Next Generation Hardware Speculation with a Technical Spin [pre E3 2019]

[function]

Honestly, SSD even with mlc or tlc seems fine with lifespan, despite all the writing involved. I've no doubt It won't be a problem on a console...

Almost all your operations on a console are reads, at least as things currently stand. You install once, but may end up reading that data hundreds or thousands of times. Games saves are typically very small.

A good SSD has an SLC cache with hundreds or thousands of times the write endurance of QLC (and better performance).

An SLC cache might also be modified with a partition for saves, or a section of scratchpad for games to use at will.

My hope would actually be for a decent chunk of SLC under developer control, and a larger section of generalised TLC or QLC that was a more automatic cache, perhaps taking hints from the game about what to cache, and in what order.

 

[Xbat]

But if they using a small scratch pad type system the writes are going to go up massively. It's possible if you play three different games in a day that could be 150GB plus a day.

 

[function]

64GB SLC and 256 GB TLC or QLC would be fine, IMO. I would expect to see 2 or 3 games in heavy rotation, at least in my experience, so that'd cut down on writes as for most loads the data would already be resident. Write endurance is typically lower on smaller drives with the same nand type, as there are physically fewer cells, but smaller drives make it easier to use nand types with higher endurance.

The SLC is just a bonus for games that want to use very large amounts dynamic, persistent data. The way games currently use persistent storage you probably wouldn't need it, saving you a bit of $$ on the SSD you chose.

Finally, a hybrid solution (SSD and HDD) wouldn't need to cache all the data to SSD. Video, music, dialogue (in all its uncompressed, multiple language glory) would be just as happy on the HDD, potentially saving many, many GB.

 

[anexanhume]

How is 1TB an acceptable size?

Mechanical HDDs only go to 2TB in the 9.5mm height category (i.e. what’s in consoles now), so I’d say it’s pretty acceptable.

Problem is a cache system is going to be hit by a lot of writing, won't that effect lifespan a lot?

Sufficiently sized, entire games are going to be sitting in there. I imagine it would only be a problem if you frequently rotated several large titles. I would bet they thought about that.

 

[BRiT]

I'm sure the entire game doesn't need to be there, on fast access. MS are, and I'm sure Sony will too, focus on chunking out the game assets more, for instance not all 12+ languages need be installed. Likewise not all game assets would be flagged for Fast access.

 

[manux]

If all these rumours pan out microsoft is going to be in software support hell. They would have roughly these hw platforms to support

xbox one, next xbox lowest end edition, xbox one x, xbox highest end edition, cloud xbox/streaming service and clients, windows.

I wonder if ms will really do above or if things like lowest end next xbox would either not happen or perhaps ms kills some of the products. Killing existing products could make existing user base feel negatively towards ms and make them consider playstation. Oh, and it might get even worse with midcycle upgrade.

Sony on the other hand has pretty clean portfolio with supposedly good difference in price&performance between different products.

 

[mrcorbo]

If all these rumours pan out microsoft is going to be in software support hell. They would have roughly these hw platforms to support

xbox one, next xbox lowest end edition, xbox one x, xbox highest end edition, cloud xbox/streaming service and clients, windows.

I wonder if ms will really do above or if things like lowest end next xbox would either not happen or perhaps ms kills some of the products. Killing existing products could make existing user base feel negatively towards ms and make them consider playstation. Oh, and it might get even worse with midcycle upgrade.

Sony on the other hand has pretty clean portfolio with supposedly good difference in price&performance between different products.

This is MS. Supporting a handful of platforms is child's play relative to what they are required to do in their other core businesses.

 

[Rootax]

Well, they will retire the xbox one and one x at some point... Like, Sony won't keep ps4 and pro alive a long time (from a first party pov / aaa games) after ps5 hit the market...

 

[BRiT]

If the difference is only for different resolutions rendered and output (1080p vs 4k), there isnt much more for them to need to support. Enable developers to have dynamic scaling solutions and you're mostly done.

 

[manux]

If the difference is only for different resolutions rendered and output (1080p vs 4k), there isnt much more for them to need to support. Enable developers to have dynamic scaling solutions and you're mostly done.

It's not that easy in practice. Each platform has to be qa'd separately. Also to give best possible experience it's probably not optimal to just change resolution as each platform and performance target is unique enough to require it's own optimizations&settings. Of course ms can half ass it but then the competing platforms will get better reviews for cross platform games.

 

[mpg1]

I see Xbox One devices being dropped fairly quickly in terms of developer support after the new consoles release. One S will be far to weak technically and One X will have far too small an install base to make it worthwhile to develop for.

 

[function]

It's not that easy in practice. Each platform has to be qa'd separately. Also to give best possible experience it's probably not optimal to just change resolution as each platform and performance target is unique enough to require it's own optimizations&settings. Of course ms can half ass it but then the competing platforms will get better reviews for cross platform games.

If you keep the underlying architecture the same you minimise need or even the options for different optimisations. Use one dev kit for both, identical API and compiler and development environment releases, and I bet large amounts of your QA testing would work either.

