I really miss the seriousness and sincerity of presentations like these. Most people probably find it boring, but I think it is the right attitude toward this kind of work.
I do fluid dynamics today, and even with modern tools its computationally intensive. I cant believe how efficient this is. Its so fascinating knowing the math behind it.
you might be the perfect person to ask, I own a 3090 but have been wanting to run 'vintage' code on it to compare it directly to super computers of past abstract papers. where would I even begin in going down that rabbit hole? Obviously there might be certain computations that suit the 3090 better than others, kinda figured fluid simulations might be one of those areas
@@burnerr That'd be good to run something that uses CUDA, which is parallel computing that takes advantage of GPU. In the video they were talking about CPU processing, and graphic cards were not a thing, so the code may only use your GPU unless it's modified. What the video doesn't tell (unless I wasn't paying enough attention) is the computational time required back then to do those simulations. You see, they run the code in a super computer, get the data, and what they present in this video are the results interpreted by visualization tools.
Physical Chemists use computational fluid mechanics modelling and simulation techniques with the Diffusion PDE to model and predict chemical transport phenomena with statistical mechanics in atomic and molecular systems.
@@burnerr The main difference is that GPUs are highly efficient for code that is written for parallel execution. Each core in a GPU is less powerful than a CPU core, but you have orders of magnitude more cores in a GPU than in a CPU. Think of a simple operation, like adding two vectors element-wise. But they each have millions of elements. You can give a chunk of of these operations to each core and with all of them working simultaneously, you can make the process faster. What I mean to say is that simply comparing computational power without any context will not lead you anywhere. You need to know exactly what kind of algorithms/mathematical operations you need for a given problem, so that you can say whether it can be made faster by this type of parallelisation. (And I'm not even going into the problem of memory management, since you need to move stuff to and from GPU memory to make these things work). Conclusion: not an easy task. Bonus comment: a 3090 (or any consumer GPU for that matter) may be fast for certain things, but bad at other (look at comparison between single and double precision operations). Most consumer graphics programs do not care about double precision. Scientific programs usually do.
A single 3090 ti is 230,526x as powerful as the cray 2. The computing power we have available today is absolutely incomprehensible compared to the systems they were using in 1989.
And if that isn't mind blowing enough, todays algorithms are an order of magnitude more efficient, both for time and for space complexity. I can't find any direct comparisons between the oldest methods and state of art, but I do know that the most recent techniques like Lattice Boltzmann already cut down the space requirements for calculations in half compared to previous techniques while there were 10s if not 100s of different optimizations discovered since 1989
@@sumofat4994 it's not a fair comparison as cray 2 was designed for a specific use by the military while 3090 ti is designed for general public use for all sorts of applications. It all comes down to optimization
@@sumofat4994 lol bro what? programmers now are idiots. tools and AI do all the work; tools and AI built on the shoulders of the giants of the 1980s and 90s.
This is such a great inspiration and what a wonderful time for CFD. Kwak & Rogers probably spinning their artificial compressibility code, somewhere over the sea in Japan, Kawamura, with his massive computers as ones seen here doing quasi DNS, or ILES something else that has seen the light in '89 thanks to Boris as well. I just started going to school that year, and the road that will make me appreciate what is shown here, still doing my best to replicate it on my own.
One of the "ads" that played before/after Bill Nye the Science Guy was some kids laying in the grass each saying what they wanted to be when they grew up. Every one said cop or baseball player... except for the one girl at the end said she wanted to be a "Computational Fluid Dynamics Engineer". Never forget.
If this was what NASA was doing in 1989 which is now mostly standard in CFD packages, I can't even imagine what they can do with today's computing power
It really is amazing that people write comments as if today's state of the art computing came out of nowhere. What you see here was built/coded based on 41 years of new developments and experience acquired from the first stored-program computer back in 1948, so these WERE the most advanced computer programs of its time. I hope people in 40 years from now will look back at what we have today, and not be as patronizing as these comments are.
