Also furthermore... thunderbolt 2 is 20 gigabit, so 16 gigabit FC is no problem. 32 gigabit would be feasible via channel bonding. If/when it comes. In the future. When we may well have 40 gigabit thunderbolt 3 anyhow.
Big assumption. It could be a disk attached to a fiber channel array, it could be via 10 GbE (yes, via thunderbolt attached NIC) SAN, it could be a PCIe based SSD in an external thunderbolt connected PCIe bay.
Why install and waste PCIe lanes on low end consumer grade SATA ports when they will be unused by 90% plus of their customer base? Thunderbolt gives you the choice of whatever you want connected externally, including things that don't exist yet.
Because they would be SATA3, and with thunderbolt you have the bandwidth to run future versions of SATA or SAS (or fiberchannel, etc.) instead, as future needs dictate.
Because only the components that 90% of people are used in the enclosure, why make a big box that will sit empty, or include things 90% will not use? All the boxes you're talking about can sit next to the desk on the floor where your regular workstation case would sit. Whilst this sits on your desk with easy access to your USB and thunderbolt ports.
Given the orientation of the heat-sink and airflow arrangement and the low speed the fan generally runs at, i suspect this machine could run fanless (via convection) well enough to alert you to the fact that the fan has failed and/or throttle itself until you fix it.
Depends how you look at the prospect of external expand-ability. If the machine had slots that were 12 inches long a 13 inch card wouldn't fit (i've run into similar issues with PC cases), and the cooling for the external devices, power consumption, etc. is entirely up to the third party OEM.
... and remote X sucks balls really bad anyway. It's passable on gigabit ethernet, anything slower than that and it is pretty horrible. Meanwhile, even RDP is usable over 64 kilobit.
Because it only works for very generous definitions of "works". If you've never used anything else maybe remote X seems like it rocks, but vs. ICA or RDP (even the versions from 1999) its performance is abysmal.
If you're too blind to see where the industry is going (hint: look at trends rather than specific OEMs), you're pretty clueless. So no, the fact that it is intel vs. Nvidia isn't really relevant. He who has the best power:watt for a complete system will win out. If NVidia get a system on a chip out that competes and can punch it out in volume, they'll survive. If not... they won't.
We're in a transitional period - broadwell should be very interesting. The edram is also usable by the CPU portion, and don't forget the discrete GPU + CPU combo consumes 1.5-2x as much power.
Performance per watt, NVidia isn't even close. I'm just waiting for multi-socket portables - you could stick 2x intel CPU/GPUs in a machine in the same thermal/power envelope as CPU + Discrete GPU, and have a lot less complexity with GPU switching for power purposes. Plus twice as many CPU threads.
Pretty much. I predict that within 3-5 years the discrete video card market will be much like the discrete sound card market is now. Very small and rarely purchased except for niche users. Nvidia should be scared, but then the writing has been on the wall for a good 5 years now already.
Slashdot Mirror
Funny, i've been using RDP over frame links as slow as 64kbit since 1999...
Also furthermore... thunderbolt 2 is 20 gigabit, so 16 gigabit FC is no problem. 32 gigabit would be feasible via channel bonding. If/when it comes. In the future. When we may well have 40 gigabit thunderbolt 3 anyhow.
Have you heard of fiber-optic cabling perchance?
Pretty much. And if they wanted local disk connectivity they'd be using SAS anyway, not SATA.
Not 20 Gb, but there are faster-than-SATA PCIe based SSDs available that you could easily hook up by a Sonnet thunderbolt->PCIe expansion box.
Uh.... any serious storage system back then used SCSI, not just apple.
Big assumption. It could be a disk attached to a fiber channel array, it could be via 10 GbE (yes, via thunderbolt attached NIC) SAN, it could be a PCIe based SSD in an external thunderbolt connected PCIe bay.
Why install and waste PCIe lanes on low end consumer grade SATA ports when they will be unused by 90% plus of their customer base? Thunderbolt gives you the choice of whatever you want connected externally, including things that don't exist yet.
Philips is also proprietary btw.
Because they would be SATA3, and with thunderbolt you have the bandwidth to run future versions of SATA or SAS (or fiberchannel, etc.) instead, as future needs dictate.
Because only the components that 90% of people are used in the enclosure, why make a big box that will sit empty, or include things 90% will not use? All the boxes you're talking about can sit next to the desk on the floor where your regular workstation case would sit. Whilst this sits on your desk with easy access to your USB and thunderbolt ports.
Thunderbolt can be channel-bonded.
10 GbE is do-able via thunderbolt, as is fiber-channel.
Given the orientation of the heat-sink and airflow arrangement and the low speed the fan generally runs at, i suspect this machine could run fanless (via convection) well enough to alert you to the fact that the fan has failed and/or throttle itself until you fix it.
Depends how you look at the prospect of external expand-ability. If the machine had slots that were 12 inches long a 13 inch card wouldn't fit (i've run into similar issues with PC cases), and the cooling for the external devices, power consumption, etc. is entirely up to the third party OEM.
... and remote X sucks balls really bad anyway. It's passable on gigabit ethernet, anything slower than that and it is pretty horrible. Meanwhile, even RDP is usable over 64 kilobit.
Because it only works for very generous definitions of "works". If you've never used anything else maybe remote X seems like it rocks, but vs. ICA or RDP (even the versions from 1999) its performance is abysmal.
... running a massive service such as that with elevated privileges...
If you're too blind to see where the industry is going (hint: look at trends rather than specific OEMs), you're pretty clueless. So no, the fact that it is intel vs. Nvidia isn't really relevant. He who has the best power:watt for a complete system will win out. If NVidia get a system on a chip out that competes and can punch it out in volume, they'll survive. If not... they won't.
You mean like the biggest gaming market, the mobile gaming market, where integrated GPU is the norm?
We're in a transitional period - broadwell should be very interesting. The edram is also usable by the CPU portion, and don't forget the discrete GPU + CPU combo consumes 1.5-2x as much power.
Performance per watt, NVidia isn't even close. I'm just waiting for multi-socket portables - you could stick 2x intel CPU/GPUs in a machine in the same thermal/power envelope as CPU + Discrete GPU, and have a lot less complexity with GPU switching for power purposes. Plus twice as many CPU threads.
Pretty much. I predict that within 3-5 years the discrete video card market will be much like the discrete sound card market is now. Very small and rarely purchased except for niche users. Nvidia should be scared, but then the writing has been on the wall for a good 5 years now already.
... OS X just keeps getting better.
On the contrary, it ADDS a second vector for malware to the machine.
You know how DPI works, right? Screen size is irrelevant as to whether or not DPI matters.