Oil-Immersion Cooled PC Goes To Retail 210
notthatwillsmith writes "Everyone's seen mods where someone super-cools a PC by submersing it in a non-conductive oil. It's a neat idea, but most components aren't designed to withstand a hot oil bath; after prolonged exposure materials break down and components begin to fail. Maximum PC has an exclusive hands-on, first look at the new Hardcore Computer Reactor, the first oil-cooled PC available for sale. Hardcore engineered the Reactor to withstand the oil, using space-age materials and proprietary oil. The Reactor's custom-manufactured motherboard, videocards, memory, and SSD drives are submersed in the oil, while the dry components sit outside the bulletproof tank. The motherboard lifts out of the oil bath on rails, giving you relatively easy access to components, and the overall design is simply jaw-dropping. Of course, we'd expect nothing less for a machine with a base price of $4000 that goes all the way up to $11k for a fully maxed out config."
Bulletproof? (Score:4, Interesting)
Bulletproof? Seriously?
Re:so much for quick repair (Score:5, Interesting)
Reminds my of Kryotech. (Score:4, Interesting)
So my first worry is upgrade path. Which my scanning of the article's many usages of the words "custom" and "proprietary" leads me to believe there really isn't one.
Kryo's problem was that while you could buy a bad-ass refrigerated system for a mere 2x the cost of a top-end system that got a good 30% more performance -- they broke the 1 GHz barrier when air-cooled athlons were still running at around 600-700 MHz -- but then six months to a year later that system was merely "top of the line", and then of course soon after that "sub-optimal". Air-cooled athlons hit 1 GHz, and of course Kryotech came out with even faster systems, but it was obvious that the advantage you were getting was temporally speaking not worth the price.
Now with a slide-out motherboard and all it seems that upgrading this thing is at least -possible-, so perhaps if the company stays in business, you could at least purchase a compatible upgrade from them. Assuming there isn't a huge premium for the upgrade parts, that could be reasonable. The main thing is to have the re-usable oil cooling system. If they could make it so it can use off-the-shelf parts, and just sold the case itself, then that would be the ultimate to me.
And..... why? (Score:2, Interesting)
I read TFA and was waiting for all the reasons this was a good idea. Like why all Buicks have three holes in each side. Or why glad bags are so much better with yellow and blue zippers. I expected to read that the machine was like, totally silent. Or that things ran, well, slicker. Maybe I missed a whole page of pluses? There gotta be a whole lot of pluses for a $4K box that you can't change the motherboard.
Versus water cooling (Score:3, Interesting)
I'm just wondering what the advantage of this thing is versus non-submersed liquid cooling.
1) Noise? They said it's quiet "for so much hardware." Yeah well considering what a normal three GPU system sounds like that's not saying a whole lot. A good pump-and-block cooled setup can run around 25 dBa which is something like a whisper at 6 feet. In both cases you still need a couple of fans running, so I imagine non-cooling factors will dominate noise.
2) Performance? The article says cooling probably won't exceed the best liquid cooled setups that focus on the CPU/GPU.
It certainly is a cool idea, but I think I'd rather pay for a normal liquid cooled setup.
Re:Reminds my of Kryotech. (Score:3, Interesting)
I've been building / hacking together machines since shortly after the words IBM and AT were used together in one breath. What I've seen, over generations of building machines with the intent of upgrading them was that if you build it using a good upgradable motherboard and power supply, you might get lucky and the machine will be good for a single generation of upgrade after 12-18 months.
- Meaning maybe roughly 12 months later it's cost effective to upgrade :
- - to a same generation CPU that is quite a bit faster (think Pentium 33MHz to Pentium 133MHz, or PII-300 to PIII-900MHz)
- - the amount of memory can generally be doubled (although given the current cost of RAM and the limit of 32 bit OS, it's almost cost effective to max it out at purchase - but with 64 bit OSs I think this will still hold)
- - a second hard drive that is twice as large as the one you bought a year earlier (effectively tripling the space)
- - maybe adding a video card that is twice as fast as the first one you had in it.
About a year after you build your ultra-upgradable machine, the architecture changes (chipsets, video card platform, memory sockets, memory type, CPU sockets, hard drive interfaces) so the upgrade options taper off fast. Six months after that you can still upgrade to faster hardware at a premium price, but in very short order it's cheaper to replace the machine w/ current generation architecture than to upgrade 2 or more sub-systems. About a year after that, the machine is relegated to door-stop duty.
