AMD IS THROWING out the code names like you knew they would at the 2010 Analyst Day. Couple that with a bunch of demos, and you have something interesting.
OK, I am overstating things. 90 minutes in, an HD video and 6 code names isn’t exactly edge-of-seat type entertainment, but then again, I am not a financial analyst, I am told they like this stuff. That said, AMD is name dropping furiously with Trinity, Krishna, Komodo, Terramar and Sepang all mentioned. On top of that, they demo’d a Valencia/Zambezi/Bulldozer chip with a Cayman GPU.
The short story is Krishna is Ontario +1 with 2-4 cores instead of 1-2. Basically it is a 28nm shrink, but AMD won’t say that yet. It’s arrival is more dictated by the process than anything else. Trinity is Llano +1, substituting Bulldozer cores for the current K10.x ones. It will be interesting to see which chip is released first, at this point in time, it could go either way. Komodo is the top of the fusion stack with 8 Bulldozer cores and a GPU on die.
Sepang is a 10 core/5 module/1 socket/extra cheese server part, and the cores are listed as “next-generation” bulldozer parts. Terramar is more than an abandoned circuit, it is two Sepangs on an MCM, and can go to 4 sockets. A little math tells you that 2 next-generation Bulldozer cores are about as big as a low end GPU.
Then it was on to the demos. AMD wowed the audience with a demo of a Cayman equipped 8 core Bulldozer playing an HD video while running MS’s CPU load meters. At the same time! No, really, they pulled it off! Wow!
Oh wait, that has been doable for several years now. With a 1 core machine armed with an IGP. Call me unimpressed, especially since they showed a Llano doing 5x that at AMD’s TFE conference 3 weeks ago. Then again, it was new running silicon.S|A
Latest posts by Charlie Demerjian (see all)
- Globalfoundries 7nm process isn’t even close to the name - Sep 26, 2016
- ARM upgrades realtime offerings to v8-R and adds Cortex-R52 - Sep 21, 2016
- Everspin and Globalfoundries team up for embedded ST-MRAM - Sep 15, 2016
- Intel’s Xpoint is pretty much broken - Sep 12, 2016
- ARM adds 2048-bit vectors to v8A with SVE - Sep 7, 2016