Do you think there will be a mid gen refresh console from Sony and Microsoft?

[I'mDudditz!]

They're showing different sets of data. One is CP2077 and the other is average RT performance across a quite of games.

Yes and the averages are based on tests all running at max settings which means textures especially will expose 10gb vram limits of the 3080 and cause to perform less than what the actual GPU is capable of. I'm amused but not surprised you're leaning on this argument. It's the completely opposite logic the overwhelming majority of posters take when discussing PS5 performance in comparison to 3070 8gb. In these situations, defendants of the 3070 point to vram limits, but I guess here we are not applying same reasoning because..... "reasons".

So we've moved from "easily surpass" to "decimate the 7900XTX" now have we?

I see no problem using these two descriptions interchangeably.

And your evidence for this is a leakers guess about an assumption mixed in with some Sony magic sauce of your own imagining?

Comments like these are the reason threads tend to go off the rails. I didn't say anything about any magic sauce, so why introduce such toxicity? I referenced a Cerny patent that suggests the inclusion of dedicated hw for ray traversal in a future iteration of a PS console. At best, you can push back that a patent isn't confirmation but even then that would be a petty rebuttal in a speculation thread.

 

[Riddlewire]

That thing is an eyesore, regardless of the side panels. The molds should be repurposed for something more appropriate.

 

[snc]

First tweet says "but not RT stuff" and then the 2nd one says "but with RT stuff"..

first says rdna3.5 is without rt improvements, second is that he guess (as me ;d) that ps5pro will be rdna3.5 but with rdna4 rt

 

[pjbliverpool]

Yes and the averages are based on tests all running at max settings which means textures especially will expose 10gb vram limits of the 3080 and cause to perform less than what the actual GPU is capable of.

The games in question are:

Control
Cyberpunk 2077
Deathloop
F1 22
Far Cry 6
Metro Exodus
Resident Evil Village
Watch Dog Legion

Do you have any evidence, at all, that any of these games (let alone all of them) are in any way limited by the 3080's 10GB VRAM... at 1080p?

I'm amused but not surprised you're leaning on this argument. It's the completely opposite logic the overwhelming majority of posters take when discussing PS5 performance in comparison to 3070 8gb. In these situations, defendants of the 3070 point to vram limits, but I guess here we are not applying same reasoning because..... "reasons".

The "reasons" would be:

1. The 3070 has 8GB. 10GB is 25% more than that and far less a limiting factor.
2. Those argument are only ever made in relation to a relatively small number games which provably are limited by VRAM capacity on an 8GB GPU at consoles settings. What proof do you have that any of the above games are limited by 10GB VRAM at 1080p?

Comments like these are the reason threads tend to go off the rails. I didn't say anything about any magic sauce, so why introduce such toxicity? I referenced a Cerny patent that suggests the inclusion of dedicated hw for ray traversal in a future iteration of a PS console. At best, you can push back that a patent isn't confirmation but even then that would be a petty rebuttal in a speculation thread.

The magic sauce reference was in relation to your wild assumption that 1. these hardware traversal engines that you assume will be part of the PS5 Pro will catapult its RT performance well beyond that of Ampere (which would be required for a 6800XT level part to "trounce the 7900XTX in RT" and 2. if they don't, PS5 special sauce API's will do that anyway.

 

[Tkumpathenurpahl]

If it’s 7nm. The only thing slimmed was everything else I guess. Curious to see the tear down here.

If they couldn’t get the “slim” variant down below 7nm, that means the cost savings are not there yet and I have Low expectations on a pro variant arriving.

I suppose it could be argued that a Pro wouldn't be as price sensitive, and so a more expensive node is more feasible.

If the thing in the pictures is right though, Sony have really fumbled. "V" is the kind of simple design aesthetic that PS2-4 Sony would've nailed, and the initial design just barely gets the job done - with many people understandably thinking it absolutely fails at even that. To shit the bed more on a redesign though? Oh dear

My disappointment with the leaked image (if true) is partially because I got carried away with the elegance of separating the console from its ODD.

They could've quite easily gone the route of an external power brick, integrated into the optional ODD attachment to mitigate the criticism that power bricks always garner. A visibly smaller console would surely be feasible with both the ODD and PSU removed in terms of volume, and the latter also in terms of cooling requirements.

