NVIDIA Ampere A100 GPU Specs, Ray Tracing, $200,000 DGX-3, & EGX | GTC 2020 Keynote Recap

Gamers Nexus

203 E megtekintés33

    NVIDIA today announced its Ampere A100 GPU & the new Ampere architecture at GTC 2020, but it also talked RTX, DLSS, DGX, EGX solution for factory automation, and more.
    Sponsor: CDPR's Cyberpunk 2077 Case Mod Contest (geni.us/ws5uKk)
    Show notes/article: www.gamersnexus.net/news-pc/3584-nvidia-gtc-2020-keynote-recap-ampere-a100-specs-ray-tracing-dgx
    NVIDIA's GTC 2020 event was originally canceled, but they decided to host it anyway -- from NVIDIA CEO Jensen Huang's kitchen. The Ampere architecture was NVIDIA's major news item, with additional coverage of the Mellanox inclusion (after its acquisition), NVLink at 600GB/s, and more. We're recapping the NVIDIA GTC 2020 keynote to get through the core items efficiently.
    00:00 - Recap of NVIDIA GTC Announcements
    02:50 - NVIDIA RTX News in GTC 2020 Keynote
    04:55 - NVIDIA Omniverse RTX 8000 Solution
    05:23 - NVIDIA Ampere GPU Specs & Instanced GPUs
    10:15 - AMD Epyc Rome in NVIDIA DGX-3 A100
    13:50 - NVIDIA EGX for Factory Automation
    The best way to support our work is through our store: store.gamersnexus.net/

    Like our content? Please consider becoming our Patron to support us: www.patreon.com/gamersnexus

    ** Please like, comment, and subscribe for more! **

    Links to Amazon and Newegg are typically monetized on our channel (affiliate links) and may return a commission of sales to us from the retailer. This is unrelated to the product manufacturer. Any advertisements or sponsorships are disclosed within the video ("this video is brought to you by") and above the fold in the description. We do not ever produce paid content or "sponsored content" (meaning that the content is our idea and is not funded externally aside from whatever ad placement is in the beginning) and we do not ever charge manufacturers for coverage.

    Follow us in these locations for more gaming and hardware updates:

    t: gamersnexus
    f: gamersnexus
    w: www.gamersnexus.net/
    Editorial, Host, Video: Steve Burke

    Közzététel: 6 hónapja


    1. Gamers Nexus

      We uploaded a recap of our 5.45GHz R3 3100 stream yesterday: hugets.info/show/r5aKabG3zIiCznY/vide.html Also learn at what point the AMD Ryzen 3 3300X bottlenecks GPUs, incl. 2080 Ti, 2060, and more: hugets.info/show/tKFjhK3MuIepo2k/vide.html TIMESTAMPS 00:00 - Recap of NVIDIA GTC Announcements 02:50 - NVIDIA RTX News in GTC 2020 Keynote 04:55 - NVIDIA Omniverse RTX 8000 Solution 05:23 - NVIDIA Ampere GPU Specs & Instanced GPUs 10:15 - AMD Epyc Rome in NVIDIA DGX-3 A100 13:50 - NVIDIA EGX for Factory Automation Show notes of NVIDIA announcement: Show notes/article: www.gamersnexus.net/news-pc/3584-nvidia-gtc-2020-keynote-recap-ampere-a100-specs-ray-tracing-dgx

      1. Rudi Ansah

        Good disk cover system data computer 👍ok

      2. De ViceCrimsin

        Ya, J took a couple of notes. And by couple I mean a few. And a few I mean he made Linus proud with that flawless execution. Shit. I want to buy one with my broke-ass and I don't even know what it's for or how to use it. So if that's how I feel ... I can only imagine the reaction of his target audience.

      3. Turtle_Online

        Solid coverage as always, I'd be a lot less informed without GN. Thank you! At about 8:04 you talked about scaling up vs scaling out and said it was outside of your coverage area. Not sure if this comes across as rude but I figure it might be worth explaining it since it's super simple concept to understand. Scaling out means you will spin more instances of a server, this is used for elastic computing, which is the ability to increase and decrease the amount of servers you're running on the fly based on current workload, this can drastically reduce operating costs but an application has to be designed with this in mind. Scaling up means you will be upgrading the existing hardware of a server, the major difference is that this cannot be done on the fly. In order to upgrade without a disruption of service requires a second server for high availability because the server has to come down in order to upgrade.

      4. Rothschild killed JFK

        Lucky you Coronavirus doesn't affect your hair style, me now i look like Mel Gibson in Lethal Weapon.

      5. LB

        UE5, lumen and nanite makes RTX useless. Nvidia primary announcement is only a 400watt, 19 teraflop GPU when AMDs MI100 GPU is already expected to be a 200watt 25 teraflop GPU. Nvidias A100 is only note worthy at tensor workloads.

