EEVblog Electronics Community Forum
Products => Crowd Funded Projects => Topic started by: JacquesBBB on February 12, 2015, 09:53:47 pm
-
https://www.kickstarter.com/projects/projectexergy/project-exergy-darehenry?ref=discovery (https://www.kickstarter.com/projects/projectexergy/project-exergy-darehenry?ref=discovery)
Obviously, these guys have no idea of what there are speaking about.
Many of the new big computer centers use very efficient free cooling systems, where the
cooling is made by direct ventilations, with eventually some water evaporation help.
As an example, I run a 2000 core system with a 1.04 PUE which means that the cooling cost is only 4% of the
effective computer cost, compared to the 40-60 % this guy is mentioning. It is some
similar efficient cooling that is used by the facebook data centers.
They also seems to underestimate that they will need to evacuate the heat he will produce with
their high temperature computer. I cannot believe he already has raised 34000$.
Hopefully, they will not get funded, but it is not sure as they seems to be quite a lot of illuminate people
all around that are easy preys for such bullshit project.
-
Why would I want this? I already have a device that uses my computer's generated heat to heat my room. It's called a CPU heatsink, and I can tell you it does a good job based on the blisteringly hot temperatures my office can reach after my PC's been on for a while.
-
Yes, he also forgot to say what does he do with all this heat in summer
when you want to cool the temperature of your house.
-
Sorry, but am I right in thinking they will save energy by using purposely inefficient computers as glorified bar heaters?
-
I wish I could find the link...someone was working on some kind of new nano wire thermo-electric generator for use in automobiles so on board electronics can be powered without a need to beef up the alternator or battery. I have to find that article.
Regardless, I DO wonder if this nano-wire tech could be used with electronics heat sources to re-capture energy. Usable electricity is WAY more useful than just heat in practical applications.
EDIT: Here it is: http://www.extremetech.com/extreme/199022-nanowire-breakthrough-could-recycle-vehicle-exhaust-improve-car-efficiency (http://www.extremetech.com/extreme/199022-nanowire-breakthrough-could-recycle-vehicle-exhaust-improve-car-efficiency)
The article does imply that it requires a lot of heat over a large surface area (the whole exhaust) system. Tiny amounts of heat from a PC just won't cut it.
-
It is so outrageous that i believe now that it is a hoax. Probably they started with 20 or 30 % of fake donations to try to grab the remaing part.
-
They are trying to put some cash in their pockets by reusing this idea:
http://www.geek.com/news/german-cloud-provider-will-heat-your-house-if-you-let-them-install-a-server-1609178/ (http://www.geek.com/news/german-cloud-provider-will-heat-your-house-if-you-let-them-install-a-server-1609178/)
-
Three comments:
- solar roadways type bullshit :bullshit:
- good marketing
- stupid market base to buy into stupid things
-
The concept of using waste heat from some useful process (eg. running a CPU) in place of pure ohmic heating as in an electric bar fire, is a good one. The total amount of heat produced from running a 500W computer is, as anyone with a basic understanding of physics knows, exactly the same as a 500W ohmic heater - so to convert electrical power straight into heat without it doing something else useful in the process is always missing out on an opportunity to get something for nothing. (Not energy, of course, but 'usefulness').
It's a shame this is so poorly understood, by the people who propose these schemes and those who back them. The lack of elementary knowledge about heat, temperature and energy here is staggering, and the fact that they'd announce it so clearly in public, embarrassing.
I had an idea years ago to use the waste heat from a desktop CPU to brew coffee. If a CPU draws 100W and can run with a water cooling system at, say, 90 deg C, then there should be enough heat available (and at a high enough temperature) to heat water for coffee via a heat exchanger. The whole lot could be built into the PC case, and would be the first genuinely fit-for-purpose "developer's" PC ever made, IMHO.
-
I could just imagine the tap/faucet being right over the keyboard. :-DD
-
TL;DR. What do they claim to do?
I stopped the video half way when the kids started dancing. The pledges give nothing real in return and yet they have $500 average pledge.
-
TL;DR. What do they claim to do?
As Mechanical Menace said, they're basically talking about purposely using an inefficient computer to generate heat, using a heat exchanger to capture it, and then pumping it through a house for central heating (or cooling, apparently?)
It's basically just doing something "useful" (number crunching, cloud computing, etc.) with that electrical power in the process of turning it to heat for heating, rather than just pumping it through an ohmic heater.
-
super energy efficient and energy productive by capturing and storing up to 95% of the heat generated by the internal components.
Where does the other 5% of the energy go? :o Heresy
And crap for efficiency too :p
http://www.walmart.com/ip/FRIGIDAIRE-FFRH12222-Window-A-C-w-Heat-12K-Btu-208-230V-G5332381/41269944 (http://www.walmart.com/ip/FRIGIDAIRE-FFRH12222-Window-A-C-w-Heat-12K-Btu-208-230V-G5332381/41269944)
11,000 btu/H or 3200 watts of heat out, with only an 1100 from the wall :p
And you actually get it instead of some vaguely related trinket!
-
DareHenry by Project Exergy Team, looks like selling magic heat to me. Can't see this making any sense at all from an energy point of view.
But if you pledge 10 grand you get:
"Corporate Sponsor - Your company logo will be prominently displayed on our website, select marketing materials and prototypes clearly showing your support of the quest to evolve the way energy is consumed. You’ll receive a commemorative signed and numbered heat exchanger ‘plaque’ suitable for display. Based on the design our partners at the Clark County Skills Center created, this plaque illustrates the power ideas and innovation have to engage people in creating solutions to some of our biggest problems. We'll want to work closely with you as the product develops to drive a sustainable and profitable future for us all."
-
DareHenry by Project Exergy Team, looks like selling magic heat to me. Can't see this making any sense at all from an energy point of view.
But if you pledge 10 grand you get:
"Corporate Sponsor - Your company logo will be prominently displayed on our website, select marketing materials and prototypes clearly showing your support of the quest to evolve the way energy is consumed. You’ll receive a commemorative signed and numbered heat exchanger ‘plaque’ suitable for display. Based on the design our partners at the Clark County Skills Center created, this plaque illustrates the power ideas and innovation have to engage people in creating solutions to some of our biggest problems. We'll want to work closely with you as the product develops to drive a sustainable and profitable future for us all."
So basically, for ten thousand dollars, you get your logo on some guy's random website that will rapidly fade into obscurity, and a sign you can hang on your wall. Sounds legit.
-
Well, you get "select marketing materials" too! :-DD
-
Well... Should I dive in here and straighten this out or are you all so KS jaded it is not worth the time?
Lawrence
-
Well... Should I dive in here and straighten this out
You mean pile on more bullshit?
Good to see that this scam was cancelled!
