interesting. the whole case is a heat sink directly applied to the cpu.
although, 100W cooling does nothing for the r9 390 which is something like 275W TDP. i have a veritable wind tunnel below my table, blowing really hot air up my shorts.
My fear with these designs is that because the chassis is the heatsink, there is the issue of the continuous deformation (expansion and contraction) of the chassis body/heatsink, and you get mechanical stresses on the mobo that eventually cause internal vias to break. So yes, it may work fine for 12 months, but the overall lifespan of the product might be significantly shortened.
As long as there's enough distributional area (if it's only warm to the touch), there shouldn't be enough heat build up to cause it to wear faster, right?
I've own and have built a number of these and not not seen this ever. The cases are much thicker and rigid than you'd expect. 5+ years for a few of them now.
Not going to happen. They're too small to expand much.
Aluminium is 22 microstrain per degree apparently. Say it raises to 60C and is 30cm long, that gives you're looking at 0.26 mm. Easily accommodated by flex in the mountings.
I was hoping for something like this with the skylake iris 550. Top of the line perf/watt (for a cpu+gpu). GPU fast enough for light gaming/webGL, or older games.
There's an intel NUC that's close (M2 slot + iris 540), but it's a little pricey at $340 and uses 15 watt version (that throttles very quickly), instead of the 28 watt version that can actually sustain a decent level of performance.
Fast, quiet, decent CPU+GPU shouldn't really be that hard.
Seems like GPUs are getting their with the gtx 1050/60/70 having tiny half length cards available, but I've seen very few cases designed for them.
I've owned a couple NUCs. One that was fanless (Gigabyte BACE-3000), but a bit slow. I upgraded to an Gigabyte Intel i5 NUC, it has a fan but I can't hear it at all. Brilliant machine.
I think the "quiet pc issue" is a solved problem provided you aren't building a massive gaming machine.
The site is currently reporting an "Error establishing a database connection." If at all possible, please try to make simple content sites like this static using a tool like Netlify or http://stout.is. Static sites are faster, more reliable and cheaper to operate.
It'd depend on the shared hosting provider and your site.
Having hundreds of MB of content that people are going to want to load, even if it's all served statically, will probably still have you in a bad way with the cheapest of hosting providers.
But it'll have a much better chance of surviving than someone running wordpress or something that's got zero caching and needs to do a ton of calls to a shared MySQL host.
Then yes, it would probably be fine, within reason.
When running a static site at very high traffic loads it becomes about how many connections the webserver can handle and how much bandwidth you have to serve the site itself, rather than pure power of the server to process all the requests a dynamic site would generate.
You can start to chew through bandwidth allocations pretty quickly when you get a couple thousand concurrent visitors, so a shared hosting plan might run out pretty quickly if the cap is small, even with a fairly small site. And something like Apache would need tweaking a fair bit to handle that number of connections without eating all the RAM. Nginx is better in that regard and could pretty easily handle thousands of concurrents.
So a small DigitalOcean VPS can easily handle millions of 'hits' per day if set up with a little care, and more than that if setup well, but you just have to watch you don't saturate the connection.
I mirrored a few sites that had faced the reddit hug of death (just to see what the load was like) and I found you can easily hit 200mbps* sustained connection requirement with all the visitors it brings just from the comment thread, which people don't enter as much as the list posting.
*(depending on the size of the page obviously, the bigger the page the more mbps it'll need to serve, average was around 20-50mbps).
A VPS is quite a bit different than regular shared hosting though.
Like the absurdly low single-digit-per-month hosting plans with 'Unlimited' everything. They're run to push the most customers onto the fewest boxes possible, and offer incredibly large feature sets.
So, while you might be tuning your static content to be as small as possible, you're probably on a box that's serving a heck of a lot poorly optimised sites. Disk IO is probably going to be an issue if you can't keep your content in memory (remembering that memory pressure is probably pretty high)
Ah, you're right, disk IO is one I hadn't considered too. I"m so used to availability of cheap VPS and SSD I'm not sure the last time I used actual shared hosting myself.
Those sorts of shared hosting plans have a funny way of disabling your site if you get a big surge of traffic, despite the 'unlimited' claims.
I was told once, "Yes, your plan is unlimited, but not infinite.' when I had a post on a long forgotten site go mildly popular.
I suppose on those types of machines the sheer number of connections would get you in hot water too, especially if they're running vanilla apache and it starts burning through memory there.
If it's static, which is really recommended for content like this, I'd opt for Netlify (we love it for our projects) or a comparable service.
A traditional shared host will likely run into trouble once you reach a significant traffic spike, which is exactly the moment when you absolutely don't want to experience an outage.
Netlify, even on their cheapest 10 bucks/month plan, will let you use their CDN network and offer performance that is far ahead of shared hosts, let alone database backed websites.
