Bridget: (origin: Gaelic.) Brighid, "fiery dart." The name of the muse who was believed to preside over poetry in pagan times, in Ireland. Brighid, in the Gaelic, also signifies a hostage, a pledge of security.

It will be well fitted when she's speeding through the martian landscape...

These improvement will depend upon to a much improved microprocessor. The vital brain to direct as fast, as smart and as efficiently all Mars' operations.

That part, MER is lacking that much power since it depends very much from Earth remote commanding.

The know most powerful microprocessor that is going to send along with MSL: RAD 750, alike to IBM/Motorola PowerPC 750 dated on the year 1998 which is still very much lagged to our present technology. (Third Generation and now the latest ones is of 5 Generation with 8.125 times faster).

Maybe, one of the most noticeable bottleneck of the space exploration advancement is the radiation-hardened process done by the BAE Systems, isn't?

Rodolfo

Space programs are fundamentally competative, which is probably a good thing. International competition has always been a bigger source of passion for space exploration than the politically-correct theory of international cooperation.

That said, the Amercians and Russians generally played by the rule that you get to brag only after you accomplish something. I'm not impressed by Chinese dictators saying they will build Moon bases. I'm not impressed by computer graphics images of spaceships that haven't been built yet. And I question whether a Mars lander is more advanced than anything NASA ever did, after it hits the planet like a bug against a windshield on the freeway.

No. Briget will end up as a crazy cat lady, hiding Martian cats away in the safety of her (fur-free) crater...

Sabrina would have been a better name, especially if powered by an RTG, but those fules know NOTHING about atomms.

Bob Shaw (Form IIIe)

You are right that the RAD-750 represents the best current hardware that is rated for missions like these but it's worth pointing out that the RAD-750 is closer to 20x slower than current generation hardware (whether x86, Power, ARM, Niagra (SPARC) or Cell). More importantly the MIPs( or FIPS)\watt numbers for some of the current gen hardware beats the RAD-750 by almost 200x. The latest 1Ghz ULV Core Duo has an average power consumpton of 0.75watt. Benchmarking comparisons are hard when the CPU architectures are as different as the Power architecture of the RAD-750 and the x86 Core Duo are but the former is rated at ~240 VAX Mips while the 1 Ghz ULV Core Duo is about equivalent to a 2Ghz P4 which is ~4500 VAX Mips. That's just shy of 20x the processing capability while eating 7.5% of the power.

It's also worth pointing out that the RAD-750 has about 20x the performance per watt of the RAD6000 that the MER's use (22 Mips peak @ 20 Watts) which makes the MER on board compute capability about 4000x worse than the current "state of the art" here on earth. For me that just shows how extremely hard space exploration actually is.

Helvick, Your comments are for Amen! Much improvement. If the MSL or ExoMars would have had it, it would be a good Mars soldier!

I think that the radiation-hardening process is a very long, expensive and complicated process. I don't know about its process but I suspect that this process takes again the same process as the original but with others material. I will try to find more information about this process since I think it is one of that is causing a BIG TECHNOLOGICAL LAGGING for any sophisticated space explorations.

Rodolfo

Of course, with custom realtime OS's - the processing overheads for your average spacecraft are only a fraction of those for the OS's used by those 'mainstream' processors. I've not actually heard of computing performance being a limiting factor for spacecraft - but I may have missed such reports.

Doug

For most spacecraft they aren't but the power consumption of the RAD6000 on the MER's is a significant percentage of the daily power budget. The numbers that I have gleaned from the various web sources are not necessarily reliable but they seem to broadly agree that the "processing" consumes about ~30% of the power budget on average and more than that for compute intensive activities like VISODM. A standard VISODM "step" is around 75cm of drive (15 seconds) followed by 2-3 minutes of computing. I think that the drive motors consume around 30W but even if they consume 100W at full tilt and the analysis only take 2 minutes then a VISODM drive segment consumes 50% more power on computing than it does on actual motion.

Did someone say RAD 750 User Manuals?

And much more....

EDIT _ DOH! The links to the manuals are broken, but the other stuff is still neat.

I'm an absolute idiot when it comes to computers, but if the RAD-750 doesn't provide all the processing power you want for MSL or ExoMars, can't you put more than one aboard a rover, each RAD-750
controlling different functions on the rover?

