Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
TB - is it the future?
- Thread starter Varmann
- Start date
- Sort by reaction score
Looks pretty nice although the storage capacity to cost ratio with SSDs is still a bit steep.
On Amazon, this is £279 in the UK. I bought a Samsung 250GB SSD not long ago for £179 so that's £100 more for the case and a TB cable. Not actually too bad I guess.
Aware of that, I guess I wasn't clear. I was implying the reason why they DID NOT ship 10GbE on the new pro. They could have, but it would have been wasted for the 90% who do not have 10GbE to the desk. Thunderbolt can provide 10GbE on that port IF someone needs it, if they don't they can use it for fiberchannel or something else.
Meanwhile, the next thunderbolt upgrade will be 40Gbit.
And you have ignored the issues of CPU overhead, protocol translation and latency, all of which USB sucks at. Badly. More bandwidth over USB will only exacerbate the problem and further the performance advantage of thunderbolt. There's no way you could effectively do a 3d accelerator over USB for example - the protocol overhead is horrendous. People seem to confuse link signalling speed with performance, and it's not the case. Thunderbolt gets close, but USB is about 50-60% of the actual wire speed once overhead is taken into account for a typical peripheral.
Compare people's SSD performance on a USB3 port vs native SATA3.
Thunderbolt is faster than SATA3, and can actually provide a SATA3 port as easily as anything else.
Depends on the industry you're in.
People seem to see the word "pro" and think heavy CPU or GPU user. A lot of people (who are professionals) aren't. I'm a professional network engineer for example. I use a Mac Pro. Could I get by with an air? Probably for most things. The reason I went pro was simply for RAM/storage capacity, screen size and gigabit ethernet. CPU is way faster than I actually need and GPU is irrelevant for me for actual work.
But if you are a professional 3d modeller, producer, etc then a laptop won't generally cut it unless you have no alternative.
If you have a desk, you wan't a desktop for more power and storage expandability.
But yes, I agree - thunderbolt GPUs will be a win.
Get rid of the discrete GPU in my machine. it's useless on battery anyway (which is what it will be running off when I'm not at a desk) as it chews too much power. Replace the consumed space with battery capacity.
A desktop GPU over thunderbolt will be faster, even if it's only got 10Gb going to it. Future thunderbolt versions will be better.
Last edited:
I guarantee you it is not gigabytes.
Because that would be 1 terabit, and that is still beyond the reach of all but the most high end of experimental backbone network interfaces.
And as far as acer dropping TB on their machines? So what. Acer are, and have been for a number of years, a producer of pretty garbage hardware.
No laptop or common garden variety desktops include 10GbE yet, even though 40GbE is here and 100GbE is coming.
10GbE has been around for well over 5 years. I guess that means high speed ethernet is a failure, right?
Last edited:
I agree, it's a safe bet that it's gigabits, but to nitpick 10 gigabytes would not be 1 terabit, it would be 80 gigabit.
Pretty sure one of the posts above was referring to usb doing "100 gigabites"
100 gigaBYTES = ~1 terabit (yes, 8 data bits per byte, but there is a bit of signalling overhead), so the actual data you consume to get 800 gigabit of throughput is closer to 1 terabit.
Yes, and my post was to demonstrate that it most likely WASN'T bites=bytes.
What is your actual point?
To me this would be the break out product thunderbolt and one I'm sort of surprised we haven't seen as yet.
Someone should make a Thunderbolt monitor with a MXM card slot for a GPU (or even a full size card slot). On something like the new MacPro it would free up the two GPU in the machine for openCL crunching. For a laptop it could give the machine the power to drive larger screens than the embedded could handle on it own.
And I agreed with you. My point was that 10GB is not ~1Tb, I stated that pretty clearly.
I can certainly see apple putting a GPU in a future version of the thunderbolt display (one big reason I've held off on buying one).
They use essentially the same enclosure for an entire iMac. There's a lot of thermal headroom in there for a fairly hefty GPU, and if they give it plenty of RAM the bus speed won't be so much of an issue, especially if it can be uplinked by multiple TB ports.
----------
And I didn't state that either.
100GB (which was the figure I was discussing) ~ 1Tb including protocol overhead.
You never mentioned the 100GB figure, 10GB is what is implied by the previous post. Also, overhead is deducted not added to the figure, an 800Gb/s signaling rate is including the over head, that is, the actual data rate would be lower.