Intel is still trying to pretend Thunderbolt is not dead, but two new chips don’t do much to change the underlying problems. The new parts fix one of many glaring problems of the first generation but still hurt the user badly.
The new parts are called the DSL4510 and DSL4410, the 2014 Haswell generation parts are called Falcon Ridge. DSL4x10 adds DisplayPort 1.2 support, Falcon Ridge ups the per-lane bandwidth to 20Gbps, but both are still broken proprietary tools designed to lock competition out. These parts are so awful that Intel wouldn’t actually brief us on them, only those they knew were bought and paid for enough to not point out the obvious flaws were briefed.
What are the problems? Well DP1.2 support in 4×10 sounds great right? What could be the problem with added features? The problem is that it isn’t a feature add, it is a crippling bug fix that should have been in all Thunderbolt chips since day one. What bug is this? Channel bonding. The old chips couldn’t do DP1.2 on Ivy even though Ivy supported it because they couldn’t bond the two channels, it was utterly broken.
You might recall that Thunderbold is a proprietary competition lock-out that Intel tried desperately to shut down USB3 in order to protect. Like Thunderbolt itself this scheme failed and USB3 is everywhere, Thunderbolt is nowhere for good reason. USB3 is a useful point to point bus, Thunderbolt isn’t, it is just a very badly designed somewhat transparent PCIe pipe. In theory it should have been able to do DP1.2 from the start, but the inability to bond channels meant that the bandwidth was far too limited to actually do so. When the bug was fixed, voila, DP1.2 support! For some reason Intel touts the 4×10 as having a new feature rather than pointing out they knowingly sold millions of absolutely broken parts, those would be the recent macbooks.
“But wait, Intel said you can channel bond in all of their talks, presentations, and the like for years” you say. They lied. Really, they lied. No press bothered to point out that the most basic function of the ‘technology’ was completely broken, it would endanger ad revenue if they did. So they turned a blind eye to the lies and try to pretend this technology wasn’t crippled.
Like the first iteration, the upcoming Falcon Ridge parts are broken too. No, not lacking optical support like the last ones, and it can probably channel bond too, just that the speed is once again way below what was promised to customers. SemiAccurate has talked to many people involved with future Thunderbolt parts and for years they said it was a 25Gbps interface. Intel is now saying 20Gbps, whoops. That footnote is nowhere in the public materials though, the ones Intel wouldn’t give us or otherwise. Any guesses why?
In the end, the new 4×10 parts are just a less broken spin of a fundamentally broken spec. Intel still will not allow anyone else to make Thunderbolt compatible chips, nor will they ever. The plan is to get it established and lock out any competition from the PC market or charge them an unconscionable fee to play ball. Luckily for the consumer, Thunderbolt has badly failed and the market has moved on from the PC. These two things are not coincidence.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