-
It was moved to IndieGoGo. As far as bullshit goes, we were final 4 out of more than 100 applicants for 3mil in DoE funding and presented in DC to a panel of industry experts. We are partnered with a national lab, the Fraunhofer institute and Notre Dame on that proposal. I'm more than happy to pass along some facts about the actual concept, less interested in jousting with a board of trolling armchair engineers...
;-)
What will it be? Do you have some actual questions about the project?
-
I've got a question.... Why did you cancel the campaign on the (atleast) semi-reputable Kickstarter and go over to the motherlode-of-all-scams Indiegogo?
-
First and foremost the campaign was more about marketing than it was about funding and it was very successful at putting the concept in the hands of the media. But as evidenced by the thread above and numerous others we've engaged since the KS campaign, this is a very technically complex concept that is likely not well suited for consumer sales. Incidentally, the German company Cloud & Heat ran into the same issue when they tried a crowdfunding campaign on Seedmatch.
Since then we have been working with focus groups, technical and non technical, to find a quicker path to comprehension but we're not quite there yet. The engagement we got on KS was actually pretty surprising but some of the interactions with the KS staff made us concerned enough about the platform to halt the campaign. It was the request of several backers that prompted us to reopen the campaign on IndieGoGo.
Beyond that, Kickstarter is primarily a platform for selling low cost, high risk of delivery trinkets and we are likely not selling a low cost consumer appliance anytime soon. We like the idea of crowdfunding campaigns but this technology is expensive enough, at least in the early stages, that it probably won't be a good fit for KS. That said, we're still in the R&D phase for likely another year or so as we develop/test the computing appliance, thermal storage and desiccant cooling systems so we haven't made a determination about the ideal launch market.
-
... it was very successful at putting the concept in the hands of the media.
Just like any other bullshit product. This statement speaks no of a campain's ability to deliver.
-
less interested in jousting with a board of trolling armchair engineers...
Ouch, that sure hurts!
But I understand, it is certainly easier and less work to fool nontechnical people.
-
"But I understand, it is certainly easier and less work to fool nontechnical people."
Still haven't heard a single technical question. Let me know if anyone here actually has an understanding of thermodynamics and interest in engaging, more than happy to educate 'technical' people.
-
"Just like any other bullshit product. This statement speaks no of a campain's ability to deliver."
We delivered everything we promised, Bud.
You obviously didn't read the campaign?
-
Let me know if anyone here actually has an understanding of thermodynamics
Most people here understand that you can heat a room with the waste heat from various bits of powered equipment. So, there is probably not much that you can contribute.
-
Still haven't heard a single technical question. Let me know if anyone here actually has an understanding of thermodynamics and interest in engaging, more than happy to educate 'technical' people.
Typical home computers use only 100-150W, so why would you bother? Do you really think people will buy 1kW computers and sell off cycles just to make this work?
How do you keep the HDDs and SSDs from cooking?
If desiccant refrigeration actually works, why don't people use it now, with cheaper heat sources?
-
Sure, Jim. You may have us confused with Nerdalize?
The first prototype runs much hotter than a typical computer, hot enough today to boil water off the GPUs. With a few more degrees of heat we can also supply desiccant driven air conditioning and absorptive refrigeration. We're also using a phase change thermal storage system to store enough thermal energy from the 1st unit in 15 gallons of liquid to supply the domestic hot water and space heat for an 1100sf house.
As I mentioned, we're partnered with the Fraunhofer Institute who are researching high temperature computing today. They are currently running ASIC at 300C - we can make process steam with 300C.
Next?
-
David,
"Typical home computers use only 100-150W, so why would you bother?" The 1st unit runs at 1600W, 150W really wouldn't do much.
"Do you really think people will buy 1kW computers and sell off cycles just to make this work?" They already do. Check out Cloud & Heat https://www.cloudandheat.com (https://www.cloudandheat.com), Qarnot Computing http://www.qarnot-computing.com/ (http://www.qarnot-computing.com/) and Nerdalize http://www.nerdalize.com/ (http://www.nerdalize.com/).
"How do you keep the HDDs and SSDs from cooking?" We run 2 seperate loops on the 1st prototype and have a stratified cooling layout that keeps the memory alive on V2.
"If desiccant refrigeration actually works, why don't people use it now, with cheaper heat sources?" They do http://www.nrel.gov/docs/fy11osti/49722.pdf (http://www.nrel.gov/docs/fy11osti/49722.pdf). If you could be paid to make the heat through distributed computation, why would you use a cheaper heat source?
The trick is to run the compute hard and hot.
-
The trick is to run the compute hard and hot.
Remarkably deep technical insight! Thanks!
Maybe if you offered some free keychains or posters left over from the KS, we would "warm up" to the idea.
PS: This may come as a surprise, but I actually heat my kitchen by operating a refrigerator in it!
-
"Remarkably deep technical insight! Thanks!"
The Department of Energy sure thinks it is, Jim.
Considering the entire industry is focused on running computers as cold as they can keep it is a pretty big departure from the norm.
-
Hi,
As you've discovered, the audience here is generally technically astute, but also deeply unforgiving when it comes to even a whiff of marketing over substance. On a bad day I'm as guilty of this as anyone.
Your idea to use the waste heat from a computer for something useful, in preference to simply generating heat with a resistor, is fine. I don't think anyone's questioning that principle. For an audience with a strictly electronics (but not thermodynamics) background, it might do you no harm to explain why running a heat source at a higher temperature can be a good thing when it comes to making use of that heat elsewhere.
Given that you obviously understand this already, you shot yourself in the foot when you mentioned Solar Roadways in your campaign. That project was the poster child for marketing over feasibility. Guilt by association, I'm afraid.
If you want to win over a technical audience, you need to explain, quantitatively, what your product actually does and how it might be useful. A simple explanation about its power consumption, MIPS per Watt, operating temperature, and the method for ultimately extracting the heat produced and making use of it, would go a long way. Above all, you just need to demonstrate that you have a strong understanding of the physics, and that you're not trying to do something for which the numbers simply don't add up.
A purposely inefficient computer would be a bad idea... no better than a laptop plus a bar heater. One which is just as good in terms of MIPS per Watt as any other server, but capable of running at a high temperature and providing a means to make good use of the heat that would otherwise be wasted, is potentially quite a good one.
I still think you should use a standard desktop CPU and make coffee with it, though!
-
Considering the entire industry is focused on running computers as cold as they can keep it is a pretty big departure from the norm.
Not really; try a search for "high ambient temperature servers".
http://www.geek.com/chips/googles-most-efficient-data-center-runs-at-95-degrees-1478473/ (http://www.geek.com/chips/googles-most-efficient-data-center-runs-at-95-degrees-1478473/)
-
Given that you obviously understand this already, you shot yourself in the foot when you mentioned Solar Roadways in your campaign. That project was the poster child for marketing over feasibility. Guilt by association, I'm afraid.
Andy is right. Solar Roadways are an industry joke. The motherload of all unfeasible projects.
Being associated with them is technical industry poison.