That's what I do, with GitLab. I chose GitLab because it allows you to use true/full https with custom domains thanks to letting you upload your own certificates.
I know you can get what looks like https with Cloudflare, custom domain and GitHub but it's not full end-to-end.
GitLab also has built in CI (rather than something external like Travis with GitHub) so you can simply push a commit and have a free 'runner' (really a Digital Ocean instance) spin up, run a build script then deploy to GitLab Pages, all for free. It's pretty amazing what you can do there to be honest.
I'm using it with Hugo but there are 'runners' for just about any SSG. I think many people have switched to using it for the ability to use Jekyll plugins unlike over at GitHub.
From the connection times I'm guessing the the servers are somewhere on the East Coast of the US (maybe they're still on Azure?) so I can hit sub second loads in Europe, the US and just about get under 1.5 secs in Australia/Asia.
It's impressive for the grand old price of 'free'!
I got so annoyed with the complexity and poor performance of web frameworks that I sat down for a day last year and made my own little database-less CMS for this kind of usecase. It's not quite as fast as a purely static solution, but I think it would be fast enough for HN:
Looks like we took the site down. I have an Intel Skull Canyon NUC and love it, will be using these from here on out. Be nice to see more companies offer fanless cases.
200 W for gaming card? Just ridiculous. Even gtx 1050 need 300 W PSU. And modern quiet fans in quiet cases are really quiet - below 20db, so it's difficult to hear it at all. My son have such with Gtx 1080 - it's the gaming card.
GTX 1050 is just 75 watts, nvidia is just being conservative because many people have crappy power supplies or tons of power consuming extras.
Assuming an intel cpu (mostly 65 watts or less), a SSD (few watts), some dims, and the normal overhead for usb/power conversion, etc. you can get away with quite a bit less than 300.
card: 75w, cpu when gaming: 80w, memory: 6w, 2 ssd drives: 6w. Total 167 W. 200 W pcu with 80% efficiency: 160 W. PCU with 95% efficiency: 190 W. Step left - step right (any gaming hard moment) and bye bye. Looks like nvidia recommendation is not so dumb. And it's for a low-level gaming card.
> 200 W pcu with 80% efficiency: 160 W. PCU with 95% efficiency: 190 W
That's not how efficiency is computed (in general, but especially in PSU).
The power rating (200W in your example) is what the PSU can provide to the system as DC power, the efficiency is the difference between AC input (wall-draw) and DC output.
An 80%-efficient 200W PSU will provide 200W internally but draw 250W from the wall-plug (and waste the extra 50W as heat).
If your 200W PSU maxes out at 160W it's not because it has 80% efficiency it's because it's a worthless piece of shit. You should throw it away as there's significant risk it will blow up (figuratively or literally) and/or catch on fire. Here's an article (in french but you can look at the pictures) checking out garbage PSU exhibiting that sort of behaviour: http://www.x86-secret.com/dossier-36-3000-Alimentation_Nonam... spoiler: shitty components, detonations and burnt smells.
I've been putting together fanless PCs for my own use for a few years now, also disliking the aircraft-taking-off sound of many desktop PCs when they are switched on. I started with the Zalman Reserator[0] water cooling tower, then moved to the Zalman TNN 500-AF[1] where the case was a giant heatsink (it was an amazing piece of kit), although more recently have been using ready-made fanless PCs (which are more widespread now most of the components run much cooler). The site at http://www.silentpcreview.com/ used to have useful information for people building quiet PCs, and https://www.quietpc.com/ still has some good components (and ready-made machines) too.
Different people have different noise threshold, and different ability to filter out noise. I couldn't sleep next to a running computer even using large low-speed fans until I built one using a Reserator. I have friends who can sleep next to what are essentially heli turbines. The big annoyance of the Reserator is it's a pain in the ass to carry around.
The TNN didn't use watercooling it used heatpipes. These are not waterblocks, they're heat absorbers. The part that's bonded to the case would be a heat spreader.
A regular case wouldn't be that good at dissipating heat though.
I can recommend the NoFan passive heatsinks. The CR-80EH costs less than many premium HSFs and will easily cope with an i7-6700. It's now possible to achieve dead silent operation without compromising on performance or spending a fortune on exotic parts.
Modern CPUs will just downclock themselves. I mean it's not insane; you won't damage the components, but you will be running A LOT slower than you could be.
Do you run tools to see how far down your CPUs frequency scaling is taking it?
I was about to write the same thing. The CPU won't be damaged, but it may underperform. The bigger concern would be other components, especially motherboard VRMs which seemingly tend not to have any thermal failsafe (on occasion, there are reports of them catching on fire).
I would be surprised if any downscaling takes place. I run some heavy computation software from time to time and compare the performance to other computers. Many of them in datacenters. I would probably notice if a machine suddenly downscales.