Generally, Doug is right. There's a lot of semi-informed speculation on this thread, less real info. The RAD750's performance is comparatively poor from two factors: first, the process changes that make its internal registers immune from radiation-induced bit flips slow down the clock speed considerably, but more importantly, external components, also rad-hard, are running more slowly, as are the busses. The RAD750 on MRO doesn't even have an L2 cache and it's using a 33-MHz PCI bus.

If you wanted a non-mission-critical computing resource that didn't have to be totally bulletproof against radiation, there are many options, including commercial processors that happen to be latchup-immune and various gate arrays. For our MSL instruments we are using Xilinx FPGAs; clocked at 40 MHz they are many times faster at doing JPEG compression than code running on a fast desktop system would be.

Rover speed is typically limited more by the capabilities of the drivetrain and the overall power budget. It's not like MER would be going 50 KPH with a faster processor. Despite what AI people will try to tell you, we don't know how to write autonomous nav software regardless of how fast our processors are.

And finally, MIPS (aka "Meaningless Indicator of Processor Speed") is a bad metric for judging computer performance in this or any other problem domain.

I was thinking that too. The AI is one of the software components which needs a fast CPU, lots of RAM in order to perform the harzard avoidance analyze more sophisticated and perform the required action with a much better performance as the MER does. A much improved AI will need much less from Earth remote direction and hence the rover will have greater autonomy to perform the core activities more productively in Mars.

It is true that the MIPS "Millions Instructions per Second" is an old comparision computing power that actualy is obsolete except it is only good to have an idea about how the younger brother computer is improved against the older brother if the model or serie is about the same.

Well, I seems like that the AI is a new field that must work harder to improve the space exploration by improving the autonomy capabiltity of probe or rover. If the microprocessor RAD750 is limited in its computing capability, so why don't put more microprocessors in parallel. The most powerfull computers work with many processors in parallel.

In few words, I think the AI is still very new and I speculate that in the future, the AI will play with a much greater importance. Imaginate that JPL tell the rover: "Please go there, over that dark spot and tell me what is that up?

Rodolfo

The Bell Labs inventor of UNIX, Ken Thompson, was once on an ACM panel discussion and the question was posed, "What is the major contribution of AI to Computer Science". Ken answered, "Fraud!". That was probably true a a few decades ago, when MIT and Stanford AI labs dominated the field. You had grandiose claims and no real results. One famous Stanford profressor, giving a demonstration of an English-language query system, accidently hit the carriage return twice -- the program printed the answer to his question, and then printed the answer to the next question, which he hadn't typed yet.

These days, I would take a look at AI development in the computer-game industry. Games, as frivolous as they may seem, are the driving market force behind a lot of elements of the computer industry. Why does the graphics card in your PC go faster than an SGI workstation? Why do the Intel and AMD processors do vector math? Why was the Cell Processor developed? PC games. That is the commercial application of megaflops.

Click to view attachment

Now let's talk radiation hardened computers. Here is a good solid Russian solution!

Three answers: mass, power, and cost. A single flight RAD750 board uses tens of watts, weighs over a kilo (just for the board, not counting card cage, etc.) and costs, last time I checked, nearly a million dollars. And we don't need more cycles anyway.



Games have driven graphics development, sure. But I would argue that there's nothing like real AI in any game out there. Real AI of a sort useful for rovers would be able to sense the environment and react/plan accordingly. Games just don't have to do that; they define the environment, there's no need to sense it.

Thanks for jumping in Mike - you are dead right on both of the above. I used Mips quite arbitrarily and without enough qualification but the intention was to find some metric that emphasized how extremely different the stuff that has to fly is from what we can put in general purpose PC's.

On the issue of rover speed I was trying to show that there are situations where the current rovers' progress is, to some degree, limited by the electrical power that the onboard computing systems consume during the compute intensive semi autonomous driving modes. I agree that any "improvement" in that would not necessarily lead to a faster rover but it would free up some power for other things.

Don:

What is it?

Bob Shaw

Good to hear your comments!

About the weigth, its is by far heavier than any normal microprocessor and its peripheral components; its price is prohibitive for any commercial applications. However, a more powerfull microprocessor will save money on the other side. It is that we are going to learn the results quicker and hence the mission won't take as long as does MER, hence saves money to the mission operations.

Hence, I see that AI is a very promisory role for future space missions and NASA must pay greater efforts on that. I have enclosed a interesting reports in which make lots of emphasis about the importance of autonomy for a greater producivity of mission. The productivity depends much by a powerfull microprocessor.