Add in the fact of moving from a failed Kickstarter to Indiegogo, you are begging to have your project not taken seriously I'm afraid.
I know that's not fair, but that's the way it is.
Thankfully engineers are easy to convince, just show them the data. Talk is cheap.
-
Thankfully engineers are easy to convince, just show them the data. Talk is cheap.
:-+
-
Is there any data on how this affects reliability, MTBF, etc. of the individual parts?
-
Is there any data on how this affects reliability, MTBF, etc. of the individual parts?
Running silicon based semiconductors hot enough to boil water is rarely beneficial for reliability :)
Surely a pretty high bandwidth internet connection is going to be required to be able to effectively sell compute cycles? Is the cost of this connection taken into account in the economics?
-
Most of this type of things are based in some fact, therefore the idea to use that fact redeemed to be wasteful and use it into something useful is valid. What most of this inventors fails to notice is the fact that the development of product is always trying to be as efficient as possible. 40 years ago in the "Computer room" you could bake a loaf of bread, now you need a whole bunch computers to make the room just "pretty warm". I'm certain that one of the main goals of the engineers that tinker with making new computer circuits is to minimize the production of heat.
My point is that this "free energy" won't last forever, the only ones that will benefit from the "investment" will be the "invertor" who will get "free" money.
-
"Typical home computers use only 100-150W, so why would you bother?" The 1st unit runs at 1600W, 150W really wouldn't do much.
How many CPUs is that? Or, are you relying on high-power GPUs?
"Do you really think people will buy 1kW computers and sell off cycles just to make this work?" They already do. Check out Cloud & Heat https://www.cloudandheat.com (https://www.cloudandheat.com), Qarnot Computing http://www.qarnot-computing.com/ (http://www.qarnot-computing.com/) and Nerdalize http://www.nerdalize.com/ (http://www.nerdalize.com/).
Those are in Europe - how can you make it work in the US, where we have hot summers, and slow Internet?
"How do you keep the HDDs and SSDs from cooking?" We run 2 seperate loops on the 1st prototype and have a stratified cooling layout that keeps the memory alive on V2.
Sounds like handwaving... what is the temperature of the low temperature loop?
"If desiccant refrigeration actually works, why don't people use it now, with cheaper heat sources?" They do http://www.nrel.gov/docs/fy11osti/49722.pdf (http://www.nrel.gov/docs/fy11osti/49722.pdf).
OK, so they write reports about it, but why don't they actually use it?
-
How many MIPS per Watt does your device achieve?
-
Hey, now things are getting interesting! Fortunately it is my anniversary so I'll be out of pocket for several days and will have to get back to the conversation early next week. Have a good weekend all, looking forward to the discussion.
-
Still haven't heard a single technical question.
You have not even answered my first inquiries. I am waiting for answers on these points.
Many of the new big computer centers use very efficient free cooling systems, where the
cooling is made by direct ventilations, with eventually some water evaporation help.
As an example, I run a 2000 core system with a 1.04 PUE which means that the cooling cost is only 4% of the
effective computer cost, compared to the 40-60 % this guy is mentioning. It is some
similar efficient cooling that is used by the facebook data centers.
They also seems to underestimate that they will need to evacuate the heat he will produce with
their high temperature computer.
He also forgot to say what does he do with all this heat in summer
when you want to cool the temperature of your house.
-
Is there any data on how this affects reliability, MTBF, etc. of the individual parts?
Running silicon based semiconductors hot enough to boil water is rarely beneficial for reliability :)
We can estimate it! (Arrhenius equation)
Those are in Europe - how can you make it work in the US, where we have hot summers...
It's from Germany, they also have decent summers
Edit:
For 60°C to 100°C I got an acceleration factor of 13.7? This would reduce 100 000hours mtbf (~11years) to 7325hours (305days)
-
Hi All, lets start here.
We are working on solutions to run computers significantly hotter and harder to specifically create heat with computation, we're hitting 200+ in the cooling loop today. That heat can be used for more than just running a boiler, we can power air conditioning and refrigeration as well with about 225-235 f. I know you are electronics guys so you will have to invest some google time looking for desiccant enhanced air conditioning and absorptive chillers/refrigeration - the building science guys get the implications pretty quickly. With the addition of refrigeration and air conditioning loads, thermal storage and proper sizing there is no room in our model for 'exhausting heat' - there is also a reason it's called project Exergy! If you want to talk about exhausting useful heat, track down Nerdalize or Cloud & Heat.
Since the biggest loads in our building currently run on heat (space and domestic hot water) or could be converted to run on heat (air conditioning and refrigeration) with well developed, existing technology, we could be getting two (or more) benefits out of the same energy that we are using to just run our heaters or computers today. (please don't tell me about the cost of natural gas, we're obviously targeting electric heating markets which are large and growing) One or the other could happen for free, from an energy perspective, if we used computers to make the heat that runs our homes/businesses/economy.
Interestingly, the majority of our US economy actually runs on heat, not electricity... but that is a far more confusing topic that I'd rather talk to economists about.
A gaming rig/laptop aren't going to make enough heat to have an impact so we aren't talking about converting current computers to a super computer/heater. We are building high temperature, modular 2-4kW computing appliances coupled with thermal storage that sit in the space a typical hot water heater sits today and performs local computation that datacenters are currently doing remotely near the arctic circle (JacquesBBB). Yes, there is an existing if small but growing market for this compute. The model is to keep the demand and compute local so we aren't shipping huge amounts of data to the arctic circle - which is pretty expensive, BTW. http://aceee.org/files/proceedings/2012/data/papers/0193-000409.pdf (http://aceee.org/files/proceedings/2012/data/papers/0193-000409.pdf) The data is also parsed across multiple machines taking care of security, redundancy and resiliency issues as well as cutting down dramatically on the need for high speed internet - at this point. Again, this already exists in some of the distributed computing projects https://en.wikipedia.org/wiki/List_of_distributed_computing_projects (https://en.wikipedia.org/wiki/List_of_distributed_computing_projects) as well as distributed rendering.
(JacquesBBB) - we don't want to build data centers, we want to eliminate them so I am not going to spend time jousting about datacenter performance. The equation is simple: datacenters buy energy for compute and cooling. We are using existing energy consumption (space/water heat - refrigeration/air conditioning), which is already paid for, to compute. Datacenters pay significant sums to transfer information to the arctic circle - we want to keep it local and distributed on short hops which are inherently faster.
-
Is there any data on how this affects reliability, MTBF, etc. of the individual parts?
There is plenty of manufacturer data on temperature and MTBF issues, each chip manufacturer publishes their heat/performance data. When we first started the prospects seemed pretty dismal given what they publish. Some R&D time figuring out how to run hot was the key, not the published data.