That said, I would be happy to measure it and post the results. Running the computations now. According to "top" it saturates 2 of my 8 CPUs. So far, "lscpu" reports the CPU at 1600 MHz. "sensors" reports the CPU Temperature is slowly rising. Was 48°C at IDLE. Is 50°C now.
15 minutes later. CPU temp was raising so slow that I stopped it now. Don't think it will get to anything critical ever. It was at 52°C. CPU speed was still at 1600MHz.
A lot of CPUs will cycle-skip which causes worse performance than just manually reducing the multiplier. Especially for real-time tasks like audio or video.
It works in the sense that the hardware will survive quite a while, but you are likely giving away a lot of performance to thermal throttling (at least if you have a mini-PC, it can be factor 5 - 10 or so)
I wouldn't call it insane (except for the cable cutting, that seems like it's just unnecessary when you can just remove the component or likely modify with software), but I also think it's ill-advised for most use cases that aren't yours, which is likely why most people are saying it's nuts.
A lot of modern computing is just greedy with resources, and will gladly eat up as much as is available, be it games or web-browsers. My first computer building experience I didn't fasten the heatsink to the mobo and it would lose it's contact with the CPU when the case was upright; powering the computer on when it was upright would allow it to load for ~ 10 seconds then shut down with no warning. CPUs and other high performance components hit their thermal limit fast, meaning they get really hot and stay really hot pretty fast without some mitigating factor. To me it seems illadvise to not give the components all the help you can to keep them cool.
Granted, my first home built was with an e8400 wolfdale, and the power and thermal efficiency of CPUs has come a long way since then, but they still get toasty without a bit of assistance. Well planned passive cooling is great - my first GPU was passively cooled and it served me well for a good 5 years until a decent night of blackjack in Vegas afforded me a new GPU. But I could still see how it struggled when the card started to reach it's thermal limits. The card ended up getting donated to a friend of mine, and I think she's still using it successfully to this day. But that being said, the passive cooling on the card was nearly triple the size of the card itself.
Funny, I was looking at these guys' latest offerings just earlier today, before the HP announcement.
This is built by the same guys that built the FitPC (http://www.fit-pc.com/). I've had one of their little 5-watt fanless systems running 24/7 in my home for a couple of years now. I love it. Never had a problem with it. It's easily one of my favorite pieces of computing equipment.
I was looking at their stuff earlier because I'm considering getting some more from them.
Checked this out last night when it was linked to in the HP post. I like the direction these fanless PCs are going in, but as others have said, it's more or less a solved problem.
Fanless laptops are where I'm now looking. The UX360CA Zenbook is fanless, low-power, and very reasonably priced (at least in the USA - still waiting on price to 'settle' over here). Silence would be very welcome for night reading and viewing.
> Fanless laptops are where I'm now looking. The UX360CA Zenbook is fanless, low-power, and very reasonably priced
The fanless Zenbooks are fantastic. But they're also really low power with the Core-M CPUs with a TDP of 4.5 Watts [0]. That's in a power budget that could fit in a tablet.
The Airtop-D has a Core i7 CPU, which has a TDP of 65 Watts (can downtuned to 37W) [1]. You can roughly expect it to have 10x+ more computational power.
You will be sorely disappointed if you expect those CPUs to have a 10x difference. First, you have a 2.75x clockspeed difference and then a double core count so 5.5 times is what you could expect if they were the same architecture but the 6Y75 is Skylake (don't be mislead by the "Core M" moniker, these the same architecture as their larger wattage brethen just with a lower TDP, they used to mark them with an Y postfix like i3-4220Y and the Y is still in the moniker) and the 5775C is Broadwell but it has eDRAM so the difference is negligible here http://www.anandtech.com/show/9483/intel-skylake-review-6700.... So in real world, you can expect to see a 3-5x speedup.
Ah, that explains that. Bearing in mind I'm thinking of upgrading from an AMD 350e Fusion CPU, so I know I'll be blown away by the performance of a new machine. It's just picking the right form factor.
I can't wait to see how quickly mutt and vim load. [Removes tongue from cheek]. In all seriousness I do run a VM, a situation I'm trying to address by writing the main software tool I use in Windows as a Vim plugin. The VM works for now, but is obviously a strain on a minimally resourced system.
And if I'm completely honest I'd love to be able to run my business and do all my work from a Raspberry Pi and/or netbook (think Chromebook pricing but entirely Linux friendly). I'm only editing text, why complicate things!?
Yes, absolutely. I'm not comparing it to the Core-M/iX range though, just making a point that in my situation either of those will be a vast improvement. Despite the fact that the E-350 does what it needs to just fine right now, it could be a touch quicker.
Yes, 3-5x (peak performance) sounds quite reasonable. But 10x is not out of the question under high load for extended amounts of time (like compiling big software) when you take the form factor thermals into account. A laptop like that will start throttling quite aggressively quite quickly.
