AMD Ryzen beats Intel Core i7 as a heater (that's also a server) French cloud concern Qarnot wants to use AMD processors to heat water. No, this isn't a cruel joke at the expense of thermally inefficient silicon. Instead it's a deliberate plan to use CPUs' heat-producing powers for good. Qarnot makes a device called the “ Q.rad” that uses three CPUs to provide heat and serve as a node in a …
I am typing this on a recent A12 crop. It is the first time ever I have a laptop which hits the 3 key points for me:
1. Proper video with full linux support
2. As fast as a decent desktop - it is within 20-30% of a desktop compiling linux kernel or openwrt.
3. Light and cold.
My previous job tried to give me a top of the line Core i7. The nvidia version simply did not work properly with Linux and the intel one was painful to work with on a 4k screen - you could see how it redraws it as if it was a 1990-es ISA VGA.
How would you say the A12 compares to the older APUs?
Performance-wize it beats the living daylights out of old laptop models - both E and laptop A series. Tested with development - both C and Java.
Its power consumption is still not as good as Intel. Some of it can be attributed to the incomplete support for the GPU power-management in Debian stable, but some of it is for real. The equivalent i5 with the same battery (both HP 15 inch) supposedly lasts longer. Not that I care - it is a desktop replacement, if it will travel somewhere it will be where there are sockets for it.
The model I used for tests is a HP 15-ba047na. Unfortunately, there are only 2-3 models to chose from on the market - this one is with the best video subsystem.
I would suggest replacing the ridiculously slow spinner HP put into it for it to really shine.
For instructions on how to change/add memory and replace the spinner search for "HP 15 AMD Service Manual". It is significantly easier than the equivalent Intel models which use a different case design (so make sure you have the one for AMD models). While the case design for Intel gives you easier access to the memory and WiFi modules, changing the drive on that requires complete disassembly.
Tools needed for the AMD version are a car trim pry tool (or a plastic kitchen "knife" for baking) and a cross-head screwdriver. It is not even using hex screws or something else like that - fully serviceable as long as you know (or guess) how to open it.
It's a good idea! IT equipment produces a lot of heat - especially in server rooms where there's a lot of kit packed in high density. I'm always amazed how few places try to utilise that heat to (for example) heat the building, rather than just using air-con to waste the heat outside. Good to see some companies at least using that heat for a beneficial purpose.
"Holy shit, was that Angelina Jolie?"
You've never watched the majesty that is Hackers? It's one of the best/worst films about, well, hackers, ever made!
The soundtrack is fantastic, and it's got Angelina Jolie and Jonny Lee Miller (yes, that's Sickboy with the brolly), who ended up getting married after the film. It's also got the obligatory psuedo VR hacking sequences that every 90's film about computers has to have, oh, and pretty much the first appearance of Wipeout.
I probably did see it at one point, won't know until I go watch a couple more clips - I almost never know who people are in films at the time unless they are already well-known. I'm a lazy bastard on things like that.
I also have no idea what most of the music I listen to is called either :)
Although not generally applicable to purpose built DCs, which tend to use inert gas fire suppressants, many of the swimming pools on the roofs of buildings, especially multi-story buildings, are usually primarily put there to act as the reservoir for the sprinkler system.
In most cases, the mains water supply has insufficient capacity for a sprinkler system, which needs to deliver a lot of water very quickly, but you can't use a pump to try to draw water from the main at a higher rate, by under-pressuring it, without the risk of collapsing the main.
Thus, for sprinkler systems, you need a large reservoir at the top of the building that can be trickle-charged, as it were, in times of non-fire and in suitable climates these can double up as swimming pools.
I had a tour of a Swiss university's new HPC machine room. The geeks wanted to do exactly that; apparently the architects asked if they could guarantee that the computer room would still be in that location for 20 or so years outputting a similar amount of heat - they couldn't make that promise so they had to build it without.
Supposedly the CS department at the Uni of Manchester (first purpose built CS dept in the country) was built with that in mind, simply recirculating the heat from the machine room. I can confirm while there in the mid nineties there appeared to be no functional heating to speak of - the implicit assumption was that computers would remain largely valve based.
Except that you really don't want the headache of water-cooling your rack-mounted servers. A single leak becomes a real headache. Besides, datacentres are curiously tetchy about piping water in and around the machine halls... You can use Fluorinert or similar, but then you can't use plastic pipes and fittings. So then you're down to dragging the heat out of the air.
I also wouldn't want to get a plumber out in an emergency. It was bad enough waiting for the IBM guy to turn up for a cooling fault. 8 days, and he still made an arse of it - put on so much thermal paste you'd have thought he was laying tiles...
But anyway - hot air -> aircon -> fluid -> heat pump to ramp up the temperature -> ??? -> profit!
