So I upgraded my HD

Sorry, posted a mistake. Congrats OP.

I replaced the hdd samsung 830-256 gb and installed the Apple SSD 256 gb in RAID 0 speed is very decent

ps2wiz said:

I always thought you need identical drives for RAID0?

Click to expand...

Identical helps. Same size is the basic requirement.

To those claiming thunderbolt is pointless and USB3 is good enough: see benchmark.

250 megabytes per sec is 2.5 gigabit (or so).

I.e., due to protocol overhead, CPU consumption, etc. and all round just cheap crappy design, USB3 is getting nowhere near its rated 5 gigabit throughput above.


Don't get me wrong, USB3 is fine for cheap peripherals that are not latency or bandwidth sensitive. But for high speed stuff it is inherently deranged, and USB 3 is just a faster version of such.

ps2wiz said:

I wish I had a Thunderbolt enclosure I could have swapped the same HD into and ran a benchmark.

Click to expand...

Yeah, i'd be keen to see it also.

Given that it is essentially PCIe on a cable, you should see native SATA3 speed, assuming your device has a SATA3 controller on the end of the thunderbolt link.

So you took the stock HD out of your iMac and replaced it with the Intel 520 SSD? Can you replace the stock HD with any SSD? I thought the 21.5" models couldn't have a SSD installed in place of the HD?

throAU said:

To those claiming thunderbolt is pointless and USB3 is good enough: see benchmark.

250 megabytes per sec is 2.5 gigabit (or so).

I.e., due to protocol overhead, CPU consumption, etc. and all round just cheap crappy design, USB3 is getting nowhere near its rated 5 gigabit throughput above.


Don't get me wrong, USB3 is fine for cheap peripherals that are not latency or bandwidth sensitive. But for high speed stuff it is inherently deranged, and USB 3 is just a faster version of such.

Click to expand...

If properly implemented USB 3.0 can outperform Thunderbolt with similar devices attached to both types of interface. I have seen 440MB/s or more on a single USB 3.0 connection with appropriate controller and SSD, which is pretty good.

But of course it is only a 5GB connection, instead of 2x10. It is also peanuts, and uses cheap passive cables, as opposed to ludicrously overpriced with even more ludicrously overpriced cables. They made a serious misjudgement in coming up with a design that needs active cabling IMHO. Especially considering doing so was completely unnecessary - quad rate infiniband carries 10GB per channel (same as TB) with passive copper cabling, and total interface bandwidth up to 300GB.

TB peripherals continue to be expensive and uncommon, compared to USB 3.0, and this fact significantly increases the risk that it may ultimately fail to establish itself as the interface of choice for high bandwidth requirements. The jury is out for me. It might well go the way of FireWire and fall into disuse having been superseded by something faster and cheaper. It would be relatively simple to give USB 3.1 a speed bump to 10 or 20 GB and if that happened in the next couple of years, TB is toast.

Edit: Well well well, I just read this:

10 GB USB 3.0 may be just around the corner:

http://www.businesswire.com/news/ho...d-USB-USB-3.0-Performance-Double-Capabilities

10 GB USB 3.0 with the same cheap USB 3.0 connectors and cables, fully compatible with USB 1.0, 2.0 and 3.0 devices. Very bad news indeed for Thunderbolt.

That said, I don't see USB 3.0 becoming the defacto high bandwidth interface. TB will have more serious threat from others in that department.

Chippy99 said:

If properly implemented USB 3.0 can outperform Thunderbolt

Click to expand...

Just... no.

Not if both are configured correctly.

USB has a heap of protocol overhead thunderbolt does not have and is a heavily CPU driven bus.

USB3 (and 4, and 5, etc) will become the "end user" standard because it is "good enough" for most people.

But it's still garbage compared to thunderbolt and not usable for the same types of things at the high end due to horrible latency, high CPU usage and generally inferior performance (which is inherent to its design as it involves protocol changes from PCIe to USB to device protocol and back rather than just PCIe -> device and back as with thunderbolt).

Thunderbolt does and always will cost more because it uses actual hardware rather than your machine's CPU to do a bunch of stuff. Its a better, more flexible design. This is reflected in the cost.

They're aimed at two entirely different market segments, and not competitors.

Comparing them and saying one will kill the other is like saying that say, the Ford Taurus will kill the Ferrari 458 because it can transport people cheaper.


Thunderbolt will be the de-facto high end interface because it is PCIe. Anything else that connects to the PCIe bus (i.e., any expansion device or peripheral, including USB3) on your motherboard can theoretically be driven over thunderbolt with the same driver. Any other high speed external bus design is going to be connected to the pc via PCIe and will have higher overhead than thunderbolt, which IS external PCIe.

Unless PCIe is replaced as a system bus, thunderbolt is here to stay for certain, and even after that happens it will remain for some time like PCI and ISA slots did when they were replaced as the main system bus in PC hardware.

10 gig USB is still playing catchup to 20 gigabit full duplex Thunderbolt 2.

Cute cat <3

I've heard that there is a need to pick an SSD with compatible temp sensor, otherwise the drive overheats/or results in the fans spinning at high speed constantly.

Was this a consideration?

throAU said:

Just... no.

Not if both are configured correctly.

USB has a heap of protocol overhead thunderbolt does not have and is a heavily CPU driven bus.

