If I've read enough between the lines on PS4 we'll have
Full Power On - Duh!
Standby 1 - Instant Game Recovery as you described. The memories are put in refresh mode, and APU put in a low power state but not shut down. 2ndary chipset still active (controller for Network, HDD, BT - everything needed to background update, wake on Remote Access, etc.)
Standby 2 - APU and GDDR powered down. 2ndary chipset still active for updates, etc.
Standby 1 would still consume more power than Standby 2, but would leave your games in last known state. Delta in power consumption of leaving the GDDR5 in Refresh mode and keeping the APU in low power state vs having this off may be as high as 20W more (educated guess only)
I've heard Sony distinctly describe both scenarios, but whether when describing what I've stated as Standby 2, they left out the APU being in low power and GDDR3 in refresh intentionally since it was a different mode or was a simple omission as they where just trying to get the features out is still to be determined.
Results 76 to 88 of 88
for some reason you're pissing me off.
why do they need a separate chip to update it in low power mode?
No freaking clue, I just remember Sony talking about it at their original conference.
I just watched Cerny's Gameplay 2013 presentation. He actually contrasted their options of using a 256bit wide GDDR5 interface w/ no eDRAM (he didn't say it but it would have been the L3 memory inside - relatively large for inside but extremely small in size compared to the 8GB of GDDR5) vs a 128-bit wide GDDR5 interface.
Basically 176GB/s GDDR5 w/ no eDRAM vs 88GB/s GDDR5 w/ an eDRAM (@1088GB/s internal.)
I could only imagine this would have been an L3 memory operating a CPU speed w/ 1024bit access to each of the 8 cores. But it would have to be limited to something on the order of 10's of MB's in size (maybe 32 or 64MB.)
Cerny's comments was they choose the wider GDDR5 as it would allow for faster out of the box performance instead of waiting for developers to figure out which items they need to optimize to use the small eDRAM.
Unfortunately he didn't go into much details regarding the quantities of eDRAM he was thinking about or the L1/L2 cache architecture. I would have loved to see that, but at this point he was apologizing to the audience for getting too technical.
Separate controller is probably on the order of a couple of watts of power consumption, why the APU even at a very low power state w/ the GDDR active is probably on the order of 10-20W. Besides the APU doesn't have a native Ethernet support so why not have a low power 2ndary device use it.
Users Browsing this Thread
There are currently 1 users browsing this thread. (0 members and 1 guests)