The Zenbook I have isn't a powerful computer by any measure. It's a fine laptop for light use but the CPU and thermal characteristics aren't suitable for intensive computation. I'd expect any fanless desktop computer to wipe the floor with it when it comes to perf.
This is why I rent a server at Hetzner. Considering you can get an i7 3770 for 33.61 EUR a month, it's pretty economical. It also serves as a backup (2x3TB) and occassional webhost.
Thanks for that. It's something I've been trying to get to the bottom of, with a very limited understanding of CPU comparisons. I understand clockspeed is not something that can be directly compared, where MIPS might be more of an apples-to-apples comparison?
In that light the M7 turbos up to 3.1 Ghz, where the i7 turbos to 3.7 Ghz. That's with double the cores/threads. On paper not a 10x improvement, but in terms of MIPS it could well be. In reviews people don't seem to be phased by the new CPUs, they say they can watch HD videos and run VMs as normal, but nobody has yet given a succinct summary of the limitations. Not to mention that in these specific 2-in-1 Zenbooks I've seen the CPUs are Core-M3s, not M7s.
Aside: I didn't know the Core-M CPUs couldn't handle more than 16GB RAM, that's an interesting stat from that spec sheet.
The decision between new desktop or laptop is still to be made. A situation where something like the Zenbook could remote into a more powerful desktop sounds good but in practice would it work - I sometimes prefer to work without an internet connection, completely remotely. Sure I could sync the machines with git, rsync or similar, but the laptop still draws me in. Particularly the 2-in-1 form factor. My Kindle (keyboard!) recently went to Davy Jones' locker and I'm quite happy reading on my netbook if the settings are right. So then the tablet + desktop combination comes to mind. I don't know. Power vs portability still an issue in 2016. Hopefully not for much longer!
> I understand clockspeed is not something that can be directly compared, where MIPS might be more of an apples-to-apples comparison?
There's a joke saying that MIPS is acronym for "Meaningless Information Provided by Salesmen".
It's just not a good measure of overall performance, because there's no guarantee that it translates to real world performance.
The only way to get good performance indicators is looking at benchmarks, and with an emphasis to benchmarks that resemble the workloads you expect to be running.
For ballpark estimates (of similar hardware), the power consumption is a good measure because the performance per watt figure is typically quite close for similar CPU architectures. It doesn't work for GPUs or CPUs of very different architectures, though.
For the comparisons you linked on the Surface, it will tell you how the CPU works on that particular form factor. But it's not a good estimate of the CPU power in general, because a small device like that is always going to be thermally constrained. You can't take this conclusion and apply it to another kind of laptop/desktop.
To go down the rabbit hole even further common benchmarks will run with one-off drivers to give the appearance of better performance.
I've seen vendors skip mem-zero, skew buffer sizes, overclock, disable thermal limits and all sorts of other shenanigans specifically targeted at gaming benchmarks.
Only real way to evaluate hardware is to run the exact loads(or simulacrum of them that's not disclosed to the vendor) which probably overkill for consumer hardware.
I find the hum of lots of computer fans comforting. I did a lot of my work at uni very late at night in the computer labs when it was empty, just the constant noise of a hundred or so desktop machines. It's my ideal working environment.
That's awesome. http://mynoise.net has similar ambient sounds, they have a section on "Transports" that contains (just a few); Spaceship, Railroads, Sailboat, Aircraft cabin, Flying Fortress.. etc
It's awesome when I'm writing short stories, they've also got stuff like; RPG Dark Forest, Dungeon, Battlefield that helps me get in the mood if I'm writing a fantasy story.
IIRC it's a sound engineers hobby side project. I've got nothing to do with the project, just been a happy user for years.
What is the use-case for this? It seems pricey for entry level, but has a lot of nifty options for power-users.. but, then why would I risk having no airflow or extra power (200watt PS) for when I get cookin' with storage or fast video?
I'm a 'fan' of silence, but a Silencio ATX case is <100 and I never notice my computer, ever. I also spent a lot of time finding just the right keyboard and mouse, a sweet-spot of solid action without too much click... us techies sure are picky now that I think about it, and for a good reason, tools of the trade.
> I'm a 'fan' of silence, but a Silencio ATX case is <100 and I never notice my computer, ever.
Sounds like you have a pretty high "silence" threshold. I have a Define R4 with undervolted front fans, and while it's good enough when awake I couldn't sleep next to it.
I love the fans in my PC, the way they make almost no noise.
But they make noise. And worse, they are a marvel of engineering. If some accident of dust or bumping were to upset their balance I fear will I be back to the loud old days.
Worst of all, they are moving parts that might simply fail. Having a thing that is fanless from the start is much less likely to suddenly find its assumptions about heat dissipation violated.
