Predict: The Next Generation Console Tech
- Thread starter Acert93
- Start date
- Status
itsmydamnation
Veteran
im just thinking , how would stuff like mega texutring compare to more traditional textureing techniques in that configuration. To me it would seem like it would have an advantage, being able to just stream the new data needed from the slow memory into the fast EDRAM. Would other more rigid formats work as well, i would expect the entire texture would need to be located in the one space ( DISK/RAM/EDRAM).
As soon as we learned that there would be 8 GB of RAM and that ESRAM would be present, the logical thing to assume would be that slow DDR memory was being used. I don't think we even need a source to tell us anything.
ImaginaryInsider
Newcomer
AndyH on Neogaf, I really don't remember him posting any info before. If Durango has 8gb, it pretty much has to have slower gddr3/4 ram. I'm curious how they might tweak the edram solution this gen, how much makes sense?
Bagel seed
Veteran
It's not really slow, just in the same overall design choice as the 360. And again they'll go with EDRAM for the framebuffer (and maybe other things) but this time more MBs.
Inuhanyou
Veteran
Which is technically not a bad set up for what they are trying to accomplish.
The implication is that ram itself is not slow, but the type is slow in comparison to what people think "ought" to be used, like GDDR5. But DDR3 is pretty cheap nowadays, and DDR4 is going to be cheaper eventually. It makes sense from a cost and heating angle to use DDR3,or DDR4. And either type of ram can work just fine in a next gen console providing it has the bandwidth to breath and supplementary EDRAM.
This would not be a wii u situation in any case as the Wii U is extremely bandwidth starved.
The implication is that ram itself is not slow, but the type is slow in comparison to what people think "ought" to be used, like GDDR5. But DDR3 is pretty cheap nowadays, and DDR4 is going to be cheaper eventually. It makes sense from a cost and heating angle to use DDR3,or DDR4. And either type of ram can work just fine in a next gen console providing it has the bandwidth to breath and supplementary EDRAM.
This would not be a wii u situation in any case as the Wii U is extremely bandwidth starved.
Bagel seed
Veteran
1gb of assets per frame (assuming at 30fps) would be what, 2 to 3 times as many assets we can realistically display on 360/PS3? What would be optimal based on a 1.5TF gpu?
arijoytunir
Regular
edit
Last edited by a moderator:
30GB of BW would only be some 3X Wii U...and only a pitiful 1/3-ish more than 360.
bkilian has put a damper on this thread for sure. But if Sony comes out with a significantly more powerful box, MS has some explaining to do RE "regulations".
If they somehow sell out the gaming box's power, in order to be classified as a "set top box", if such a thing is possible?. And Sony doesn't? They are gonna get a beat down...
I still think a 1-1.2 TF box will be pretty fun though. And bkilian is right at the price. Something that say comes out at 299 will sell a lot. All that said, something beefier at 399 will do better, in the long run for sure and probably even in the short. It's never the price, it's whether the price is justified. All the great pricing in the world never saved gamecube.
Could they be releasing relatively modest consoles which can be extended by cloud computing? Instead of say releasing a game which runs entirely locally or entirely from the cloud they can solve the latency problem by splitting the work between local and cloud computing with the former being for latency sensitive code relating to the user interactions?
It would certainly be interesting to see what sorts of things devs would justify/work out pushing to a high-latency processor system. Pathing AI for drunk NPCs? Vaguelly real-time global illumination ish information? Large-scale aspects of a water simulation?
I guess it depends on what kinds of things are relatively latency tolerant but compute heavy or different ways of thinking about solving problems are. If the consoles have relatively large memories then perhaps they can use part of it for precomputed algorithms where much of the legwork has been done off site.
Say for instance you can have global illumination pre computed for the specific time of day and then sent to a console with only the dynamic lights being handled locally. That could be the best of both worlds between pre-baked and dynamic lighting.
http://semiaccurate.com/forums/showpost.php?p=174466&postcount=67
(Not saying it must be real but that's funny responds)
Btw i saw HD8970M is 384bit,is that true?because i can't find any news about HD8970M yet
http://en.wikipedia.org/wiki/Sea_Islands_(GPU_family)
This should be quite usable if game uses proper virtual texturing, even without pre-baking.
No chance of that with an AMD CPU unless AMD have seriously and I meen seriously boosted there IMC ( I'm only getting just over 30Gb/s and I'm running triple channel 1866Mhz DDR3 )
Realistically I wouldn't expect more then 20-25Gb/s from a console using DDR3.
- Status
Similar threads
- Locked