In general cool computers are happy computers and heat can dramatically reduce the lifespan of a computer if not managed correctly. It took us some time to figure out the major culprits in heat related failure, the biggest killer is electromigration. Heat and high current loads actually pick up and carry the conductor in chips, diminishing the conductor's capacity to transport electricity. There are several ways around this problem, the expensive solution is to move to a tungsten based conductor on an SoI chip. We're working with the Fraunhoffer Institute to attract some research funding to look into the macro economics of high temperature computing with SoI. They already have a series of SoI ASIC that run well and for typical lifespans at 300c - we can make high temperature steam with 300c.
The less expensive fix extends the life of the chip and is what we have been doing on the first prototypes; modulating the clock speeds and voltage based on temperature work like a charm. Lower voltage at higher temperatures reduces electromigration issues and allows things to live longer. More importantly, we're only designing these things with a 2 year lifespan in mind. The economics given the reduced electricity cost by distributing the compute to places with an existing heat load look pretty good. Two years is a little shorter than the average server replacement schedule but it is not far off. We're hoping to run them hard enough that, after their service life, the wheels are falling off and they'll have little resale value. If you stack up the ancillary benefits you can probably point to ROI on the systems that are well under two years. We'll rebuild the unit and cross ship it to the facility.
MTBF in high temp computing is a design issue that we, and others, are working on. If there is a market, it will be overcome.
-
Given that you obviously understand this already, you shot yourself in the foot when you mentioned Solar Roadways in your campaign. That project was the poster child for marketing over feasibility. Guilt by association, I'm afraid.
Andy is right. Solar Roadways are an industry joke. The motherload of all unfeasible projects.
Being associated with them is technical industry poison.
Add in the fact of moving from a failed Kickstarter to Indiegogo, you are begging to have your project not taken seriously I'm afraid.
I know that's not fair, but that's the way it is.
Thankfully engineers are easy to convince, just show them the data. Talk is cheap.
Only time will tell if Andy is right.
Solar Roadways is technically feasible and I'm more than happy to joust on that one as well. They are not, however, economically feasible. That doesn't mean they will be in the future or ever, that depends on factors that are related to but not dependant on economic feasibility. Faraday was also considered a loon in his day for many of the same reasons - how many electric motors power our life?
The planet?
Time will tell, not Andy. (nothing personal Andy!)
;)
-
Is there any data on how this affects reliability, MTBF, etc. of the individual parts?
Running silicon based semiconductors hot enough to boil water is rarely beneficial for reliability :)
Surely a pretty high bandwidth internet connection is going to be required to be able to effectively sell compute cycles? Is the cost of this connection taken into account in the economics?
Agreed - you have to go through some serious machinations to get semiconductors to live at high temps. It took two years to get to a stable 180-200f and MTBF is still a question! As far as bandwidth is concerned, it really depends on the proximity and the load itself. Meshnets and the proliferation of fiber/high speed wireless should solve for this. We're not building for today's network, we're building for tomorrow's network.
There are some pretty interesting development happening around high speed, peer to peer networks. http://www.researchgate.net/publication/228550567_Reducing_Bandwidth_Utilization_in_Peer-to-Peer_Networks (http://www.researchgate.net/publication/228550567_Reducing_Bandwidth_Utilization_in_Peer-to-Peer_Networks)
-
Most of this type of things are based in some fact, therefore the idea to use that fact redeemed to be wasteful and use it into something useful is valid. What most of this inventors fails to notice is the fact that the development of product is always trying to be as efficient as possible. 40 years ago in the "Computer room" you could bake a loaf of bread, now you need a whole bunch computers to make the room just "pretty warm". I'm certain that one of the main goals of the engineers that tinker with making new computer circuits is to minimize the production of heat.
My point is that this "free energy" won't last forever, the only ones that will benefit from the "investment" will be the "invertor" who will get "free" money.
A very good point, as chips get smaller and faster they generally produce less heat! But... they are smaller and faster so you can fit many more of them into the same space and typically generate the same heat and far more compute. Computers are pretty efficient heaters so the same 2-4kW of compute will likely still produce 2-4kW of heat, less the data shipped out of the computer back across the network as a result of the computation. This means the free energy will last until computers stop making heat, right?
As far as the 'Inventor who will get free money', that is rarely the case. The pioneers get the arrows and the settlers get the land. There is usually little monetary upside to pioneering.
-
"Typical home computers use only 100-150W, so why would you bother?" The 1st unit runs at 1600W, 150W really wouldn't do much.
How many CPUs is that? Or, are you relying on high-power GPUs?
"Do you really think people will buy 1kW computers and sell off cycles just to make this work?" They already do. Check out Cloud & Heat https://www.cloudandheat.com (https://www.cloudandheat.com), Qarnot Computing http://www.qarnot-computing.com/ (http://www.qarnot-computing.com/) and Nerdalize http://www.nerdalize.com/ (http://www.nerdalize.com/).
Those are in Europe - how can you make it work in the US, where we have hot summers, and slow Internet?
"How do you keep the HDDs and SSDs from cooking?" We run 2 seperate loops on the 1st prototype and have a stratified cooling layout that keeps the memory alive on V2.
Sounds like handwaving... what is the temperature of the low temperature loop?
"If desiccant refrigeration actually works, why don't people use it now, with cheaper heat sources?" They do http://www.nrel.gov/docs/fy11osti/49722.pdf (http://www.nrel.gov/docs/fy11osti/49722.pdf).
OK, so they write reports about it, but why don't they actually use it?
1) We're using GPUs today
2) look up the thread - there are solutions for summer and we would have no economic reason to target regions with slow internet. The best distributed compute loads would be located in areas with high speed internet.
3) It's currently a separate loop, we can run it at whatever temperature we want. The real problem will be how to keep the next prototype's memory cool when it is all submerged in the same case and coolant - that will take some interesting engineering. We're already exploring ways of stratifying the coolant and have some good ideas, we'll see how they work at high temps over the next year.
4) There are a handful of companies licensed and building the tech now. It's pretty irrelevant, chillers run on the same temps. DEVAP is just far, far more efficient and an interesting application for DoE funding.
-
As die temperature increases, its common for CPUs to become increasingly less efficient, requiring more power to do the same operations. Do the benefits of this system outweigh the extra power consumption?
(http://i272.photobucket.com/albums/jj163/idontcare_photo_bucket/Intel%20Core%20i7-2600K/TempvsPowerfor3GHzat1491V.png)
The increase in this example, for instance, is larger than the power consumed by a more powerful cooler so one could actually be wasting more energy by using less cooling power. Which isn't directly related to this particular use, but you're still drawing more power than you would otherwise, which is not "free".
-
Hi,
As you've discovered, the audience here is generally technically astute, but also deeply unforgiving when it comes to even a whiff of marketing over substance. On a bad day I'm as guilty of this as anyone.
Your idea to use the waste heat from a computer for something useful, in preference to simply generating heat with a resistor, is fine. I don't think anyone's questioning that principle. For an audience with a strictly electronics (but not thermodynamics) background, it might do you no harm to explain why running a heat source at a higher temperature can be a good thing when it comes to making use of that heat elsewhere.