    2. Rafa RS

      No way I'm getting rid of my GT630

    3. Total GAMIX

      Bet this card can render a movie in a week

    4. Monkey D. Luffy

      The real question can you game on it

    5. Rudi Ansah


    6. swimler

      does it play minecraft?

    7. Cody Gibson

      I wonder if they will do a DGX A200 with 16 of these things. Like going from the 1 to 2.

    8. Dyortos

      If it was Intel and not amd first off then we'd be saying the same thing about Intel towards amd.. Just be happy you'll have both options in the future.

    9. John D.

      Why did he film it in his kitchen and why was it in the oven? I've been interested in the news reports but don't know much about chips but I do understand the numbers involved so I know it's important.

    10. Timothy

      Hard to listen to this guy.

    11. Jacob Morris

      can i use it for gaming

    12. Papi Simba

      Ok but, can it run Crysis Remastered?

    13. Soohoran

      10:55 What does he do with all that scrapers?

    14. Mugen

      can it run fortnite?

    15. Mohamed Osama

      I loved the whole video. Thank you gamers nexus. This is always better than Linus' normie content

    16. Ted McFadden

      You just know, behind the scenes, some engineer was sweating bullets while Mr. Huang was handling that assembly with oven mitts. ESD, strain relief, cold joints... what could POSSIBLY go wrong?

    17. cfb36

      If you listen closely you can hear the CCP climax at the possibilities in the background.

    18. noko59

      When? Date release?

    19. Sky Nabess

      oh man, that oven gag tho

    20. ColdieHU

      RTX cards are still dying rather fast. Had 2 brand new ones dying on me in the timeframe of 6 weeks.

    21. Joel Doxtator

      Can't wait to find A100 GPU's in the E-waste bins 20 years from now.

    22. Karthik Marath

      Accurate price 😂😢

    23. Gondo

      Would be nice if AMD can get there drivers and shit together and give Nvidia a run like they are doing with Intel. Win win for all of us if that happened

    24. Sovsefanden

      i hope rdna2 is not garbage ... but i will go with 3080 or 3070 myself.

    25. herewestand

      200k a pop, cloud companies can attach the gpu compute to vm's, then sell the usage to customers. probably get the cost of each very quickly lol

    26. Ace of Hearts

      AMD's CPU is so good even Nvidia has to use it lel. Damn that price. I remember when SGI was selling these super expensive GPU systems

    27. Leo Comerford

      When a man pays for a solid marble ventilator hood with elaborate detailing and a relief carving, d*mned if he's going to pass up any chance to show it to the world. (A relief carving directly behind a gas hob though? I'd hate to be the poor cleaner who gets to degrease that with washing-up liquid and a set of toothbrushes.)

    28. Iam_Dunn

      I personally welcome our AI overlords and hope they inherit an Earth that will sustain them for Billions of years. :)

    29. matthew bailey

      What's it like trying to maintain the A100? Oh geez...

    30. tudomerda

      Nvidia is merely "amping" AMD with the kitchen stunt, they're saying to AMD "anything you do we can do better", I have a message for Nividia, "if AMD releases a card which can do 4k 120Hz at reasonable temperatures ( < noise) I'm going with team red irrespective of what you release Nvidia, I'm tired of your incessant price gouging". I couldn't care less if Nvidia releases a card which is +10% > AMD's for a zillion dollars, shove it Nvidia.

    31. llamapi3

      DLSS 2.0 is really good.

    32. Kylo Ren

      Could GA-100 graphics card run Cyberpunk 2077?

    33. KuroBushi

      I wish your videos was zoom to fill friendly like Linus's videos 😟 my note 9 wants to use the full 6.4 inch screen man and especially since I don't have no notch or something 😂

    34. Matt King

      Scale up = More Performance from 1 instance. Scale out = Increasing instances - Also, not sure if I heard you mention it, Steve, but they are also using AMD EPYC "Rome" CPUs with the 64 cores (I think).

    35. Sina Madani

      Interesting facts regarding the BMW partnership

    36. Dragon Offski

      Volta, Ampere...Please tell me the next architecture it going to be called Wattnow?

    37. Mr.GamingMonk

      0:22 *HOLY SHIT... LOOK AT THE SIZE OF THAT CRYSTAL?!?!?!?!?* 😲

    38. DD

      All the OG Intel's founders have either left or passed away and look what happens to Intel today. Nvidia is Jensen's baby - As Apple is Steve's baby - the difference of having a Founder/Owner running the company vs a bean counter/stockholder is NIGHT & DAY.

    39. Marc Wolfe

      The coolers on the DGX GPUs get me aroused lol

    40. SpacebarWASD

      What happened to their self driving project? It was so hyped before, and now they stopped talking about it...

    41. Hayder Abdulridha

      So should I buy the 2070s or should I wait?

