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

Status
Not open for further replies.
I think this best sums it up....


I believe PS4 will keep Sony's gaming division healthy during these losses for the next 3-4yrs (or at least until shrinks arrives and/or BoM becomes significantly cheaper) after PS5 arrival. And I'm pretty sure Sony has been socking away a certain amount of PS4 hardware/software/service profits towards such PS5 losses.
So basically just over a year or 2, into next gen.
Obviously if you start the 3-4 years from now.
 
The average gamer is 34 years old and 72 percent are age 18 or older. http://www.theesa.com/about-esa/industry-facts/

"New details on the primary audience for the PlayStation Network in the US has revealed that the average user is 28 years old and is likely to be in the "middle- to higher-income range." https://www.gamesindustry.biz/articles/psn-average-user-age-is-28-in-the-us

I wouldn't say average console gamer would be turned off by a $499 machine if the value is there, not even if an alternative $299 weaker box is available. The average gamer can pay, and he wants value.
 
Last edited:
  • Like
Reactions: JPT
So basically just over a year or 2, into next gen.
Obviously if you start the 3-4 years from now.

The PS2 run was 13yrs long, and supported by Sony for 7yrs after the PS3 launch. And the PS3 run was 11yrs long, and supported by Sony for 4yrs after the PS4 launch. If anything, the PS4 run may out last the PS2 run. So nothing odd about those statements. If something is generating revenue and profit, you keep supporting it until it doesn’t. Makes business sense to me…
 
Obligatory DF Article about latest circus ... https://www.eurogamer.net/articles/digitalfoundry-2019-ps5-tech-demo-footage-surfaces

The presentation is based on Marvel's Spider-Man, developed by Insomniac Games. It starts off by showing a series of 'fast travel' warps around the game's New York map, with the PS4 Pro taking 8.1 seconds to complete a single jump. By contrast, the next generation hardware completes the same process in just 0.83 seconds. Roughly speaking, loading times are improved by an order of magnitude.

It's interesting to see the demos described in the Wired article actually running, though I suspect we are seeing a slightly different presentation. Spider-Man loading times for PS4 Pro in the Wired piece was described as 15 seconds with interstitial screens, and Spider-Man himself present on-screen. The demo here has no screens and no Spider-Man, but completes in just eight seconds. However, the next-gen score of 0.8 seconds is consistent.

Beyond that, no further details on the next generation project were revealed. However, a slide from the firm's presentation reiterates the key points made in the Wired article: an all-new CPU/GPU, solid-state storage, 3D audio, back-compat, disc support, 8K compatibility and - yes - ray tracing.

---

 
Again why would Sony take a loss when they could sell you a streaming option??? The only reason to take a loss would be for a stop gap solution until cheaper option is available in bulk.

Because they can't sell you a streaming option. Early adopters are the "hardcore" gamers and they want an actual console.

All this streaming talk like the latency issue and image quality has been solved. Also early adopters play a lot so the need to have a minimum of a TB data cap a month will be a problem for lots of people around the world.

In fact in my opinion the streaming option isn't the cheap option not yet at least. It will be a percentage of the early adopters who are interested in the tech and have uncapped fibre.
 
Put my thoughts together last week, didn't see a dedicated thread so following others. Predictions:

Lockhart
  • 5.4TF Navi (15% Raw improvement over polaris)
  • 12GB GDDR6
  • Zen 2 8c16t 2.8Ghz
  • 750GB SS (Could they really go 500-640GB?) Games sizes not increased as much due to non duplication and higher compression due to gpu decompression. Was in favour of tiered storage on the scorpio, but may just use some form of fast start while copying from external to SS when using external.
  • $300 & 350 ($300 is a discless version)
  • Runs 1X in compatibility mode, for XO, OG & X360
Getting to 5. 4TF shouldn't be a problem CU & Ghz wise so cooling and power delivery should be cheaper also.


Anaconda
  • 12TF Navi
  • 16GB GDDR6
  • Zen 2 8c16t 3.0Ghz
  • 1TB SS
  • $500
  • Runs 1X in compatability mode, maybe forced 16x AA, OG & 360 games 16* resolution boost instead of 9
Be nice if released an updated Xbox sdk to support Anaconda for 1X games, to allow simple doubling of resoultion for games releasing this year etc. So no post patching required for games without Scarlett dev box's etc.

PS5
  • 10TF Navi
  • 16GB GDDR6
  • Zen 2 8c16t 2.8Ghz
  • 1TB Big push on SS, as xbox 1P limited by pc so they have greater scope of how they can use it above and beyond simply loading and streaming assets faster.
  • $450
  • PS5 boost mode.
Not sure I think that Anaconda is worth $50 more in terms of mindshare so will be tough to compete, but lockhart is good value IMO.