For starters there are presumably use cases for which silence is of critical importance. The computer at the production desk in a broadcast or recording studio, church, or theatre comes to mind as one example.
I have been using a laptop and a docking station for a while now. I think this is the optimal configuration.
- no / very little noise
- no need to synchronize between work / home (Given that I also have a docking station at home, I essentially use my laptop as a portable hard drive with a screen)
- I can take my current work to meeting rooms anytime.
Doesn't that heavily depend on usage? At least I've never had a laptop which wouldn't turn into a noise generator once you start compiling, computing or heavy imaging/editing.
But yes a single machine + docking station at home and one on the job isn't a bad solution
> Doesn't that heavily depend on usage? At least I've never had a laptop which wouldn't turn into a noise generator once you start compiling, computing or heavy imaging/editing.
Exactly. For the past decade I've built desktop computers that are silent no matter the load (I don't do any GPU heavy work). I really cannot say that for my work laptops.
I recently got my parents a fanless desktop computer. There a lot of options, but I went with a cheap Shuttle XPC slim barebone [1]: It is tiny, has a really nice industrial design and two front facing COM ports (!).
"Two serial ports: Many PCs do not have these legacy ports any longer, since they have been superseded and replaced by USB for most consumer applications, but they are still commonly used for applications such as industrial automation systems, scientific analysis, POS systems and other such fields of application."
Yes, I noticed that to. I just ended up going for a cheap ~6 Euro USB adapter. On the plus side you get an SD card reader. (And besides Bluethooth it has great connectivity: 2x LAN + WLAN with external antennas.)
On the other hand, there are no fans pushing dust inside that PC(in fact, the whole inside of this thing can be sealed airtight) and the amount of dust that naturally settles on the radiators outside should be much smaller than on any active-cooled PC.
These (AirTop, Fit-PC, etc.) are completely sealed machines so no dust enters. It does build up on the case - depending on your environment - but doesn't effect the thermal characteristics at all.
It looks like the airtop has open air pipes that work like chimneys, letting the heated air inside flow up and out of it. These ought to be easy to clean with compressed air though.
Arguably they've been around for a while now (I have two fanless shuttle mini PCs running on my desk, and one laptop made fanless by stabbing a screwdriver through the vent holes into the fan). It's just that whenever you want the last three inches of extra performance, you pay a heavy price with respect to thermal power.
There are fanless NUC cases available. I have one from Akasa, it's a beautiful piece of kit. Solid black metal with heatsink ridges on the top, it looks and feels like a hi-fi amplifier.
The only NUC that I've heard that's loud under load is the new skull canyon which has the IRIS 580 and looks like a black plastic GPU, and sometimes sounds like a GPU.
The "normal" NUCs that are approximately mac mini in size seem quite in a wide range of use cases.
I am using this https://www.amazon.com/Qotom-Q190G4-Celeron-Processor-Barebo... machine as a Linux router and also as a media player. It's astonishingly cheap for what it is and is not an Atom which CPUs I really dislike because of their incredibly slow single thread performance. This is not fast either but it's about on par with a i3-4020Y so don't be put off by the "Celeron" marking. It's very similar to the Shuttle @dpfu mentions in this thread except a little different in ports and appearance and 20% cheaper. The Shuttle one has two DIMM slots while this only has one.
Qotom boxes are great! I run one as a router with pfSense, and it gives me no trouble. If I ever start using my TV again, I plan to get another for media center use.
One note of interest: depending on the shipper, you may need to pick it up in person, as photo ID may be required if it comes through customs. This was no issue for me since the DHL depot at BWI is less than a mile from the light rail stop - I even got to see a couple of 777 and 747 freighters up close! - but may be inconvenient depending on shipper and depot location. Still very much worth it IMO - the price/performance of Qotom hardware is extremely impressive, and nothing I could more easily source compares.
Yeah when I looked at the price I was extremely skeptical of getting four Intel Ethernet controllers but there they are... LOL I remember when a four port Ethernet card costed more than this whole machine :D
I'll second the Qotom boxes. I use it as a firewall and run Sophos UTM Home. It's fanless and the integrated heatsink gets kinda warm to the touch but with some air flowing over it (ceiling fan) it works great.
If this thing can dissipate 200W of heat, that I must say it's truly impressive. I cannot say I want to buy one (I'm not that sensitive about noise), but from technical point of view it seems like a little marvel.
Does anybody understand the copper heatpipes? It mentions virtual vacuum, but my understanding is that heatpipes are typically filled with a liquid that boils and turns to gas as it heats, thus moving up along the pipe and carrying heat further away to allow for a larger contact surface with the heatsink. Are airtop's heatpipes somehow different? What "virtual vacuum" and how does it work?
