RX 480 slide leaked:
http://videocardz.com/60752/amd-rade...cations-leaked
Biscuit (31-05-2016)
http://videocardz.com/60373/amd-pola...s-on-june-29th
Also RX480 meant to be $200:
http://forums.hexus.net/pc-hardware-...480-199-a.html
$200 with VAT would be £165. If AMD can deliver R9 390X/Fury level performance under £200 that would be a decent increase indeed!!
watercooled (31-05-2016)
BTW,this thread is now 4 years old and nearly has had nearly 300000 views. Surely,this must have the highest views/year of any Hexus thread??
I think it must be one of the longest threads too, if not THE longest.
TBH it seems like it's become more of a 'random processor chitchat too small to put into its own thread' thread. I know some mods on other forums seem to dislike long threads and force 'part 2' threads; is there a reason for that, maybe just a legacy one which no longer applies?
The pricing rumours seem to indicated a $200 to $250 price for the RX 480 which would indicate a £165 to £210 price in the UK with VAT added. If that is R9 390X level or GTX980 level performance for that price as hinted in The WSJ article,it is going to make the GTX1070 look overpriced very quickly! You would be gaining 25% extra performance for at least a 50% increase in price,maybe more.
Something just occurred to me; the 1070 is roughly 3/4 the 1080p in terms of processing components, and I know it's not as simple as this in reality because of e.g. memory controllers, caches, etc which stay put regardless, but if you take 75% of the 1080's die size you get ~229mm2, very similar to Polaris10's rumoured 232mm2.
Again, I'm completely aware that's a very imperfect comparison, just something to ponder. GPU core count vs die size do seem to scale relatively well though, e.g. using Maxwell for comparison, the GM204 has 2/3rds of the GM200's cores with very close to 2/3rds of the die size. The GM 206 is a bit larger than that scaling would imply but it's still only ~10% off.
The thing is the rumours say Polaris 10 is a 2516 shader chip with 2304 shaders enabled.
The shaders aren't necessarily comparable to older GCN versions though are they? Plus Polaris will likely clock higher. The core count on the 1080 is much lower than the Titan X after all.
But the cores clock very much higher. Remember the Titan X has 20% more cores than a GTX1080. However,the Titan X probably Turbos to 1.2ghz to between 1.3ghz and the GTX1080 to 1.8GHZ or more.
So you are probably looking at a 40% to 50% increase in clockspeeds stock for stock,and when you take that into account,it is not surprising the GTX1080 is around 20% to 30% faster overall.
Considering that Pascal was a recent addition to the Nvidia roadmap,ie,originally it was Maxwell to Volta,it looks very much like Pascal is a slightly rejigged Maxwell with some added features and a massive clockspeed bump.
P10 clockspeed is 1.266GHZ:
https://www.techpowerup.com/223043/a...-above-1-2-ghz
Zen briefly shown off:
http://www.anandtech.com/show/10391/...-ridge-silicon
Bristol Ridge announced:
https://www.techpowerup.com/223041/a...-a-series-apus
True. However I still think it's important to wait for real benchmarks before jumping to conclusions about performance besides ballparking it.
Assuming it's legit (they often are) the benchmark at Videocardz puts it above a 980 and close to a Fury: http://videocardz.com/60253/amd-rade...k11-benchmarks
The codename matches the more recent leak too: http://videocardz.com/60819/amd-rade...polaris-67dfc7
>980 performance would be good for sure!
But either way it should be good for their mobile business as Nvidia don't have anything to compete yet.
There are currently 23 users browsing this thread. (0 members and 23 guests)