 

[manux]

Well, they will retire the xbox one and one x at some point... Like, Sony won't keep ps4 and pro alive a long time (from a first party pov / aaa games) after ps5 hit the market...

I believe sony might keep pro around because those products populate different pricepoints and hence are unique selling to people sensitive to various price points. Going to end of 2020 I wouldn't be surprised to see 199$ ps4, 349ps4 pro, 499$ ps5. Mid cycle refresh might be 599$ to spread the product range further and allow some improvements via increase in BOM.

Having BC+crossplay creates pc like ecosystem where there is more choice for consumer. Likely the higher end products sell a lot less and the cheaper models are mainstream where masses are. I suspect ps4 and pro will be around for a very long time still. That is unless consumers stop buying them but that seems unlikely as ps5 is expensive and is not going to come to 199$ any time soon if ever.

 

[manux]

If you keep the underlying architecture the same you minimise need or even the options for different optimisations. Use one dev kit for both, identical API and compiler and development environment releases, and I bet large amounts of your QA testing would work either.

What I'm trying to say is that even if architecture is same the optimal experience probably isn't resolution scaling. Different optimizations make sense for 1080p compared to 4k. To make experience optimal more than just changing resolution and framerate is needed. And even if one only scales resolution/fps the QA process for each platform needs to be done separately.

 

[turkey]

What I'm trying to say is that even if architecture is same the optimal experience probably isn't resolution scaling. Different optimizations make sense for 1080p compared to 4k. To make experience optimal more than just changing resolution and framerate is needed. And even if one only scales resolution/fps the QA process for each platform needs to be done separately.

That's already the case with PS4 pro on its own at this stage. The games have to adapt to screen resolution choice so you have a 4k QA path and a 1080 path. This would be similar on Xbox but with two different hardware SKU's.

 

[manux]

That's already the case with PS4 pro on its own at this stage. The games have to adapt to screen resolution choice so you have a 4k QA path and a 1080 path. This would be similar on Xbox but with two different hardware SKU's.

Plus pc and cloud for microsoft. It's quite many permutations to support even if it would turn out to be not trying to maximize every platform and just do qa effort to run through each game on every platform and do minor tweaks/fixes based on findings. There could easily be things like memory running out on lower end platform causing crash/corruption. Lowest work effort is to limit game features based on lowest end platform and leave performance on higher end platforms on table. These kinds of errors might happen on corner cases and if QA is not done proper for every platform gamers will get incomplete buggy games. That's a very short route to extreme consumer dissatisfaction. Especially when majority of your gamers are on the lower end platform.

Software is not easy problem that just works.

 

[BRiT]

Even some mostly-Sony developers have to QA for PC platforms too when they release for WinOS.

 

[AlNom]

I believe sony might keep pro around because those products populate different pricepoints and hence are unique selling to people sensitive to various price points. Going to end of 2020 I wouldn't be surprised to see 199$ ps4, 349ps4 pro, 499$ ps5. Mid cycle refresh might be 599$ to spread the product range further and allow some improvements via increase in BOM.

Having BC+crossplay creates pc like ecosystem where there is more choice for consumer. Likely the higher end products sell a lot less and the cheaper models are mainstream where masses are. I suspect ps4 and pro will be around for a very long time still. That is unless consumers stop buying them but that seems unlikely as ps5 is expensive and is not going to come to 199$ any time soon if ever.

Maybe it's a discussion for a different thread, but at the same time, they've already hit 90M+ sales with the vast majority of that on the base SKU. It's debatable if they can cost reduce to be able to comfortably hit $199 MSRP to make it a worthwhile (i.e. profitable) proposition for late gen adopters who may spend less anyway. The 4Pro Cer'nly may have some breathing room for a lower price cross-gen "entry" option.

 

[manux]

What I was trying to say is that to me the rumored microsoft hw stack doesn't make sense. Something is off and probably ms will simplify their product stack. This to me feels opposite to sony whose hw stack looks like something that could live for very long time without having to kill existing products as long as consumers keep buying. BC being there should keep ps4 viable as whatever investment consumer does for games/online friends/... carries over to the eventual ps5 upgrade. This is especially true if sony can drop ps4 price to 199$ or less. PS4 is still plenty expensive and I suspect there is casuals that would jump in if price was in impulse buy range.

In essence bc and cross play allows to not have to wait until price comes down. Compatibility should also make it less likely introduction of new consoles kills old one as any investment to old console will carry over. This is all new dynamics we are seeing in console market now. Old rules don't apply.

 

[BRiT]

The specific rumors about the HW stack dont make sense, but as a concept it could work if they kept things scaled appropriately where the difference is based on resolution needs.

You'd keep the same Storage I/O characteristics for latency, random reads and random writes.
You'd keep the same base RAM characteristics.
You'd keep nearly the same CPU in order to hit the same frame-times based on Frame Resolution targets.
You'd scale the Video RAM characteristics based on Frame Resolution targets.
You'd scale the GPU based on Frame Resolution targets.

Physically you wouldnt have separate VRAM and general RAM pools, but you'd have the quantity of it proportional to what's needed for the Frame Resolution targets.