News & Rumours: Playstation 4/ Orbis *spin*

Status
Not open for further replies.
Translation from Gaf

http://www.neogaf.com/forum/showpost.php?p=170430251&postcount=720

Here some quick translation of that guy's teardown:

New power supply ADP-200ER replaced ADP-240CR and is 80g lighter.
New BD drive.
New 8Gb GDDR5 chips (Samsung K4G80325FB-HC03)
New main board SAC-001 replaced SAB-001 and has 60mm less width.
The APU CXD90037G remains the same.

Following chips existed at previous models are now missing:
- Fujitsu USB 3.0-SATA bridge
- Marvell Ethernet LSI
- Genesys Logic USB 3.0 hub controller LSI

The secondary processor CXD90036G has direct connection to all of them (the guy speculates Sony may have eliminated USB3.0 - SATA conversion).

HDMI transmitter is MN864729 by Panasonic.

Boot system LSI is changed to A00-C0L2 (there seems to be some new patent regarding system control)

Blu-ray drive control subboard is now missing and controller LSI R9J04G011FP1 & motor driver BD7764MUV are now on mainboard.
Cooling fan remains the same (KSB0912HE).

Heatsink's base is changed from aluminum to steel.
The heatpipe existed at heatsink's base remains, but one existed at aluminum fins is now missing.
Click to expand...
 
Seems to be both a new APU and a new secondary processor:

Launch CUH-1000 PS4:
APU: CXD90026G
secondary: CXD90025G

CUH-1100 PS4:
APU: CXD90026AG (minor revision A)
secondary: CXD90025G (same)

CUH-1200 PS4:
APU: CXD90037G (major revision 37)
secondary: CXD90036G (major revision 36)

Could it be 20nm? Probably not, it would be physically smaller...
 
20nm may not be worth it for so large a chip. Given TSMC's track record for 20nm, it'd likely be far better to stick to 28nm yields, and wait for 14/16nm to ramp up in volume (after Apple takes it's share of the uh... pie).
 
Chinese article mentions that APU still has same same area size. 28nm remains, but since new chip has new code name, it was probably new more efficient stepping.
 
It's a really small motherboard. GDDR5 doesn't need trace length matching, that's really helping.

This is not perfect scale, but good enough for drama. :oops:

DieH@rd said:
For clarity sake, thats WiiU, PS4 v2 and Xbone.
Click to expand...

G5vZjKy.jpg
 
Last edited:
The board area dedicated to memory on the Xbox one is .... not small.

Suddenly I have an urge to hold a Duke controller.
 
It will be interesting to see mobo of the first console that will utilize silicon interposer and HBM... although I presume that MS will manage to design even that to be big and clunky with zero internal airflow pathways.
 
It probably couldn't be much worse than the 360 Slim. As already mentioned, the biggest issue is DDR3 trace length matching. HBM obviously doesn't have that problem.
 
It is very interesting that the "Gflops/watt" of PS4 GPU is relatively high in 28nm GPUs. In CUH-1200 PS4 system power is about 125W, so the GPU power is around 80W. It means that "Gflops/watt" is more than 20, which can beat most GPUs of R9-300 series. In other words PS4 is very power-efficient. How does SONY or AMD achieve that?
 
RDGoodla said:
It is very interesting that the "Gflops/watt" of PS4 GPU is relatively high in 28nm GPUs. In CUH-1200 PS4 system power is about 125W, so the GPU power is around 80W. It means that "Gflops/watt" is more than 20, which can beat most GPUs of R9-300 series. In other words PS4 is very power-efficient. How does SONY or AMD achieve that?
Click to expand...
800Mhz, lower voltage.

Fury Nano will do same thing.
 
Better binning, lower temperature, less leakage = higher performance per watt.

Anandtech talked a bit about it their latest Fury X review.
 
Status
Not open for further replies.
Back
Top