Middle Generation Console Upgrade Discussion [Scorpio, 4Pro]
- Status
To each his own. MS is waiting till Fall 2017 or even winter to launch Scorpio. I question why ? If its just Polaris with Puma then i'm sure they could launch in 2016 since both are avalible . It could be Vega , but little vega is said to be pushed up to this fall . So again they might be able to launch this year with Puma / Vega or even Zen / Vega. They could even launch Zen / Polaris since I believe Zen is this year . So why wait ? Better yields on little vega ? Zen + ? Big Vega ? Maybe Little Navi ? Maybe better memory ?
I dunno but we got a year to wait
Where from? Because your source contradicts all the others reporting numbers. eg. 18/19 million beginning of this year.
eastman must be talking about 'hopefully by the time Scorpio comes out' 
Anyway, I have a really basic/stoopid Q - All this talk of a fancy CPU sounds quite logical (due to timescales) however, am I wrong in thinking it would be wasted due to the (relatively) low bandwidth? And the bandwidth is one for the factors that holds this back from being true 'next gen'?
Anyway, I have a really basic/stoopid Q - All this talk of a fancy CPU sounds quite logical (due to timescales) however, am I wrong in thinking it would be wasted due to the (relatively) low bandwidth? And the bandwidth is one for the factors that holds this back from being true 'next gen'?
I know some people hate upscaling with a passion, but is that still the case from upscaling from 1440/2160p to 4k?
Or is that a less of a problem for those people?
Moving forward I think we're only going to get more alternative rendering methods like the checker board method, and if a multi plat is using such a method, how much if any benefit is there in rendering it at 4k?
Easy marketing, wasteful use of resources?
Although I think game implementation, and actual machine marketing are two different things.
Or is that a less of a problem for those people?
Moving forward I think we're only going to get more alternative rendering methods like the checker board method, and if a multi plat is using such a method, how much if any benefit is there in rendering it at 4k?
Easy marketing, wasteful use of resources?
Although I think game implementation, and actual machine marketing are two different things.
Personally I think it's other things that stop it being 'next gen'.
hardware compatibility, eco systems, software compatibility.
was Wii not next gen compared to game cube?
If ps4 could not play neo games (not talking about exclusives), and it had a different eco systems etc, that would make it a next gen system. Just a weak jump to next gen that's all.
I maintain the best estimate of XBO shipped numbers is to just double USA sales. This would put them around 24 million I guess (Aqua's latest table had XBO at 12.02 million in USA through May). Shipped. Sold through will be a little less. 21/22 million maybe. I dont see that much point distinguishing shipped to retail from sold. It'll be much harder to definitively ascertain the latter anyway, for marginal benefit.
30 million thus is definitely out of the question for what that's worth.
lol, yea brain fart2160p upscales extremely nicely to 4k. I doubt anyone will notice the difference.
2160p is 4K! 3840 x 2160
As for 1440p > 4k, that's basically equivalent to upscaling 720p > 1080p. It's not a nice ratio so will be blurry, but it'll depend on viewing distance and screen size. If people are using 4K TVs in the same FOV as their 1080p sets, they probably won't see much difference between 1440p and 2160p native. But if, as is expected, they are using the screens to provide a larger FOV, the blurriness may well be equivalent.
I'd say that upscaling detriment will be less on the whole than 720p > 1080p because I expect most 4K sets not to be 4x the area of the 1080p sets, so there will be higher resolution density and less difference between 1440p and 2160p verus the old HD upscale.
I'd say that upscaling detriment will be less on the whole than 720p > 1080p because I expect most 4K sets not to be 4x the area of the 1080p sets, so there will be higher resolution density and less difference between 1440p and 2160p verus the old HD upscale.
Que? We're talking about rendering resolution and upscaling to create that 4K output.
