PCWorld had an interesting interview with Raja Koduri. I guess this is what we expected at ‘architecture preview’ few days ago. It’s not very informative, but we can conclude what is important for Radeon Technology Group in the coming year.
Below you can find my notes, this is not a transcript. Just a highlight of all questions and answers.
[ 1:15 ] One of the big things in Vega architecture preview were High-Bandwidth Cache & Controller. What sort of benefit can gamers expect from this technology?
[ From gaming perspective, they noticed that memory utilization in games is usually not higher than 50%. Vega will essentially be doubling your memory capacity ]
[ 3:00 ] If game requires 4GB of VRAM, in reality it will only use 2GB. Does it mean VEGA will actually allocate 2GB?
[ Yes, if game developers take advantage of that, they will be able to use higher resolution textures ]
[ 4:15 ] If memory efficiency is increasing, does it mean in the future gamers will need less frame buffer?
[ explains why bandwidth is more important than frame buffer size ]
[ 6:40 ] Asking how to re-educate consumers to understand high-bandwidth cache
[ Raja explains how memory change affects the market, consumers need to look at the bandwidth ]
[ 8:40 ] Will games at day one (Vega launch) consume less memory with Vega?
[ explaining how memory swapping speed will increase with Vega ]
[ 10:00 ] Why does Raja think there’s not much of a difference between ultra and normal settings?
[ games are developed for default settings, changing settings to ultra only makes artwork up-scaled ]
[ 12:00 ] What will it take to make developers develop texture for ultra settings?
[ it’s very hard to notice the difference in textures at higher resolutions, says Christmas 2017 will be interesting time, Raja’s goal for 2017 is to bring 4k fluidity to games ]
[ 14:00 ] Jokes about Raja’s 2017 ‘resolution’ being 4k. Is Raja’s goal to bring 1000$ PC to play 4K games?
[ explains what is needed to offer good 4k experience ]
[ 16:00 ] What is the overclocking potential of VEGA?
[ They are figuring it out themselves, we will know in the coming weeks ]
[ 16:20 ] Vega was built using the same Infinity Fabric as Ryzen is based around, and they already talked a little about their overclocking capabilities, will any of that knowledge translate into Vega?
[ CPU design and GPU are different ]
[ 17:00 ] What is Infinity Fabric?
[ It’s the way we connect different function blocks (IPs) ]
[ 18:00 ] Asking about improving efficiency using new technologies: Programmable Geometry Pipeline (removing polygons that are not visible in the game) Draw Stream Binning Rasterizer (removing unused pixels from the scene). How much performance could these technologies bring?
[ The benefit could be quite large. Vega is a big step in dealing with pixels that don’t matter. Their concern is how to be more efficient to offer better experience ]
[ 21:20 ] How consoles are changing PC gaming (Mantle and DirectX12). Is 2017 is going to be a big breakout for DirectX12? Why is AMD doing so well in DX12 compared to NVIDIA?
[ The most important reason is game optimization (especially console ports) ]
[ 25:25 ] DirectX11 was on you and your drivers, and DirectX12 is on the game developers?
[ Yes ]
[ 25:27 ] The only demo you let people see was Doom where it was beating 1080, but that was a Vulkan API. Would it still beat the pants off 1080 if that was running DirectX11?
[ Raja says doom only has OpenGL and Vulkan, but he thinks it WOULD at that resolution since it’s more GPU bound. Doom has some console optimizations, which are not available for OpenGL ]
[ 27:00 ] That was running Vega 10? Was it high-end Vega10 or entry-level Vega10?
[ Raja is laughing, but admits it’s Vega 10, he’s not ready to answer this question in detail ]
[ 27:37 ] NVIDIA is going away from multi-GPU SLI, but is AMD still committed to multi-GPU?
[ They definitely think there is room for it in the future. DX12 and Vulkan gives better access to multi-GPU ]
[ 29:00 ] Will VEGA work with FreeSync?
[ Yes it supports FreeSync ]
[ 29:50 ] Implementation of HDR standards is all over the place, is there a secret sauce added to FreeSync2 that will improve that HDR experience?
[ FreeSync2 has no secret sauce to make everything simpler ]
[ 31:50 ] One of the main features of FreeSync2 is to reduce lag, but will developers need to use any special API to make it seamless?
[ It will depend on what parameters they want to control, but yes, developers need to take responsibility for FreeSync2 ]
[ 32:35 ] If game support HDR, but it lacks FreeSync2 implementation, will we see still any benefits from FreeSync2?
[ Yes, we will see benefits in some cases ]
[ 32:50 ] FreeSync monitors are cheaper than G-Sync, does it matter to AMD how many units are sold vs competition?
[ We shouldn’t look at numbers, but they are happy about it. Overwhelmed by support for FreeSync ]
[ 34:50 ] NVIDIA is still not interested in implementing FreeSync. Does Raja think that the consumers who own FreeSync monitor and NVIDIA GPU will have a simpler choice to switch to AMD?
[ That’s what he would do, Vega is coming .. “your FreeSync panel investment will pay off” ]
[ 35:50 ] Do you think NVIDIA will support FreeSync at some point?
[ <claps hands uncontrollably and laughs> ]
[ 36:00 ] FreeSync2 monitors will be more expensive, so probably not as popular as FreeSync. What kind of share does AMD expect to gain with FreeSync2 over FreeSync?
[ There will be few key monitor announcements in Q1 2017, but it should help panel manufacturers to give them some marketing value ]
[ 38:20 ] Can you say where is VEGA being manufactured?
[ He can’t answer this question ] [ Raja says that PC should stand for Performance Computer ]
[ 40:00 ] Are enthusiast important to AMD right now?
[ Absolutely.. he mocks NVIDIA saying that they don’t like PC as much as AMD does ]
=== end ===