Name: 英偉達 2025 年 Computex 主題演講：19 分鐘內公佈所有資訊 (Nvidia's 2025 Computex Keynote: Everything Announced in 19 Minutes)
Uploaded: 2025-05-19T14:13:02.000Z
Duration: 19 min 54 s

So one of the most important industries, of course, is telecommunications.

是以，最重要的行業之一當然是電信業。

程度副詞

Just as the world's cloud data centers have now become software defined, it stands to reason that telecommunication should also be software defined.

正如世界雲數據中心現在已成為軟件定義的數據中心一樣，電信也理應成為軟件定義的數據中心。

And so that's the reason why we've taken now some six years to refine and optimize a fully accelerated radio access network, RAN stack, that does incredible performance.

是以，我們花了大約六年的時間來完善和優化完全加速的無線接入網絡（RAN 棧），使其具有令人難以置信的性能。

For data rate per megawatt, or data rate per watt, we are now on par with the state-of-the-art ASICs.

就每兆瓦數據傳輸率或每瓦特數據傳輸率而言，我們現在與最先進的 ASIC 不相上下。

And so once we could do that, once we could achieve that level of performance and functionality, then after that we can layer on top AI.

是以，一旦我們能做到這一點，一旦我們能達到這樣的性能和功能水準，之後我們就可以在人工智能的基礎上進行分層。

是以，我們在這裡有很好的合作伙伴。

You could see SoftBank and T-Mobile, Indosat and Vodafone are doing trials.

你可以看到軟銀和 T-Mobile、Indosat 和沃達豐正在進行試驗。

Nokia, Samsung, Kyocera are working with us on the full stack.

諾基亞、三星、京瓷都在與我們合作開發全套產品。

Fujitsu and Cisco are working on the systems.

富士通和思科正在開發這些系統。

And so now we have the ability to introduce the idea of AI on 5G or AI on 6G, along with AI on computing.

是以，現在我們有能力引入 5G 人工智能或 6G 人工智能以及計算人工智能的理念。

我們正在通過量子計算實現這一點。

Quantum computing is still at the noisy intermediate state, intermediate scale quantum called NISQ.

量子計算仍處於嘈雜的中間狀態，中間尺度量子被稱為 NISQ。

However, there are many, many good applications we could already start to do.

不過，我們已經可以開始做很多很多很好的應用。

是以，我們對此感到非常興奮。

We're working on a quantum classical or quantum GPU computing platform.

我們正在開發一個量子經典或量子 GPU 計算平臺。

We call it CUDA-Q, working with amazing companies around the world.

我們稱其為 CUDA-Q，與世界各地的優秀公司合作。

And so GPUs could be used for pre-processing and post-processing, for error correction, for control.

是以，GPU 可用於預處理和後處理、糾錯和控制。

And so in the future, I predict that all supercomputers will have quantum accelerators.

是以，我預測未來所有的超級計算機都將擁有量子加速器。

And so a supercomputer would be a QPU with QPUs and GPUs and some CPUs.

是以，超級計算機將是一個帶有 QPU、GPU 和一些 CPU 的 QPU。

And that would be the representation of a modern computer.

這就是現代計算機的表現形式。

In the last several years, we've been working on a new computer to make it possible for us to do inference time scaling, or basically thinking incredibly fast.

在過去的幾年裡，我們一直在研究一種新的計算機，使我們能夠進行推理時間擴展，或者說基本上是以令人難以置信的速度進行思考。

Because when you think, you're generating a lot of tokens in your head, if you will.

因為當你思考的時候，你的腦子裡會產生很多代幣，如果你願意的話。

And you iterate in your brain before you produce the answer.

在得出答案之前，你要在大腦中反覆推敲。

So what used to be one-shot AI is now going to be thinking AI, reasoning AI, inference time scaling AI.

是以，過去的一次性人工智能現在將成為思考型人工智能、推理型人工智能、推理時間擴展型人工智能。

And that's going to take a lot more computation.

And so we created a new system called Grace Blackwell.

是以，我們創建了一個名為格蕾絲-布萊克韋爾的新系統。

格蕾絲-布萊克韋爾會做幾件事。

Scale up means to turn what is a computer into a giant computer.

擴大規模是指將計算機變成巨型計算機。

Scale out is to take a computer and connect many of them together and let the work be done in many different computers.

所謂 "擴展"，就是將一臺計算機連接多臺計算機，讓工作在多臺不同的計算機上完成。

Building larger computers that is beyond the limits of semiconductor physics is insanely hard.