So AI is still a novel and with a radiation hardened microprocessor up to date will help to improve the AI.

A Naturally Inspired Robot MarExo

The problem resides of a very long time lagging between the new microprocessors and the radiation-hardened ones. Mike, do you know why it is?

Rodolfo

It's the program timing unit that controls the course-correction engine in Mars, Venera and Fobos probes.

Brits Unveil Latest Robot To Search For Life On Mars

British scientists on Monday took the wraps off a prototype craft to search for signs of life on Mars, hailing it the smartest piece of equipment ever designed for exploration of the red planet.

Details about Bridged

• 1 Is an autonomous robotic scientist
• 2 The project costs $US 910 millions
• 3 Measures 3 meters by 1.8-meter
• 4 It costs $US189 millions
• 5 It will cover more ground (range) than MERs.
• 6 It will have incorporated smarter sensors and adjustment for a safer landing.
• 7 It will be a solar powered.
• 8 It will be tested in Spain and Tenerife to prove its capabilities.
• 9 It is expected to weigh 150 kg in Earth (lighter than MER with 186 kg).
• 10 It will have a life marker chip
• 11 It will incorporate a micro seismometer to facilite the water search.
• 12 It will have a long drill tube of 2 meters.
• 13 It includes an orbiter and a descent module
• 14 The orbiter will operate as a data relay satellite.

Then wait beyond than 2011 for knowing the happening news.

Rodolfo

I must say that I'm fully in agreement with Mike on the liklihood of "AI" being of any benefit to rovers any time soon. More computing power will help some limited functionality and the power consumption\weight savings that might be made by space rating current computing tech would be worth some effort but the end result will not be AI or anything close.

I've worked on and with such systems in the past and they have always disappointed. Fully autonomous AI is still 20+ years out even here on earth with effectively unlimited power and size constraints.We still do not have even the beginnings of the theoretical foundations of what will be needed to build an AI.

Assisted rather than artificial intelligence is something that has made significant strides but the only places where Artificial Intelligence has made any progress have been in well defined domains (e.g. Chess playing systems like Fritz) but even those are really just variants on assisted intelligence where the only brains belong to the developers or users. "Expert" systems, genetic algorithms, simulated annealling, neural networks, bayesian filters and the rest are useful in extremely well defined problem domains but each one has been over hyped.

What's your thoughts on the DARPA challenge? AI or not, they did essentially do what the MERs do, except at 30+ miles per hour, usually using multiple high-end machines to manage the sensors and AI. With MRO quality imagery and a Mars GPS system, rovers going 1-5MPH on their own should be reasonable.

We don't have a Mars GPS system, and we're highly unlikely to have one within a couple of decades - however some of the hazard avoidance and image interp. of the Darpa winners is fantastic and I'm hoping that the US Military doesnt keep it all to itself and some algorythms can make it through to potential planetary rovers in the future.

But - and it's a big but - and likely to be so for a very long time - those DARPA vehicles are using essentially mobile super-computers compared to anything put into space. Maybe it goes around in circles a little bit

No powerfull space-suitable CPU's available....so no high processing requirements ever made
No requirement thus no real shotcoming in the availability of more powerfull processors etc etc.

Doug

It would save a lot of time and planning if a rover was even smart enough to take commands from Earth like "Turn 35 degrees and go 250 meters, avoiding obstacle".

Computer vision is notoriously unrobust. Stereo fusion (calculating a depth map from two camera views) has been around for decades, but it can be fooled by unusual textures or visibility features (something one camera sees but is hidden to the other). You definately want to believe input from cat whiskers or inclinometers more than you believe the vision algorithm, or the rover will end up driving off a cliff.

You're pushing the limits of robust AI algorithms to plan a path around an obstacle and still try to reach the target coordinates. And if anything hairy happens, it should stops and call Earth for help. Recognition and path planning tasks performed routinely by a rodent are well beyond what a super computer can do today.

The issue isn't processor speed, it is the primitive state of the art in AI algorithms. It's an unfortunately feature of academic AI culture to exagerate that state of the art, so be skeptical until you see a rover really doing what is promised on realistic outdoor terrain, not a white floor with colored cubes.

Don - have you seen the results of the Darpa challenge? It couldn't be any less lab-conditions - it was quite an achievment. Loads of the entries were complete failures - but a few were superb and completed a complex course over terrain both rough and smooth, with plenty of obstacles, totally unassisted.

