Nar..the secret sauce is the 40m+ user base
PS4 Pro Speculation (PS4K NEO Kaio-Ken-Kutaragi-Kaz Neo-san)
- Thread starter mpg1
- Start date
-
- Tags
- sony sony roach motel
- Status
Please... Don't rile up the forum nubs. They won't catch the sarcasm; they'll just take your word as gospel and repeat it for all eternity, or well, half of it anyway.
After all, that's roughly how long it took the 14+4 rumor to die out...Twins, Basil. Twins.
yeah polaris not looking great in terms of heat/power consumption...at least compared to pascal
I think the thinking was..."if this doesn't at least match a 970 in reviews then we're fucked."
It's too high for 14 nm process. But how about TSMC 16nm?
If fact with TSMC 16nm we may really have a 5.5TF neo at 100 W power consumption.
Love_In_Rio
Veteran
Yes.There were rumors of waiting for a next year launch to wait for a 5,5 Tflops console to be cheaper and more affordable.This fits with waiting for TSMC chips and save in the cooling solution thanks to less consumption.
I think it's clocked high for the fact that it's being used for VR & so on that's using the GPGPU functions.
New upscaling method patented by Sony on January 2016. They call it "uprendering". That's bold. Credit to Thuway on GAF. Could it be Neo secret sauce?
http://www.freepatentsonline.com/20160005344.pdf
http://www.freepatentsonline.com/20160005344.pdf
Lol. Dunno why that made me laugh.
Anyway great to see them take it in this direction. Nothing pains me more to see 4x more power to increase the native output of pixels on screen.
Sent from my iPhone using Tapatalk
If you keep reading all the patents from SCE/SIE you'll go mad.
The best source I think is freshpatents, it is updated more timely than freepatentsonline....
http://stks.freshpatents.com/Sony-Computer-Entertainment-Inc-nm1.php
http://stks.freshpatents.com/Sony-Computer-Entertainment-America-Llc-nm1.php
The best source I think is freshpatents, it is updated more timely than freepatentsonline....
http://stks.freshpatents.com/Sony-Computer-Entertainment-Inc-nm1.php
http://stks.freshpatents.com/Sony-Computer-Entertainment-America-Llc-nm1.php
Rikimaru
Veteran
Yes. This is about PS2 emu.
Why? I don't mind either way but if the choice is 4.2TF November and 5.5TF 6 months later it seems a no-brainer - especially will all the Scorpio talk.
There was a rumour on GAF that they will announce a slim and neo at the same time, not sure if it's still going round but it might explain all the confusion around the '2 options'?
I'm eager to hear what Sony have to say about Neo. There are many unknown leaving the TFLOPS figure aside. The system, has to have more than two shader engines to reach 36 CUs. With the release of Polaris we have now a better view about what the transition to 14nm brings on the tables wrt power consumption and performances.
Sony target for its GPU throughput sounds right. The bandwidth figures are low as the RX480 (with faster clocks but also faster V-RAM but the same 32 ROPs) seems to benefit nicely from memory overclock. With the bandwidth figure, 32 ROPs make sense (though I would favor 48 ROPs). Supposedly Sony stuck to 8GB of RAM and a 256 bits bus.
I see three possible set-up for the ROPs and the inner layout of the system:
1) 4 shader engines each consisting of: 9-10 CUs and 8 ROPs + geometry engine
2) 4 shader engines each consisting of : 9-10 CUs and 16 ROPs + geometry engine
3) 3 shader engines each consisting of 12-13 CUs and 16 ROPs + geometry engine
The PS4 is 2 shader engines consisting of 10 CUs and 16 ROPs + geometry engine
To me the layout that looks like the easier match to existing hardware is the choice number three. It is also in "line-ish" with the increase in memory bandwidth and measured improvement Delta Color Compression provides and it would allow the 48 ROPs to operate as well as the 32 ones in the original design. I would discard the choice 2 as the number of ROPs (64) does not ring true with the bandwidth figures we heard. The layout 3 would also allow the use of salvaged parts (disabling one shader engine altogether and some CUs in the remaining 2).
It could prove a much better design than Scorpio makes it looks, Rx480 is not 4K proof and power hungry, the performances Sony target are sound in the context of existing shipping GPU parts.
Sony target for its GPU throughput sounds right. The bandwidth figures are low as the RX480 (with faster clocks but also faster V-RAM but the same 32 ROPs) seems to benefit nicely from memory overclock. With the bandwidth figure, 32 ROPs make sense (though I would favor 48 ROPs). Supposedly Sony stuck to 8GB of RAM and a 256 bits bus.
I see three possible set-up for the ROPs and the inner layout of the system:
1) 4 shader engines each consisting of: 9-10 CUs and 8 ROPs + geometry engine
2) 4 shader engines each consisting of : 9-10 CUs and 16 ROPs + geometry engine
3) 3 shader engines each consisting of 12-13 CUs and 16 ROPs + geometry engine
The PS4 is 2 shader engines consisting of 10 CUs and 16 ROPs + geometry engine
To me the layout that looks like the easier match to existing hardware is the choice number three. It is also in "line-ish" with the increase in memory bandwidth and measured improvement Delta Color Compression provides and it would allow the 48 ROPs to operate as well as the 32 ones in the original design. I would discard the choice 2 as the number of ROPs (64) does not ring true with the bandwidth figures we heard. The layout 3 would also allow the use of salvaged parts (disabling one shader engine altogether and some CUs in the remaining 2).
It could prove a much better design than Scorpio makes it looks, Rx480 is not 4K proof and power hungry, the performances Sony target are sound in the context of existing shipping GPU parts.
- Status