Fascinating documentary. CFD - Computational Fluid Dynamics is a very integral branch of Fluid Mechanics which is a major branch of Applied Mathematics and Industrial Mathematics / Mathematical Modelling which has applications in Chemical Engineering, Chemical Physics and Physical Chemistry.
The gods of the RUclips algorithm recommending this to me in 2022 means I have either succeeded in life and am on par with Einstein, or that I have watched every video available and this video is a warning to log out and take a break from the internet...
really impressed by the video, really magnificent simulation tech indeed in 1989. I also find it interesting to watch these old videos from 80 or 90s, it shows the beauty of science and engineering.
I had a copy of FAST for my SGI back in the 90s. I was really sad after 9/11 to have found it removed from distribution to the public. Now that IRIX is 20 years EOL I think it would be very cool if you released, or even open sourced, this software to the public.
See the book "GPU COMPUTING Gems - Emerald Edition" (GPU Gems" is a different series by a different publisher). It has 3 chapters on using GPUs for CFD simulations of turbines.
Why is no one talking about the music on the video here? I really wish I could present CFD results with a soundtrack like this. My next goal is to convince my colleagues and supervisor this is a good idea. One can only dream...
I'll do a quick comparison to put things in perspective The Cray-2: A Raspberry Pi 4: 190 billion operations per second 268 billion operations per second 4 cores 4 cores cost $32 million costs $100-$200
@@Hyraethian Thanks for that comparison ! I have had an interest in CFD since the mid 1980s. However, I am a visual learner, and also short on money. I have seen a CFD program called OpenFOAM that is free, but still seems to require a knowledge of writing code and such. Is there now an open source ( or low cost ) CFD program / (" app" ? ) that has an easy to understand GUI ? ( visual tabs and buttons, rather than blocks of text to input ) I have built 3D models in Softmage XSI, so anything up to that level of learning difficulty is what I am after. Years ago, Autodesk released a CFD program called "Project Falcon" It was very basic, but got me super excited, because I was even able to run it on a first gen Surface Pro tablet. It gave info such as cD and lift, and all within less than an hour of render time. It seemed to be geared towards kids, since it was extremely basic and had silly things you could run tests on, like a sci fi spaceship and even a oair of sunglasses. It seemed created for touch control as well, ( and worked great on the Surface Pro touch screen ) I had high hopes that a phone app would be created and that it would be offered to the masses for cheap. However, that was all dashed when Autodesk killed it and made it Autodesk Flowworks ( or some other name like that ) and now markets it for thousands of dollars.
@@DennisZIyanChen I just pulled stuff from Wikipedia. I tried to read through the PDF of the CRAY-2 Computer Systems Functional Description. But it was too dense for me and I just decided to quote the figures other people had worked out.
Can anyone help me figure out what do you call the introduction music to this video. That kind of music was ubiquitous in these kinds of documentaries. thank you
whats crazy is that they effectively combined thousands of PCs into one simulation database from MSFS 2020. A hivemind of computing. There is actual fluid dynamics so the military, aviation businesses will be able to use the flight data which has accurate simulation variables.
Cool but the format is very dry which is a given for the time period this was made in. Great information, we have progressed much and still rely on a ton of simulation data which has become highly advanced since then.
I really miss the seriousness and sincerity of presentations like these. Most people probably find it boring, but I think it is the right attitude toward this kind of work.
dont you want to hear about black transexuals
One half the US suffers from ADHD, the other half is permanently on potent pharmaceutical drugs. No wonder it all goes down the drain.
todays kids attention span is very low
Seems like those shorts, like TikTok, are destroying out generation’s attention span
all this work just for this technology to be used in some anime porn game
I do fluid dynamics today, and even with modern tools its computationally intensive. I cant believe how efficient this is. Its so fascinating knowing the math behind it.