    42. waspoza

      0:46 wow, good Polish pronunciation of "cdprojekt". tip_hat.gif

    43. eric peterson

      Did anyone else hear him say CD Proyaks Red?

    44. David

      Please learn to say Ampere

    45. Davey

      Can I get a card without RTX so I dont spend money on something I wont ever use?


      But can it run crysis?

      1. Iam_Dunn

        GENIUS LOSER ....well the second part of your user name seems appropriate.... Can we let the 10yr old meme die peacefully?

    47. Sam Brandt

      just in time manufacturing is just a buzz word isn't it?

    48. nes flo

      Why you show us ground truth for dlss 2.0? You should cut 720p dlss 1.0 and 1080p dlss 2.0 frames from GTC video!!

    49. Michael Springstroh

      Not sure how I missed this during the announcement.... but why does Jensen Huang have a goldfish bowl full of silicone spatulas on his counter? There has to be close to 20.

    50. damysticalone87

      What we want are the Z490, X570 and the TRX40 fanless passively cooled AoRUS Xtreme boards, but with AMD's and these specs: - 10Gb Lan, - Wlan G6, - Pcie-4.0, 5.0, 6.0, - Usb-4.0, - Displayport 2.0 (for 16k), - Hdmi-2.2 / 3.0 (for 16k), - 7nm Cpu (x2+) Zen3, 5nm Zen4, - 7nm Gpu (x2+), - Ddr5 Ram.

    51. damysticalone87

      More importantly and the most important 4sure right here: Check that out, sort by "Newest first": hugets.infodiscussion Attention! It looks like the HUgets-App doesn't show the right tab, open the link with a browser like Chrome!

    52. JayJapanB

      Jensen reballing his Ampere gpus.

    53. Pezzy Da Franchize

      i don't know what he's talking about.. but I like it #subscribe


      Anyone desperately buying RTX cards for ray tracing all i gotta say is wait 2 years and u will recieve what u are expecting :).

    55. TechFan

      Sounds like the GPU pc will address the cards in containers and run individual processes or custom parallelism and put of band processes where to cards could contact each other with custom software

    56. Belfor09

      You know... It's hard to screw 1080p native image more than what they showed in the "demo". It was blurred so much by the TAA, that DLSS looked mighty impressive in comparison. I'm not telling that DLSS is crap. It's not crap. It is impressive. But comparing it to the blurry outcome of a terrible TAA implementation is just bad.

    57. 29110sc

      where is the 3080

    58. Harry M

      Actually, FP64 performance is 9 TFLOPS in A100, not 20. This is an incemental improvement over 7 TFLOPS with V100. What they show in the graph is FP64 TENSOR performance, which is 20 TFLOPS, if you change your code to take advantage of it. Scientists that do matrix multiplications can indeed take advantage of this in CUDA 11 with special libraries.

    59. davidbrennan5

      I think it would be cool to show older computers/gaming computers in your backdrop on videos.

    60. Vyshu Siroi

      But will it run crisis at 60?

    61. Raul S

      Size of a100 IS 825mm2 , even on 7nm . Wow. Wonder a102 that will power rtx3080ti be smaller or same size , if it is , new card will cost at least $2000.00 .

    62. Tommy2Gunz

      Lol, passing judgement on Nvidia's poor production value with spliced audio as you are reading your scripts from papers on your table...

    63. Jay Stannard

      I saw a bunch of nVidia cards in the back of that Jensen presentation. What an nVidia shill, where's the objectivity?


      Has to wear that Leather jacket at home, What a pogchamp!

    65. Sir Monstras

      First of I really enjoy this channel, but I can't be the only one that puts these video's on 1.25 speed, because that makes his voice and speech so much easier to listen to.

    66. Richard Wilkinson

      Will the new DLSS 2.0 work on first gen RTX cards?

      1. FBC Aware

        It already does.


      @4:17 But... TO BE FAIR, the horse has to come before the cart: You don't invest into implementing technologies in your games for which there is no hardware to drive it. The dumping on RTX the reviewer class [everyone] gave, feeding the AMD fans red meat to claim it was a fraud in some way, or a re-branded Pascal, was wayyyy disproportional. It was exactly what it was expected to be: a 1st gen implementation of new cutting edge techs!

    68. Adam W

      Where's the segment tracker usually on the left? I like that in GN news videos. It makes it easy to go back and watch a section later again.

    69. Mohammed Almansour

      Wow! How many spatulas are in that bowl???

    70. Space Ghost

      Can I fold on it?

      1. Iam_Dunn

        mophomanners ..Yup, it will even iron it all first.... LOL :)

    71. Abdullah Almaosharji

      The funny thing is that these products will probably have more effect on our daily lives as consumers than 'consumer' targeted GPUs.