 

[GhostofWar]

Looking at the recent info and discussions around the console cpus being far less performant than most of us expected, I have some questions. If all the pie in the sky predictions about using zen4 and rdna4 are accurate how much of a bottleneck is memory bandwidth going to be? RT is very bandwidth hungry right? Will a zen4 cpu crush zen2 when both are using gddr memory because we see in the pc space the uplift in cpu performance was the most noticable with ddr5 and/or with tight timings.

Wish this was alot closer to release, fascinated to see what ends up being possible.

 

[DuckThor Evil]

Do you have any evidence, at all, that any of these games (let alone all of them) are in any way limited by the 3080's 10GB VRAM... at 1080p?

You could just use the 3090 in your argument. It's a stronger GPU than the 3080, does not have the potential VRAM issue and is behind the 7900XTX in the average graph.

 

[Seanspeed]

If it’s 7nm. The only thing slimmed was everything else I guess. Curious to see the tear down here.

If they couldn’t get the “slim” variant down below 7nm, that means the cost savings are not there yet and I have Low expectations on a pro variant arriving.

Well this revision is supposed to come within like the next couple months, while the Pro is slated for sometime in 2024 according to rumors, which are certainly sounding quite credible by now. So not really on the same timescale.

And unless this new revision comes with a reduced pricetag(I dont think it will), I'm expecting the Pro to be like $700. So Sony can make up for cost aspects there. A proper Slim will come, though. I'm pretty confident of that if they're going with a Pro model. The cost aspect will eventually make sense given everything else they can reduce on the system and packaging.

 

[Globalisateur]

If it’s 7nm. The only thing slimmed was everything else I guess. Curious to see the tear down here.

If they couldn’t get the “slim” variant down below 7nm, that means the cost savings are not there yet and I have Low expectations on a pro variant arriving.

I think from others rumors they already stopped manufacturing 7nm APUs for PS5. Current 1200 models are using 6nm. 2000 models should logically use 5nm. 2024 Pro could use 4nm.

 

[Seanspeed]

Looking at the recent info and discussions around the console cpus being far less performant than most of us expected, I have some questions. If all the pie in the sky predictions about using zen4 and rdna4 are accurate how much of a bottleneck is memory bandwidth going to be? RT is very bandwidth hungry right? Will a zen4 cpu crush zen2 when both are using gddr memory because we see in the pc space the uplift in cpu performance was the most noticable with ddr5 and/or with tight timings.

Wish this was alot closer to release, fascinated to see what ends up being possible.

The biggest performance jump from Zen 2 to Zen 3 came from the doubled L3 available to each core. The way the PS5's(and Xbox's) current APU is configured, this isn't really possible since there's basically two separate 4 core complexes end-on-end on the side of the die. This is useful for space-saving when you want to maximize room for a chunky and blocky GPU on the same die, but they will need to rethink this kind of configuration if they want to enable a unified L3 setup, which will be difficult when they're trying to pack in an even larger GPU.

They could potentially use the higher density libraries that they're using with Zen 4c to help things out there.

Either way, I dont think making the CPU as powerful as possible is really a super high priority. The existing Zen 2 CPU will likely still surprise us in terms of what it's capable of in time. It's an area where 'console optimization' still tends to be notable. Just look at what devs did with those low speed, pitiful Jaguar cores last generation. So any improvements moving to Zen 4 cores(which I see no reason to call this a 'pie in the sky' prediction at all) would be more welcome rather than critical.

 

[Seanspeed]

I think from others rumors they already stopped manufacturing 7nm APUs for PS5. Current 1200 models are using 6nm. 2000 models should logically use 5nm. 2024 Pro could use 4nm.

Some people will say 7nm even if they mean 6nm. Same with 4nm and 5nm. I'll often just say '7nm family' or something like that even if talking about 6nm because it's mostly the same thing.

Also, I think it's pretty obvious the new '2000' model coming this year isn't gonna be 5nm, or else they could have done a proper Slim model, not just a revision.

 

[McHuj]

Also, I think it's pretty obvious the new '2000' model coming this year isn't gonna be 5nm, or else they could have done a proper Slim model, not just a revision.

I don't think that's true at all. What you're seeing is the reduced scaling capabilities of new nodes. The PS5 already started at a very high power consumption compared to previous generations and comes with an expensive cooling setup. It makes zero sense for Sony to continue with that by minimizing enclosure size, it's all about cost reduction. A 5nm PS5 will still end up in the ~150 W power consumption range.

For PS5 Pro, I don't think 5nm offers enough gains.

For next-gen, I hope they delay enough until a 2nm process is ready because they will need it.

