Thunderbolt (as much as I love the idea) is made and operated by Intel, and Intel wants to charge a premium for it. Finally it will be an open standard that multiple manufacturers can adopt. And because it is already standardized and cheap technology it will be easier and cheaper to implement. Chip makers will not have to add a new and expensive tech to the device, the connection will just siphon off lanes that are already there. The external PCIe will be cheaper and easier because it will be hosted directly by the mobo chipset, or the CPU by tapping into already existing hardware. but we have not seen that happen over the copper version that was released. It can send PCIe, it can send DP, it was supposed to be able to send USB 1/2/3, as well as Ethernet, all over the same wire. You think the specification will somehow make things cheaper? Nope.thunderbolt is a medium that allows for multiple interconnect protocols to be transmitted over the same wire. Our numbers require more Mac-specific testing, since we don’t know whether it was the update or our system to blame. Promise tells us it sees slightly better performance on PCs, but the delta should only be a few percent at most. However, since we "upgraded" to a 13.3” MBP, our new results seem to suggest a ceiling around ~800 MB/s. Prior to that, we were able to achieve close to 920 MB/s in RAID 0 on our 15” MacBook Pro. However, in the summer of 2011, Apple and Intel decided to put device I/O on its own channel in order to preserve the display's signal integrity under heavy workloads. Originally, Thunderbolt allowed device I/O and display signaling to share bandwidth over both bi-directional channels. This is the only device we're reviewing with performance characteristics that look a little different under OS X than Windows, and there's a reason why. We see sequential reads top out around ~780 MB/s in that arrangement. RAID 5 is a fair compromise, enabling block-level striping with distributed parity. It falls just 50 MB/s short of the 1 GB/s barrier at a queue depth of 16! If redundancy is more important to you than blistering speed, RAID 1E operates effectively as three RAID 1 arrays, which is why sequential read performance falls to ~315 MB/s. The R6 really stretches its legs when we measure sequential read performance in RAID 0.
0 Comments
Leave a Reply. |