AMD Vega Hardware Reviews

Note that on Pascal cards, at least vith MSI afterburner, overcklocking also implies some form of undervolting.
Afterbuner deals with a voltage / frequency graph. + 50 Mhz is equivalent to increasing the frequency by 50 on _all_ the voltage steps. Hence after +50 is applied, the card will try to achieve say 1600 mhz using the voltage that was previously required for 1550 mhz. (if it fits in the power limit).
Let me rephrase that: No, we didn't do any manual undervolting, what MSI Afterburner does "behind the curtains" automatically is a different thing.
 
Just tried to clarify to readers that some undervolting is actually hapenning anyway.

Further and since is the case, manual undervolting might not get you far. Or anywhere, for that matter
 
It's the same size as the one in Hawaii/Fiji.
It have the same size of geometry engines, but the engines are totaly different. Now you have Draw Stream Rasterizer, another Setup of AC Units and some more. What i finde intresting ist that AMDs Vega have the Same Performance at the front end like GP102 with 2 Geometry engines more.
 
Half a billion triangles per secons is laughable for modern graphics chips in the three-digit price range. More important is, what is done with all these tris. If I can early discard 95 % of them before being shaded, then hey, throw ten times the number at my Vega.
 
Half a billion triangles per secons is laughable for modern graphics chips in the three-digit price range. More important is, what is done with all these tris. If I can early discard 95 % of them before being shaded, then hey, throw ten times the number at my Vega.

Discarging is a week Point at Vega without primitive shader -.- Thats why i'm asking for. If it Comes i think we will see a huge benefit.

@Rys
Maybe you can say us when primitive shader will come?
 
Maybe you can say us when primitive shader will come?
Rys's work is devrel, IIRC, even if he knows the answer it's not going to be his department to announce unannounced things on forums. :p

Why? Because games render tons of triangles that contribute nothing to the scene?
Sometimes. Open-world games can have huge overdraw from vegetation and stuff like that.
 
Sometimes. Open-world games can have huge overdraw from vegetation and stuff like that.
Primitive shaders won't help you there. Will a primitive shader rasterize triangle to figure out if all its pixels are covered to determine that it doesn't need to rasterize?
It has been said by AMD very specifically what this tech is able to cull. Yet even here on a 3D tech forum we continue to have this magical ideas...
 
Last edited by a moderator:
No but open world have a lot of hidden triangles, or triangles which are smaller than a Pixel.
Again: how do you determine that your 3 vertices will produce a hidden triangle without rasterizing it? You can check depth at 3 vertices and see that they are behind already rendered geometry. Does this mean you can discard the triangle? No! A number of pixels inside the triangle can still be visible because only the 3 samples at vertices positions are hidden. Can you discard a triangle whose 3 vertices are all of screen? Again no! One vertex to the left, one to the right and one above. Your triangle can still cover your entire screen with pixels.
Yes, you can compute that a triangle will not cover a single sample and then you can reject them. Games do use LOD though. You don't render 1 million triangle model 10 miles off. Unless stuff is done badly in game...
 
Again: how do you determine that your 3 vertices will produce a hidden triangle without rasterizing it? You can check depth at 3 vertices and see that they are behind already rendered geometry. Does this mean you can discard the triangle? No! A number of pixels inside the triangle can still be visible because only the 3 samples at vertices positions are hidden. Can you discard a triangle whose 3 vertices are all of screen? Again no! One vertex to the left, one to the right and one above. Your triangle can still cover your entire screen with pixels.
Yes, you can compute that a triangle will not cover a single sample and then you can reject them. Games do use LOD though. You don't render 1 million triangle model 10 miles off. Unless stuff is done badly in game...

https://www.gamersnexus.net/guides/3010-primitive-discarding-in-vega-with-mike-mantor
And at that point when you finally have the final position of the vertices of the triangle, is one point where we can always find out whether or not the triangle is inside of the frustum, back-faced, or too small to hit. From frustum testing, there’s a mathematical way to figure out whether or not a vertex is inside of the view frustum. If any one of the vertices are inside of the view frustum, then we’ll know that the triangle can potentially create pixels.

Thats the key function of primitive shader ;)
 
You really don't get what I'm talking about do you? Which category does a tree behind a mountain fall into: inside of the frustrum, back-faced or too small to hit?
 
Anyway, TBDRs have been throwing away all opaque overdraw since the 1990s. It's quite possible to do.
Primitive shaders are not TBDR. Also TBDR's haven't been throwing away all opaque overdraw since the 90's. In their simplest form they just take advantage of on chip bandwidth (speeding up overdraw), at there best they reduce the number of shader invocations by deferring shading work until visibility is determined.

Also @MDolenc do you see any practical performance benefit from primitive shaders? If not, why?

edit - also if you don't think so then why did AMD bother with them?
 
Last edited:
Back
Top