Most expensive CPU in the world, Curiositys RAD750.

[soccerballtux]

^looks like I need to read

 

[PlasmaBomb]

I don't get it. why don't they just enclose an arm bases sbc inside a leaded (radiation proof) box?

According to this really old paper (Phys. Rev. 79, 293–296 (1950) The Penetration of Particles Associated with Cosmic-Ray Stars) even 8" of lead shielding isn't enough to stop the really energetic electrons which represent 1-2% of the cosmic radiation found at ~3500m.

 

[ShintaiDK]

According to this really old paper (Phys. Rev. 79, 293–296 (1950) The Penetration of Particles Associated with Cosmic-Ray Stars) even 8" of lead shielding isn't enough to stop the really energetic electrons which represent 1-2% of the cosmic radiation found at ~3500m.

Exactly. Not to mention the cost of bringing the extra weight to space.

 

[aigomorla]

*looking for the macintosh logo*

its a powerpc right?

 

[soccerballtux]

Exactly. Not to mention the cost of bringing the extra weight to space.

at $1000/kg (cost of next moon-capable SpaceX rocket) you could bring 200kg worth of metal box along with you for the cost of this processor

 

[soccerballtux]

According to this really old paper (Phys. Rev. 79, 293–296 (1950) The Penetration of Particles Associated with Cosmic-Ray Stars) even 8" of lead shielding isn't enough to stop the really energetic electrons which represent 1-2% of the cosmic radiation found at ~3500m.

cool!

 

[Idontcare]

They wont. Also you would be surprised how much is even made on 180nm today. besides Intel, highend process nodes only account for a few % of the capacity at foundries. While I dont have the current numbers. I dont even think 65nm and below accounts for more than 25% at TSMC for example.

Even for Glofo, only ~25% is 45nm or lower.

3" is also easily penetrated.

Not disagreeing with the spirit of your post, just throwing this out there because I happen to have it compiled already. The basic premise remains true, a large amount of TSMC's wafer volume (the above graph is revenue volume, not wafer volume, and higher revenue for newer nodes will weight the smaller wafer volume accordingly) comes from >150nm processes.

 

[Gigantopithecus]

*looking for the macintosh logo*

its a powerpc right?

Yes, it's similar to the chip in the original iMac.

http://www.baesystems.com/download/BAES_052281/Space-Products--RAD750-component

 

[soccerballtux]

^^500nm lol

 

[Tuna-Fish]

Well, I think someday all manufacturing will have advanced beyond 90nm.

ShintaiDK and IDK sort of covered this already, but I just have to add that generally, process capacity isn't shut down, at all, ever. By die size, microprocessors are very small part of the total semiconductor industry, things from sensors to solar panels are made from the stuff. As the leading edge foundries like Intel and TSMC move to new process, they sell their old, outdated tools to some other company that does something different from silicon. The total amount of wafer area produced steadily grows over time, and there's someone operating somewhere at basically all nodes below a micron. (And quite a few above a micron.)

 

[PlasmaBomb]

at $1000/kg (cost of next moon-capable SpaceX rocket) you could bring 200kg worth of metal box along with you for the cost of this processor

Your cost per kilo is out... putting the initial 24 satellite GPS constellation into orbit cost $12billion. Early satellites were ~750 kg later ones were heavier at ~1500-1600kg. So lets pick 1000kg for simplicity - that's a cost of $500,000 per kilo, which means any "paltry" savings on a $200k circuit board and chip would not buy you sufficient shielding to guarantee the satellite would work.

http://en.wikipedia.org/wiki/GPS_(satellite)
http://nation.time.com/2012/05/21/how-much-does-gps-cost/

 

[alyarb]

i'm not getting that. the block 1 satellites were about $20mil each. an atlas rocket is about $10 million... so 30x24 is not even a billion. and you say 12 billion to lift these vehicles? I think you are throwing in just a little extra R&D cost and maybe 5 years of payroll.

 

[Idontcare]

Yeah I think plasmabomb forgot to pull out all the admin overhead in that $12B figure, surely that wasn't just the launch cost for 24 satellites...then again, when you are spending other people's money, $12B might be a downright bargain

All joking aside, this is why the CPU is priced to be $200k and not $20k or $2k...it is priced at what the market will bear, not what it costs to design/manufacture.

There are a lot of ways to make the seemingly benign cost orders of magnitude more than its weight in gold...and going to the expense of putting the thing in space is one of those ways.

It is not "what is this rad-hardened CPU worth"; rather, it is "what is this rad-hardened CPU worth to you (NASA, etc)?"