Doug

Jim Bell said that during certain points of the day, especially at sunrise
and sunset, the Mars Rovers would sometimes refuse to move ahead.

Turns out they interpreted their shadows as pits in the ground and
did not want to fall into them. They had to be "told" that shadows
were okay to drive on.

About a decade ago, a robot car designed by the US military being
tested on a regular automobile road kept swerving over to the other
side of the lane for seemingly no reason.

Turned out that a row of trees had their shadows falling across the
road and the robot car interpreted them as obstacles and obeyed its
programming by trying to avoid them.

This does make one admire the abilities of what the brain can do
in such a compact package.

And on a road.

To be fair - it was hardly a beautiful tarmac highway....

http://www.darpa.mil/grandchallenge05/gran...05/dsc_3925.jpg

There are bits of the floor of Gusev crater, and almost all of Meridiani where I would rather drive my car than on that road

Doug

Most of the teams preprogrammed the entire route from airphotos/satellite images and could have (or did) dead-reckoned nearly the whole way on GPS without even having vision or laser-scanning systems. And the vision systems were highly optimized to find the road edges.

I looked at this fairly extensively a few months back, and in my opinion the applicability to planetary rovers is pretty low. I won't even discuss the relative power density between gasoline and solar or RTG systems. Between lidar and racks of processors, the GC vehicles were burning through kilowatts of electricity.

Oh - I quite agree ( and mentioned earlier ) there are few parallels

The simplest way to do this sort of thing is to put a human brain in the loop.

Doug

Sure, the control remote range of present technology won't be practical beyond of Mars. That is that any kind of robot (rover, aerobot, plane) on any Gallilean and Saturninian moons won't be easy without a well developed AI along with plenty peripheral sensors and powerfull computer system to advance the scientific mission not so longer time than the MER's does in Mars.

At the present technology, to rover in a real time in Moon is feasible, in Mars, only with remote command up to 95% and 5% of hazards avoidance. For Mars and Venus, the robot technology areas needs to work harder in both ways: Improve the AI and peripheral sensors and hence, this demand will develop a new market so that, I think, BAE will justify it as a good business before selecting a more powerfull microprocessor and its peripherals (RAM, EPPROM, ROM, bus, etc.) to be radiation-hardened.

Rodolfo

P.D.Now there are soccer game among mini-robots (very funny).

That's what I was getting at. It is a successful but special-purpose solution. I do think it is feasible to get a rover to avoid obsticles with occasional calls for help. But that takes another special-purpose solution that is pushing the state of the art. The rover is not going to be "smart" in any sense.

News articles about these kinds of things always exagerate, both because the journalists don't understand the science and because the academic culture has evolved to speak very aggressively and compete for precious small grant money. There is a natural tendancy to anthropormorphize, and you see blatent attempts to encourage that with projects like these. They are fun to check out, but what you see is misleading.

What biological brains do is indeed remarkable, and the robots you see in movies are pure science fiction. Nobody really knows how smart a computer could be if it was programmed correctly. Maybe a high-end PC could be as smart as a human, but the breakthrough in software technology has not happened yet.

Here's the cover of ESA BUlletin we talked about ( FREE copies available via ESA publications )
http://www.esa.int/esaMI/ESA_Publications/index.html

Actually, the issue here is navigation. Avoiding obstacles is only a very tiny part of that.

In that respect Meridani and Gusev are not really very challenging sites and Spirit and Opportunity not really very representative of the kinds of rovers that will be needed to traverse them. Both sites are largely open plains where for the most part obstacles are few and far between and those which do occur a rover can generally (the sandtraps Opportunity keeps getting itself mired in are an important exception) see coming for dozens of yards if not a mile or two off, and thus can identify them (and work out a way around them) long before it actually encounters and has to deal with them. Even the dune/ripple fields Opportunity is currently traversing are no real obstacle. Not only can it see over their tops, when it comes to an end of a trough instead backtracking and going around to another it generally simply rolls over a ripple to a neighbouring trough. That sort of solution would have been far less viable, if not downright impossible, had it been confronted by (say) the kind of rock-filled obstacle course Sojourner faced at its site.

