Originally Posted by cssorkinman
I've got a few applications that cry out for a mini setup like that. It would be easiest to use a touch screen but I'd like the touch screen to connect wirelessly to the mini. I've got the wireless VGA part covered, but do you know of a ultra dependable ( its for industrial control) wireless usb to use for the touch screen inputs?
Originally Posted by Mega Man
also i would like to add has anyone tested the 295x2s pcie2.0 vs 3.0 i think that would be interesting
Same CPU to GPU traffic.
GPU to GPU traffic wouldn't even make it out of the card.
Originally Posted by Liranan
Originally Posted by KyadCK
Ya no... That is simply not true. PCI-e 3.0 is not just a doubling of the speed, and with AMD's new Crossfire system it's far more important than people give it credit for.
Just picked me up a pair of these; http://www.newegg.com/Product/Product.aspx?Item=N82E16856501007
I would have loved a Jaguar based version, but they either don't exist, don't meet the spec I need, or are too expensive.
I did have a fun told-you-so moment though. Someone was pestering me to get these for the past month (they won't be needed for another few weeks), and I said that it's better to wait if you have that long because who knows when a deal will pop up. 20% off is nice, no?
The latency reduction of PCI-e 3 is truly that great that it improves multi card setups by a noticeable margin?
How is Bay Trail for every day use? I need a laptop replacement and was thinking a BT tablet over another pretty heavy laptop but the lack of RAM has put me off so far. I've tried the Transformer T100 and it felt pretty smooth but I wasn't able to test it with lots of tabs.
It's not about that specifically.
PCI-e 2.0 x16 has a theoretical bandwidth of 8GB/s. It is encoded at 8/10, meaning 8 of every 10 bits are packet data. This results in a true max bandwidth of 6.4GB/s.
PCI-e 3.0 x16 has a theoretical bandwidth of 15.75GB/s. It is encoded in 128/130. This results in a true max bandwidth of 15.5GB/s.
In reality, PCI-e 3.0 is 2.4 times as fast as the same number of lanes in 2.0. In a situation where a card is not bandwidth bottlenecked, the amount of time taken up by transferring data is reduced by 58%. This does not sound important, but at higher framerates or higher resolutions in AMD's crossfire, it can make a difference. The more limited you are by bandwidth and not by GPU power, the larger the impact.
More importantly than speed however, is cost. Adding more lanes to a motherboard increases the number of layers it must be and it's complexity. The more complex a motherboard is, the more it cost. The more lanes you need to have on your chipset, the more complex, larger and more powerful (and power hungry) that chipset becomes. Larger is more expensive to make. Complex is more expensive to develop. Power consumption requires larger (and more expensive) heatsinks to cool them.
There's a reason AMD's 990FX and Intel's X-series cost more than 970 and Z-series.
I'll let you know when I get them. These will be my first BayTrail, but it theoretically competes with Jaguar well, and I do own that.