Middle Generation Console Upgrade Discussion [Scorpio, 4Pro]

Status
Not open for further replies.
Most people were (and most rumors were pointing to Q4 2016), until Scorpio's specs started to leak and Sony decided to go completely silent about it on E3.
Now.. no one seems to know anymore.
Yes i believe so, the only thing that I find hard to grasp, it's that this holidays the push was VR VR and VR , and introducing another sku.would complicate things

If they manage to do it, my hat's off to sony
 
But one thought keeps popping on my mind, why Holidays 2017?

Phil stated they could have had a more powerful console than XBO for Holiday 2016, but they wanted to hit 6 TFOPs (due to developer feedback), and that wasn't possible for Holiday 2016. I believe Fallout VR was stated as one example where they would need 6 TFOPs.

I'm guessing they could have had something similar to PS4 NEO with a Polaris based GPU for 2016 if they'd wanted, but they took a pass on that.

Regards,
SB
 
Last edited:
A tin foil hat.
Do we wear other types at B3? Lol
Yes we do. The dunce cap. Everyone.

dunce_cap.jpg
 
Well I'm confused. Microsoft passed on using a the 36CU Polaris based gpu this year like neo but yet are using GDDR5(X) memory so it couldn't be Vega? So what is it?
 
There's no rule that they need to match any specific discrete GPU, and we do not have a full picture of what is in the IP pool(s) this time around.
The current gen console GPUs are lumped into the same IP pool as GCN 1.1 or (GCN2), but they have obvious differences and custom elements.

We could take each console maker's current IP level, current IP+customization, Polaris, Polaris+customization, GCN3, GCN3+customization, Vega, Vega+customization, and try mixing and matching scenarios.
 
More or less although I've never seen a '4K version' of a PC game, they are designed to scale from minimum requirements upwards. But this re-enforces the statement from a few days back and this bodes well that in addition to upscaling, downscaling/sampling may be an option for 1080p TV owners. Render high, down-sample for a nicer output.
or people just could buy a 4K TV.

TV LED 40" - Samsung, UHD 4K, HDR, Flat. 688€ (529£ UK)

upload_2016-6-18_1-21-48.png

http://tiendas.mediamarkt.es/p/tv-led-40-samsung-40ku6000-uhd-4k-h-1324312

http://www.rgbdirect.co.uk/Products/Television/LED-TV/SAMSUNG/UE40KU6000
 
With a 40" 4k, how close do you have to be to see the 4k resolution?

For most living rooms, below 84" looks like a waste of money.
I agree. A 40" TV is about bedroom size, not living or rec room size.
 
With a 40" 4k, how close do you have to be to see the 4k resolution?

For most living rooms, below 84" looks like a waste of money.
Computer monitor? I think so! :)

1 arms length baby!


Sent from my iPhone using Tapatalk
 
At 40", unless it's being used similar to a monitor (IE - just sitting 2-3 feet away) I'd rather have a 1080p set and downscale 4K content for free SSAA. Hell, 40" in a typical living room? I'd go for a 720p set and downscale for some really high quality free SSAA. 1080p is mostly wasted at that size. But I don't think you can even buy 720p TV sets anymore.

Regards,
SB
 
BTW, seems there's no perfect place to put this but, doesn't Scorpio/Neo mean the end of "too the metal" coding? Dont they basically have to go Direct X only now?

And, it seems to me like it's not a big loss and wasn't being used anyway. I guess over time as hardware got more powerful and complex, the benefits of low level coding got less and less, in relation to effort expended. You probably optimize more at almost a high level, flowchart, focus on memory optimization, level now, and that's where the big gains are.
 
With a 40" 4k, how close do you have to be to see the 4k resolution?

For most living rooms, below 84" looks like a waste of money.
wondering how it would you work for you if you played games relatively close to the TV. Most of the 4K TV sets I've seen IRL weren't large at all, I mean, not larger than 48" or so. The detail was very crisp anyways, looking close to them. At which distance from the TV do you usually play?

upload_2016-6-18_5-33-0.png
 
And 13m sounds like a lot but a metric ton of those 'VR capable PCs' are going to include PCs which are not used for VR, or even for gaming. Let me know when a PC VR-only game sells more than a million copies. 500,000 will be impressive. Having aspirations of 100m is great but again, many of those will be machines not used for gaming but for CAD, rendering or Photoshop because more GPU power is how to make Photoshop faster. :yep2:
That's a lot of sales, but it may depend on the market. China for instance would have a lot of potential users, but are likely far more price conscious than Western markets. That may work for hardware sales, but I'm not sure how well games from Western developers transfer into that market.

Vega seems to be Polaris + HBM2, in the same way that architecturally Fiji was Tonga + HBM.
Unless the rendered PCB shown in Scorpio's announcement is just a placeholder (not impossible), we're looking at a SoC with 384bit RAM (1666MHz or 6.66GT/s GDDR5) with no HBM on sight.

So in a way, if anything the GPU will be Polaris based.
It wouldn't be surprising if pictures were withheld as nobody knows the approximate size of Vega last I checked. Not sure it would matter at this point, but the size likely corresponds to one of the upcoming Vega chips. XB1 also had that ESRAM which might be interesting to design around. With the capacities required for 4k/VR that would likely look more like a single stack of HBM. Zen to my understanding was designed around a MCM, so it seems likely Scorpio would be CPU+GPU+at least one stack of HBM.
 
Well I'm confused. Microsoft passed on using a the 36CU Polaris based gpu this year like neo but yet are using GDDR5(X) memory so it couldn't be Vega? So what is it?
I doubt GDDR5X will be completely abandoned in the Vega lineup. HBM2 will be for their most expensive models in their lineup. The cost of HBM2 will still be quite expensive in 2017 late. Cheaper HBM1 cannot even be momentarily considered for Scorpio because of its 4GB memory limit.

GDDR5X currently reaches 10Gbps per pin but there JEDEC roadmap plans over the next few years to roll out 12Gbps, 14Gbps and eventually 16Gbps per pin modules.
 
Last edited:
Just did some simple math, fyi a 40CU polaris (with 2560 shader cores) can reach the 6 TFLOP number if it is clocked at 1.17Ghz, which we still don't know what sort of temps and power draw it's reaching to do that (stock RX480 is running at 1.266Ghz). This seems more plausible to me rather than the Vega simply because both the CPU and GPU have to be crammed in a single soc and the bigger die usually means more heat + more space + higher power draw, three things you want to avoid when building a console.
 
Last edited:
Status
Not open for further replies.
Back
Top