I am a game dev mate. Due to the stream processors now found on modern gpu's, They can be used for a LOT more than just graphics work. For example, The PhysX api can run on a modern 8 series gpu or above, Because of these stream processors and their programability. Methods that were once limited to applications such as maya and 3dsmax, Can now be done at a good 60FPS using modern GPU's. Thats why crysis looks so good. Also offline rendering is done on the cpu, not the gpu, So processor is key.
So basically, a CPU can do one thing at a time, compared to a GPU that can do hundreds of thinks at a time. I'm just guessing, cause the demonstration didn't really explain the difference very well, but it was soooooo cool. :P
CPUs actually can do several things at a time, that's why they are multi-core. But only in the order of tenths or hundreds. In turn, GPUs have thousands of cores
Some stuff can't be done parallel - for example. the calculation of pi. Each step needs the result of the step before it. Stuff like this would be EXREMELY slow on a GPU. (Unless you want to calculate it a thousand times - then GPU would probalby be done before the CPU, because it does those thousand calculations at once, and not after each other, like the CPU)
Same, I’d be bricked up as soon as I saw them. And then I would work sooo hard the rest of the day. Just solid blood flow to the main vein all class. Man that would be dope
I still remember when there was so much hype when new tech pieces (both hardware and software) were announced. Nowadays, it's just meh, not that exciting, and very obvious they're just trying to make more money, get richer, rather than help improve technology and help the human society.
How come CPUs still cannot render frames with ease? There is more then just 200pixels at a time. GPU can handle mutliple framebuffers (offline rendering), and can automatically add occulsion, culling, aa, af, textures. That cannot be easily handled by a CPU, on a limited bus, with limited parrall processing capability. Multiplied by millions of times over, the CPU just loses out. It cannot outdo the amout of raw processing the GPU was designed for.
There will always be a slight difference in the shots. BUT the human eye cannot see it normaly. It is also possible that the gas had less travel distance to the bottom tubes, and therefore shot them first. I would like to apologize for my crappy english, but i'm from norway. I hope this has Enlightened you, S1meN.
The GPU made it all happen in under 1 second genius. The CPU shot the paintballs out individually. The GPU calculated the air pressure, time, etc. so the paintballs would fire simultaneously and faster than any gun could fire them, without the whole thing exploding.
this is why 3D renders take so long, sadly. Nvidia came out with Some software that allows the GPU to do the actual rendering now in conjunction with AutoDesk's 3D studio
The video is slowed down so much that the eye would normally see it being fired simultaneously. If two people decided to fire there paintball guns at the same time, they would have the ability to fire really close together, but still it would not be at the exact same time. The rest of the explanation is under here.
The gpu can perform any highly parrallel mathematical computation, Not just graphics specific. So physics, AI flocking, Anything that just requires fast calculations rather than any conditional statements, Which is where the cpu comes in.
4:06 Adam: "We're gonna need - we realized - a bigger bot." Ppl: "Yeaaaahh" Jamie: "Yeah we need a bigger bot. Then I started to think we need a REALLY BIG bot." Ppl: "YYYYEEAAAAAAHHHHHHUUUUUYYYY!!!!11oneoneeleven" :D
Yes CPU is single thread but have more options than a GPU. A GPU only process graphics and is limited for that and all graphics derived works. But CPU can do a much more variety of works. In the representation you can see it can walk and aim in range, the other one is more heavy and only can aim parallel. In a real CPU and GPU the like happens.
I never said cpu's could render 3d frames with ease. All i was saying is that what nvidia showed here could easily have been done on the cpu, even in realtime. 1100 pixels is not a lot, especialy when ur not applying fragment shaders to each and every one, Comparing to say a resolution of 1280 x 1024, totaling 1310720 pixels, in SCREEN SPACE, then you have to account for all the fragment and vertex calculations being done in object space, Theres a big big diffrence.
as anyeos said GPU is for graphics and derivated work.. do you have any idea how long does it takes to render a 3d scene .....i tell you a lot ...from hours to days
Maybe fft would be another bad demonstration, But So is what they are doing here. I mean when u open a photograph in photoshop etc, Its not like you see the image scaling across the screen pixel by pixel. IF they really wanted to show parralel processing, They should have used the same idea, BUT rendering something on the canvas in realtime. Now that would be impressive!
Why in the ASSS would you need anything more than a simple CPU to do that? If you like told it what to paint in realtime, Then sure, But that is obviously pre-processed.
