**Luke Lafreniere** (0:00)
Howdy, shoppers. Fire up that propane grill and settle in for a new season of Hulu's hit animated series, King of the Hill. Hank and Peggy are staying active in retirement.
**Linus Sebastian** (0:08)
That's a dog.
**Dan** (0:09)
You will not have to lift a finger.
**Luke Lafreniere** (0:11)
Bobby and Connie hard-launched their relationship.
**Linus Sebastian** (0:13)
Now that we're dating, it's important your parents know I'm a man.
**Luke Lafreniere** (0:17)
And the alley guys are, well, they're still the alley guys.
**Dan** (0:20)
Yep.
**Luke Lafreniere** (0:21)
Watch the new season of the Hulu original series, King of the Hill, now streaming on Hulu, and Hulu on Disney Plus for bundle subscribers. Terms applied, this has been a paid advertisement.
**Linus Sebastian** (0:30)
Friday, and welcome to The WAN Show. We've got a great show lined up for you guys this week, actually. There's lots of fun stuff. Nvidia has provided new details on their Vera data center CPUs, and this raised Luke's eyebrows just a bit. Yeah, see, he did it. You couldn't see it because it was the one that was closer to me, but he did it. He did it again.
I am really, really excited about them and what they might mean for the future of CPU microarchitectures that will, I will lay out a case for this, hopefully make their way to gamers and be a major benefit. In other news, ooh, Framework had a bit of a rough week. Investment disclosure.
With respect to pricing of memory modules in their laptops, we'll get into that a little bit more later. What else we got?
**Luke Lafreniere** (1:26)
Microsoft introduces XBox Backwards Compatibility Program on PC. Exciting, interesting.
**Linus Sebastian** (1:34)
For XBox games.
**Luke Lafreniere** (1:35)
Kind of confusing. We'll get into it.
Also, Fallout 5, Fallout 3, New Vegas, Remasters, also a new Obsidian Fallout. All of this stuff was announced. Maybe Bethesda will use an IP for the first time in, I don't know, over a decade, I think.
**Linus Sebastian** (1:53)
Cool.
**Luke Lafreniere** (1:53)
I heard you're supposed to do that as a company. Yeah.
**Linus Sebastian** (1:56)
I heard you're supposed to IP freely. Oh, I pressed the wrong one. What the?
Oh, good. The foot cam is there.
**Luke Lafreniere** (2:11)
I just got jumpscared by my own feet.
**Linus Sebastian** (2:18)
My feet are so touching the floor. The show is brought to you by Shokz, AMD, Squarespace and XSplit alongside our rap partner Dbrand, our laptop partner Razer and our chair partner who is also Razer. Why don't we jump right into our headline topic for the day, which is, of course, I actually forget which one we picked.
**Luke Lafreniere** (2:38)
Nvidia CPU.
**Linus Sebastian** (2:39)
Oh, yes. I want to talk about that. Okay. So the original source here was actually some analysis from my old buddy, Mr. Ryan Shroud, who used to run PC Perspective and then was working at Intel for a little while, and then now is kind of doing this whole, like, analyst thing. And so I was looking at his analysis, and he kind of laid out a bunch of the details that Nvidia is providing about their new Vera CPUs. Vera is built on the Olympus Core, which is Nvidia's first custom core. So, while Grace was a custom designed CPU, it was a partnership between Nvidia and MediaTek, the actual cores themselves were off the shelf pretty much ARM cores. It's a matter of laying them out and, you know, figuring out all of your IO and how to integrate it into your bigger system. This time around, we're looking at a custom core. This is an Nvidia design from the ground up, and this is what gets me excited. It's 88 cores on a monolithic compute die, running 176 threads using what Nvidia calls spatial multi-threading.
Okay. Where is he going with this? Yeah. Meanwhile, companies like AMD have upcoming epic CPUs that are going to have as much as 256 Zen 6C cores. What could possibly make me excited about 88 cores on some ARM CPU that is destined for the data center?
You still don't know where I'm going with this.
**Luke Lafreniere** (4:25)
This is still the question I had.
**Linus Sebastian** (4:27)
Wonderful. Okay.
What I'm excited about is that for the last 20 years, okay, everything has been okay. We'll give you some single threaded improvements, but really what we're building is better power efficiency and better multi-core for the data center so that they can take all those cores and they can slice them all up into cloud instances for all of their customers. Big fat f**king deal for my counter-strike frame rate. Nvidia is coming in and going, no, what the market actually needs is big, strong cores and not as many of them. Now, the reason is not one that's likely to excite you guys. The reason is that Nvidia is looking for huge performance uplifts in agentic AI.
234 more minutes of transcript below
Try it now — copy, paste, done:
curl -H "x-api-key: pt_demo" \
https://spoken.md/transcripts/1000651996090
Works with Claude, ChatGPT, Cursor, and any agent that makes HTTP calls.
From $0.10 per transcript. No subscription. Credits never expire.
Using your own key:
curl -H "x-api-key: YOUR_KEY" \
https://spoken.md/transcripts/1000778293230