Yes, partial differential equations, for the most part, are computational intensive.
you might be the perfect person to ask, I own a 3090 but have been wanting to run 'vintage' code on it to compare it directly to super computers of past abstract papers. where would I even begin in going down that rabbit hole? Obviously there might be certain computations that suit the 3090 better than others, kinda figured fluid simulations might be one of those areas
@@burnerr That'd be good to run something that uses CUDA, which is parallel computing that takes advantage of GPU. In the video they were talking about CPU processing, and graphic cards were not a thing, so the code may only use your GPU unless it's modified. What the video doesn't tell (unless I wasn't paying enough attention) is the computational time required back then to do those simulations. You see, they run the code in a super computer, get the data, and what they present in this video are the results interpreted by visualization tools.
Physical Chemists use computational fluid mechanics modelling and simulation techniques with the Diffusion PDE to model and predict chemical transport phenomena with statistical mechanics in atomic and molecular systems.
@@burnerr The main difference is that GPUs are highly efficient for code that is written for parallel execution. Each core in a GPU is less powerful than a CPU core, but you have orders of magnitude more cores in a GPU than in a CPU.
Think of a simple operation, like adding two vectors element-wise. But they each have millions of elements. You can give a chunk of of these operations to each core and with all of them working simultaneously, you can make the process faster.
What I mean to say is that simply comparing computational power without any context will not lead you anywhere. You need to know exactly what kind of algorithms/mathematical operations you need for a given problem, so that you can say whether it can be made faster by this type of parallelisation.
(And I'm not even going into the problem of memory management, since you need to move stuff to and from GPU memory to make these things work).
Conclusion: not an easy task.
Bonus comment: a 3090 (or any consumer GPU for that matter) may be fast for certain things, but bad at other (look at comparison between single and double precision operations). Most consumer graphics programs do not care about double precision. Scientific programs usually do.
A single 3090 ti is 230,526x as powerful as the cray 2. The computing power we have available today is absolutely incomprehensible compared to the systems they were using in 1989.
And if that isn't mind blowing enough, todays algorithms are an order of magnitude more efficient, both for time and for space complexity. I can't find any direct comparisons between the oldest methods and state of art, but I do know that the most recent techniques like Lattice Boltzmann already cut down the space requirements for calculations in half compared to previous techniques while there were 10s if not 100s of different optimizations discovered since 1989
and?
Too bad the software is 1,000,000 x slower because programmers are 2,000,000x worse than they used to be.
@@sumofat4994 it's not a fair comparison as cray 2 was designed for a specific use by the military while 3090 ti is designed for general public use for all sorts of applications. It all comes down to optimization
@@sumofat4994 lol bro what? programmers now are idiots. tools and AI do all the work; tools and AI built on the shoulders of the giants of the 1980s and 90s.
My left ear loves this
I was a fluid dynamics specialist on the first Avatar. This is fascinating to see cfd’s early days!
A great time capsule of a video. I'm impressed they had the capability to render that with the kind of computing power they were limited to back then.
This is such a great inspiration and what a wonderful time for CFD. Kwak & Rogers probably spinning their artificial compressibility code, somewhere over the sea in Japan, Kawamura, with his massive computers as ones seen here doing quasi DNS, or ILES something else that has seen the light in '89 thanks to Boris as well. I just started going to school that year, and the road that will make me appreciate what is shown here, still doing my best to replicate it on my own.
I love how the voice is on the left ear but the music is on the right
One of the "ads" that played before/after Bill Nye the Science Guy was some kids laying in the grass each saying what they wanted to be when they grew up. Every one said cop or baseball player... except for the one girl at the end said she wanted to be a "Computational Fluid Dynamics Engineer". Never forget.
Beautiful. CFD has come a long way ...
If this was what NASA was doing in 1989 which is now mostly standard in CFD packages, I can't even imagine what they can do with today's computing power
This video is pure gold. I love to watch in couple with new CFD techniques described in "two minutes paper" channel
I just started a career in this field. I also run simulations from a desktop machine from home. I absolutely love it!
It really is amazing that people write comments as if today's state of the art computing came out of nowhere. What you see here was built/coded based on 41 years of new developments and experience acquired from the first stored-program computer back in 1948, so these WERE the most advanced computer programs of its time. I hope people in 40 years from now will look back at what we have today, and not be as patronizing as these comments are.