Thats funny cause the HDD is fast enough. Th RAM is just there so running app data can be accessed immediately. this is why 3D artists need a lot of it. Is because it is easier to access data that right in front of you. Th HDD is fast enough, but it is more efficient if you can access it in the RAM. also your HDD uses a series of data, while the RAM is parallel and different pieces of data can be accessed at once. this is why you can be talking to some one while rendering a 3D movie.
if it`s a very small scene or if you have a big render farm .... and about video games ..i think you are trying to say that in 3d video games everytime you enter that game the scene it`s rendered...if it`s that so you are wrong ... the scene it`s just loaded from pc , that`s why some games like gta 4 or sacred 2 are so big ....
Holy mother of failbots... The "Uber GPU" Has to fire all the "paintball guns" at once! Thatd be amazing. Only one thing... if you had 1100 paintball guns lined up, and had each tied to a piece of sting, yanked the string, you could paint the mona lisa too! Gpu has to do zilch but pull the string -.- Thats NOT parralel tasking or multitasking or w/e they want to call it. Thats something well frankly I think the closest thing youd need to a GPU is a pressure gauge, and air compressor/pump.
Think of the cpu as a college professor and the gpu as the students he teaches. The professor can solve problems the students simply can't, and can solve individual problems faster. But when it comes to more basic topics, a class of 100 students can solve a 100 question exam much faster than the professor ever could by giving 1 student per question. So it's that the gpu can do more basic operations at once, but the cpu solves more complex individual linear problems faster.
NVIDIA RULES and so does the myth busters put them together and you get something epic! GPUs can do more faster the a CPU keep CPUs for math leave graphics to NVIDIA!!
Liked it and there is a big difference in the CPU and GPU the CPU does all the operations and basic imaging, the GPU then takes that image and renders it making the image manytimes better, some grapgic cards completly skip the CPU anf do all the imaging itself allowing your computer to run at a higher level
since when did you get an education!? That is just amazing. I didn't think 2 yr olds could get educations. Technology has come a long way. And go ahead and laugh. Watch me care. I am SOOOOOOO offended. YAWN! NEXT!
it says that, in the gpu each pixels u see on your screen is processed in parallel to each other, compared to cpu each pixels needed to be processed one by one... making GPU process faster than CPU... but still, i dont think the comparison is good enough though... however, that's the simplest way to explain... or at least what i think...
I am a game dev mate. Due to the stream processors now found on modern gpu's, They can be used for a LOT more than just graphics work. For example, The PhysX api can run on a modern 8 series gpu or above, Because of these stream processors and their programability. Methods that were once limited to applications such as maya and 3dsmax, Can now be done at a good 60FPS using modern GPU's. Thats why crysis looks so good. Also offline rendering is done on the cpu, not the gpu, So processor is key.
So basically, a CPU can do one thing at a time, compared to a GPU that can do hundreds of thinks at a time. I'm just guessing, cause the demonstration didn't really explain the difference very well, but it was soooooo cool. :P
CPUs actually can do several things at a time, that's why they are multi-core. But only in the order of tenths or hundreds. In turn, GPUs have thousands of cores
My GTX280 upgrade from my really cheap card felt to about this comparison. It's that good. OC's impressive on air
I'm from the future. RTX 2080 isn't that good.
@@freshbakedclips4659 LOL! Loving my 3080 though, we've come a loooong way.
well that is not going to happen again, a 4090 has 16'000 cores now. not enough paintball tubes to fit on stage for that.
Some stuff can't be done parallel - for example. the calculation of pi. Each step needs the result of the step before it. Stuff like this would be EXREMELY slow on a GPU. (Unless you want to calculate it a thousand times - then GPU would probalby be done before the CPU, because it does those thousand calculations at once, and not after each other, like the CPU)
best comment 11 years later , nowadays that everything is machine learning , deep and so on , GPUs are way better for that but in case of PDEs depends
If I would have those guys as teachers, I would go to school with a big smile everyday
Same, I’d be bricked up as soon as I saw them. And then I would work sooo hard the rest of the day. Just solid blood flow to the main vein all class. Man that would be dope
Looking back from the year 2022, audience vibes here were very distinctly 2008 lmao. I just don't see this specific flavour of enthusiasm anymore.
I still remember when there was so much hype when new tech pieces (both hardware and software) were announced. Nowadays, it's just meh, not that exciting, and very obvious they're just trying to make more money, get richer, rather than help improve technology and help the human society.
That's pretty cool to watch at the end where it plays in slow motion
I'm from the future, RTX 2080 isn't that good compared to RTX 3090.
I like when he mentions robots you hear the crowd cheer but one over the rest you can clearly hear him say, "YEA! ROBOTS!"