Given that you obviously understand this already, you shot yourself in the foot when you mentioned Solar Roadways in your campaign. That project was the poster child for marketing over feasibility. Guilt by association, I'm afraid.
If you want to win over a technical audience, you need to explain, quantitatively, what your product actually does and how it might be useful. A simple explanation about its power consumption, MIPS per Watt, operating temperature, and the method for ultimately extracting the heat produced and making use of it, would go a long way. Above all, you just need to demonstrate that you have a strong understanding of the physics, and that you're not trying to do something for which the numbers simply don't add up.
A purposely inefficient computer would be a bad idea... no better than a laptop plus a bar heater. One which is just as good in terms of MIPS per Watt as any other server, but capable of running at a high temperature and providing a means to make good use of the heat that would otherwise be wasted, is potentially quite a good one.
I still think you should use a standard desktop CPU and make coffee with it, though!
The idea isn't actually to reuse waste heat from a computer. The idea it to generate primary heat from computation, there is a pretty significant difference. Waste heat from computers is typically too low in temperature to be able to do anything too terribly useful. Increasing the maximum operating temperature of computing is the interesting proposition. Even if the compute used to make heat were slow and hot it would still be delivering a useful benefit beyond just heat.
At a higher level, far more of the energy we pull out of the ground to run our economy today is used to make heat than it is to make electricity. With climate change concerns and the rapid shift to distributed, renewable generation sources - wind and solar specifically - there will be a shift in the way we make heat. Combining computation and high temperature heat generation produces at least two primary benefits from the same energy source and far too many ancillary benefits to list.
MIPs per usable BTUh is likely a better measure of the effect we are looking for and the focus of the next stage of development. MIPs per Watt is a pretty irrelevant design feature at this stage, we have a few more important hurdles than optimizing MIPs per Watt. I'm sure you can understand that I'm not terribly interested in telling a forum too many specific details about how we are going to generate, capture and store the heat.
Hopefully you have enough information at this point to understand that Project Exergy is not 'another Bullshit "save the world" project' at this point and that the approach is viable and well within the realm of technical feasibility? If not I'm certainly willing to answer some more questions if you think this is more 'marketing vs substance'.
-
As die temperature increases, its common for CPUs to become increasingly less efficient, requiring more power to do the same operations. Do the benefits of this system outweigh the extra power consumption?
(http://i272.photobucket.com/albums/jj163/idontcare_photo_bucket/Intel%20Core%20i7-2600K/TempvsPowerfor3GHzat1491V.png)
The increase in this example, for instance, is larger than the power consumed by a more powerful cooler so one could actually be wasting more energy by using less cooling power. Which isn't directly related to this particular use, but you're still drawing more power than you would otherwise, which is not "free".
The economics are the opportunity. With today's chips, the system may not produce the optimal MIPs/usable BTUh. If there is enough of economic incentive, considering 2% of US electricity used for Data Centers vs ~20 used for heat alone in buildings, we're pretty sure there will be significant interest in optimizing chips to run hot. Fraunhofer estimates the cost to optimize chips to run hot and fast are by no means insurmountable.
-
It seems a bit unrealistic to expect industry to change that much for the sake of this. From what I've understood from your comments it can work very well, provided that the industry moves to a different type of process (tungsten for interconnects) for semiconductors that will be used exclusively* for a new heat-to-MIPS computer class. This would also require a lot of other processes and techs that are meant exclusively for that class of computers - high temperature PCBs, high temperature misc support components...
I'm quite sure it's possible, but basically this would only get you a relatively small increase** in computation power during winter in certain areas of the world.
* The reasoning behind the exclusively part is that pretty much everyone in the market wants more power efficient devices rather than purposefully wasteful devices.
** Until the next generation of computational devices set the bar higher by conventional, energy efficient means.
-
It seems a bit unrealistic to expect industry to change that much for the sake of this. From what I've understood from your comments it can work very well, provided that the industry moves to a different type of process (tungsten for interconnects) for semiconductors that will be used exclusively* for a new heat-to-MIPS computer class. This would also require a lot of other processes and techs that are meant exclusively for that class of computers - high temperature PCBs, high temperature misc support components...
I'm quite sure it's possible, but basically this would only get you a relatively small increase** in computation power during winter in certain areas of the world.
* The reasoning behind the exclusively part is that pretty much everyone in the market wants more power efficient devices rather than purposefully wasteful devices.
** Until the next generation of computational devices set the bar higher by conventional, energy efficient means.
Thanks for the input - we'll stick with the recommendations from the DoE/ARPA-e and continue pursuing the project. It was a bit unrealistic for Henry Ford to make a car that everyone could afford... he still did.
-
I hope you don't imagine that appeals to the past inspires any sort of confidence, since this is generally the first resort of the nutcase. "THEY LAUGHED AT EINSTEIN!!121231"
-
Thanks Nerull,
We're really not looking for approval, just hoping to pass along some real information about the project.
Any more technical questions, Guys?
-
An issue is that directly converting electricity to heat is not the best way to use electricity to heat a house.
You can be 100% efficient with a simple heater (or CPU), but with a heat pump, you can actually be several times more efficient. The idea is you use the electricity to pump heat from the outside in, so you heat up your house not only by the waste heat of the pump, but by the actual heat that was pumped into the house. getting maybe 200% efficiency.
It's like installing your air conditioner backwards. Heat from electricity + heat plus pumping is more than you can get just by running a CPU or a space heater etc.
-
Thanks Nerull,
We're really not looking for approval, just hoping to pass along some real information about the project.
And we all know that smugness is the best way to save the world :)
It will be nice to get that real information though.
-
So, instead of putting these in houses, why not target them towards data centers? Use the heat to create high temperature steam, drive a turbine generator and use that to power the systems. Obviously you can't self-power the system this way, unless you've also broken some thermodynamic laws, but it would go a long way towards reducing the power consumption of the data center. Plus you wouldn't need all the cooling capacity.
Sent from my Tablet
-
A domestic hot water cylinder sits at around 70 deg C.
Why all the work on high temperature semiconductors, when you could build a drop-in replacement for an electric immersion heater out of a CPU with a small pump and a heat exchanger?
What's the compelling application for the higher temperature?
-
An issue is that directly converting electricity to heat is not the best way to use electricity to heat a house.
You can be 100% efficient with a simple heater (or CPU), but with a heat pump, you can actually be several times more efficient. The idea is you use the electricity to pump heat from the outside in, so you heat up your house not only by the waste heat of the pump, but by the actual heat that was pumped into the house. getting maybe 200% efficiency.
It's like installing your air conditioner backwards. Heat from electricity + heat plus pumping is more than you can get just by running a CPU or a space heater etc.