Uhm, so 320GB/s is now considered low bandwidth? What bandwidth level are thinking is needed to be "next gen" enough? Over 9000?eastman must be talking about 'hopefully by the time Scorpio comes out'
Anyway, I have a really basic/stoopid Q - All this talk of a fancy CPU sounds quite logical (due to timescales) however, am I wrong in thinking it would be wasted due to the (relatively) low bandwidth? And the bandwidth is one for the factors that holds this back from being true 'next gen'?
It's relatively low for the resolution. If we think 150 GB/s is enough for 1080p (current consoles), linear scaling would suggest 600 GB/s is needed for the same fidelity at 4k (which I'm going to call 2160p from now on because it's more accurate and a comparative term). DF have an article on a 2160p GPU, the Titan 1080 GTX, which is 320 GB/s uncontested VRAM, with additional CPU bandwidth of DDR main RAM.
So Scorpio, and very much more Neo, are low on BW for 2160p rendering, similar to PS3 being relatively low for 1080p rendering. I think that's a fair assertion.
Yeah, it wasn't a dig - just from a very basic level the increase doen't seem big enough to mean a 'true next gen' experience...not to say the difference will be very noticable.
Yah, but as AlNets would say, they can just turn off AF completely to save bandwidth because it doesn't matter anyway.
So there is 0 possibility that MS does not included embedded ran onto this SOC?
I mean they just seems like they are 2 for 2 so far. Adding some would make BC trivial.
Being a closed system makes development easier.
And embedded ram has proven very effective at adding additional bandwidth for cheap.
XBO memory structure was hamstrung by the fact that 85% of bandwidth was sitting at 32MB. But that doesn't need to be the case, what if they could come up with proper ratios.
It shouldn't be too hard to figure out at what stage of the rendering pipe requires the most concurrent bandwidth and what stage it requires single direction. If post processing is the culprit for requiring 600 GB/s, embedded ram seems a good fit for this and still save costs.
I wouldn't put this past Microsoft, for all the shit I gave them earlier above that Jay and Eastmen called me out on, would they be willing to waste their investment in R&D for esram techniques and tools?
They may not want to market embedded ram given the negative stigma they've been getting from the Internet. Perhaps they are only marketing the shared pool numbers.
Sent from my iPhone using Tapatalk
I mean they just seems like they are 2 for 2 so far. Adding some would make BC trivial.
Being a closed system makes development easier.
And embedded ram has proven very effective at adding additional bandwidth for cheap.
XBO memory structure was hamstrung by the fact that 85% of bandwidth was sitting at 32MB. But that doesn't need to be the case, what if they could come up with proper ratios.
It shouldn't be too hard to figure out at what stage of the rendering pipe requires the most concurrent bandwidth and what stage it requires single direction. If post processing is the culprit for requiring 600 GB/s, embedded ram seems a good fit for this and still save costs.
I wouldn't put this past Microsoft, for all the shit I gave them earlier above that Jay and Eastmen called me out on, would they be willing to waste their investment in R&D for esram techniques and tools?
They may not want to market embedded ram given the negative stigma they've been getting from the Internet. Perhaps they are only marketing the shared pool numbers.
Sent from my iPhone using Tapatalk
Last edited:
They probably have enough bandwidth to maintain BC, don't know how much affect latency will have, may end up being a wash.
It did have a cost, that being die space.
I think they ended up with 32Mb more so out of size of chip than anything else. They thought with new rendering techniques they could get away with it. (had to balance cu's with esram)
Turns out, they may not have been to far wrong in terms of size, they ended up lacking in overall TF imo.
32MB was small and made development harder though, but lots of these new rendering methods seems to help with fat gbuffers.
I personally wouldn't be against a smaller pool of ram, to help with contention and put the speed right where it's required. Have no idea what would be a reasonable size though. It's one of the reasons I wouldn't have minded it being a hbm and ddr4 based machine.
PS : wasn't calling you out on your overall view, just the list made me jump ship :smile2:
- Status