This is not all that new of an idea. I had one of these around 2005/2006: https://www.quietpc.com/hfx-mini-metal - completely fanless and (virtually) silent - back in those days there were no SSDs, so one had to content onesself with slow spindle speed HDDs mounted on soft rubber to minimise vibrations.
Despite the fact that the mfr appears to think it's a "revolutionary" idea, I'm not really seeing any material difference between that and the case above, from over 10 years ago. Perhaps I'm wrong though?
I had a couple of machines from Hush Technologies (long since out of business), also ~10 years ago. They were passively cooled through a finned chassis, connected to the CPU and other hotspots by heat conducting pipes.
Like yours, sounds like the same sort of thing as these Airtop machines.
151 comments
[ 4.7 ms ] story [ 214 ms ] threadalthough, 100W cooling does nothing for the r9 390 which is something like 275W TDP. i have a veritable wind tunnel below my table, blowing really hot air up my shorts.
Aluminium is 22 microstrain per degree apparently. Say it raises to 60C and is 30cm long, that gives you're looking at 0.26 mm. Easily accommodated by flex in the mountings.
There's an intel NUC that's close (M2 slot + iris 540), but it's a little pricey at $340 and uses 15 watt version (that throttles very quickly), instead of the 28 watt version that can actually sustain a decent level of performance.
Fast, quiet, decent CPU+GPU shouldn't really be that hard.
Seems like GPUs are getting their with the gtx 1050/60/70 having tiny half length cards available, but I've seen very few cases designed for them.
I think the "quiet pc issue" is a solved problem provided you aren't building a massive gaming machine.
Google Cache for those who want it:
http://webcache.googleusercontent.com/search?q=cache:EHdeB5i...
Having hundreds of MB of content that people are going to want to load, even if it's all served statically, will probably still have you in a bad way with the cheapest of hosting providers.
But it'll have a much better chance of surviving than someone running wordpress or something that's got zero caching and needs to do a ton of calls to a shared MySQL host.
When running a static site at very high traffic loads it becomes about how many connections the webserver can handle and how much bandwidth you have to serve the site itself, rather than pure power of the server to process all the requests a dynamic site would generate.
You can start to chew through bandwidth allocations pretty quickly when you get a couple thousand concurrent visitors, so a shared hosting plan might run out pretty quickly if the cap is small, even with a fairly small site. And something like Apache would need tweaking a fair bit to handle that number of connections without eating all the RAM. Nginx is better in that regard and could pretty easily handle thousands of concurrents.
So a small DigitalOcean VPS can easily handle millions of 'hits' per day if set up with a little care, and more than that if setup well, but you just have to watch you don't saturate the connection.
I mirrored a few sites that had faced the reddit hug of death (just to see what the load was like) and I found you can easily hit 200mbps* sustained connection requirement with all the visitors it brings just from the comment thread, which people don't enter as much as the list posting.
*(depending on the size of the page obviously, the bigger the page the more mbps it'll need to serve, average was around 20-50mbps).
Like the absurdly low single-digit-per-month hosting plans with 'Unlimited' everything. They're run to push the most customers onto the fewest boxes possible, and offer incredibly large feature sets.
So, while you might be tuning your static content to be as small as possible, you're probably on a box that's serving a heck of a lot poorly optimised sites. Disk IO is probably going to be an issue if you can't keep your content in memory (remembering that memory pressure is probably pretty high)
Those sorts of shared hosting plans have a funny way of disabling your site if you get a big surge of traffic, despite the 'unlimited' claims.
I was told once, "Yes, your plan is unlimited, but not infinite.' when I had a post on a long forgotten site go mildly popular.
I suppose on those types of machines the sheer number of connections would get you in hot water too, especially if they're running vanilla apache and it starts burning through memory there.
I know you can get what looks like https with Cloudflare, custom domain and GitHub but it's not full end-to-end.
GitLab also has built in CI (rather than something external like Travis with GitHub) so you can simply push a commit and have a free 'runner' (really a Digital Ocean instance) spin up, run a build script then deploy to GitLab Pages, all for free. It's pretty amazing what you can do there to be honest.
I'm using it with Hugo but there are 'runners' for just about any SSG. I think many people have switched to using it for the ability to use Jekyll plugins unlike over at GitHub.
From the connection times I'm guessing the the servers are somewhere on the East Coast of the US (maybe they're still on Azure?) so I can hit sub second loads in Europe, the US and just about get under 1.5 secs in Australia/Asia.
It's impressive for the grand old price of 'free'!
https://github.com/muellermichel/guetzli
Assuming an intel cpu (mostly 65 watts or less), a SSD (few watts), some dims, and the normal overhead for usb/power conversion, etc. you can get away with quite a bit less than 300.
That's not how efficiency is computed (in general, but especially in PSU).
The power rating (200W in your example) is what the PSU can provide to the system as DC power, the efficiency is the difference between AC input (wall-draw) and DC output.
