|
|
英偉達(dá)產(chǎn)品路線圖曝光,事關(guān)Blackwell和Rubin 在 Hot Chips 2024 展會(huì)開始之前,Nvidia 展示了其 Blackwell 平臺(tái)的更多元素,包括正在安裝和配置的服務(wù)器。這是一種不太含蓄的說(shuō)法,表明 Blackwell 仍在發(fā)展中——?jiǎng)e介意延遲。它還談到了現(xiàn)有的 Hopper H200 解決方案,展示了使用其新 Quasar 量化系統(tǒng)進(jìn)行的 FP4 LLM 優(yōu)化,討論了數(shù)據(jù)中心的溫水液體冷卻,并談到了使用 AI 來(lái)幫助構(gòu)建更好的 AI 芯片。它重申,Blackwell 不僅僅是一個(gè) GPU,它是一個(gè)完整的平臺(tái)和生態(tài)系統(tǒng)。 Nvidia在 Hot Chips 2024 上展示的大部分內(nèi)容已經(jīng)為人所知,例如數(shù)據(jù)中心和 AI 路線圖顯示 Blackwell Ultra 將于明年推出,Vera CPU 和 Rubin GPU 將于 2026 年推出,隨后是 2027 年的 Vera Ultra。Nvidia 早在 6 月份就在Computex上首次確認(rèn)了這些細(xì)節(jié)。但 AI 仍然是一個(gè)大話題,Nvidia 非常樂意繼續(xù)推動(dòng) AI 的發(fā)展。 據(jù)報(bào)道,Blackwell 的發(fā)布推遲了三個(gè)月,但Nvidia 既沒有證實(shí)也沒有否認(rèn)這一信息,而是選擇展示正在安裝的 Blackwell 系統(tǒng)的圖像,以及展示 Blackwell GB200 機(jī)架和 NVLink 交換機(jī)中更多內(nèi)部硬件的照片和渲染圖。除了該硬件看起來(lái)可以消耗大量電量并且具有相當(dāng)強(qiáng)大的冷卻功能之外,沒有太多可說(shuō)的。它看起來(lái)也很昂貴。 Nvidia還展示了其現(xiàn)有 H200 的一些性能結(jié)果,運(yùn)行時(shí)使用和不使用 NVSwitch。它表示,與運(yùn)行點(diǎn)對(duì)點(diǎn)設(shè)計(jì)相比,推理工作負(fù)載的性能可以提高 1.5 倍——這是使用 Llama 3.1 70B 參數(shù)模型。Blackwell 將 NVLink 帶寬翻倍以提供進(jìn)一步的改進(jìn),NVLink Switch Tray 提供總計(jì) 14.4 TB/s 的總帶寬。 由于數(shù)據(jù)中心的功率需求不斷增加,Nvidia 也在與合作伙伴合作以提高性能和效率。其中一個(gè)更有希望的結(jié)果是使用溫水冷卻,其中加熱的水可以再循環(huán)用于加熱以進(jìn)一步降低成本。Nvidia 聲稱,使用該技術(shù)可以使數(shù)據(jù)中心的用電量減少 28%,其中很大一部分來(lái)自移除低于環(huán)境溫度的冷卻硬件。 上面是 Nvidia 演示文稿的完整幻燈片。還有一些其他值得注意的有趣內(nèi)容。 為了準(zhǔn)備 Blackwell,現(xiàn)在增加了原生 FP4 支持,可以進(jìn)一步提高性能,Nvidia 一直致力于確保其最新軟件從新硬件功能中受益,而不會(huì)犧牲準(zhǔn)確性。在使用其 Quasar 量化系統(tǒng)調(diào)整工作負(fù)載結(jié)果后,Nvidia 能夠提供與 FP16 基本相同的質(zhì)量,同時(shí)使用四分之一的帶寬。生成的兩個(gè)兔子圖像可能在細(xì)微方面有所不同,但這對(duì)于 Stable Diffusion 等文本到圖像工具來(lái)說(shuō)非常典型。Nvidia 還談到了使用 AI 工具來(lái)設(shè)計(jì)更好的芯片——AI 構(gòu)建 AI,一路向下都是烏龜。Nvidia 創(chuàng)建了一個(gè)內(nèi)部使用的 LLM,有助于加快設(shè)計(jì)、調(diào)試、分析和優(yōu)化。它與用于描述電路的 Verilog 語(yǔ)言一起工作,是創(chuàng)建 2080 億個(gè)晶體管 Blackwell B200 GPU 的關(guān)鍵因素。然后,這將用于創(chuàng)建更好的模型,使 Nvidia 能夠在下一代 Rubin GPU 及以后的產(chǎn)品上工作。[此時(shí),您可以隨意插入您自己的 Skynet 笑話。] 總結(jié)一下,我們對(duì) Nvidia 未來(lái)幾年的 AI 路線圖有了更高質(zhì)量的了解,該路線圖再次將“Rubin 平臺(tái)”與交換機(jī)和互連定義為一個(gè)整體包。Nvidia 將在下周的 Hot Chips 會(huì)議上介紹有關(guān) Blackwell 架構(gòu)、使用生成 AI 進(jìn)行計(jì)算機(jī)輔助工程和液體冷卻的更多細(xì)節(jié)。
參考鏈接 https://www.tomshardware.com/tech-industry/artificial-intelligence/nvidia-shows-off-blackwell-server-installations-in-progress-ai-and-data-center-roadmap-has-blackwell-ultra-coming-next-year-with-vera-cpus-and-rubin-gpus-in-2026 END
Nvidia shows off Blackwell server installations in progress — AI and data center roadmap has Blackwell Ultra coming next year with Vera CPUs and Rubin GPUs in 2026
When you purchase through links on our site, we may earn an affiliate commission. Here’s how it works. Nvidia Hot Chips 2024 Much of what will be presented by Nvidia at Hot Chips 2024 is already known, like the data center and AI roadmap showing Blackwell Ultra coming next year, with Vera CPUs and Rubin GPUs in 2026, followed by Vera Ultra in 2027. Nvidia first confirmed those details at Computex back in June. But AI remains a big topic and Nvidia is more than happy to keep beating the AI drum. While Blackwell was reportedly delayed three months, Nvidia neither