AWS Trainium, every mention

22 scenes, the whole family · ← back to AWS Trainium

mentionsepisodesper episode tap a year for its mentions

Verbatim, from the transcripts: the passages where AWS Trainium comes up

loading…

▶ 49:56 Alex Kantrowitz I guess Amazon to a degree has Tranium, but the TPUs have shown that they can be used to build real state-of-the-art models, and if you look at Google, so Google's market cap as we're speaking today, you know, is four trillion dollars, and…

▶ 2:54 Alex Kantrowitz Uh, has been, has put out, uh, and Tranium, which is Amazon's version of it, um, you know, two of the top three models in the world, Claude and Gemini were trained on it, and what does that mean for NVIDIA going forward, right? 3 times in the scene

▶ 14:09 Alex Kantrowitz They have, you know, marched, marched people around their, you know, their new facility that they've built with their Tranium chips, talking about how their, you know, deep partnership with Anthropic has been what's enabled them to build…

▶ 23:17 Reed Albergotti And I think the rumor, I don't think it's been confirmed that, uh, OpenAI is going to actually use the Tranium. 3 times in the scene

▶ 43:02 M.G. Siegler Because obviously Amazon is a huge, uh, shareholder in Anthropic and has that big partnership around their, uh, Tranium chips, right? 2 times in the scene

▶ 1:00:17 Ed Zitron Someone, it might have been anthropic, forgive me if I'm wrong, but they're using 500,000 Tranium GPUs, their own. 2 times in the scene

▶ 48:40 Alex Kantrowitz Uh, Amazon says they'll buy as basically as many GPUs as they can get from NVIDIA, but they also talk about their Tranium chips and you know, that it's a balance.

▶ 53:32 Reed Albergotti They acquired this Israeli company, Annapurna Labs, 10 years ago, and, you know, this is their second generation of the Tranium chips, Tranium II, and this chip is, I mean, it's not as good as, like, an NVIDIA chip on a one-to-one basis,… 2 times in the scene

▶ 2:15 Matt Garman We're building together with them what we call Project Rainier, uh, and so it's using our next generation Tranium II chips, and this, uh, cluster that we're building for them, uh, in twenty-twenty-five will be five times the size, the… 2 times in the scene
▶ 9:20 Matt Garman And, uh, you know, we're, we've made a big bet on Tranium as a, an additional option for customers, you know, the vast majority of, we should define it. 3 times in the scene
▶ 10:09 Matt Garman And we launched Tranium one, uh, a couple of years ago in 2002.
▶ 10:14 Matt Garman And, um, are just GA-ing Tranium two here at reInvent. 4 times in the scene
▶ 11:00 Matt Garman And just learn what's working, what's not, what are the things you need accelerated so that training three and training four and training five and training six can all get better as we, as we continue to go.
▶ 11:00 Matt Garman And just learn what's working, what's not, what are the things you need accelerated so that training three and training four and training five and training six can all get better as we, as we continue to go.
▶ 11:00 Matt Garman And just learn what's working, what's not, what are the things you need accelerated so that training three and training four and training five and training six can all get better as we, as we continue to go.
▶ 15:47 Matt Garman Number one is, I think Tranium II will be a material impact there, and 4 times in the scene

▶ 36:49 Jack Clark We are also, um, you know, publicly we've stated that we're working on Tranium chips. 2 times in the scene

▶ 22:44 Swami Sivasubramanian And on the, their end, Anthropic will also use AWS, uh, Tranium, uh, and Inferentia tips to build, train, and deploy their future models and, uh, as well.
▶ 23:19 Alex Kantrowitz First of all, you talk about terrarium chips, which you said can train a model four times faster than the industry standard. 6 times in the scene
▶ 24:02 Swami Sivasubramanian Uh, Tranium two chips, uh, are designed with the sole purpose of being able to train these large language models, and highly optimized, uh, for Tranium use case. 3 times in the scene

▶ 3:43 Pat Gelsinger For the training application, you now have what Nvidia does, accelerators like what we're doing with Gaudi, but also then the TPU from Google, you know, the Tranium from Amazon, what Microsoft just announced with Maya, what AMD announced.

▶ 41:05 Matt Wood But with Tranium and with Inferentia, those are totally custom designed.

Turn any episode into a week of clips.