An 80%-efficient 200W PSU will provide 200W internally but draw 250W from the wall-plug (and waste the extra 50W as heat).
If your 200W PSU maxes out at 160W it's not because it has 80% efficiency it's because it's a worthless piece of shit. You should throw it away as there's significant risk it will blow up (figuratively or literally) and/or catch on fire. Here's an article (in french but you can look at the pictures) checking out garbage PSU exhibiting that sort of behaviour: http://www.x86-secret.com/dossier-36-3000-Alimentation_Nonam... spoiler: shitty components, detonations and burnt smells.
[0] https://www.quietpc.com/reserator1-v2
[1] https://www.quietpc.com/tnn500af
A regular case wouldn't be that good at dissipating heat though.
http://www.nofancomputer.com/eng/products/CR-80EH.php
1) I buy any desktop I like for its form factor
2) I either use the onboard graphic or a fanless graphics card
3) I set the threshold for all fans that are accessible via fancontrol to a level that keeps them permanently off
4) I pull or cut the cables of any fans that remain spinning
People always tell me I am "insane" to do so. But I have been doing this for years now and never had a problem.
Do you run tools to see how far down your CPUs frequency scaling is taking it?
That said, I would be happy to measure it and post the results. Running the computations now. According to "top" it saturates 2 of my 8 CPUs. So far, "lscpu" reports the CPU at 1600 MHz. "sensors" reports the CPU Temperature is slowly rising. Was 48°C at IDLE. Is 50°C now.
15 minutes later. CPU temp was raising so slow that I stopped it now. Don't think it will get to anything critical ever. It was at 52°C. CPU speed was still at 1600MHz.
A lot of modern computing is just greedy with resources, and will gladly eat up as much as is available, be it games or web-browsers. My first computer building experience I didn't fasten the heatsink to the mobo and it would lose it's contact with the CPU when the case was upright; powering the computer on when it was upright would allow it to load for ~ 10 seconds then shut down with no warning. CPUs and other high performance components hit their thermal limit fast, meaning they get really hot and stay really hot pretty fast without some mitigating factor. To me it seems illadvise to not give the components all the help you can to keep them cool.
Granted, my first home built was with an e8400 wolfdale, and the power and thermal efficiency of CPUs has come a long way since then, but they still get toasty without a bit of assistance. Well planned passive cooling is great - my first GPU was passively cooled and it served me well for a good 5 years until a decent night of blackjack in Vegas afforded me a new GPU. But I could still see how it struggled when the card started to reach it's thermal limits. The card ended up getting donated to a friend of mine, and I think she's still using it successfully to this day. But that being said, the passive cooling on the card was nearly triple the size of the card itself.
This is built by the same guys that built the FitPC (http://www.fit-pc.com/). I've had one of their little 5-watt fanless systems running 24/7 in my home for a couple of years now. I love it. Never had a problem with it. It's easily one of my favorite pieces of computing equipment.
I was looking at their stuff earlier because I'm considering getting some more from them.
Fanless laptops are where I'm now looking. The UX360CA Zenbook is fanless, low-power, and very reasonably priced (at least in the USA - still waiting on price to 'settle' over here). Silence would be very welcome for night reading and viewing.
The fanless Zenbooks are fantastic. But they're also really low power with the Core-M CPUs with a TDP of 4.5 Watts [0]. That's in a power budget that could fit in a tablet.
The Airtop-D has a Core i7 CPU, which has a TDP of 65 Watts (can downtuned to 37W) [1]. You can roughly expect it to have 10x+ more computational power.
It's just a whole different category of devices.
[0] http://ark.intel.com/products/88199/Intel-Core-m7-6Y75-Proce... [1] http://ark.intel.com/products/88040/Intel-Core-i7-5775C-Proc...
I can't wait to see how quickly mutt and vim load. [Removes tongue from cheek]. In all seriousness I do run a VM, a situation I'm trying to address by writing the main software tool I use in Windows as a Vim plugin. The VM works for now, but is obviously a strain on a minimally resourced system.
And if I'm completely honest I'd love to be able to run my business and do all my work from a Raspberry Pi and/or netbook (think Chromebook pricing but entirely Linux friendly). I'm only editing text, why complicate things!?
The Zenbook I have isn't a powerful computer by any measure. It's a fine laptop for light use but the CPU and thermal characteristics aren't suitable for intensive computation. I'd expect any fanless desktop computer to wipe the floor with it when it comes to perf.
In that light the M7 turbos up to 3.1 Ghz, where the i7 turbos to 3.7 Ghz. That's with double the cores/threads. On paper not a 10x improvement, but in terms of MIPS it could well be. In reviews people don't seem to be phased by the new CPUs, they say they can watch HD videos and run VMs as normal, but nobody has yet given a succinct summary of the limitations. Not to mention that in these specific 2-in-1 Zenbooks I've seen the CPUs are Core-M3s, not M7s.
