AMD is going to be launching Tahiti 2 in the near future, at least according to almost every board maker in Taipei. T2 to T1 in silicon isn’t anything like T2 was to T1 in the movies, but there is a lot of good here.
The short story is that Tahiti 2, aka Tahiti XT 2 in ASIC speak, isn’t actually new silicon, they are just binning better. If you thought that Tahiti/HD7970 was binned really conservatively for yield, you would have been right. Think of it as the polar opposite of Nvidia’s GTX680 footgun binning method.
End result, well over a GHz, but likely to be called HD7970 GHz edition as is all the rage nowadays. People are telling SemiAccurate that final silicon will run from about 1050-1075MHz, so about a 15-20% boost over Tahiti. Since T2 is binned for low leakage, it should run at the same TDP or less than current Tahitis. That means no board changes, no cooler changes, and likely no memory changes either, but all of those things are just fine the way they are now.
As we mentioned earlier, T2 is a real product, unlike GTX680 which was never actually meant to be real according to Nvidia sources. We have good technical reasons to believe Tahiti 2 will have great yields and ship in more than enough quantity, conversely GTX680 never will.
Also, if you are waiting for New Zealand, aka the dual GPU 7990, this ASIC is what AMD was waiting for. While there are other, non-AMD related issues with launching that card, the silicon should no longer be a hold up now.
Last but not least, there will be a new binning of Tahiti Pro, now called Tahiti Pro 2. Sources did not tell us the specs to that part, but it will likely be nothing more than a mild update to better utilize the ASICs that don’t end up being called Tahiti 2. So there you have it, the brand ‘new’ Tahiti 2.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