 

[iroboto]

I think from others rumors they already stopped manufacturing 7nm APUs for PS5. Current 1200 models are using 6nm. 2000 models should logically use 5nm. 2024 Pro could use 4nm.

I'm not entirely convinced that these have occurred. I don't doubt that yield is likely to have improved, but the cost to move to new nodes is super expensive. I think PS5 seeing more yield is not a function of moving to newer and smaller nodes, but just a function of having a better developed yields. Until someone brings out real receipts that nodes have dropped, I'm not going to believe it.

PS4 had multiple revisions as well, and only moving to slim did node changes occur the cost is enormous to take on moving to a new node, so the cost per transistor has to be able to make up that difference for them to want to do it. That being said, I don't think they'll be moving to 3 separate nodes, 2 of which are silent changes.

 

[Globalisateur]

I'm not entirely convinced that these have occurred. I don't doubt that yield is likely to have improved, but the cost to move to new nodes is super expensive. I think PS5 seeing more yield is not a function of moving to newer and smaller nodes, but just a function of having a better developed yields. Until someone brings out real receipts that nodes have dropped, I'm not going to believe it.

PS4 had multiple revisions as well, and only moving to slim did node changes occur the cost is enormous to take on moving to a new node, so the cost per transistor has to be able to make up that difference for them to want to do it. That being said, I don't think they'll be moving to 3 separate nodes, 2 of which are silent changes.

Even if the cost makes it as expensive to produce each APUs compared to previous nodes they'll still save money on cooling, size (weight of plastic and board components) and shipping (reduced size and weight).

 

[iroboto]

Even if the cost makes it as expensive to produce each APUs compared to previous nodes they'll still save money on cooling, size (weight of plastic and board components) and shipping (reduced size and weight).

Outside of having more yield has there be any changes to ps5 cooling since ? Is it still using Liquid Metal? Etc

 

[Flappy Pannus]

Either way, I dont think making the CPU as powerful as possible is really a super high priority. The existing Zen 2 CPU will likely still surprise us in terms of what it's capable of in time. It's an area where 'console optimization' still tends to be notable. Just look at what devs did with those low speed, pitiful Jaguar cores last generation. So any improvements moving to Zen 4 cores(which I see no reason to call this a 'pie in the sky' prediction at all) would be more welcome rather than critical.

Agreed, it's more in the 'nice bonus' camp rather than something that should be expected for a midgen upgrade - especially if there's a decent implementation of the mythical FSR3 which could help those CPU limited scenarios.

 

[Globalisateur]

Outside of having more yield has there be any changes to ps5 cooling since ? Is it still using Liquid Metal? Etc

Still T3000 cooling but they use smaller heatsink / less copper in both new models.

 

[iroboto]

Still T3000 cooling but they use smaller heatsink / less copper in both new models.

That’s not a particularly strong sign they moved to a smaller node. They may just realize that they overcompensated and now pulled back.

 

[DuckThor Evil]

That’s not a particularly strong sign they moved to a smaller node. They may just realize that they overcompensated and now pulled back.

I think this is the original source for the 6nm shrink. It was widely reported almost a year ago and generally accepted afaik. I haven't really seen anyone question the validity of it before you.
https://www.angstronomics.com/p/ps5-refresh-oberon-plus

 

[Seanspeed]

I don't think that's true at all. What you're seeing is the reduced scaling capabilities of new nodes. The PS5 already started at a very high power consumption compared to previous generations and comes with an expensive cooling setup. It makes zero sense for Sony to continue with that by minimizing enclosure size, it's all about cost reduction. A 5nm PS5 will still end up in the ~150 W power consumption range.

For PS5 Pro, I don't think 5nm offers enough gains.

For next-gen, I hope they delay enough until a 2nm process is ready because they will need it.

I'm quite confident this new PS5, if it turns out as it is currently being shown and rumored, is not a 5nm shrink. While there's technically 'reduced scaling' with new process nodes nowadays, the jump from 7nm to 5nm is still a proper jump and quite significant. We haven't hit any hugely damning diminishing returns just yet.

Even if we're only talking a reduction to 150w, that's still less than an Xbox One X could draw, and that was quite a compact box.

Also, minimizing enclosure size *is* a big part of reducing overall costs. You can use a smaller and simpler motherboard, smaller PSU, smaller cooling hardware, and less plastic with the casing. And with all this, you can make the overall packaging and retail box smaller and lighter, which is pretty significant in terms of reducing costs for warehousing and shipping.