Salient point. If we are just looking at the most efficient way to heat a house then solar or geo thermal with storage is the hands down winner and heat pumps (space heat and DHW) are a the most efficient uses of electricity to heat. But looking at it from a macro level, 2-3% of US (slightly less global) electricity is consumed in datacenters, 30-60% of that is used to eliminate waste heat by pumping it top the atmosphere. The growth of data collection and number of data centers is growing rapidly and the growth has no end end in sight so, by some accounts, this is a problem that is compounding. http://www.sciencedaily.com/releases/2013/05/130522085217.htm (http://www.sciencedaily.com/releases/2013/05/130522085217.htm)
Other side of the coin, 30% of the total energy we consume is used to heat buildings. Combining the two eliminates the wasted energy pumping heat to the atmosphere and displaces energy we were using to do just one thing - compute or heat. So the macro economic potential is very large.
-
Thanks Nerull,
We're really not looking for approval, just hoping to pass along some real information about the project.
And we all know that smugness is the best way to save the world :)
It will be nice to get that real information though.
I count 3 pages of smugness before I entered the conversation. If you have a need for information then feel free to ask a question?
-
So, instead of putting these in houses, why not target them towards data centers? Use the heat to create high temperature steam, drive a turbine generator and use that to power the systems. Obviously you can't self-power the system this way, unless you've also broken some thermodynamic laws, but it would go a long way towards reducing the power consumption of the data center. Plus you wouldn't need all the cooling capacity.
Sent from my Tablet
The designs are modular so they could be used in a datacenter but, you are right, converting heat back to electricity would be a loss vs using the heat for heat. We have looked at Organic Rankine Cycle generation but the numbers favor creating heat where we need it.
-
A domestic hot water cylinder sits at around 70 deg C.
Why all the work on high temperature semiconductors, when you could build a drop-in replacement for an electric immersion heater out of a CPU with a small pump and a heat exchanger?
What's the compelling application for the higher temperature?
The need for heat is pretty seasonal so increasing temperature and running absorptive chillers/refrigeration or regenerating desiccant for evaporative space cooling will extend the system's benefits through the summer months and increase the benefits. It may not make the economic break even, that's why we are doing the research. It is an absolute slam dunk for DHW, much cheaper system.
And I think it would be pretty cool to cook my thanksgiving turkey with compute one day... 8)
-
But looking at it from a macro level, 2-3% of US (slightly less global) electricity is consumed in datacenters, 30-60% of that is used to eliminate waste heat by pumping it top the atmosphere. The growth of data collection and number of data centers is growing rapidly and the growth has no end end in sight so, by some accounts, this is a problem that is compounding.
You keep saying that data centers use 30-60% of power to eliminate wasted heat, unless they go to the Artic where data transmission become a problem.
But although these figures can be reached by old generation data centers, there are some alternative ways of reducing
power overhead that is used in modern data centers.
For example, the large Facebook data center of Prineville has a PUE of only 1.07 (only 7% overhead)
https://www.facebook.com/notes/facebook-engineering/designing-a-very-efficient-data-center/10150148003778920 (https://www.facebook.com/notes/facebook-engineering/designing-a-very-efficient-data-center/10150148003778920)
In the computer center that I run, with more than 2000 Xeon cores, we have a PUE of less than 1.04 (and this is at latitude 45 deg, not in the Artic)
right where we work. The cooling is made, as in the Facebook data center with direct air cooling and water evaporation for
the warmest days.
Fraunhofer estimates the cost to optimize chips to run hot and fast are by no means insurmountable.
You keep mentioning your Fraunhofer contract as a validation for your project, but it is very unclear what exactly is your Fraunhofer contract.
It is well established that there is a current research for increasing the running temperature of the computer chips. This is a way to reduce the cost
of cooling the computer centers by letting them run at higher temperature. This has nothing to do with your present project, which is running on purpose the
cpu to high temperatures to heat habitations.
-
What about the problem of heat requirements fluctuating over short periods of time? For example, my family's greenhouses need almost no additional heat during even the coldest winter days due to incoming solar radiation, but may need massive heat (as much as 1MW) on a cold winter night Could this match any kind of computation demand?
Even my own well-insulated house built in 1998 has a heat loss in excess of 15KW at the lowest outdoor temperatures. Do you plan to stage computation up and down to match heat requirements?
-
But looking at it from a macro level, 2-3% of US (slightly less global) electricity is consumed in datacenters, 30-60% of that is used to eliminate waste heat by pumping it top the atmosphere. The growth of data collection and number of data centers is growing rapidly and the growth has no end end in sight so, by some accounts, this is a problem that is compounding.
You keep saying that data centers use 30-60% of power to eliminate wasted heat, unless they go to the Artic where data transmission become a problem.
But although these figures can be reached by old generation data centers, there are some alternative ways of reducing
power overhead that is used in modern data centers.
For example, the large Facebook data center of Prineville has a PUE of only 1.07 (only 7% overhead)
https://www.facebook.com/notes/facebook-engineering/designing-a-very-efficient-data-center/10150148003778920 (https://www.facebook.com/notes/facebook-engineering/designing-a-very-efficient-data-center/10150148003778920)
In the computer center that I run, with more than 2000 Xeon cores, we have a PUE of less than 1.04 (and this is at latitude 45 deg, not in the Artic)
right where we work. The cooling is made, as in the Facebook data center with direct air cooling and water evaporation for
the warmest days.
Fraunhofer estimates the cost to optimize chips to run hot and fast are by no means insurmountable.
You keep mentioning your Fraunhofer contract as a validation for your project, but it is very unclear what exactly is your Fraunhofer contract.
It is well established that there is a current research for increasing the running temperature of the computer chips. This is a way to reduce the cost
of cooling the computer centers by letting them run at higher temperature. This has nothing to do with your present project, which is running on purpose the
cpu to high temperatures to heat habitations.
Thanks for the thoughts Jacques and great news on you 1.04 PUE! That is really impressive and certainly shows some datacenter efficiency chops. The problem is a 1.04 PUE is a unicorn in the overall data center market. The vast majority of datacenters are nowhere near that operating efficiency as evidenced by this really well researched and cited report by the NRDC - take a look at page 31 and you'll see that vast majority of energy consumption in Datacenters is in the 1.8 to 2.0 PUE. Perhaps you can explain what a 2.0 PUE actually represents in operating efficiency to the forum? That might be the quickest way to illustrate the 30-60% number, PUE is a pretty hard metric to understand for the uninitiated. NRDC: https://www.nrdc.org/energy/files/data-center-efficiency-assessment-IP.pdf (https://www.nrdc.org/energy/files/data-center-efficiency-assessment-IP.pdf)
Thank you for bringing up Prineville - that is (was) a truly cutting edge data center. My former company, PECI, was part of the prineville project and my good friend and former colleague authored and presented this paper on the project - I'd have to ask but I'm pretty sure he presented with the Facebook engineers: http://aceee.org/files/proceedings/2012/data/papers/0193-000421.pdf-000421.pdf (http://aceee.org/files/proceedings/2012/data/papers/0193-000421.pdf-000421.pdf)[/url] At the time Facebook's leased data centers had an average PUE of 1.5 and best practice and state of the art were running 1.3 PUE and 1.2 PUE, respectively. So it is really impressive that you are hitting a 1.04.