I WANNA SEE THEM DO THIS WITH 1920x1080 PAINTBALLS!!!!
At 60 repaintings per second.
@@Aranimda wwith 4 billion different colours
Adam and Jamie are awesome.
You my friend, just made my night, in the least expected way, I couldn't even describe to you.
if the ball color was better the final picture could be even better
Best. Product. Demo. EVAH!!!
I'm detecting that lag may be an issue with this setup.
+19Luke89 Needs more Titans
Yes a GPU can be really powerful, but a GPU can't do everything by itself, the CPU helps it render more than you think.
They invited all of the Intel graphics people to this conference to sabotage their employees
This is so outrageously inaccurate
I luv mythbusters and this vid!
How come CPUs still cannot render frames with ease? There is more then just 200pixels at a time.
GPU can handle mutliple framebuffers (offline rendering), and can automatically add occulsion, culling, aa, af, textures.
That cannot be easily handled by a CPU, on a limited bus, with limited parrall processing capability. Multiplied by millions of times over, the CPU just loses out. It cannot outdo the amout of raw processing the GPU was designed for.
There will always be a slight difference in the shots. BUT the human eye cannot see it normaly. It is also possible that the gas had less travel distance to the bottom tubes, and therefore shot them first.
I would like to apologize for my crappy english, but i'm from norway. I hope this has Enlightened you, S1meN.
Fun fact - NVIDIA stock was at $0.18 when they paid mythbusters for this demonstration. Today it’s valued over $100 for a 555X
The GPU made it all happen in under 1 second genius.
The CPU shot the paintballs out individually.
The GPU calculated the air pressure, time, etc. so the paintballs would fire simultaneously and faster than any gun could fire them, without the whole thing exploding.
this is why 3D renders take so long, sadly. Nvidia came out with Some software that allows the GPU to do the actual rendering now in conjunction with AutoDesk's 3D studio
Very true, paintball will never be the same again ^^
the slow motion looks awesome xP
The video is slowed down so much that the eye would normally see it being fired simultaneously. If two people decided to fire there paintball guns at the same time, they would have the ability to fire really close together, but still it would not be at the exact same time. The rest of the explanation is under here.
check out that slow motion at the end
The gpu can perform any highly parrallel mathematical computation, Not just graphics specific. So physics, AI flocking, Anything that just requires fast calculations rather than any conditional statements, Which is where the cpu comes in.
8:17 cue Feel Good Inc - Gorillaz
... that was damn awesome...
that was indeed the concept of GPU's...
Aaah ok it is like some "co-processor floating point unit"? :P
Not bad but I have an old GPU that is not util for me.
Sorry about the mix up in my comments......
That was awesome!!!
Very creative comparison.
I wonder how many frames per second you can get on that? :P
I would've laughed so hard if there was a barrell break or chop xD
Nvidia reuploaded this with comments turned off
8bit mona lisa! :)
the mona lisa pic. looks kinda like the old mario graphics
4:06
Adam: "We're gonna need - we realized - a bigger bot."
Ppl: "Yeaaaahh"
Jamie: "Yeah we need a bigger bot. Then I started to think we need a REALLY BIG bot."
Ppl: "YYYYEEAAAAAAHHHHHHUUUUUYYYY!!!!11oneoneeleven"
:D
Yes CPU is single thread but have more options than a GPU. A GPU only process graphics and is limited for that and all graphics derived works. But CPU can do a much more variety of works. In the representation you can see it can walk and aim in range, the other one is more heavy and only can aim parallel. In a real CPU and GPU the like happens.
DOOD slo mo is soooooo SICKKKK
If only i bought nvda stock back then
if every barrel had a machine gun mechanism you could get at least ten. You would also need to work ou t a way of replacing the backdrop
where to get original video???
if you know ....TELL ME
writen by Adam Savgae
writen
>writen
I never said cpu's could render 3d frames with ease. All i was saying is that what nvidia showed here could easily have been done on the cpu, even in realtime. 1100 pixels is not a lot, especialy when ur not applying fragment shaders to each and every one, Comparing to say a resolution of 1280 x 1024, totaling 1310720 pixels, in SCREEN SPACE, then you have to account for all the fragment and vertex calculations being done in object space, Theres a big big diffrence.
as anyeos said GPU is for graphics and derivated work.. do you have any idea how long does it takes to render a 3d scene .....i tell you a lot ...from hours to days
The Mythbusters are great! BUT....I was waiting for them to blow the canvas up...oh well
"I LOVE FREE STUFF!!!!!!"