Aside: I didn't know the Core-M CPUs couldn't handle more than 16GB RAM, that's an interesting stat from that spec sheet.
The decision between new desktop or laptop is still to be made. A situation where something like the Zenbook could remote into a more powerful desktop sounds good but in practice would it work - I sometimes prefer to work without an internet connection, completely remotely. Sure I could sync the machines with git, rsync or similar, but the laptop still draws me in. Particularly the 2-in-1 form factor. My Kindle (keyboard!) recently went to Davy Jones' locker and I'm quite happy reading on my netbook if the settings are right. So then the tablet + desktop combination comes to mind. I don't know. Power vs portability still an issue in 2016. Hopefully not for much longer!
EDIT: Found a rudimentary comparison based on the Surface. It doesn't show even a 3x improvement: http://www.lovemysurface.net/surface-pro-4-core-m3-vs-i5-vs-...
There's a joke saying that MIPS is acronym for "Meaningless Information Provided by Salesmen".
It's just not a good measure of overall performance, because there's no guarantee that it translates to real world performance.
The only way to get good performance indicators is looking at benchmarks, and with an emphasis to benchmarks that resemble the workloads you expect to be running.
For ballpark estimates (of similar hardware), the power consumption is a good measure because the performance per watt figure is typically quite close for similar CPU architectures. It doesn't work for GPUs or CPUs of very different architectures, though.
For the comparisons you linked on the Surface, it will tell you how the CPU works on that particular form factor. But it's not a good estimate of the CPU power in general, because a small device like that is always going to be thermally constrained. You can't take this conclusion and apply it to another kind of laptop/desktop.
I've seen vendors skip mem-zero, skew buffer sizes, overclock, disable thermal limits and all sorts of other shenanigans specifically targeted at gaming benchmarks.
Only real way to evaluate hardware is to run the exact loads(or simulacrum of them that's not disclosed to the vendor) which probably overkill for consumer hardware.
https://www.youtube.com/watch?v=NeESf9aCZHQ
https://www.youtube.com/watch?v=Gy_RHdE7zsc
:-)
https://www.youtube.com/watch?v=ZPoqNeR3_UA
It's awesome when I'm writing short stories, they've also got stuff like; RPG Dark Forest, Dungeon, Battlefield that helps me get in the mood if I'm writing a fantasy story.
IIRC it's a sound engineers hobby side project. I've got nothing to do with the project, just been a happy user for years.
Sounds like you have a pretty high "silence" threshold. I have a Define R4 with undervolted front fans, and while it's good enough when awake I couldn't sleep next to it.
I love the fans in my PC, the way they make almost no noise.
But they make noise. And worse, they are a marvel of engineering. If some accident of dust or bumping were to upset their balance I fear will I be back to the loud old days.
Worst of all, they are moving parts that might simply fail. Having a thing that is fanless from the start is much less likely to suddenly find its assumptions about heat dissipation violated.
- no / very little noise
- no need to synchronize between work / home (Given that I also have a docking station at home, I essentially use my laptop as a portable hard drive with a screen)
- I can take my current work to meeting rooms anytime.
- ...
Doesn't that heavily depend on usage? At least I've never had a laptop which wouldn't turn into a noise generator once you start compiling, computing or heavy imaging/editing.
But yes a single machine + docking station at home and one on the job isn't a bad solution
Exactly. For the past decade I've built desktop computers that are silent no matter the load (I don't do any GPU heavy work). I really cannot say that for my work laptops.
[1] http://www.shuttle.eu/fileadmin/resources/download/docs/spec...
They'll come in handy!
But seriously, what's the motivation behind putting those on the machine, or wanting them?
"Two serial ports: Many PCs do not have these legacy ports any longer, since they have been superseded and replaced by USB for most consumer applications, but they are still commonly used for applications such as industrial automation systems, scientific analysis, POS systems and other such fields of application."
https://www.scan.co.uk/products/akasa-newton-h-aluminium-fan...
The "normal" NUCs that are approximately mac mini in size seem quite in a wide range of use cases.
One note of interest: depending on the shipper, you may need to pick it up in person, as photo ID may be required if it comes through customs. This was no issue for me since the DHL depot at BWI is less than a mile from the light rail stop - I even got to see a couple of 777 and 747 freighters up close! - but may be inconvenient depending on shipper and depot location. Still very much worth it IMO - the price/performance of Qotom hardware is extremely impressive, and nothing I could more easily source compares.
What are your ambient temperatures?
Despite the fact that the mfr appears to think it's a "revolutionary" idea, I'm not really seeing any material difference between that and the case above, from over 10 years ago. Perhaps I'm wrong though?
Like yours, sounds like the same sort of thing as these Airtop machines.