As far as Fraunhofer is concerned, we are teamed on a couple of R&D proposals with them, NREL and Notre Dame - we are not under any contract, kind of a weird assumption. No, you can't see the proposals and again... we're really not looking for your approval. :D
Can you point me to some information on your Datacenter performance or methodology? I'd be really interested to see how you are managing a 1.04 PUE!
-
What about the problem of heat requirements fluctuating over short periods of time? For example, my family's greenhouses need almost no additional heat during even the coldest winter days due to incoming solar radiation, but may need massive heat (as much as 1MW) on a cold winter night Could this match any kind of computation demand?
Even my own well-insulated house built in 1998 has a heat loss in excess of 15KW at the lowest outdoor temperatures. Do you plan to stage computation up and down to match heat requirements?
Great question. We're using a phase change thermal storage system that acts like a thermal battery to help ride through these typical fluctuations. This way you can run the computation at a relatively flat performance profile and use the stored thermal energy to meet fluctuating demands. Coupled with refrigeration/AC and DHW loads we're hoping to see a pretty even seasonal compute load.
-
Great question. We're using a phase change thermal storage system that acts like a thermal battery to help ride through these typical fluctuations. This way you can run the computation at a relatively flat performance profile and use the stored thermal energy to meet fluctuating demands. Coupled with refrigeration/AC and DHW loads we're hoping to see a pretty even seasonal compute load.
Really. How many megajoules per cubic meter do you think you can store? At what cost?
-
Can you point me to some information on your Datacenter performance or methodology? I'd be really interested to see how you are managing a 1.04 PUE!
The technology is SGI ICE cube air containers.
http://www.ansys.stuba.sk/2011/pdf/SGI/ICE%20CUBE%20AIR_4274.pdf (http://www.ansys.stuba.sk/2011/pdf/SGI/ICE%20CUBE%20AIR_4274.pdf)
Our PUE is even better than 1.04 over the full year. I did not check the latest data.
-
Great question. We're using a phase change thermal storage system that acts like a thermal battery to help ride through these typical fluctuations. This way you can run the computation at a relatively flat performance profile and use the stored thermal energy to meet fluctuating demands. Coupled with refrigeration/AC and DHW loads we're hoping to see a pretty even seasonal compute load.
Really. How many megajoules per cubic meter do you think you can store? At what cost?
The fatty acid/fatty ester material we are using currently stores ~62.5 kBtu/gallon - we're under NDA with the manufacturer so we can't say much more about the material.
-
Can you point me to some information on your Datacenter performance or methodology? I'd be really interested to see how you are managing a 1.04 PUE!
The technology is SGI ICE cube air containers.
http://www.ansys.stuba.sk/2011/pdf/SGI/ICE%20CUBE%20AIR_4274.pdf (http://www.ansys.stuba.sk/2011/pdf/SGI/ICE%20CUBE%20AIR_4274.pdf)
Our PUE is even better than 1.04 over the full year. I did not check the latest data.
Yep, evaporatively cooled - no magic there unless you are doing something special with the DX system?
The reason I've mentioned data centers sited near the arctic circle (not hyperbole!) is, if you are running an ultra efficient datacenter (1.04 counts) that is not using 100% free cooling then you are likely are using potable water (your tech looks way too cookie cutter to be reusing wastewater?) to run the system. This practice is still contributing to the large and growing problem associated with water cooled chillers and evaporative cooling.
By using the rapidly shrinking reserves of potable water on the planet and relying on the undervaluation of the water/energy nexus. Unfortunately your 1.04 PUE is really relying on economic efficiencies that are directly related to devaluing water and ignoring its energy content based on its requirement for life and health as a subsidized and regulated utility. Pumping, treating, storing and transporting water have very high energy cost that are not yet being passed down to the end user and, therefore, not well accounted for in PUE or the broader economy.
If you want to get closer to numbers that represent the full impacts of computing I would suggest trying Energy Reuse Effectiveness (ERE) instead of PUE. See page 8 of this Green Grid paper: http://www.thegreengrid.org/~/media/WhitePapers/ERE_WP_101510_v2.pdf (http://www.thegreengrid.org/~/media/WhitePapers/ERE_WP_101510_v2.pdf) an page 11: "The concept and application of PUE has been beneficial to the industry but also challenging on occasion, due to misuse. ERE is a more precisely defined metric than PUE was in its first white paper." ERE helps define the value of used and recovered energy on and off the datacenter, NREL and Notre Dame have both been involved in the development of ERE.
CA datacenters are already under fire for water consumption, here is a (somewhat) positive article on the topic, there are many negative articles begining to circulate: http://www.bloombergview.com/articles/2015-06-30/let-data-centers-have-their-water (http://www.bloombergview.com/articles/2015-06-30/let-data-centers-have-their-water) I can't imagine how datacenters are going to come under scrutiny in parts of Asia/India where potable water is already a major issue.
By comparison, the system we envision does not use water for cooling - it uses existing heating loads for cooling - so there is no associated water use in the design. There is also less water use in the macro because we can eliminate the single use of electricity (and the power plant cooling costs associated) for compute or heat. There is actually a reason it is called Project Exergy.
-
Really. How many megajoules per cubic meter do you think you can store? At what cost?
The fatty acid/fatty ester material we are using currently stores ~62.5 kBtu/gallon - we're under NDA with the manufacturer so we can't say much more about the material.
One presumes that the phase change temperature of this material is one of the reasons that you need your computation devices to run hot.
-
Taken from actual prototypes that made the ultimate sacrificed in the quest for change! As we melt them, you get them based on the time you pledge. Mounted, signed and dated by the Team, these parts should make one heck of a conversation piece!
So a melted computer part... for $1000. :o (It'll make a great conversation with my pet rock lol :'( )
When I could buy two entire life works of salvaged parts for that price (and some of them work, too!)...
(referring to https://www.youtube.com/watch?v=x8nbHYOc8ns (https://www.youtube.com/watch?v=x8nbHYOc8ns) EEVBlog #737 - World's Biggest Collection Of Electronics Components)
-
Really. How many megajoules per cubic meter do you think you can store? At what cost?
The fatty acid/fatty ester material we are using currently stores ~62.5 kBtu/gallon - we're under NDA with the manufacturer so we can't say much more about the material.
One presumes that the phase change temperature of this material is one of the reasons that you need your computation devices to run hot.
dfmischler,
The temperature we're looking to achieve is related to the heat loads we're targeting, not the phase change material. The PCM manufacturer can tailor the temperature to meet a fairly broad set of temperature requirements.
