TR Forums

As there are now Windows motherboards with Thunderport port, guess this can be used by Mac and PC guys.

Got my Seagate Thunderbolt Adapter working with a Samsung 830 series 512GB SSD, write is 328MB and read is 365MB, reliable with external boot. It needs some modifcations to make this work. I thought this may interest some people on this forum.



As this is a bit complicated and a few pages long with a couple of pics, full instructions on my blog

http://wolfgangtechnology.blogspot.co.at/2012/06/mac-with-seagate-thunderbolt-adapter.html

Hope I do not violate any rules on this board with this link to my blog

Happy reading, Wolfgang

That's slower than I would expect. eSATA would be faster, assuming HDTune is even remotely comparable.

The most interesting there here is just how much space and bandwith is required for various video formats. Some of those high-end color depth and resolution settings are astounding in how much MB/s is used.

I'll be in my bunk.

I do not have a boot camp W7 installed, so I can not test with HD Tune to compare. Have only W7 in VM, and this is definitely a lot slower, but for a windows installation good enough. I agree this would be interesting to compare. As Black Magic Speed Test uses video files for testing, a HD what uses data compression would show less speed with Black Magic Speed then HD Tune or others, as the video files are basically not compressable.
The speed test with VM is one of my next projects when I have time. For my knowledge HD Tune and the other windows disk test tools need a drive letter for testing an do not accept a path (from the shared disks). This means It would be necessary to make for each internal and external disk a vmdk file, copy it then on all HD you want to test and then edit manually the vmx file that Fusion can use this disks with drive letters, then the Windows speed test tools could be used. Sure the speed will include the speed loss of the virtual disk, CPU performance etc. but would be interesting to see. Well a project when I have more time.
//Wolfgang

that is really impressive, now that is it coming out, talk about backup drive quicker than internal hehe. even ssd's will have some competition, time will tell.

Don't do performance testing via VM. With VMware products there are a minimum of two layers of handling/redirection, at best case, and I'd imagine the other virtualization suites are similar. Expect VMs to achieve 90% of native I/O performance, best case.

There are many reasons why enterprise-scale virtualization generally involves iSCSI/other SAN technologies as the backend, and VM I/O penalties are those reasons.

derFunkenstein wrote:

That's slower than I would expect. eSATA would be faster, assuming HDTune is even remotely comparable.

The most interesting there here is just how much space and bandwith is required for various video formats. Some of those high-end color depth and resolution settings are astounding in how much MB/s is used.

Well, you're assuming the adapter is not SATA2, which, if it is, is maxing the bus out on that.

Forge wrote:

Don't do performance testing via VM. With VMware products there are a minimum of two layers of handling/redirection, at best case, and I'd imagine the other virtualization suites are similar. Expect VMs to achieve 90% of native I/O performance, best case.

There are many reasons why enterprise-scale virtualization generally involves iSCSI/other SAN technologies as the backend, and VM I/O penalties are those reasons.

I also skimmed and didn't notice he was doing it with a VM. Yeah VM IO kind of sucks. Its 'good enough' to run a server on, but not good enough for anything truly disk intensive.

Corrado wrote:

Forge wrote:

Don't do performance testing via VM.

I also skimmed and didn't notice he was doing it with a VM. Yeah VM IO kind of sucks. Its 'good enough' to run a server on, but not good enough for anything truly disk intensive.

He's not yet, but he's talking about it, so that he can run the Windows-only tests.

Corrado wrote:

Well, you're assuming the adapter is not SATA2, which, if it is, is maxing the bus out on that.

That doesn't make it not slow. Regardless, it's a waste of an expensive cable and enclosure.

regarding the speed discussion, based on the comments, I guess the purpose and functionality of Thunderbolt is not fully understood. Thunderbolt is not a competing I/F to eSATA, USB3 and others.
Thunderbolt is just a encapsulation of display-port and the PCIe Bus for your PC or Mac, this means it can only drive a display-port Monitor directly, without any chips, the important 2nd part of Thunderbolt is the encapsulated PCIe bus, this means one or many PCIe cards or chip sets can be in a external box of your PC or laptop (in this box can be additional multiple SATA I/F, display cards(with some limitations) USB3 controllers etc). If you connect to a external HD with Thunderbolt, then the external box has a PCIe SATA chip set to connect the disk, this means a disk connected directly via eSATA to PC will always be faster, because there is no overhead for the Thunderbolt protocol to encapsulate the PCIe bus. But on the other hand, you can have multiple PCIe cards or chip sets in the external box and then reach via a single cable transfer rates of up to 2x 10Gbs, combined for all I/F in this external box. To have faster disk access, you will need to have Fiber Channel controller in the external Thunderbolt box, but same as before, the Fiber Channel controller in the server itself will be faster.
For a desktop or server PC means this, connect your disks to eSATA ports or internal PCIe cards for maximum speed, but if your PC does not have any free PCIe slots anymore (and a laptop at normal does not have any PCIe slots), you can add now multiple PCIe cards/chip sets external. For a laptop it means you can connect like a docking station multiple high speed devices via only one Thunderbolt cable. There is definitely NO speed advantage when using Thunderbolt for a single native I/F.

To get the understanding what overhead is needed by Thunderbolt for the PCIe encapsulation see now the speed measurement of same Samsung 830 series 512Gb SSD when installed as main disk in MBP (SATA 3 I/F), just for info, the SATA controller in the Seagate Thunderbolt adapter is also a SATA 3 controller.

Samsung 830 as internal boot drive in MBP (filesystem is HFS+)



Samsung 830 installed in Seagate TB adapter (filesystem is HFS+)



So you can see you loose about 20% performance via Thunderbolt. When now the internal speed test from Black Magic is now compared to HD Tune values for Samsung 830 512GB, then the ratio for comparison should be approximate clear.

Hope this helps, for what applications Thunderbolt can be used, or not. (regarding HD speed test in VM's, HD Tune, ATTO and others when used in VM's, seem to show wrong results (a lot to high values, higher then the native speed) this needs further investigation how to make proper speed measurements (seems all disk I/O is buffered), so only real files copied can be used)

//Wolfgang