USB3 (and 4, and 5, etc) will become the "end user" standard because it is "good enough" for most people.

But it's still garbage compared to thunderbolt and not usable for the same types of things at the high end due to horrible latency, high CPU usage and generally inferior performance (which is inherent to its design as it involves protocol changes from PCIe to USB to device protocol and back rather than just PCIe -> device and back as with thunderbolt).

Thunderbolt does and always will cost more because it uses actual hardware rather than your machine's CPU to do a bunch of stuff. Its a better, more flexible design. This is reflected in the cost.

They're aimed at two entirely different market segments, and not competitors.

Comparing them and saying one will kill the other is like saying that say, the Ford Taurus will kill the Ferrari 458 because it can transport people cheaper.


Thunderbolt will be the de-facto high end interface because it is PCIe. Anything else that connects to the PCIe bus (i.e., any expansion device or peripheral, including USB3) on your motherboard can theoretically be driven over thunderbolt with the same driver. Any other high speed external bus design is going to be connected to the pc via PCIe and will have higher overhead than thunderbolt, which IS external PCIe.

Unless PCIe is replaced as a system bus, thunderbolt is here to stay for certain, and even after that happens it will remain for some time like PCI and ISA slots did when they were replaced as the main system bus in PC hardware.

10 gig USB is still playing catchup to 20 gigabit full duplex Thunderbolt 2.

Click to expand...

I have two words to say to you:

Sony
Betamax

You seem to imagine that technical superiority (assuming there is any, which I do not actually accept anyway) is a magical guarantee of success, when of course that is untrue.

I understand USB and Thunderbolt have different design goals and target use-cases, of course. Intel are not idiots and they knew what USB is about when inventing Thunderbolt. But do you think the average buyer knows this? The fact is - as opposed to our opinions - that USB 3.0 has completely thrashed Thunderbolt in terms of sales and market acceptance. It was earlier to market, it is vastly cheaper, more flexible (in terms of things you can hang off it), has wildly more manufacturers supporting it and is pretty fast. And by the way, I stand by my performance assertion. Show me a single SSD on any Thunderbolt device achieving 440MB/s read speeds. If Thunderbolt is technically better then that is not being translated to any meaningful performance advantage in most use cases.

When USB goes to 10 GB/s later this year, it will make matters only worse for TB adoption. 99% of the market are not buying fast raid arrays, they are buying a USB flash drive or 1TB 2.5" backup disk, and for that USB is already plenty fast. As more and more move to SSD, and SSD gets faster, 10 GB USB will be plenty fast for that as well. Who gives a toss if Thunderbolt is "doing a bunch of stuff" that USB isn't doing? The mass buying public vote with their wallets and will buy the cheapest product that offers them what they want/need. So Thunderbolt does things for you that otherwise your CPU has to do. (Does it??? Like what???) Anyway, who cares, CPU cycles are cheap as chips and getting ever cheaper.

Imho Thunderbolt would have already have been on the edge of extinction as far as the mass market is concerned, had Apple not decided to back it. Maybe it would have had more chance if it was not so very late in terms of peripherals availability and not so expensive and didn't need $40 cables.

But anyway, luckily for TB Apple are backing it, so we will see if it manages to survive.

But as I have already posted, it's a tough gig. Most consumers will continue to buy USB 3.0 because it is cheaper, available and does what they want. The Data Centre has Infiniband and 100G Ethernet and I don't see any openings for Thunderbolt there. Its slower than both (dramatically so in the case of Infiniband) and has zero market presence or support there.

As I type this, I am actually becoming even more sure that TB is ultimately doomed. There is not enough space left for it in the market. Apple will struggle on with it as they did with FireWire, but ultimately it's a dead duck.

I agree about the moggy though ;-)

Chippy99 said:

I have two words to say to you:

Sony
Betamax

Click to expand...

Pretty good analogy actually.

Betamax stayed relevant at the high end for recording as I understand it.

But most end users don't need it and won't pay for it.


I never claimed thunderbolt will be for the mass market. But technically speaking, between USB and thunderbolt - they aren't even in the same league.

TB still stay around, because for external PCIe expansion, there is no other option, and USBx will NOT do what you can do with thunderbolt. But.... I agree, most users won't need it (and their ports will never be used). But the ports will still be fitted to higher end machines.

Most users don't need fiberchannel, RAID or 40 gig ethernet (yes, I have servers with it) either, but those are all still on the market and doing well.

"If properly implemented USB 3.0 can outperform Thunderbolt with similar devices attached to both types of interface. I have seen 440MB/s or more on a single USB 3.0 connection with appropriate controller and SSD, which is pretty good."

It's taken a little while, but finally folks are coming round to this reality.

The fact is that for most people (not all), USB3 is a far more important feature on Macs (in particular) and computers in general than is Thunderbolt. USB3 was Apple's most significant upgrade in the last five years.

Thunderbolt may offer advantages that USB3 can't (extension of the internal PCI bus), but the advantages it offers are required by very few users....

----------

"USB has a heap of protocol overhead thunderbolt does not have and is a heavily CPU driven bus."

Nope -- entirely wrong.

Refer to "USAP" insofar as it applies to USB3.

USB2 had "overhead". USAP eliminates the overhead by implementing SCSI protocols on the USB bus that relieve the processor of the tasks that were demanded of it using USB2.

USB4 wil be even faster.