-
Taken from actual prototypes that made the ultimate sacrificed in the quest for change! As we melt them, you get them based on the time you pledge. Mounted, signed and dated by the Team, these parts should make one heck of a conversation piece!
So a melted computer part... for $1000. :o (It'll make a great conversation with my pet rock lol :'( )
When I could buy two entire life works of salvaged parts for that price (and some of them work, too!)...
(referring to https://www.youtube.com/watch?v=x8nbHYOc8ns (https://www.youtube.com/watch?v=x8nbHYOc8ns) EEVBlog #737 - World's Biggest Collection Of Electronics Components)
I wonder how many rocks could you have bought for the $4 you paid for your pet rock? Gary Dahl sold 1.5 million of them at roughly $4 each (in 70s dollars!) by the time the craze ended, BTW... amazing.
Selling melted computer parts is not the point, but I'm sure you knew that.
Did you have a technical question?
-
While the physics of utilizing the heat from processing is sound, the practical marketability and implementation is not.
When you consider what a data center actually does for it's customers, only a small percentage of it is directly relates to CPU cycles.
Storage and network bandwidth/latency are the major components and at present this product does not appear to provide anything positive in those areas with it's suggestion of moving the processing into peoples homes.
You cannot move the storage and you cannot run the necessary level of high bandwidth/low latency network infrastructure into everyone's home hot water system (or wherever you want to utilise the heat).
This is just another scam.
-
Did you have a technical question?
How many MIPS per kW do your machines produce?
-
While the physics of utilizing the heat from processing is sound, the practical marketability and implementation is not.
When you consider what a data center actually does for it's customers, only a small percentage of it is directly relates to CPU cycles.
Storage and network bandwidth/latency are the major components and at present this product does not appear to provide anything positive in those areas with it's suggestion of moving the processing into peoples homes.
You cannot move the storage and you cannot run the necessary level of high bandwidth/low latency network infrastructure into everyone's home hot water system (or wherever you want to utilise the heat).
This is just another scam.
Thanks for your undoubtedly learned opinion. Of course you can move the storage, Maidsafe is already beginning to transform that market. Read the thread - all of this is covered.
:blah:
I've wanted a reason to use this icon, this seems like the perfect opportunity.
-
Well guys, it has been a couple week of fun but I am going to bow out of the discussion at this point since the technical questions seem to have reached a finish and cycled back to repeating the same. I'd encourage the 'technical' people to reach out through the website if you have some real questions that aren't answered in the thread - please don't bother with miles per gallon and MIPS/kW, they are irrelevant design features at this stage of development.
Feel free to join me in London at the DEVCON Conference the week of the 9th and the DCD Converged conference the 18th - 20th, I'll be on a panel discussion of this topic as well as presenting the project - I'm looking for a few good hecklers.
Andy, you seem to have a pretty good grasp on things - would be great to bump into you when I'm in the UK.
Thanks
-
please don't bother with miles per gallon and MIPS/kW, they are irrelevant design features at this stage of development.
Absolutely not, that's the key to this whole thing. If this machine takes 2kW to do the same amount of work as an ordinary 100W desktop then the entire project is a waste of time. The users would be better served buying an ordinary desktop for their computation needs and a standard 2kW resistive heater for their heating needs.
If your machine actually does 2kW of work (similar MIPS/FLOPS per watt to a modern efficient processor) then the project might actually have some merit.
-
While the physics of utilizing the heat from processing is sound, the practical marketability and implementation is not.
When you consider what a data center actually does for it's customers, only a small percentage of it is directly relates to CPU cycles.
Storage and network bandwidth/latency are the major components and at present this product does not appear to provide anything positive in those areas with it's suggestion of moving the processing into peoples homes.
You cannot move the storage and you cannot run the necessary level of high bandwidth/low latency network infrastructure into everyone's home hot water system (or wherever you want to utilise the heat).
This is just another scam.
Thanks for your undoubtedly learned opinion. Of course you can move the storage, Maidsafe is already beginning to transform that market. Read the thread - all of this is covered.
:blah:
I've wanted a reason to use this icon, this seems like the perfect opportunity.
Do you even know what maidsafe is? It's not not relevant to the bulk of corporate data centre usage. No sane company would trust their corporate date to that. It shows that you really don't comprehend the market's. No surprise really.
Don't let the door hit you on the arse on the way out.
-
If this company hires people to do product engineering as well as they hire people to do PR outreach, they're doomed. Why on earth would anyone give money to someone who's that much of a dick.
-
Thanks Nerull,
We're really not looking for approval, just hoping to pass along some real information about the project.
Any more technical questions, Guys?
You're not looking for approval, you're just begging for money. If you think that insulting everyone you come across is a good way to do it, you're in for a shock.
-
Hey there gents. Just a last quick follow up after the DCD London conference to let you know that it went stunningly well. The attendees were very engaged and had an immediate recognition of the merit and potential of combined heat and computation. We'll be talking with IBMs OpenPOWER foundation in the next couple of weeks about the project and have already been invited to present at next year's event.
I do want to thank you for the conversation, it did solidify a number of thoughts I'd had prior to engaging at the conference. Jacques, you'll be happy to know that a research paper has spawned after conversation about developing a more holistic and realistic accounting for water/energy as it relates to datacenter energy consumption. There was a number of fairly pointed conversations about shifting energy off the books to create the impression of a lower PUE during the conference. I don't think anyone thought it was nefarious in intent but it seems the value of a more realistic PUE number - and the dollars that represents in fair marketing - is at least a serious conversation.
Nerull, I'm going to leave you to your echo chamber and wish you good luck in all endeavors. Just keep in mind we're still not looking for your approval. :-+
-
How.
Many.
MIPS.
Per
kW?
:-//
-
How.
Many.
MIPS.
Per
kW?
:-//
please don't bother with miles per gallon and MIPS/kW, they are irrelevant design features at this stage of development.
In other words, he doesn't give a shit if this contraption is even a reasonably viable alternative to a typical resistive heater, he's just after the paycheck.
-
How.
Many.
MIPS.
Per
kW?
:-//
Okay, persistence wins, I'll give it one last shot.
MIPS/kW are, by and large, an irrelevant measure of compute performance that are generally disregarded for several reasons.
- Firstly they vary dramatically by workload vs system architecture, this is obvious when you consider high performance ASIC are optimized for specific workloads and virtually useless for anything but that workload. Given this MIPS are not, at all, an effective measure of the usability of a generic machine that will see different workloads that may or may not be optimized for its architecture.
- The system we are designing can run many different processors so MIPS/kW is - again- a design feature and one we are not at all focused on at this time. How many MIPS do you want to make? Tell me the processor you want to use and the workload you will be using and the system can be optimized for those parameters... it is a design feature.
- the system varies voltage and clock speeds depending on temperature making the MIPS/kW performance a curve based on those variables - this is why we are focused on optimizing the thermal systems and not irrelevant design features - the more efficient the thermal systems the more variability we'll have for optimizing compute performance
Thanks-