 Shuttle was presenting a Barebone and let them run 24/7 as a firewall!
Could this be something interesting related to this thread?Shuttle DS57U5
- passive cooling without turning parts
- Intel Core i5-5200U with 2 Cores (2,2 GHz, Turbo: 2,7 GHz) HD-5500-GPU
- 2,5" SSD/HDD
- 2 x SO-DIMM
- mSATA SSD miniPCIe slot
- 2 × Gigabit LAN Ports (Intel i211 & i218LM)
- 2 x USB 3.0
- 2 x RS-232
- WLAN (WiFi)
- Price ~550 € because brand new it will going down
Shuttle will be offering a lower cost model, that works with an Core i3-5005U
soldered onboard, the price therefore is not given to the public at this time.One hint, that the colling system is working well, the barebone must be stand vertical!
Shuttle DS57U5 (This CeBit news are only in german!)
-
@gonzopancho:
TDP is useful, all other things being equal. SSDs take more power than eMMC. Powering USB takes more power than you might think. Yes more RAM -> more power.
But the CPU (assuming you're using it) will still be the largest consumer.
Your NUC "at idle" isn't really using the CPU much. It's mostly in C states. If you're using your NUC that little for pfSense, I wonder why you're using it.
So you run your pfSense box(es) with the CPU pegged at 100% at all times? Or the overwhelming majority of the time? With no headroom to spare? That's the only time TDP is really useful, when the CPU is maxed out. TDP's usefulness is directly proportional to CPU load over time (and for cooling design obviously).
I can see where, if your traffic is much more predictable and deterministic, and likely future growth has been accounted for, you could run your hardware in such a state. But, it's actually pretty hard to hit true TDP 100% of the time, particularly with modern complex CPUs like current Intel x86 stuff. In the DIY computer world (overclocking in particular), people use artificial benchmarks like prime95 and FurMark to push their hardware to the theoretical max to ensure adequate cooling. But in reality, even the most demanding application won't hit that actual theoretical max, even if the CPU is at "100%".
But what if your load is not deterministic? And I'll argue most aren't. What if you foresee some future growth (faster Internet connection, more users, need for VPN, etc)? What if the nature of your traffic is very bursty? I.e., long-periods of little-to-no load, with short periods of very intense load/demands.
If you say that it's pointless to have NUC-like hardware at idle most of the time, you're missing my point. For a general purpose computer (such as a media/home theater PC), the NUC is wonderful: I can leave it on all the time with a negligible hit to my electric bill, but have a lot of power when I need it. That's generally h264 video decoding now, but what about the next fancy codec? What if I decide to run a game simulator on this PC too? What if I also want to use the machine for some software development? Or even for my desktop: I don't need a 95W TDP i5-2500 quad core to compose this message; I could just as well do it on my cell phone's ARM CPU. So is my desktop CPU a waste? I'm sure the HD encoding I regularly do that maxes out my CPU would take just a bit longer on my cell phone's CPU.
To me, it's the best of both worlds: lower electrical power consumption, but high computational potential when I need it. Are you suggesting this isn't a valid use-case for a network device?
Say you're an independent consultant, who mostly works from home, but occasionally travels for business to do on-site work (say a sales pitch, or critical client-support work), and your work requirement calls for a fast, reliable connection back to your home system. Here is a case where your network device might be mostly idle (or at least near-idle), but will need some decent computational power from time-to-time. What if your network connection deals with financial markets? These are extremely bursty. What if you're hosting an E-commerce site? Even a huge site like Amazon surely has enough redundancy and over-capacity to support inevitable failures and peak-load times like Christmas.
In all these cases, idle load will dominate. Or, maybe not idle, but certainly not-peak power draw. So the long-term power consumption will be dominated by a number that is considerably less than TDP.
@gonzopancho:
So the 8 core variant of the RCC-VE, at full tilt, with the USB, miniPCIe, and m-sata all running, will run off a 36W power supply.
This figure includes headroom (because you don't want to actually pull 36W from a 36W power supply for too long.)That's great, now what is the actual draw from the outlet? A 36W PSU provides enough power plus headroom. How much headroom is there? And what is the efficiency of the PSU? And, what about idle consumption at-the-wall? I understand you think this may not be relevant, but for me it is.
@gonzopancho:
At idle, the 2 core doesn't reliably register on a Kill-a-Watt. The Kill-a-Watt is hardly an example of state-of-the-art measurement technology, especially at low loads.
Who suggested the Kill-a-Watt is SotA? I'm not looking for ultra-precision here. This isn't a legal audit, I'm looking for a reasonable prediction of the impact to my electrical bill. If the measurement tool can give me that (say <5 Watts, or 5-10W, or 10+), I'm happy.
I understand, from a commercial perspective, you probably can't get by quoting Kill-a-Watt numbers. But I'm the consumer, not the seller, and like I said, looking for reasonable ballpark numbers, +/- a couple watts.
Edit: Perhaps some of the disagreement over the usefulness of TDP comes from CPUs that don't have the newer power-saving features. For example, the Atoms prior to Centerton: there was very little difference in power consumption between idle and maxed-out; the TDP was a useful proxy for making power consumption predictions. But the latest generation Atoms, and of course the desktop/server range of Intel CPUs, have a massive amount of computational potential at electrical power levels far below TDP.
-
Hi all,
For those interested: I have just ordered a Shuttle DS67U with Intel Celeron, the version that Shuttle already hinted on being released. Offcourse I will post details on the journey later on.
Datasheet here: http://www.shuttle.eu/fileadmin/resources/download/docs/spec/barebones/DS67U_e.pdf
With support for AES-NI, QuickSync, dual Intel-based LAN, it looks like a winner to me :)
(now all I have to do is find out if the RealTek WLAN NIC will work … but that wasn't a dealbreaker for me.) -
I have a MiniITX board with dual intel NICs:
pfSense i7-4510U + Dual Intel 82574 Gigabit MiniPC Build: https://forum.pfsense.org/index.php?topic=113610.msg631641#msg631641
Brand Name: HAMSING
Processor Main Frequency: 1.8GHz(Tubo 3.0GHz)
Processor Model:Intel I7 4500U
Model Number: HS-4500I
Hard Drive: Transcend 64GB SATA III 6Gb/s MSA370 mSATA Solid State Drive
RAM: 8GB 1600MHz DDR3L PC3-12800 ECC CL11 1.35V SODIMM
Video: VGA+HDMI
Audio: Realtek ALC6662
Network: Intel 82574 21000M
USB : 6usb2.0 2USB3.0
RS232: 6RS232
WIFI: 300M -
I have a MiniITX board with dual intel NICs
Do you have any figures on power usage? Power usage is one of the reasons I chose this setup, combined with routing power, as it will be connected to 500mb/500mb fiber uplink.
-
I have a MiniITX board with dual intel NICs
Do you have any figures on power usage? Power usage is one of the reasons I chose this setup, combined with routing power, as it will be connected to 500mb/500mb fiber uplink.
I'll hook it up to my zWave network sometime this week and get you the stats. I'm guessing it is in the 35W range
-
I have a MiniITX board with dual intel NICs
Do you have any figures on power usage? Power usage is one of the reasons I chose this setup, combined with routing power, as it will be connected to 500mb/500mb fiber uplink.
I am running at cMAX (turbo) for this processor. My z-wave device is reporting 12-14W of usage. I also just added a mini PCI card from Jetway with 2x Intel i350 ethernet ports: https://forum.pfsense.org/index.php?topic=113610.msg643350#msg643350