As for the rovers themselves, the task of navigating Spirit and Opportunity is done almost entirely by minds back on Earth. For example, Opportunity does not decide for itself which sand trough to travel down. Its human babysitters decide for it. In that respect nothing much has really changed since the days of the Soviet lunar rovers of the 1970s and it seems unlikely to change any time soon; and even if it could change it needs to be remembered that a rover is really only a kind of proxy explorer for its human controllers on Earth. The latter will want to decide for themselves where their proxy is going. That inevitably is going to slow rover progress down to the speed the humans can get pics and other information back from the rover to Earth, make a decision, then upload the next batch of instructions. Not to mention limiting it to how far the humans can see.


No existing PC, high-end or otherwise, would be able to run such software--because no PC yet invented can match the speed of the human brain. Individually, neurons are certainly slow-coaches compared to even the slowest electronic CPU, but when they are harnessed in parallel, as the human brain does, they can process information at blinding speeds. You have only to consider how fast your own brain can identify obstacles in front of you and get you to react in some appropriate fashion then compare it to the time it takes Spirit or Opportunity to decide that the rock in front of them is an obstacle they have to go round rather than over.

Hardware breakthroughs as well as software ones will be needed before electronic brains became as smart as human ones. (And even then do not expect to see them being placed inside rovers and rocketed off on one-way trips to Mars. The creation of AI's is going to pose all kinds of ethical dilemmas when they do eventuate. For if computers ever do become as smart as human beings one issue that is inevitably going to be raised at some point is whether they should be accorded the same rights as human beings. That would presumably include not being sent off to other planets on what would amount to suicide missions.)

======
Stephen

Actually - that's not quite fair - Sojourner and MER were both able to be given a target point, and make progress toward that target point, and would avoid obsticles in the way, navigate around them and return to the target point. There was one great example where Spirit actually gave up and drove backwards around an obsticle early on.

So yes - you couldn't say to Spirit "go to the top of Husband Hill " from the rim of Bonneville..it still requires people in the loop on a daily basis - BUT - it's a lot smarter than you give credit for really.

Doug

Ah! So when it finally happens, we can replace all politicians by a low-end PC.

Cugel, we're talking about artificial *intelligence*... that last word has very little to do with politicians.

-the other Doug

Has anyone proposed a manned mission for the purpose of controlling one or more rovers from orbit but without attempting to land human beings on the surface? I'd think we'd be able to get a lot more out of the rovers if they were controlled from no more than a few light seconds distance.

I could imagine this working for Mars or Venus, and I'd think that the cost would be a lot less than a mission that aimed to put people on the surface, but I've never seen it discussed anywhere.

I've heard people on message boards like these suggest something similar, with a Phobos base. I'm sure the agencies have thought about these kind of things like they have thought about lunar bases etc.
I'm afraid we have to be patient for now.

That is one of the weakest point of NASA's research fund programs which is to improve the capability of robots for unmanned explorations. We are still using Pentium III alike for MSL and MRO. That is still backward. No much work on the interface between smart sensors, computer and software. Now, the Japan is leading on that field. We might send the Asimo, Honda's Humanoid Robot. That robot can walk and salute as any human.

http://asimo.honda.com/inside_asimo.asp?bhcp=1

Excitement fills the theater as guests witness ASIMO maneuver through a home nvironment using its amazing mobility capabilities such as walking forward and backward, climbing and descending a flight of stairs and even dancing!


However, the objective of Asimo design is to imitate as close as possible to humanoide action. Later, there were others incorporations such as the reasoning to solve problems.

Rodolfo

Actually the CPU's in most modern spacecraft are more like 1/10th the performance of a Pentium 3....and what's more, there's not much requirement for anything better.

Now - you could argue that it's cyclical - the need for more on orbit computing power has not arisen because people have programmed for what is available and that's tended to be 'enough'. Also - spacecraft are tending to become little centres for distributed computer, with each instrument having it's own processor dedicated to the aquisition, compression and storage of it's own data - it leaves the CPU of most spacecraft doing the comparatively simple task of attitude control, data management, and streaming stuff through to telecoms etc.

I'm sure if there were something 10x faster availabel for on orbit computing, it would be utilised...but the fact that such a processing system isn't in place perhaps suggests it isn't really that necessary.

You drop the bloated OS, the graphics and so forth, dedicate the use of your CPU to on orbit computing, and actually, the mathematics behind a spacecraft are comparatively simple.

