RTG 2016 update - AMD Radeon Polaris Architecture Preview
Click here to post a comment for RTG 2016 update - AMD Radeon Polaris Architecture Preview on our message forum
zer0_c0ol
https://www.youtube.com/watch?v=5g3eQejGJ_A
vbetts
Moderator
So then Polaris is a new GCN based architecture?
ruiner13
Guessing much of that power savings is due to the 14mm lithography. Once nVidia shrinks their dies and gets HBM2, they will also have substantial power savings. Hilbert, I do appreciate you calling out the apples to oranges comparison they did to the nVidia 950... "our 2016 models are better than the competitions 2015 models (which aren't substantially changed from the 2014 models)!" I should hope so!
Denial
Yeah it would seem that way. Which is honestly good -- AMD doesn't need to waste it's resources redesigning an architecture. GCN is fine. It just needed to bring power consumption down. 14/16nm will obviously help that but the architecture itself needs to bring improvements as well if they want to compete with Nvidia, which I'm sure RTG is doing with Polaris.
I find it really weird that they are dual-sourcing parts out of GF/TSMC. I wonder if 14nm will be restricted to APU stuff. It seems weird that they would have pay the overhead of dual sourcing in their main GPU line.
schmidtbag
I agree. Hell, GCN is good enough that AMD is still rebranding GCN 1.0 GPUs (even though nobody seems to really like it when they do that).
As long as both AMD and Nvidia can't figure out how to get decent FPS for 4k screens, working on power efficiency is the next best thing. Both companies have GPUs that are more than good enough for 1080p and sufficient for 2k, so as long as they keep that performance while lowering the watts, I'm happy.
BColt
So basically we will see 5% increase in FPS and some % lower TDP, is all I'm getting out of this.
zer0_c0ol
tdp is not the overall system power consumption
Twiddles
System used: i7-4790K 4x4GB DDR4. Wait whut?!
Source: http://content.hwigroup.net/images/articles/AMD%20Polaris%205.jpg
BColt
And I am talking about the AMD part power consumption.
I believe for now the only hope for performance improvements are dual GPU cards and DX12/Vulcan.
Noisiv
And getting murdered, is what you forgot to mention
Having to use big-small Tonga die and feature-less Pitcairns to fight mid-sized GM206 - is anything but GOOD ENOUGH
But it's actually good compared to the state of AMD high-end GPU market share.
[spoiler]
Top Five Video Cards Ranked by Sales (Total Revenue)
EVGA GeForce GTX 970 4 GB PCI-e 3.0 x16 Superclocked ACX 2.0 Video Card
PNY Quadro K4200 4 GB PCI-e 2.0 x16 Workstation Video Card
EVGA GeForce GTX TITAN X 12 GB PCI-e 3.0 Superclocked Video Card
EVGA GeForce GTX 980 4 GB PCI-e 3.0 x16 Superclocked ACX 2.0 Video Card
EVGA GeForce GTX 980 Ti 6 GB PCI-e 3.0 x16 SC+ w/ACX BP Video Card
Quote:
Top Five Video Cards Ranked by Volume (Units Sold)
EVGA GeForce GTX 970 4 GB PCI-e 3.0 x16 Superclocked ACX 2.0 Video Card
EVGA GeForce GTX 960 2 GB PCI-e 3.0 x16 SuperSC ACX 2.0+ Video Card
PNY Quadro K620 2 GB PCI-e 2.0 x16 Workstation Video Card
PNY Quadro K4200 4 GB PCI-es 2.0 x16 Workstation Video Card
PNY Quadro K2200 4 GB PCI-e 2.0 x16 Workstation Video Card
[/spoiler]
http://blog.neweggbusiness.com/news/best-selling-video-cards-of-2015/
Noisiv
AMD labs strikes again 😀
GTX 950 - 140W LMAO
85W averaged across 9 reviews, G3D including
http://www.3dcenter.org/artikel/launch-analyse-nvidia-geforce-gtx-950/launch-analyse-nvidia-geforce-gtx-950-seite-2
Even custom OC-ed GTX 950 peak gaming power does not reach 140W
https://www.techpowerup.com/reviews/EVGA/GTX_950_SSC/28.html
zer0_c0ol
that was the overall system power consumption that was given
The purpose of this demonstration for RTG was threefold: to showcase that a Polaris GPU was up and running, that the small Polaris GPU in question could offer performance comparable to GTX 950, and finally to show off the energy efficiency advantage of the small Polaris GPU over current 28nm GPUs. To that end RTG also plugged each system into a power meter to measure the total system power at the wall. In the live press demonstration we saw the Polaris system average 88.1W while the GTX 950 system averaged 150W. Meanwhile in RTG’s own official lab tests (and used in the slide above) they measured 86W and 140W respectively.
i really would like for people to start to read the info first post second
BColt
People buy more quadros than 370/380's? Anyone else thinks this is strange?
Noisiv
except that it wasn't
says:
"Polaris card on med 1080p scored 60fps and consumed 86W"
4790k rig is around 60-ish watts in idle
86W while gaming rly?
Noisiv
it's proly skewed by god-knows-what, but it's still telling
zer0_c0ol
did you even read the post i have given you, here it is again
The purpose of this demonstration for RTG was threefold: to showcase that a Polaris GPU was up and running, that the small Polaris GPU in question could offer performance comparable to GTX 950, and finally to show off the energy efficiency advantage of the small Polaris GPU over current 28nm GPUs. To that end RTG also plugged each system into a power meter to measure the total system power at the wall. In the live press demonstration we saw the Polaris system average 88.1W while the GTX 950 system averaged 150W. Meanwhile in RTG’s own official lab tests (and used in the slide above) they measured 86W and 140W respectively.
http://www.anandtech.com/show/9886/amd-reveals-polaris-gpu-architecture
Noisiv
I did now. And it sounds very impressive.
OTOH 86W at the wall.....thats like 70W real consumption. Which means card + CPU are pulling 10-20W watts above idle.
I'd like to be wrong, but I think someone there needs to retake their electronics lab class 😀
edit: on a 2nd thought, it is a double node jump. so it's quite possible :banana:
edilsonj
Total system consumption.
Noisiv
yeye we covered that already. Looks like Polaris is around 30W
For GTX 950:
140W - 90W(ingame) = 50W (idle)
Polaris:
85W - 50W(idle) = 35W (Polaris+CPU)
Denial
My 4790K stress tests at about 45-50w. A GTX950 has a TDP of 90w.
It means the Polaris GPU is running at about 35-40w which is definitely impressive but it is on a smaller node. So we have no idea how much of that is architecture improvements vs node shrink. You also have to remember that AMD's architecture is heavily favored in Battlefront 3. For example, a 950 outperforms a 370 in nearly every other title, by about 20+%, but in BF3 the 370 ties the 950.
