In the past weeks Intel had to face a lot of negative coverage for their upcoming Haswell processor which is due for introduction later this year and to be branded as 4th generation Core processor. One of the unofficial claims was that there was a serious bug with the accompanying Lynx Point „8 series / C220“ chipset that causes USB devices to disappear after the computer transitioned into standby mode (specifically S3).
Now Intel steps in and – albeit quietly – confirms the chipset issues. In a Product Change Notification (PCN) they announce a stepping conversion from C1 to C2 which is going to fix this bug. Intel had to change metal layers to remedy the „USB SuperSpeed device remuneration erratum“. Other characteristics of the chipset are not changed. Intel recommends not to run the full suite of electronical regression tests but only a subset focussing on the USB 3.0 issue.
The document lists the following chipsets that get a replacement with the new stepping: Q87, Q85, H87, Z87, B85 (desktop chipsets), C222, C224, C226 (workstation/server chipsets), QM87, HM87, HM86 (mobile chipsets).
According to the PCN the new stepping will be sampled to customers starting April 19th. Final product will be available starting July 15th. Intel also makes clear that customers must be ready to receive the new stepping by July 31st. In other words, this means by then the old stepping will be phased out and gradually replaced by the new one. Due to the minor change needed to fix the problem, Intel doesn't expect a major impact on customers. The new stepping is basically a drop in replacement.
While an official release date for Haswell hasn't been given yet, Intel is expected to introduce the new chip before the availability of the new chipset stepping. Reportedly Intel let customers sign an agreement to acknowledge the presence of the bug if they wanted to obtain Haswell APUs for launch. The bug in question causes nuisances like USB devices to disappear after entering standby. Detachable devices would have to be reconnected to be found again.
We don't know whether the company is able to provide a workaround for affected chipsets. Based on the description only SuperSpeed (USB 3.0 that is) ports are affected, which means that using the slower USB 2.0 ports would be one way to circumvent the problem. On desktop boards manufacturers could use 3rd party chips which would increase the BOM slightly but would not be hampered by this issue. Given the additional power such a companion chip consumes as well as the space needed on the PCB such a measure is less likely on notebooks and particularly Ultrabooks where space is very precious.
This is not the first chipset issue at Intel in recent history. In 2011 their Cougar Point „6 series“ chipset accompanying Sandy Bridge had a flaw in the Serial ATA controller that caused ports to degrade and subsequently fail over time. Back then the issue was only caught after affected mainboard went on sale. Rectifying the mistake set Intel back by a cool $1 billion.
technical core competence is not only not easy to get, but just as hard to maintain. If you are not constantly in the market for those chips, starting from scratch is definitely not easy.
That comes to ARM integration solutions: it is very hard for anybody to build ARM integrated solutions. Linking the ARM cores together with the GPU is the easy part. If you don't have core competence in 'cheap' radios, ala you are not already in the market selling those cheap radios, it is very hard to just hire the design engineers and build them from scratch. Add to that the radios are moving targets with the ever evolving standards. This is true for nvda or aapl or samsung or intel or mediatech or rockchip or anybody.
That is why intel paid $1.2B for a radio, qcom paid $3.1B for wifi/athero, samsung bought a small wifi guy, nvda buying Icera etc. etc.
As of now, there are only two companies with the complete set of radios to do a fully integrated solution.