Doug

Yep, that's a pretty clear way to make the most of human brains & robot brawn. Let the humans stay in orbit with a nice big station that you don't need to land, and let them run remotely operated robots all over the place that can stay out there all the time, don't need to come in at night (or in winter), don't need the constant "maintenance" that us humans need. This is considered to be the only feasible way to do "human" exploration of Venus. It was also one of the motivations behind our original "Red Rover, Red Rover" project with LEGO. And we still have the pipe dream of a project in which we'd build some kind of nanorovers that could be landed on the Moon -- and then operated by members of the public, for no other reason than how cool it would be for you to be able to sit down at your computer and drive a rover that was actually on the Moon. Wouldn't that be neat? Someday...

--Emily

I confess...I've driven every functional RRGTM station I could find - and even made a little mosaic from one of them (long lost...but I might try and make another one )

Doug

(PS - attached one from the Davis Creek Elementary site....the TPS one worked for a bit, then the top half of the interface wouldn't refresh If I can get it working again, I'll do a pan from there as well....but I have to say, the calibration process is shocking )

Beautiful work, Doug!

To get the TPS one to work better I have to give a kick in the pants to the guy who is SUPPOSED to restart its computer daily...who usually remembers to do so for a couple of days and then, well, it doesn't happen anymore...sigh...

--Emily

Yes, sure that NASA has selected rightly the capability of microprocessors for the missions of MRO and MSL since they aren't going to need a more powerfull microprocessors to support the mission cores that is mostly dependent by remote control principally to MSL. If NASA has put more effort about the improvement navigation autonomy of robot, sure MSL will need a more powerfull microprocessor to depend less from Earth remote control. Hence, the geology and biology scientific missions would be more productive with improvement advancement and faster return of results.

Rodolfo

Question -- would faster microprocessors also require more power?

--Emily

Definitely, yes. That is one of the engineering concerns.

However, there is a new variety of microprocessors which are energy efficient, inclusive much energy economy than RAD750 and are much more powerfull such as the Intel Centrino of last genertion which as capable as the last model of Pentium IV.

Rodolfo

Well, I wouldn't say definitely. In general, power = C*v^2*f where C is a constant depending on the number of gates and the process (how small the gates are), v is the supply voltage, and f is the switching frequency. You can bump up the frequency and keep the same power by improving process and reducing supply voltage.

Terrestrial bound commercial microprocessors have kept pushing process improvements (lower switching and quiescent power per gate), and supply voltage reductions. Analagous improvements in rad-hard microprocessors are more difficult and have been slower-paced since reducing gate geometries and switching threshold voltages typically makes them more susceptible to particle induced single event upsets. And there is less economic demand for rapid improvements in rad-hard proceessors than in commercial processors. Rad-hard processors are thus more likely than commercial processors to require more power for higher switching frequencies.

100% true however it is worth pointing out that as the technologies improve the general trend is towards (much) more computing power per watt within similar processors.

The MER RAD6000 from BAE consumes 20watts @ 20Mhz. The RAD750 comsumes 5-14watts @ 132Mhz. The 750 in it's most stringent radiation hardened mode is at least 20x more efficient (in terms of instructions/watt) than the 6000 used on the MER's.

Thanks for your comments which are good.

About the power consumption of microprocessors depends what you mentioned (more gates or transitors and frequency, leads greater temperature due to greater power consumption in Watts). There is a limit of temperature that the semiconductor material becomes unstable its electrical conducting properties (leakages currents). That is the Moore's law.

The other factor that influences the consumption of watts is related to the type of material (Bipolar versus CMOS). Historically, there is a growing power consumption when the frecquency and number of circuits grows until a change of material technology, drops the power consumption. As an example: Bi-polar material was requering lots of much power energy versus CMOS. Now there is a new variety of CMOS which needs less power than the original CMOS for the same frequency and density of circuit.

So I was saying the previous post as the general principle. However, a new semiconductor material technology helps to consume less power for the same computing capacity.

Rodolfo

Sounds like a different Moore than the one I am familiar with. Something may be getting lost in translation here. What you refer to as material, e.g. bipolar vs. cmos, I would refer to as design. Or in your other use of material, I would use process technology. Detailed discussion of semiconductor physics and fabrication are getting somewhat OT here. Suffice to say that there will be continued improvements in computational power per watt in processors, whether commercial or rad-hard. Care should just be taken in comparing performance / power of commercial vs. rad-hard. Somewhat different beasties.