建造超越半導體物理學極限的大型計算機難上加難。

格蕾絲-布萊克韋爾就是這樣做的。

Grace Blackwell broke just about everything.

格蕾絲-布萊克韋爾幾乎打破了一切。

And all of you in the audience, many of you are partnering with us to build Grace Blackwell systems.

在座的各位，你們中的許多人都在與我們合作建立格蕾絲-布萊克韋爾系統。

I'm so happy to say that we're in full production.

我很高興地說，我們正在全力製作。

But I am also, we can also say it was incredibly challenging.

但是，我也可以說，我們也可以說，這是一個令人難以置信的挑戰。

Although the Blackwell systems based on HGX has been in full production since the end of last year and has been available since February, we are now just putting online all the Grace Blackwell systems.

儘管基於 HGX 的布萊克韋爾系統自去年年底就已全面投產，並已於今年 2 月開始供應，但我們現在才剛剛將所有格雷斯布萊克韋爾系統投入使用。

They're coming online all over the place every single day.

It's available in CoreWeave now for several weeks.

現在，CoreWeave 中已經提供了幾周的時間。

And now you're starting to see it coming up from everywhere.

現在，你開始看到它從各處湧現。

Everybody's starting to tweet out that Grace Blackwell is in full production.

每個人都開始在推特上說 Grace Blackwell 正在全力製作。

In Q3 of this year, just as I promised, every single year, we will increase the performance of our platform every single year like rhythm.

今年第三季度，正如我所承諾的那樣，每一年，我們都會有節奏地提高平臺性能。

And this year, in Q3, we'll upgrade to Grace Blackwell GB300.

今年第三季度，我們將升級到 Grace Blackwell GB300。

The GB300 will increase the same architecture, same architecture, same physical footprint, same electrical mechanicals, but the chips inside have been upgraded.

GB300 將增加相同的架構、相同的結構、相同的物理尺寸、相同的電氣機械，但內部芯片已經升級。

It has upgraded with a new Blackwell chip.

It's now one and a half times more inference performance, has one and a half times more HBM memory, and it has two times more networking.

現在，它的推理性能提高了一倍半，HBM 內存提高了一倍半，網絡性能提高了兩倍。

And so the overall system performance is higher.

是以，系統的整體性能會更高。

All of that is so that we could build a very large chip.

所有這些都是為了讓我們能製造出一個非常大的芯片。

And NVLink and Blackwell, this generation, made it possible for us to create these incredible systems.

而這一代的 NVLink 和 Blackwell 使我們有可能創建這些令人難以置信的系統。

Here's one from Pegatron and QCT and Wishtron and WeWin.

這是 Pegatron、QCT、Wishtron 和 WeWin 的作品。

This is from Foxconn and Gigabyte and Asus.

這是富士康、技嘉和華碩的產品。

And you could see the front and the back of it.

And its entire goal, its entire goal is to take these Blackwell chips that are, you know, you can see how big they are, and turn it into one massive chip.

它的全部目標，就是把這些你可以看到有多大的布萊克韋爾芯片，變成一個巨大的芯片。

Now, the ability to do that, of course, was made possible by NVLink.

現在，NVLink 的出現當然讓這一功能成為可能。

But it understates the complexity of the system architecture, the rich software ecosystem that connects it all together, the entire ecosystem of 150 companies that came together to build this.

但它低估了系統架構的複雜性，低估了將其連接在一起的豐富的軟件生態系統，低估了由 150 家公司共同打造的整個生態系統。

This architecture and the entire ecosystem in technology, in software, in industry, has been the work of three years.

這一架構以及技術、軟件和工業領域的整個生態系統歷時三年。

And now, we would like to make it possible for anybody, anybody who wants to build data centers.

現在，我們想讓任何人，任何想建立數據中心的人，都有可能做到這一點。

It could be a whole bunch of NVIDIA GB200s or 300s and accelerated computing systems from NVIDIA.

這可能是一大批 NVIDIA GB200 或 300 以及 NVIDIA 加速計算系統。

And so today, we're announcing something very special.

是以，今天我們要宣佈一件非常特別的事情。

NVLink Fusion is so that you can build semi-custom AI infrastructure, not just semi-custom chips, because those are the good old days.

NVLink Fusion 可以讓你構建半定製的人工智能基礎架構，而不僅僅是半定製芯片，因為那是美好的過去。

You want to build AI infrastructure, and everybody's AI infrastructure could be a little different.