Fascinating documentary. CFD - Computational Fluid Dynamics is a very integral branch of Fluid Mechanics which is a major branch of Applied Mathematics and Industrial Mathematics / Mathematical Modelling which has applications in Chemical Engineering, Chemical Physics and Physical Chemistry.
Gotta love those NEC 15inch CRT monitors. Color too !
Those were Silicon Graphics.
Im impressed they could emulate this stuff back then.
The gods of the RUclips algorithm recommending this to me in 2022 means I have either succeeded in life and am on par with Einstein, or that I have watched every video available and this video is a warning to log out and take a break from the internet...
They dreamed of a computer with 1 Teraflops of computational power. PS4 has 6 Teraflops. The progress that technology has made is incredible.
This is better than nature sounds to relax :)
They had this in 89? DAMN
I can't believe all these are from 1989. Fabulous!!!
Man we reach the moon in 1960s...
The narrator's voice give me Forensic Files vibes, and I like that.
really impressed by the video, really magnificent simulation tech indeed in 1989. I also find it interesting to watch these old videos from 80 or 90s, it shows the beauty of science and engineering.
After this video finished I turned around my headset and played it for my right ear.
20:00 amazing visuals
it's so incredible to me that computers could render these kinds of simulations before the USSR had even collapsed. I am clearly a zoomer
I had a copy of FAST for my SGI back in the 90s. I was really sad after 9/11 to have found it removed from distribution to the public. Now that IRIX is 20 years EOL I think it would be very cool if you released, or even open sourced, this software to the public.
Is it easier to use than OpenFOAM? That is royal pain in the ...
To think that much computer power is now found in a wrist watch is amazing.
NOW we're talking! 👍
See the book "GPU COMPUTING Gems - Emerald Edition" (GPU Gems" is a different series by a different publisher). It has 3 chapters on using GPUs for CFD simulations of turbines.
Forget the amazing science and physics. Look at the rad 80’s aesthetic!
Was a boring "for the dumb public piece" until 19:00 . The entropy plot was very interesting.
Just wait til the wind tunnel union catches wind of this...
I'm not sure if this was before or after the release of Rockwell Automations' Retroencabulator.
Definitely before. The ability to provide inverse reactive current while effectively preventing side fumbling makes this whole process obsolete.
16:09 oh that old thing 😛
great video!
Great video. Thank you
I wonder if PDE was use since the maths behind it is seriously very intense.... I admire those days how people got into serious business of wonder.
Yes, probably in those simulations they are using the Finite Volume Method to solve the Navier- Stokes equations (a PDE)
WONDERFUL
Just look at the evolution of technology in 40 years !
Applies to psychology apps as well. And to think of the finite element analysis program, FEPC used back in the day.
Wow 1989!
Pure Gold love from BR
6:48 lol he was just hovering over Copy
중간에 나온 곽도찬 박사님 찾아보니 엄청난 분이시네..
That’s so cool
Great video!
NASA, with all that technology the sound on this video couldn't be normalized?
Beautiful...
My gosh. Unsteady NS in 1989?!
thanks for the asmr content
incredible documentary!
The heart is like a car engine without a spark plug
i mean that's completely untrue but like whatever
@@Danuxsy its got pistons, and so does an engine
@@olympus_drops5500 no it doesn't, it has chambers, the heart squeezes the blood out. You are wrong just admit it.
@@Danuxsy Engine does the same thing, it has chambers and it squeezes (compresses) the fuel. Excluding the sparkplug.
working fine thank you
Why is no one talking about the music on the video here? I really wish I could present CFD results with a soundtrack like this. My next goal is to convince my colleagues and supervisor this is a good idea. One can only dream...
As long as you don't have talking on one side and music on the other, like it is here.
what was the spec of the two machines running CFD at that time? I heard the name but didn't have any luck finding their references.
Cray-1 and Cray-2
I'll do a quick comparison to put things in perspective
The Cray-2: A Raspberry Pi 4:
190 billion operations per second 268 billion operations per second
4 cores 4 cores
cost $32 million costs $100-$200
@@Hyraethian Thanks for that comparison ! I have had an interest in CFD since the mid 1980s.
