Baseless Next Generation Rumors with no Technical Merits [post E3 2019, pre GDC 2020] [XBSX, PS5]

Its only January, but looks like the dumbest statement of the year competition is effectively over, thanks all for playing, see you in 2021

 

There are quite a few people who feel that an alternative approach to calculating some lighting effects is primarily going to be important to forum warriors.

 

Some kind of semi-desperate move ? Yes it can be..... I've this feeling too....

 

It was the only thing MS could do with the Xbox One when it became clear they were going to be walloped in a specs comparison, increase the clocks for the GPU and CPU. MS could do it because the size of the box, big fan and external PSU meant it could deal with the increased thermals. It's the main reason I want to see what the PS5 form factor is. If they did re-jig the clocks maybe they also changed the box and cooling to accommodate it.

 

Nah. I think the RT section of RDNA2 and/or Sony's implementation for PS5 will be some Wizardry, hint, hint.

 

That's not how product development work.

A change in clock once they start getting yield data means the yield is better than expected. A reworked chip (respin) is to solve issues, improve power consumption, increase clock, etc...

The slightly upgraded clock of xb1s was to allow HDR patches without impacting performance. It was not a reaction either.

 

The original Xbox One had both CPU and GPU clocks increased over what was first announced just before launch though:

https://arstechnica.com/gaming/2013/09/xbox-one-gets-a-cpu-speed-boost-to-go-with-its-faster-gpu/

This was back in 2013, so nothing to do with the bump the One S got.

 

Even a 8 to 9TF PS5 might do that, some are having fire on their fingertips already now

 

Gigabyte RX 5700 XT Gaming OC review

https://www.tomshardware.com/reviews/gigabyte-rx-5700-xt-gaming-oc-8g/

Average clocks - 1826MHz
Average wattage - 186W

Not bad for 40CU card. Note that, compared to say RX580 (PC equivalent found in XBX), its still pulls less from the wall.

I am starting to believe 1800MHz for 36CU chip is relatively easy to hit, especially on N7P. 2.0GHz remains high target, but lets wait and see what end of the year brings.

EDIT

Which now made me think...On N7P, 1.8GHz will probably be a sweet spot. Current 36CU based 5600XT clocked at 1730MHz will pull around ~140W. This is with less RAM, so we can add ~5W if it had additional 2 chips for 8GB. Thats fairly low and makes it highest performing GPU per watt on the market.

This makes me wonder, if we are to assume PS5 is 2.0GHz, why would MS go with ~1675MHz? Allegedly 20CUs more, yes, but still...300MHz less? That means Sony is maximizing their silicon much more, and this is critical in semi custom SOCs.

I see only 3 posilibilies :

1. XSX is 1675MHz with 56CUs, PS5 is 1.8-1.9GHz with 36CUs
2. XSX is 1675MHz with 56CUs, PS5 was 1.8GHz with 36CUs but was pushed through revisions to hit considerably more. This would put XSX some 100MHz under sweetspot, but with 56CUs it is understandable. PS5 would in this case end up 100+MHz over sweetspot.
3. XSX is 1675MHz with 56CUs, PS5 is 1600-1700MHz with 56CUs

First point is by far most logical one. PS5 would have 10-13% higher clocks, but also 35% less CUs. This would in turn result in 25-30% less TF. Not saying I believe it, but this would still make it fairly competitive chip (~15% faster then Vega 64), and would fit into console form factor/die size/TDP limit like a glow.

Second point doesn't make alot of sense and this is why I doubt retail PS5 is 2.0GHz. In Github leak, 2.0GHz was mentioned along ThermTripLimit for Oberon A0. I am not sure if they were shooting for highest possible clocks during native regression test, but if they are going for 2.0GHz they are probably going for performance at efficiency loss. This is not regular occurrence in console space, so I am doubtful.

Third point...I dunno. We have no data for any chip related to Sony that has 50+CUs so I am doubtful. Oberon is pretty clearly pointing at 36CU chip, it is chip used in Flute and in comparison to Ariel it seems to be the chip going for production. Inclusion of 256bit bus would also point to a chip that was never meant to provide double figures (IMO).