你想建立人工智能基礎設施，而每個人的人工智能基礎設施都可能有些不同。

Some of you could have a lot more CPUs, and some of you could have a lot more NVIDIA GPUs, and some of it could be somebody's semi-custom ASICs.

有些人可能擁有更多的 CPU，有些人可能擁有更多的英偉達™（NVIDIA®）圖形處理器，還有些人可能擁有半定製的 ASIC。

And those systems are so insanely hard to build.

而這些系統的建立又是如此之難。

And they're all missing this one incredible ingredient, this incredible ingredient called NVLink.

而它們都缺少一種不可思議的成分，這種不可思議的成分就是 NVLink。

NVLink so that you could scale up these semi-custom systems and build really powerful computers.

有了 NVLink，你就可以擴展這些半定製系統，構建真正強大的計算機。

And so today, we're announcing the NVLink Fusion.

是以，我們今天宣佈推出 NVLink Fusion。

NVLink Fusion 的工作原理是這樣的。

This is the NVIDIA platform, 100% NVIDIA.

You got NVIDIA CPU, NVIDIA GPU, the NVLink switches, the networking from NVIDIA called Spectrum X or InfiniBand, NICs, network interconnects, switches, and all of the entire system, the entire infrastructure built end-to-end.

英偉達™（NVIDIA®）CPU、英偉達™（NVIDIA®）GPU、NVLink 交換機、被稱為 Spectrum X 或 InfiniBand 的英偉達™（NVIDIA®）網絡、網卡、網絡互聯、交換機以及整個系統、整個端到端基礎設施。

Now, of course, you can mix and match it if you like.

And we now today make it possible for you to mix and match it even at the compute level.

如今，我們甚至可以在計算層面進行混合和匹配。

This would be what you would do using your custom ASIC.

這就是您使用定製 ASIC 所要做的事情。

And we have great partners, I'll announce in a second, who are working with us to integrate your special TPU or your special ASIC, your special accelerator.

我們有很好的合作伙伴，我稍後會宣佈，他們正與我們合作，為您集成特殊的 TPU 或特殊的 ASIC，以及特殊的加速器。

And it doesn't have to be just a transformer accelerator.

它不一定只是一個變壓器加速器。

It could be an accelerator of any kind that you would like to integrate into a large scale-up system.

它可以是任何類型的加速器，您希望將其集成到大型擴展系統中。

It's basically a switch that abuts right up to your chip.

它基本上是一個開關，直接與芯片相連。

There's IP that will be available to integrate into your semi-custom ASIC.

您可以將 IP 集成到您的半定製 ASIC 中。

And then once you do that, it fits right into the compute boards that I mentioned, and it fits into this ecosystem of an AI supercomputer that I've shown you.

一旦你做到這一點，它就能與我提到的計算板相匹配，並與我向你展示的人工智能超級計算機生態系統相匹配。

Now, maybe what you would like is you would like to use your own CPU.

現在，也許你想使用自己的 CPU。

You've been building your own CPU for some time, and maybe your CPU has built a very large ecosystem, and you would like to integrate NVIDIA into your ecosystem.

您製造自己的 CPU 已經有一段時間了，也許您的 CPU 已經建立了一個非常龐大的生態系統，您希望將 NVIDIA 集成到您的生態系統中。

And now we make it possible for you to do that.

現在，我們可以讓您做到這一點。

You could do that by building your custom CPU.

您可以通過定製 CPU 來實現這一目標。

We provide you with our NVLink chip-to-chip interface into your ASIC.

我們為您的 ASIC 提供 NVLink 芯片到芯片接口。

We connect it with NVLink chiplets, and now it connects and directly abuts into the Blackwell chips and our next generation Rubin chips.

我們用 NVLink 芯片將其連接起來，現在它可以直接連接到 Blackwell 芯片和我們的下一代 Rubin 芯片。

And again, it fits right into this ecosystem.

同樣，它也能很好地融入這個生態系統。

DGX Spark will be ready, will be available shortly, probably in a few weeks.

DGX Spark 將準備就緒，很快就會上市，可能在幾周內。

We have tremendous partners working with us.

我們有非常多的合作伙伴與我們合作。

Dell, HPI, ASUS, MSI, Gigabyte, Lenovo, incredible partners working with us.

戴爾、惠普、華碩、微星、技嘉、聯想，與我們合作的合作伙伴都令人難以置信。

這實際上是一個生產組織、部門。