stick with the very last word you said. ps4 hardware back compat means they can only have a certain number of CU's unless they figure out a software solution like MS has. This makes no sense what so ever, ps4 has 18 CU and pro has 36, so clearly they are able to emulate the ps4 on the pro. No...
www.resetera.com
Someone over on Era linked to this. "Interesting" indeed.
Crazy fantasy time, but could be looking at a base + Pro PS5 scenario at launch? Maybe VFXVeteran was onto something, who knows. $399 base PS5 with 8.2-9.2TF and 16GB GDDR6, $599-$649 PS5 Pro with 14TF and 24-32GB GDDR6?
Either way it'd solidify Oberon, but give a Pro model with two of them in a Crossfire-like setup. Keep it at 36CUs active for optimal yields (32CU active each GPU for Pro model). Early adopters would probably go for the Pro model, so favor that in the split say 75/25 favoring PS5 Pro. Keep yields and power dissipation good for mass production on Pro model (and a good ratio in terms of TF to memory bandwidth).
Would give a 256-bit memory bus on base PS5 and "512-bit" memory bus via 2x 256-bit buses on each of the Oberon GPUs in Pro model. 24GB as an even mix of 8x 1GB and 8x 2GB chips. Could theoretically go for 32GB with all 2GB chips but would probably push the MSRP up north of $600.
Still not as good plan as just going base PS5 at launch and a Pro model 2/3 years down the road, I just thought it interesting the 5700 was mentioned here since Oberon fits its profile very clearly.
If such a scenario played out it'd be interesting what MS's response would be in pricing for XSX. I highly doubt it will go for MSRP above $499 but in this scenario they could probably even try for $450 especially if Lockhart is not coming at launch.
So does this confirm PS5 will be RDNA2.0? Since from RDNA2 seems to be the only that can do RT natively?
RDNA2 doesn't mean it's 7nm EUV. They could be on RDNA1 process node but have implemented many RDNA2 features onto their custom APU such as RT, VRR and VRS.
In fact I think that's exactly what MS has done, i.e also on 7nm and not EUV. RDNA2 doesn't require EUV and it's not even suggested it needs a big chip. AMD will be making many mobile APUs in the near future for example and would want to bring features like RT and VRS to those produces in iGPUs of their own, so it would make 100% sense for RDNA2 features to be implemented on smaller GPUs as well as the bigger ones.
If not then AMD would've botched up their approach big time. And I doubt they'd make that type of mistake.