To which NASA responded - "mmm, compared to the cost of entombing a $1k off-the-shelf CPU in 20kg of lead shielding and sending it into orbit, your rad-hardened CPU is worth about $200k to us"

 

[soccerballtux]

Your cost per kilo is out... putting the initial 24 satellite GPS constellation into orbit cost $12billion. Early satellites were ~750 kg later ones were heavier at ~1500-1600kg. So lets pick 1000kg for simplicity - that's a cost of $500,000 per kilo, which means any "paltry" savings on a $200k circuit board and chip would not buy you sufficient shielding to guarantee the satellite would work.

http://en.wikipedia.org/wiki/GPS_(satellite)
http://nation.time.com/2012/05/21/how-much-does-gps-cost/

cost of next moon-capable SpaceX rocket
I'm off a bit. He's quoted as saying $500/lb is doable
it's more like $1.8k/kg currently
http://en.wikipedia.org/wiki/Falcon_(rocket_family)#Launch_vehicle_comparisons

 

[soccerballtux]

Yeah I think plasmabomb forgot to pull out all the admin overhead in that $12B figure, surely that wasn't just the launch cost for 24 satellites...then again, when you are spending other people's money, $12B might be a downright bargain

All joking aside, this is why the CPU is priced to be $200k and not $20k or $2k...it is priced at what the market will bear, not what it costs to design/manufacture.

There are a lot of ways to make the seemingly benign cost orders of magnitude more than its weight in gold...and going to the expense of putting the thing in space is one of those ways.

It is not "what is this rad-hardened CPU worth"; rather, it is "what is this rad-hardened CPU worth to you (NASA, etc)?"

To which NASA responded - "mmm, compared to the cost of entombing a $1k off-the-shelf CPU in 20kg of lead shielding and sending it into orbit, your rad-hardened CPU is worth about $200k to us"

if they just fabbed more of these and gave them out, then we exchange them for gold in weight, then we all rich

 

[HondaCop]

But the question that needs to be asked is... Is the IHS soldered to the die?

 

[PlasmaBomb]

Yeah I think plasmabomb forgot to pull out all the admin overhead in that $12B figure, surely that wasn't just the launch cost for 24 satellites...then again, when you are spending other people's money, $12B might be a downright bargain

All joking aside, this is why the CPU is priced to be $200k and not $20k or $2k...it is priced at what the market will bear, not what it costs to design/manufacture.

There are a lot of ways to make the seemingly benign cost orders of magnitude more than its weight in gold...and going to the expense of putting the thing in space is one of those ways.

It is not "what is this rad-hardened CPU worth"; rather, it is "what is this rad-hardened CPU worth to you (NASA, etc)?"

To which NASA responded - "mmm, compared to the cost of entombing a $1k off-the-shelf CPU in 20kg of lead shielding and sending it into orbit, your rad-hardened CPU is worth about $200k to us"

Quite possibly - the article states that GPS costs $2 million a day to run...

The European Galileo project is supposed to cost ~$25 billion, isn't complete and last I heard it was running over budget.

Don't forget that GPS isn't operating at normal near earth orbit heights either, which is what at lot of your typical launch costs/kg are looking at (AFAIK).

So yes if you wanted to rip out the existing CPU and add a lot of shielding (and weight), you aren't going to get a cheap launch, especially if you find you need to design a new launch vehicle with the correct payload abilities

 

[fthoma]

The RAD750 was reverse engineered from IBM's PowerPC750, using their design data and verification programs and test patterns. For the radiation hardening there were special circuit design techniques used to allow operation in high radiation environments. The project included its own semiconductor manufacturing line, as well as design of supporting interface chips such as the memory and buss interfaces. The end cost might seem high at first blush, but the development cost had to be amortized over a relatively small number of end products. The processor has been used starting with Deep Impact through the Mars Orbiter and Observer, the rovers, GRACE, and several hundred other space missions. The chief radiation hazard is not electrons but protons, which are difficult to shield against. The feature size of the critical nodes has to be large enough so that a proton will not discharge the node and not allow recovery within the processor clock cycle, which is reflected in the slow clock cycle compared to present processors. The original space station processors were Intel 386 processors for that reason. The design team that designed the replacement processor had the good fortune to have a good set of circuitry they could use that provided the proper attributes to allow smaller feature sizes, such as circuits that will not propagate a disturb. These techniques are much larger in terms of area to implement, so die sizes tend toward the larger size. It takes a couple hundred people's innovation and effort to design and manufacture a chip. Just the base mask set exceeded $1 million in the early 2000's, one bad connection and start over.

 

[Smoblikat]

I thought intel had knights corner? Isnt that a quarter of a million dollars?

I used to be an intern at BAE, they have some REALLY stupid people working for them.

Apparently I posted in this thread 4 years ago. If only I could go back and tell myself about things.........

 

[aigomorla]

Talk about a Necro....