    72. Georg F

      Hey! Huang managed to use a proper camera in his home keynote. Unlike the Xbox crew who filmed on a set of potatoes.

    73. Bionic0ne

      NVIDIA TURNED OFF COMMENTS ON THEIR RTX VIDEO LOL. They know RTX is about to be as dead as PhysX as soon as Windows drops their own and they don't want people pointing it out.

    74. LegendarySkypenis

      Please benchmark your cat.

    75. StopaskingformynameYouTube

      That GPU-stack thingy looks really REALLY cool!

    76. attq

      Hey, Steve, could you put the "waifu" GPU somewhere in the scene?

    77. Anthony Quinn

      So we just gonna ignore that Jensen Huang has like 85 spatulas?

    78. draeh

      6:15 The 1M drill holes isn't all about mounting. Don't forget PCB VIAs which are probably the bulk of all the drill holes.


      even 2080ti is old and outdated and uses more than 300w -_- for the low price of 1100

    80. babis8142

      I can't get over the pronunciation of projekt red lol; it might be realistic but damn I've been saying it the english way for years (or decades)

    81. Justa Name

      I was sceptical about RTX, because I had to pay a thousand bucks for my 2080 😪

    82. Simon West

      He didnt do another retarded 'buy more, save more' bit?

    83. Joe G.

      Yes... that is all well and good, but will it run Crysis?

    84. S B

      "We know there's a few of you" Thanks Steve

    85. Daniel Hetorilla

      oh no Steve@Gamers Nexus didn't expect for you to get caught regurgitating marketing hype. I personally can't see how you'd even paint a car every 56 seconds let alone make an entire one. Someone looked at the wiki article and did some very creative math. Just because a factory makes 1,500 cars a day doesn't really mean a car is made in 56 seconds.

    86. landoishisname

      I thought maybe they generated jensen's voice with AI during the non-live segments

    87. woolfel

      The new tensor float and sparse network optimization is freakin awesome. When mixed precision and bfloat was released, I thought tensorflow and pytorch would see a big speed up, but that didn't happen. It was kinda hard for the framework developers and the learning curve is STEEP. The new tensorfloat will make it so you don't need to be an elite CUDA + cuDNN developer to take advantage of it. The sparse network optimizer is huge for ML developer. Most NN models like faster-rcnn 150, ssd mobilenet and other large convolutional networks are sparse. That means the next version of DLSS will take less time to train and end up more efficient.

    88. Valfaun

      he didn't share the recipe for that freshly baked A100 board, i assume?

    89. anonymous

      i know this isnt for games but i would like to see how i performs in games for fun

    90. StrykerDelkar

      Waiting for Linus' video "10 Editors, 1 GPU".

    91. Just Jordan

      Damn does he always wear that Leather Jacket even when Cooking?

    92. Andre Pt

      When 3060 guys???

    93. Hugsy Malone

      What's insane is that the A100 with the largest ever 6,912 Cuda Cores, is still not the full GA100 chip. And it's rather amusing that Xeon has been dropped from the DGX in favor of Epyc. I imagine that contributed significantly to the performance density increase.

    94. John Doe

      Ah, Nividia's boss finally found a use for his kitchen which has never been used.

    95. Henrijs Zaķis

      Oh boy.... i will need to sell my kidney for sweet 3000 series gpu. Cant wait :)

    96. KoTu

      ngl I'm surprised every time you read CD Projekt like it is in Polish... (we read "j" like "y" in yes... letter "dż" is what you usually say as english "j")

    97. FalconEagle

      Thanks Steve needed a summary of Nvidia keynote.

    98. shane eslick

      G'day Steve, Just to add Fuel to the Performance Speculation leaks, I heard a Rumour this is the RTX3050 Founders Edition & it is going to MSRP $100 😁, On another thought, I still don't get why the new nVidia line has to be RTX3000 instead of RTX2100, I have been told it is because of: (a) Node Shrink: Happened between GTX200 & GTX1000 without the need for 1000 Generation Naming Jump, (B) Architecture Name change: Happened between GTX200 & GTX1000 without the need for 1000 Generation Naming Jump, How much did everyone complain about GTX1000 to GTX1600 instead of GTX1100 saying how Stupid the naming scheme was, the expectation was only a 100 Generation Naming Jump even though it had a Node Shrink & Architecture Name Change (Yet following nVidia's conventional naming scheme they would have got to GTX1600 anyway) Seems to me if nVidia does follow the jump to 3000 like the people speculating are calling it, in a 6-7 years they will get to the same position as Intel with the 10000 series naming everybody hates & is saying "Why use this Intel it is Stupid"

    99. Dan Clift

      time stamp 11.06 - how many spatulas does one man need ?

    100. Rabid Guinea Pig

      I have no damn clue what i just saw with that DGX A100 but dear lord how does one even design that. Imagine trouble shooting that massive thing