Maybe fft would be another bad demonstration, But So is what they are doing here. I mean when u open a photograph in photoshop etc, Its not like you see the image scaling across the screen pixel by pixel. IF they really wanted to show parralel processing, They should have used the same idea, BUT rendering something on the canvas in realtime. Now that would be impressive!
Why in the ASSS would you need anything more than a simple CPU to do that? If you like told it what to paint in realtime, Then sure, But that is obviously pre-processed.
woooow!
i still dont get it?
like kcowneduall said?
it just shoots paintballs faster?
Might be done if we use MISD (Multiple Instructions Single Data)
Thats funny cause the HDD is fast enough. Th RAM is just there so running app data can be accessed immediately. this is why 3D artists need a lot of it. Is because it is easier to access data that right in front of you. Th HDD is fast enough, but it is more efficient if you can access it in the RAM. also your HDD uses a series of data, while the RAM is parallel and different pieces of data can be accessed at once. this is why you can be talking to some one while rendering a 3D movie.
God Job Guys
That was awesome!
if it`s a very small scene or if you have a big render farm .... and about video games ..i think you are trying to say that in 3d video games everytime you enter that game the scene it`s rendered...if it`s that so you are wrong ... the scene it`s just loaded from pc , that`s why some games like gta 4 or sacred 2 are so big ....
Can you make another planet like that ?
did the mythbusters acc build that thing? wow
And the slow motion shows the diffirent packages of balls arriving with delay, microshutter.
Does this mean I can take my 260GTX+ to the local field with me and not suck anymore?
That was awesome.
lol.. "im a little hard of hearing.."
funny thing is he actually is ^_^
Ladies and gentlemen; science class is now over!
i don't get it. how does this compare to an actual gpu?
GPU my ass
it was just a giant paintball gun aimed at paper with all the paintballs PRESET in the barrels
"- good job - nice work ":D
Holy mother of failbots...
The "Uber GPU" Has to fire all the "paintball guns" at once! Thatd be amazing.
Only one thing... if you had 1100 paintball guns lined up, and had each tied to a piece of sting, yanked the string, you could paint the mona lisa too!
Gpu has to do zilch but pull the string -.-
Thats NOT parralel tasking or multitasking or w/e they want to call it. Thats something well frankly I think the closest thing youd need to a GPU is a pressure gauge, and air compressor/pump.
AWESOME!!!!!
Could some one explain why we don't use Parallel Processing for everything.
Think of the cpu as a college professor and the gpu as the students he teaches. The professor can solve problems the students simply can't, and can solve individual problems faster. But when it comes to more basic topics, a class of 100 students can solve a 100 question exam much faster than the professor ever could by giving 1 student per question.
So it's that the gpu can do more basic operations at once, but the cpu solves more complex individual linear problems faster.
awesome!
you break it, you buy it
NVIDIA RULES and so does the myth busters put them together and you get something epic! GPUs can do more faster the a CPU keep CPUs for math leave graphics to NVIDIA!!
Have u seen their Quadro FX 5800 card? Its a 4 gb monster!
but,
Then going to ALL my vids and posting stuff there too. That is just creepy.
Liked it and there is a big difference in the CPU and GPU
the CPU does all the operations and basic imaging, the GPU then takes that image and renders it making the image manytimes better, some grapgic cards completly skip the CPU anf do all the imaging itself allowing your computer to run at a higher level
these science stuff is cool
cool stuff.
that was great!
why not hd?
Wish PS4 presentation had sth like this ^^
because it looks cool.
how do u know?
8:49 = fuckin epic :)
Indeed
since when did you get an education!? That is just amazing. I didn't think 2 yr olds could get educations. Technology has come a long way.
And go ahead and laugh. Watch me care. I am SOOOOOOO offended.
YAWN! NEXT!
pointless how? its a good demonstration..
EPIC
holy shittt!
lol
the guy hollored
I WANT FREE STUFF
Imagina fazer isso com as novas rtx
That's fucking awesome! :P
Okay so this is what caused nvidia to lose funds for the fermi project? LOL
it says that, in the gpu each pixels u see on your screen is processed in parallel to each other, compared to cpu each pixels needed to be processed one by one... making GPU process faster than CPU...
but still, i dont think the comparison is good enough though... however, that's the simplest way to explain... or at least what i think...
my nvidia just broke.... Now I can't play on my computer...
wow it's awesome!!!!
OpenCL FTW!
@zlazher because some people are too stupid to get it by words, and this makes lots of money
@cneyr
thanks !