PlayStation 4 Pro SoC Discussion

[Sweepr]

How PlayStation Neo and the original PS4 will co-exist

We've already described how Sony is looking to offer higher resolution support for 4K screens with the Neo hardware, and how higher frame-rates, more stable frame-rates, improved graphics fidelity and further visual features are encouraged. It's a topic we'll be returning to in due course as more information comes in. However, In terms of hard and fast technical guidelines for Neo rendering, Sony offers this:

- Games running in Neo mode must operate at a native rendering resolution of 1920x1080 (1080p) or higher.
- A game's frame-rate must meet or exceed its equivalent performance level on base PlayStation 4 hardware.

It seems that there are no new guidelines for ensuring rendering or performance standards on the existing PlayStation 4. One worry is that emphasis may shift to Neo, resulting in poorer experiences for the older hardware. We hope that the sheer weight of the user base ensures that the 'base' PS4 continues to get the care and attention it deserves.

www.eurogamer.net/articles/digitalf...ion-4k-neo-and-the-original-ps4-will-co-exist

 

[Sweepr]

New leaked documents:

http://wccftech.com/playstation-4-neo-leaked-document-reveals-specs-developer-guidelines

 

[magomago]

New leaked documents:

http://wccftech.com/playstation-4-neo-leaked-document-reveals-specs-developer-guidelines

so the real question: is the xbox rev better than whats listed here?

 

[NTMBK]

New leaked documents:

http://wccftech.com/playstation-4-neo-leaked-document-reveals-specs-developer-guidelines

I found this one most interesting:

"Checkerboard rendering"... i.e., not actually rendering at 4K. More like 2K in terms of actual unique pixel count.

 

[Sweepr]

so the real question: is the xbox rev better than whats listed here?

Looks like it is. They won't hit >6 TFLOPs of GPU power using Polaris when RX 480 alone already goes north of 150W under load, so the custom iGPU is probably based on Vega.

- >6 TFLOPs Vega based vs 4.23 TFLOPs Polaris based iGPU
- >320 GB/s (384-bit? GDDR5) vs 218 GB/s (256-bit GDDR5)
- 12/24GB vs 8GB
- 8x Jaguar/Zen? vs 8x Jaguar @ 2.1GHz

 

[therealnickdanger]

I'm not sure I believe there's a cut down Vega in either console. I'm more inclined to believe we're seeing modified Polaris + CPU TFLOPs or possibly even some conflated numbers utilizing some additional processing unit or compression method.

Maybe it will be Vega, but that just doesn't seem cost effective for a console. How much are the Xbone 2 and PS5 going to cost? Still $399? $599?

It's a sure bet that neither console will do actual 4K gaming outside of simpler arcade style games. 1080p/1440p upscaled is what we'll see. Still better than 900p upscaled.

 

[Sweepr]

I'm not sure I believe there's a cut down Vega in either console. I'm more inclined to believe we're seeing modified Polaris + CPU TFLOPs or possibly even some conflated numbers utilizing some additional processing unit or compression method.

Of course not a regular Vega 10, but probably a custom Vega based iGPU with more CUs than P10 at lower lockspeeds. At the announcement video they said 6 TFLOPs of GPU power. No way they're using the same 36 CUs Polaris-based config as Sony in late 2017. They would need +1.3 GHz (above RX 480), not to mention the fact that they probably want to beat the competitor on the specs sheet.

Maybe it will be Vega, but that just doesn't seem cost effective for a console. How much are the Xbone 2 and PS5 going to cost? Still $399? $599?

These are premium consoles, Sony itself describes Neo as a high-end PS4. Regular PS4/X1 are still available on the cheap, capable of running the same games.

 

[Sweepr]

Announced today:

- November 10 Launch Date
- MSRP U$399
- More than doubled GPU power
- Polaris architecture
- Boosted CPU clocks

No specifics given. To me this pretty much confirms the leaked specs from March are the real deal, and they are avoiding Scorpio comparisons.

 

[imported_bman]

Finally have console good for 1080p. I will get one for Bloodborne, The Last of Us remaster, and FFXII remaster.

 

[2blzd]

I thought 4k gaming was still not going to be possible and every marketing thing I read says otherwise.

I thought it was just going to use aspects of 4k, like HDR

 

[imported_bman]

They are going to make use of greater than 1080p resolutions and make use of scaling and temporal filter techniques. Given the distances people sit away from TVs, modern anti-aliasing techniques, and having game elements like the UI render at 4k the final output will look fine.

 

[Sweepr]

Specs are out:

http://www.jp.playstation.com/ps4/hardware/cuh7000bb01.html

Eurogamer was spot on. Some scary parts, it's bigger than the regular PS4 and listed as up to 310W power consumption.

 

[imported_bman]

310W! Is this thing still on 28nm?

 

[swilli89]

310W! Is this thing still on 28nm?