"headache of water-cooling" you don't actually have to pass a river over your CPUs you could consider a different cooling fluid internally and pass to an external heat exchange to the water.
But then again having your gaming multicoloured rig visible through a perspex lined river (complete with fish) would be a talking point. Now I just need an old mill house with a decent internet connection
Except that you really don't want the headache of water-cooling your rack-mounted servers.
See for a good example of watercooling servers.
BTW CSCS are very open to giving tours; if you're in the area then their setp is very, very impressive.
Cool the aircon heat exchanger.
At any reasonable cost, you'd lose a lot of the heat, and there's a problem with the grade of heat.
As the nominal efficiency of a heat exchanger rises, the output recovered heat temperature declines. The value of heat depends on having reasonable quality waste heat to fit in with existing use cases. So either you have a highly efficient heat exchange producing low grade heat that there's only so much use for, or you recover reasonably useful grade of heat, but have low recovery efficiency.
As a demonstration project, or a subsidy funded toy, you can work around these, but if you're doing this commercially, recovering waste heat with an average temperature of 40C is of little use. Most building heating systems use much higher temperatures even on the return flow, and hot tap water needs to be kept at 60C to prevent legionella. Auxiliary use for heating a swimming pool is fine, but typically you'll add a lot of complexity to a system that still depends most of the time on a fossil fuel boiler.
@ Ledswinger 40C for free home heating for low income families means they are not having to use costed services as much if at all. There has been free local house and water heating from power stations successfully for years outside of the UK, so efficient transfer systems do exist.
All they need do now is remove the standing charge for the meters and most homes would have a much lower heating bill. That they wouldn't have to live off charity anymore in the UK would also be a good thing.
On energy meters why are they allowed to charge forever, I have never had one break down during my life so they clearly last for decades and at whatever it is a month they charge they have had their return over and over again.
Apparently more people than just me have considered this...
Firstly, full disclosure - I water-cooled my home desktop for years and liked it. A lot. Sadly it's all aluminium parts, so I can't really extend it or update any part of it so I eventually had to retire it when things stopped being able to be bodged into place. I'd do it again, but better things to do with money now that I have kids...
Directly water-cooling rack-mounted servers means that you have to have some way of coiling the pipes up so that you can draw the servers out of the rack on rails and not have to unplumb them - that's what I mean about the headache. A teeny leak can cause havoc, and good luck getting a warranty on that kit. And when you first connect it, you'll end up with bubbles somewhere. Again, a headache.
I did mention Fluorinert, which could then be exchanged to water at the back of the chassis. Reasonably sane, but you still have the extendable plumbing issue. At least internally it can be a sealed system with no bubbles!
Yeah - water-cooled cabinets. Looked at the sheets, but never seen one in the flesh so I couldn't possibly comment more.
Low-grade heat. Ledswinger has it. There's not many uses for it, and you can't transport it far without losing the heat in it. Under-floor heating? Great. Pool heating? Great. Domestic hot water? Not so much. I don't think it _needs_ to be kept above 60 degrees all the time, but it needs to be up past there for at least a couple of hours a day for legionella (I am not a plumber, but I have spoken to many), so it adds up to much the same thing. Wall-mounted radiators need 60 degrees, or have to be massively oversized. So until everyone gets a pool or under-floor heating then we're rather stuck for what to do with it. That's why I suggested banging a heat-pump in the middle. You'll get a lower flow-rate, but higher temperature.
And the datacentres that we're located in would still go apoplectic if you started running water into the halls. That said, there's a moderately-sized DC in central London that had water fire suppression last time I was there (about 2 years ago). Maybe they'd let you hook in...
There's a large DC on the outskirts of London owned by a major bank that has water cooled cabinets housing the blade chassis with a few CO2 cooled cabinets (TROX equipment) scattered about as well.
(BTW: If the DC you're referring to is on the river, opposite Blackfriars then it does still have a water sprinkler system)
40C for free home heating for low income families means they are not having to use costed services as much if at all
You evidently know little of the complexities and immense costs of district heating systems, nor of the very limited uses of very low grade heat. If these "low income families" were paying for the full system, they'd be paying a lot more than any conventional approach to heating. Maybe they're just paying their additional fuel costs, but even so, SOMEBODY is paying for a very expensive bit of eco-bling.
Heat recovery is like recycling. Just because you can ALWAYS do it at a technical level, doesn't mean that it is either economic, or environmentally sound.
You evidently know little of the complexities and immense costs of district heating systems, nor of the very limited uses of very low grade heat.
But I'm not, I live in a house using district heating, powered by burning wood pellets to generate electricity to heat the water. The source water for that comes in at (say) 10 C. Would we use more or less wood pellets if the water instead came in at 40 C from the large DC that is also served by the district heating? Would that make it cheaper or more expensive?