However, I am a visual learner, and also short on money.
I have seen a CFD program called OpenFOAM that is free, but still seems to require a knowledge of writing code and such.
Is there now an open source ( or low cost ) CFD program / (" app" ? ) that has an easy to understand GUI ? ( visual tabs and buttons, rather than blocks of text to input )
I have built 3D models in Softmage XSI, so anything up to that level of learning difficulty is what I am after.
Years ago, Autodesk released a CFD program called "Project Falcon"
It was very basic, but got me super excited, because I was even able to run it on a first gen Surface Pro tablet.
It gave info such as cD and lift, and all within less than an hour of render time.
It seemed to be geared towards kids, since it was extremely basic and had silly things you could run tests on, like a sci fi spaceship and even a oair of sunglasses.
It seemed created for touch control as well, ( and worked great on the Surface Pro touch screen ) I had high hopes that a phone app would be created and that it would be offered to the masses for cheap.
However, that was all dashed when Autodesk killed it and made it Autodesk Flowworks ( or some other name like that ) and now markets it for thousands of dollars.
@@Hyraethian how did you find out the specification of the Cray-2? is there a formal reference I can go to?
@@DennisZIyanChen I just pulled stuff from Wikipedia. I tried to read through the PDF of the CRAY-2 Computer Systems Functional Description. But it was too dense for me and I just decided to quote the figures other people had worked out.
now this sounds like vapor kong
So cool.
Oh I thought it was the hdtv review guy
I'm amazed how photoshop has evolved enormously since this
I'm afraid of that picture... from 1989
Время, когда компьютеры использовали не только для VK и FB.
то время, когда инженеры понимали как ставить компьютеру задачи руками
многие ученые не знают их кода. Эта волшебная коробка. извините, я говорю немного по-русски! я не ум американец
I love the music at the start, who is the singer
Dr. Tirta
8:28 윤석권 박사님, 11:24 곽두찬 박사님 두분이나 한국이름 이시네요!!
Hey you guys have my pictures from what was it June 4th 1989 that's not fair
Can anyone help me figure out what do you call the introduction music to this video. That kind of music was ubiquitous in these kinds of documentaries. thank you
0:43 amog
もう、何の記事だったかは覚えていないけど。コンピューターは7色あれば十分って言っていたどこかの教授がいたな。
Fan of nastran solver
2023 ,,,
yes
Would love to see the code for this
they did all this on computers that are comparable to nintendo 3ds
I don't hear jack. Must be a left ear exclusive.
was this classified back then ?
오홍...
recomendations
?
2:44 F35B prototype????
dr. Tirta ?
22:22 a ps5 is around 10 teraflops just fyi
iPhone 14 pro is around 1000x faster than that Cray 2. Pretty crazy
whats crazy is that they effectively combined thousands of PCs into one simulation database from MSFS 2020. A hivemind of computing. There is actual fluid dynamics so the military, aviation businesses will be able to use the flight data which has accurate simulation variables.
선생님, 회사를 만들었으면 엄청난 부 자가 되실수 있는 건였는데
Is my iphone a cray 2 computer?
The fluid dynamics i use are non-computational.
I am responsible for this software
Network ideas...
Karen worked 34 yrs 6 mos at NASA.
I know what happened lol oh boy do I got a story to tell
석환이 몇살이누
dr tirta
Cool but the format is very dry which is a given for the time period this was made in. Great information, we have progressed much and still rely on a ton of simulation data which has become highly advanced since then.
yeah If I lived during that time I would have hanged myself
😚
8:25
뭐야. 한국인이야?
이게 왜 추천?
NASA 😂😂😂😂😂😂😂😂😂😂😂😂😂😂😂😂😂😂😂😂😂😂😂😂😂😂😂😂😂😂😂😂😂😂
Meanwhile In africa, Stone Spear is invented
Oh.
back when NASA wasn't a total meme
but can it run minecraft?
sorry, had to, it's the internet
But can it run crysis*