Really cannot wait to see what they cooked up...I think MS prepared themselves very well not to lose power crown this time around, but I also think Sony came to play completely different game. MS went with 320bit bus which gives them option to counter anything Sony did with speed of their chips on 256bit bus, but also in case Sony went with 320bits (and nothing points to that direction).

Having 56CUs at relatively conservative clocks to hit 12TF means they could, in theory, up the clocks just enough to bridge the gap that might have existed in case Sony were shooting for the sun, which I think they haven't.

This is possibility that opens up only if they are going with 2 SKU strategy. Having 400mm2 chip means it will be expensive and power hungry (for console standards). It will also mean they will be able to produce considerably less chips then if they went with die size that was ~20% smaller. To counter that, they would need SKU that would go opposite direction and which they would be able to produce in bigger numbers then if they went with 60% bigger chip (say 200mm2 v 330mm2). Average of these two SKUs would result in enough console in stocks, without having to worry about power or price, as you got both end of spectrum covered.

Which brings me to PS5...If PS5 goes ABOVE of what PS4 and PS4 Pro did on their release (only on CPU/GPU side) , 8-9TF would fit very well. Considering Sony has concentrated their talk on RT, 3D Audio, haptic controller and bespoke SSD, I think they were looking to gain competitive advantage in other areas as TF increase they required would not fit console form/TDP factor. I think they came to play completely different game. 12-13TF console doesn't fit. It doesn't fit at all. Even ~9TF Navi is very much pushing the boundaries. But...if you go with 2 SKU strategy, suddenly its really not that big of an issue. You will cover your bases, be able to produce enough consoles and won't have to worry about die size. Your premium SKU will aim for enthusiast market, while cheap one will be bought en masse.

Very, very interesting prospect. If Sony can price this "hypothetical" console at $400, they can win. They would probably be losing money on it, but would eat costs (say $50-$60). MS would be firmly in $500 place, while offering same graphics with 30-40% higher resolution (similar to XBX v Pro, except PS5 would be a match and perhaps better in quite a few areas bar GPU).

For people quoting me with arbitrary conditions such as

Don't.

36CU chip they would be using on 7nm node, even without RT space on die, is around 10% bigger then variant they used for PS4Pro on 16nm node. This hypothetical GPU will also use more watts then GPU found in Pro, let alone PS4. RDNA CUs =/= GCN CUs.

If you are into arbitrary conditions, feel free to disprove NG consoles using CPUs with "same", 8 cores, and "only" 2x increase in RAM.

 

Great work as usual. ThermTripLimit. Does that mean 2ghz is a boost clock?

i.e 1.8ghz base/game clock, 2ghz boost.

 

Unfortunately, I have absolutely no idea. I am merely speculating. Exact name of file is following :

 

If that gigabyte card is a binned sample. It doesn't necessarily translate to mass production.

 

No biggie, 5600XT is 36CU chip that will not pull more then 150W at ~1700MHz. I think 5600XT cards might be on N7P, not sure though. It is currently best perf per watt card on market.

 

Yup. X1X proved that being most powerful wont win if your base console is weaker. I said it along, the biggest minefield (and also possible opportunity) for MS is Lockhart. They could end up with 4TF base, 12TF Pro, and 9TF PS4 beating them all. Just like this gen MS had 1.3TF base, 6TF Pro, and 1.8TF PS4 emerging the winner. There's going to be so many possible different variables though, like release dates, prices, whether Lockhart or Anaconda is perceived by consumers as the "base" console, whether Lockhart is even released, the list goes on, very hard to predict, it's not directly analogous to current gen, because in this case the "pro" Xbox would apparently launch first, the "base" lockhart could be substantially cheaper than base PS5 where base XBox cost roughly the same as base PS4, the list of variables is long etc...

Just saying, dont count out 9TF PS5.

 

Investigating The Potential Xbox Series S APU | Playstation 5 APU Spec & Rework Discussion

 

Assuming Lockhart launches the same time as XSX it will be interesting to see how the market reacts to having the choice of three different machines . Particularly if each machine is ~$100 apart.

 

Efficiency? Yield? Wattage design limits?