Doesn't matter at all as long as its cooled efficiently. My plasma TV uses about 800 Watts!

 

[imported_bman]

Noise though, I like my consoles to be quite.

 

[DTN107]

310W is similar to my old PS3 60GB console.

 

[railven]

310W! Is this thing still on 28nm?

I had the same question, and after reading AT's article, now wondering how much of Polaris is in it.

 

[Sweepr]

On the one hand, this highlights a fundamental weakness with PlayStation 4 Pro. Even though developers have access to another 512MB of RAM (presumably swiped for the vast 3GB system reservation), it's not enough to accommodate the 4K texture options that developers are increasingly offering with their titles. So in the case of Rise of the Tomb Raider then, it's no surprise that the PS4 Pro version offers assets in line with the existing Xbox One game, equivalent to the high texture quality setting on the PC build. It's a definite limitation - especially as Project Scorpio from Microsoft seems set to feature 12GB of memory - but on the plus side, in terms of this title at least, the downgrade is only really an issue in cut-scenes.

www.eurogamer.net/articles/digitalfoundry-2016-rise-of-the-tomb-raider-ps4-pro-vs-pc-comparison

 

[alcoholbob]

Kinda ironically though, the gap between the PS4 Pro and the top PC GPU is even bigger than the gap that existed in 2013 when the PS4 originally debuted.

The fastest PC GPU in 2013 was the 780 Ti, and it was about 2.25 times faster than the PS4 GPU at the time.

Now the fastest GPU in 2016 is the Titan X (Pascal), and it's something like 3.4 times faster than the PS4 Pro GPU.

 

[Dresdenboy]

Kinda ironically though, the gap between the PS4 Pro and the top PC GPU is even bigger than the gap that existed in 2013 when the PS4 originally debuted.

The fastest PC GPU in 2013 was the 780 Ti, and it was about 2.25 times faster than the PS4 GPU at the time.

Now the fastest GPU in 2016 is the Titan X (Pascal), and it's something like 3.4 times faster than the PS4 Pro GPU.

If there are no price limits, why don't you compare PS4 Pro to a SLI config then?

Comparing to pre built PC gaming systems at similar prices (and say w/o Win license and w/ comparably slow CPU) would make a little bit more sense (comparing apples to pears instead of pineapples).

 

[alcoholbob]

If there are no price limits, why don't you compare PS4 Pro to a SLI config then?

Comparing to pre built PC gaming systems at similar prices (and say w/o Win license and w/ comparably slow CPU) would make a little bit more sense (comparing apples to pears instead of pineapples).

I think it's relevant because Consoles up until the current generation were sold as a loss at launch with bleeding edge hardware, often something comparable the fastest GPU at the time. I'm comparing purely apples to apples from a pure technical standpoint. Also SLI is becoming very irrelevant these days due to poor scaling and due to the preponderance of APIs, often don't scale properly, or flat out don't work at all on the fastest API path in many AAA games. Why do you think SLI is relevant?

 

[Dresdenboy]

Eurogamer was spot on. Some scary parts, it's bigger than the regular PS4 and listed as up to 310W power consumption.

How does that number compare to PS4? How has it been measured?

 

[jpiniero]

How does that number compare to PS4? How has it been measured?

It's the rating of the power supply. The original came with a 250W one, and later was reduced to 230 W. The Slim ships with a 160 W.

 

[Dresdenboy]

I think it's relevant because Consoles up until the current generation were sold as a loss at launch with bleeding edge hardware, often something comparable the fastest GPU at the time. I'm comparing purely apples to apples from a pure technical standpoint. Also SLI is becoming very irrelevant these days due to poor scaling and due to the preponderance of APIs, often don't scale properly, or flat out don't work at all on the fastest API path in many AAA games. Why do you think SLI is relevant?

Was this true for XBox and PS since their first incarnations? Before Wii did Nintendo contribute to this "trend"? The Sony CEO changed and with him the goal to make losses first.

I'd compare at similar price points (whole system) instead. Also I brought up SLI just to up the performance while ignoring the price, too. If SLI doesn't work, why did some complain if it's not supported in some of the latest GPUs?

And are we talking GFLOPS or game performance at same settings? Nvidia brough up this topic at several occasions, but seemingly related to GFLOPS:

.

It's the rating of the power supply. The original came with a 250W one, and later was reduced to 230 W. The Slim ships with a 160 W.

Not knowing, whether they're going to reduce the 300W a bit (could also be related to component costs - cheaper ones might need higher ratings), we're talking about +20% to +30%. Hmm, how much of that might go to the different components and what would be the actual delta at the SoC. They might have tuned the SoC TDP a bit due to better cooling options in the larger case.

 

[Zstream]

It's the rating of the power supply. The original came with a 250W one, and later was reduced to 230 W. The Slim ships with a 160 W.

For the millionth time. The PSU rating is no indication of usage. Stop spreading this fud, it's been talked to death.