By the way, I find the district heating a lot more efficient than my previous boiler/radiators. It's roughly the same price, but I'm warmer in winter and have virtually infinite hot water, whilst before I had enough for a shower, and to the do the washing up, each day
Under floor heating typically uses temperatures of 25c.
That would normally be the surface temperature, not the heat transfer circuit temperature. In theory you can use a circuit temperature of 25 C, but in practice the heat transfer rate (with a delta T to room temperature of as little as 4C) would be absolutely appalling, meaning the system would have very low controllability, and slow warm up times. That's basic physics, no amount of clever design can alter that. So a 25C circuit is fine, if you've got a house with excellent insulation, a heat system designed for a very low temperature input, a property with very high thermal mass (inertia), and you keep it at a constant single temperature (very inefficient unless the house is occupied 24/7).
Most underfloor heating will have an absolute minimum circuit temperature of 30C, but it will perform a whole lot better with a circuit temperature of 40-50C, and that'll usually produce a surface temperature in the 25-32C range, where higher is much better for comfort and control.
This post has been deleted by its author
(literally) to get some PR for people to install their devices. As Ledswinger points out, for many situations the yield is very poor. Heating water from <= 10°C might work but either you don't do a lot of this or you need a much bigger unit. But, even in France, heating with electricity is considered inefficient.
Would someone help me understand what Q.Rad really is? I did not understand. It says:
"Qarnot’s cloud is based on a disruptive infrastructure: the Q.rad, a heater embedding three CPUs as a heat source. We reuse the heat they generate to heat homes and offices for free. Q.rad is connected to the Internet and receives in real time workloads from our in-house computing platform.
Our main business at Qarnot is to provide a cloud computing platform which is able to send heavy workloads to any kind of computer, server or heater."
How does it transmit the heat?
Used computers as heaters many times, even server rooms are sources of heat. I know of several places where the location of the condenser was inside so it would help with heating. In the summer they open the room with the condenser to the outside but most of the year all the heat is kept in the building.
The foolishness of politics and the inability of people to think for themselves is shown by such discussions. There is little to no waste when the building needs to be heated. In more than one case that I know of lighting was upgraded to be more "efficient", including replacing 500 and 1000W incandescent lights and in another area replacing MV lighting only to have to then add in electric heaters due to those areas being cold. The install costs alone was orders of magnitudes higher than operation costs, and electrical consumption did not go down.
But hey LED's are green!
The efficiency equation when heating an area is not the same as if you are cooling the area. Often the Engineers try to tell companies that but it's too complicated to understand and is in conflict with the daily propaganda from green suppliers.
One of the last places I worked, before I quit the trade forever; I drew up a proposal to replace all the 1KW exterior/workbay halogen lighting - which was on 24/7, 6 1/2 days per week; with low pressure sodium.
The proposal had a breakdown of the current cost of running the halogen, the cost of replacing the elements and light boxes (usually had to do the latter because the heat melted the alu boxes), and compared them with the cost of the LPS boxes, expected lifetime of the boxes and the elements, and the cost of running them.
The conclusion was that the total cost of buying and installing them would pay for itself in SEVEN months.
The boss looked at it, called me a smartarse, and threw it in the bin.
We produced so much heat, that even in the coldest winter, we had all the windows open and worked in shorts; yet the offices used electric heaters and electric water heaters.
The reg is quoting the Frencies as saying "Ryzen Pro is producing the same heat as the equivalent Intel CPUs we were using while providing twice as many cores."
Same heat = equal as a heater; provides twice the cores = beats it as a server. The headline should be "AMD Ryzen beats Intel Core i7 as a server (that's also a heater)".
“Ryzen Pro is producing the same heat as the equivalent Intel CPUs we were using while providing twice as many cores.”Qarnot's tests also showed “a performance gain of 30 to 45 per cent compared to the Intel i7 we were previously using.”
And yet you just HAD to make the heading sound like Ryzen was more inefficient than Intel's offerings...
.... my 14U rack of networking gear, short 1U servers and NAS are in my utility room next to one of the intake vents for my heat recovery ventilation (and the washer and (heat pump) dryer).
Any outside temperature above freezing and the combination of MHRV and circa 500W of continuous heat being pumped off that rackmount gear (+the fridge and my own body heat) keeps my ~100 sq m home at 21C.
I meet the average power consumption with solar PV on the roof too...
In my old place I did the same but without the MHRV to redistribute the heat I had:
1. A lot more equipment in a 42U rack
2. ...to leave all the doors open.... or on particularly cold days close the doors... strip to my undies and spend the day in the computer room.
3. fewer visitors of the female sex...
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