confirmed nor denied that information, instead opting to show images of Blackwell systems being installed, as well as providing photos and renders showing more of the internal hardware in the Blackwell GB200 racks and NVLink switches. There's not much to say, other than the hardware looks like it can suck down a lot of juice and has some pretty robust cooling. It also looks very expensive. Nvidia also showed some performance results from its existing H200, running with and without NVSwitch. It says performance can be up to 1.5X higher on inference workloads compared to running point-to-point designs — that was using a Llama 3.1 70B parameter model. Blackwell doubles the NVLink bandwidth to offer further improvements, with an NVLink Switch Tray offering an aggregate 14.4 TB/s of total bandwidth. Because data center power requirements keep increasing, Nvidia is also working with partners to boost performance and efficiency. One of the more promising results is using warm water cooling, where the heated water can potentially be recirculated for heating to further reduce costs. Nvidia claims it has seen up to a 28% reduction in data center power use using the tech, with a large portion of that coming from the removal of below ambient cooling hardware. Above you can see the full slide deck from Nvidia's presentation. There are a few other interesting items of note. To prepare for Blackwell, which now adds native FP4 support that can further boost performance, Nvidia has worked to ensure it's latest software benefits from the new hardware features without sacrificing accuracy. After using its Quasar Quantization System to tune the workloads results, Nvidia is able to deliver basically the same quality as FP16 while using one quarter as much bandwidth. The two generated bunny images may very in minor ways, but that's pretty typical of text-to-image tools like Stable Diffusion. Nvidia also talked about using AI tools to design better chips — AI building AI, with turtles all the way down. Nvidia created an LLM for internal use that helps to speed up design, debug, analysis, and optimization. It works with the Verilog language that's used to describe circuits and was a key factor in the creation of the 208 billion transistor Blackwell B200 GPU. This will then be used to create even better models to enable Nvidia to work on the next generation Rubin GPUs and beyond. [Feel free to insert your own Skynet joke at this point.] Wrapping things up, we have a better quality image of Nvidia's AI roadmap for the next several years, which again defines the "Rubin platform" with switches and interlinks as an entire package. Nvidia will be presenting more details on the Blackwell architecture, using generative AI for computer aided engineering, and liquid cooling at the Hot Chips conference next week.
關(guān)于我們 北京漢深流體技術(shù)有限公司是丹佛斯中國(guó)數(shù)據(jù)中心簽約代理商。產(chǎn)品包括FD83全流量自鎖球閥接頭,UQD系列液冷快速接頭、EHW194 EPDM液冷軟管、電磁閥、壓力和溫度傳感器及Manifold的生產(chǎn)和集成服務(wù)。在國(guó)家數(shù)字經(jīng)濟(jì)、東數(shù)西算、雙碳、新基建戰(zhàn)略的交匯點(diǎn),公司聚焦組建高素質(zhì)、經(jīng)驗(yàn)豐富的液冷工程師團(tuán)隊(duì),為客戶提供卓越的工程設(shè)計(jì)和強(qiáng)大的客戶服務(wù)。 公司產(chǎn)品涵蓋:丹佛斯液冷流體連接器、EPDM軟管、電磁閥、壓力和溫度傳感器及Manifold。
|
|