The problem is - the year over year increases are so steep that this happens no matter where on the curve you are when you buy. Think back - the premium for the DX2-66 was about $400 over the DX2-50, the premium for 16M of 72-pin SIMM memory was about $800 over 4M of 72-pin SIMM memory - but two years later it really didn't matter because with the P2-300MHz machines coming with 64M of PC100 memory - your two year old box wasn't fast enough regardless of whether you payed the $1,200 premium to get those two bumps or not. You could always pay another $1,200 premium for that next machine and get the P2-500MHz w/ 192M of memory, but two years later when the P4 based machines running 1.8GHz and 512M of memory - it really didn't matter whether you splurged for the extra umph or not - the box needed to be upgraded.
Is $11k out of line for a machine that's going to be on-par with the next generation of hardware, and obsolete in three years regardless? Probably. Unless you have a business reason for it, I'd say yes. I said the same thing about LCDs when IBM was selling their 16" LCD for $1,600, and SSDs were $100 per Gig. Today - both are reasonably priced, and maybe these guys will pioneer the path towards the next revolution in hardware platforms. Have to admit this much - of the $11k, how much is actual material cost and how much is 'OMFG 1337 haxor' premium? That's the margin that these machines will come down in about three years, making the platform affordable (or not.)
In my opinion there are two places that this machine makes sense - high end CAD where a company is paying $30k per seat to license the software and $125k per year for the guy behind the keyboard (25% faster machine = 1/4 fewer seats) or working on hard-duration projects (finish designing the next Space Shuttle by June 1st and get a $5M bonus.) Other than that - and the obvious rich gamer - I'd say a given day's task set would be better served by a couple three or four desktops all coming through to a single multi-LCD machine that rdesktop's to all the others, allowing the user 1 machine per LCD and the ability to mouse from machine to machine and control all of them from a single keyboard / mouse. RAIC - redundant array of inexpensive computers - it gets normal multi-tasking 'work' done faster.
Oil is nuts, What we had was so much better. (Score:4, Interesting)
This is from a now dead startup I did from 2002 to 2005. 100% totally silent High end PC's.
http://www.silentcomputing.com/i.html [silentcomputing.com] Look at the last photo.
We had an all aluminum design as well as a water cooled design. I also came up with an advanced Carbon Fiber material with 4x the thermal conductivity of Copper that was light weight.
These systems provide much more cooling then oil could ever do.
The system was sealed, 100% total silent and easy to disassemble and re-assemble. Even easier to work on then a regular PC...
We even had the hard drives in a thermally conductive rubber allowing them to run cooler then in a normal system with fans.
The main system was sealed and designed to run with a descant and dry gas like argon inside so when overclocking using peltier thermoelectric coolers on the CPU where wouldn't be any condensation. Condensation is a major problem for overclockers that up the voltage and have to supercool the CPU.
We are still planning to open source the designs.
We never were able to raise the money to start production of these. To be honest 1/2 our problem was management wanted to court Intel,HP, SUN, and they just didn't get it. The large companies really weren't interested in something that didn't conform to what they already were doing.
We never did talk to enough small investors, and finally we ended up with a bad apple in the company that try to do a hostile take over and killed the company when he failed...
Finally the last 20 polished heat sinks I had were stolen out of my garage 2 weeks ago :( Some idiots problem going to get standard aluminum recycling prices for them too, considering each one cost me over $100 each!
I always felt doing oil was just idiotic and still do.
Too bad being an entrepreneur isn't as easy as programming.
If anyone is still interested in this tech, let me know. I have 3 years invested in it, and we were partners with NASA for much of it.
There's more than one kind of "quality" (Score:3, Interesting)
Mark my words, these will be bought for high-profile uses at these kinds of places and almost certainly will get featured in at least one televison show or movie, probably several. All they need to do is a seed a few at Rodeo Drive stores or the offices of some casting agency and they'll be in like Flynn.
Same thing done in the early 1900s for ... (Score:4, Interesting)
I see a few problems (Score:5, Interesting)
Problem 1: Fire codes. To date, nobody has come up with a non-flammable insulating oil to replace PCBs (carcinogenic, nasty stuff). Oil is flammable and, in combination with electrical equipment, a very bad idea. Generally, it is not allowed within occupied ares outside of fireproof vaults.
Problem 2: Specific heat content of oil. Its not as good as water by a long shot. So, for equipment that operates below 100C, water can transport much more heat than oil. CPU and GPU water coolers are common technology. So what good is oil?
Problem 3: This doesn't eliminate fans, pumps, heat sinks, etc. that eventually move the heat into the surrounding air. That stuff still makes noise. Apart from some custom systems that plumb coolant to remote heat exchangers where the noise isn't a problem, this thing will still have fans. So what did we gain?
Problem 4: How much power do those kewl blue LEDs consume? Here's an idea ....
Tried this before, didn't work so well (Score:2, Interesting)
No, you're wrong (Score:1, Interesting)
The more things change, the more they ... (Score:2, Interesting)
Re:I'm thinking bar fridge or small freezer (Score:2, Interesting)