-
Ever wanted an RSS feed of all your favorite gaming news sites? Go check out our new Gaming Headlines feed! Read more about it here.
-
We have made minor adjustments to how the search bar works on ResetEra. You can read about the changes here.
PS5 Speculation |OT12| - Aw hell, Transistor's running this again?
- Thread starter Mecha Meister
- Start date
- Status
Yes 380-400mm2 was the number speculated here.
And 56 active CUs because of github.
Bless
edit: Mesh shading confirmed hell yeah. RDNA 2 killing it!
Holy shit.
For people who can't watch the vid:
Inside Xbox Series X: the full specs
This is it. After months of teaser trailers, blog posts and even the occasional leak, we can finally reveal firm, hard …www.eurogamer.net
380 Gigarays/sec????
Wouldn't that be stupidly close to full scene RT?
So komachi is thinking that the 80CU rumor was false and Navi 21 is 56CU hmm.
That is not gigarays at all. That is aabb traversals/intersections per second. There are multiple, like 3 to 4 needed per ray and then the triangle intersection after that.
As I view it, yes. Same with the one from January of 2019.
the die size is smaller than expected. Regardless excellent specs.
the 360mm2 die size might indicate, ps5 might be closer in power even if the die size is around launch ps4 size.
the 360mm2 die size might indicate, ps5 might be closer in power even if the die size is around launch ps4 size.
Yes I think they must have reduced the size of the CU of RDNA2. Possibly because they removed some general compute HE that is better suited for CDNA application space?
Smaller = cheaper
I will be talking about this in another video, but based upon what I have seen, no.
The specs are less interesting than the ways they're explaining how they came to those specs, and the improvements they've made throughout the pipeline - clearly having solid profiling data helped a lot
the SSD with hardware decompress, combined with direct storage and updates to texture memory fetching will help effectively give devs full access to their gamedata as though it was just right there in ram (with obvious caveats but it looks really well thought through)
the SSD with hardware decompress, combined with direct storage and updates to texture memory fetching will help effectively give devs full access to their gamedata as though it was just right there in ram (with obvious caveats but it looks really well thought through)
Doesn't nvidia also accelerate more of the raytracing pipeline via fixed function units? .
No NV accellerates the exact same thing. Triangle intersections and the BVH traversal.
NV does have the int32 fp32 split thing which helps RT performance in general though.
The crazy thing is I dont think anything he said has been wrong so far when it comes to specs
Soooo maybe I would take a look at what leaks nailed the 12.1TF number, with the CUs, and maybe pay more attention to those...
Can't wait for this video. I suppose you will be evaluating RT performance?
wowowow at that minecraft ray tracing demo! Its like the super duper pack is finally coming to fruition!
Right, so CU disable for yields are essentially useless and don't contribute to performance, only a lager die. All consoles have a certain amount disabled. No one would be discussing disabled CU in relation to chip performance.
Obviously it would be terrible PR for Sony to reveal anything now during a pandemic, or so I was told on this forum. /s
When it comes to XBX nope .
We will have to see for PS5 but the man guess if they were that was on point.
He had the PS5 pegged at 52CU's @ 2.0ghz = 13.3 TF.
Hard to imagine Sony have clocked theirs so high, but then again, anything's possible at this point.
Interesting. I didn't initially believe the github leak, but now it looks more credible. We'll find out soon enough, I guess. Surely Sony won't remain quiet for much longer.
Just checked his other posts, got the XSX spot on and said that the PS5 is 13.3TF, hot damn.
- Status