For multi-plats 2TF difference will probably equate to nothing depending on how things like RTRT is implemented. So MS would need to fund some graphically impressive games.
They would also benefit from being lead platform, then down rezing to PS5 would be slight benefit (DF). When PS5 lead platform then games would just end up the same.

All monolithic, possibly Scarlett is MCM to allow cloud blades to change IO & memory controller to use HBM which would benefit running costs etc

All custom GPUs, and labeled as custom zen CPU but won't be much customization there at all.
 
Last edited:
Btw, do you think avx2 can be used in games in a usefull way now that zen2 have it ?
whats your actual question

will games use 128bit AVX2 instructions instead of SSE4.x probably.
will games use 256bit AVX2 instructions instead of SEE4.x , who knows

but how much of game simulation is using SIMD int and not SIMD fp ( jaguar supports AVX)*

*yes i know AVX2 brings some FP/general ops but its mainly int based.
 
My question was does avx2 can help in gaming scenarios.

The biggest change is micro-architectural, Zen 2 has full width 256bit FP units, doubling the FP throughput compared to Zen. It also has full width datapaths to/from D$.

The biggest change going from AVX to AVX2 are the gather and permute capabilities of the FP units.

Cheers
 
Which one should be better for games: optane-like SSD with granular access or regular nvme SSD with 4GB DDR4 caching? Let's say both have raw speed of 4GB/s read.
Probably the 4GB DDR4 caching, in terms of performance, not sure in terms of pricing.
 
4gB for read/write caching is not a lot. On a console I doubt they will enable write caching anyway, because of possible dataloss. For read only... I don't know 4gB won't have a lot of cache hit imo.
 
Put my thoughts together last week, didn't see a dedicated thread so following others. Predictions:

Lockhart
  • 5.4TF Navi (15% Raw improvement over polaris)
  • 12GB GDDR6
  • Zen 2 8c16t 2.8Ghz
  • 750GB SS (Could they really go 500-640GB?) Games sizes not increased as much due to non duplication and higher compression due to gpu decompression. Was in favour of tiered storage on the scorpio, but may just use some form of fast start while copying from external to SS when using external.
  • $300 & 350 ($300 is a discless version)
  • Runs 1X in compatibility mode, for XO, OG & X360
Getting to 5. 4TF shouldn't be a problem CU & Ghz wise so cooling and power delivery should be cheaper also.


Anaconda
  • 12TF Navi
  • 16GB GDDR6
  • Zen 2 8c16t 3.0Ghz
  • 1TB SS
  • $500
  • Runs 1X in compatability mode, maybe forced 16x AA, OG & 360 games 16* resolution boost instead of 9
Be nice if released an updated Xbox sdk to support Anaconda for 1X games, to allow simple doubling of resoultion for games releasing this year etc. So no post patching required for games without Scarlett dev box's etc.

PS5
  • 10TF Navi
  • 16GB GDDR6
  • Zen 2 8c16t 2.8Ghz
  • 1TB Big push on SS, as xbox 1P limited by pc so they have greater scope of how they can use it above and beyond simply loading and streaming assets faster.
  • $450
  • PS5 boost mode.
Not sure I think that Anaconda is worth $50 more in terms of mindshare so will be tough to compete, but lockhart is good value IMO.

For multi-plats 2TF difference will probably equate to nothing depending on how things like RTRT is implemented. So MS would need to fund some graphically impressive games.
They would also benefit from being lead platform, then down rezing to PS5 would be slight benefit (DF). When PS5 lead platform then games would just end up the same.

All monolithic, possibly Scarlett is MCM to allow cloud blades to change IO & memory controller to use HBM which would benefit running costs etc

All custom GPUs, and labeled as custom zen CPU but won't be much customization there at all.

My only reservations about Lockhart being what amounts to a 1X with a better CPU is that I feel like that's a tough sell to anyone who already owns an X. I know 1X owners would be the minority of Xbox One owners, but I would also argue that they are probably the most likely to be early adapters, and if you can't show they how much better a new generation is they aren't going to make the leap.

Perhaps Lockhart launches later, with Anaconda launching alongside PS5 getting the "most powerful" good press and Lockhart launching later as the value option a little later. Sort of like video cards are marketed.
 
Optane is (at least) over $1-4 per GB where NAND is less than ten cents per GB. For the price of 1 TB NAND you can get <100GB Optane, but then you need some other form of bulk storage, adding at least $30.

It's not even a choice.

As for performance, a typical NAND controller has eight channels each running @400MB/s (ONFI 4), and in the future 600/800MB/s (ONFI 4.1). Optimize the read accesses for the NAND page size, and you'll see 3-6 GB/s bandwidth.

Use main memory for caching NAND

Cheers
 
Status
Not open for further replies.
Back
Top