It looks like Nvidia (NASDAQ:NVDA) is keeping to their usual form with Tegra, the schedule just slipped a year. With that we welcome the new chip in the bunch, T35, and the attendant schedule waterfall.
If you are not up to date on the current Tegra roadmap, up until a few weeks ago, it looked like this for Tegra 3, 3.3 and 4, with Tegra 5 and Denver following. Before we move on, a quick update. Insiders tell us that Denver is the T50 core, T50 is the chip family based around it, not the other way around. Or at least it was before everything slipped a few weeks ago. And that brings us back to the current situation, and the ‘new’ chip.
The T30 and T33 speedbump chips are still intact, with chips going on sale to the public at the end of September according to Nvidia’s 2011 Computex announcements. That may not have quite panned out as everyone paying even a bit of attention knows, but the chips should ship by the end of the year. T33 will follow in a bit, it is just a mild reworking/spin of T30, and it’s purpose is to lose less to Qualcomm’s Krait. It won’t, but let’s not shatter any dreams yet, Qualcomm should have an answer to keep the benchmark status quo.
From there we come to the new chip on the block, or at least the newest resurrected chip on the on block, T35. If you recall, T30/33 is a 40nm quad-ish core A9, and T40 was originally meant to be a shrink of that to 28nm. A15 based parts were going to be skipped in order free up engineering resources with an ultimate goal of bringing the A8/ARM-64 ISA based Denver to market first. Nvidia was publicly promising financial analysts that Denver based parts would ship in late Q4/12 long after they knew there was no chance of that being possible. Short term stock bumps seem to matter more than engineering at some Santa Clara based chip companies, legal concerns be damned. Consistency is a an admirable trait.
This was a decent idea, medium-term pain to get long-term gain, but the execution was not close to feasible. Part of the problem was Dear Leader’s constant micro-managing of technical specs, and the attendant wholesale changes. Moving Denver from Kepler to Maxwell cores assured that there was no way it would be out in 2012, and put 2013 as a stretch, but that somehow didn’t make it to the IR briefings. Curious that.
With Denver slip-sliding away, there was a little more breathing room for the beleaguered engineers. Between that and the woefully uncompetitive nature of the A9 cores in the A15 world of 2012, something had to be done. Again, Dear Leader rightfully deemed the A9 cored T40 to be too slow, and single handedly ordered it changed to an A15 design. Fair enough, unless you are working on the chip, then life sucks.
This change made T40 much more competitive on paper, but pushed the schedule out quite a bit. This now extended gap between T30/T33 and T40 is not a good thing, especially in light of the competition. Because those lazy engineers at Nvidia could not meet Dear Leader’s 180 degree change in vision while keeping the original schedule, it slipped more. No shame in that, the fact that there are still employed T40 engineers is testament to their strong will and perseverance.
What does a company do when they are facing a year+ without a competitive core? Resurrect a stopgap, and call it T35. This ‘new’ part is simply the old 28nm shrink of Tegra 3/T30 that was called T40 before the A15ing. Fair enough, and it is a good stopgap, but what do you call it? Tegra 4 of course.
Yes, Tegra 4 is now a 28nm Tegra 3 once again, and the A15 based T40 is now Tegra 5. Because word of massive schedule slips would likely hit management in the portfolio, this change is now deemed to be a silent one. T40 is now Tegra 5, Denver/T50 is now Tegra 6, and T35 was meant to be there all along, honest. Nothing to see here analysts, smile and move along.
In the end, the Tegra roadmap just slipped a year. Nvidia faces a rather painful 2012 and likely early 2013 with a core that is a generation behind, so expect lots of self-congratulatory press releases. In the mean time, Qualcomm is releasing Krait, TI has OMAP5, and Nvidia has hot air. Anyone want to join our pool for when the stock bubble bursts?S|A
Latest posts by Charlie Demerjian (see all)
- ARM adds CMN-600 interconnect and DMC-620 memory controllers - Sep 27, 2016
- 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