AMD Polaris Thread: Radeon RX 480, RX 470 & RX 460 launching June 29th

[Armsdealer]

AotS is a DX12 benchmark, AMD engineers talked about a new command processor, which is supposed to improve CU utilization under legacy loads. So I would expect different results from previous gcn architectures.

That's fine, but the results imply that p10 is gimped in dx12 vs 11. The aots (dx12) benchmarks are implying way way worse performance vs both amd AND nvidia cards. Than these leaks.

Previously we saw amd dx12 vs dx11 was superior vs nvidia, and that nvidia basically performed the same in both. P10 -if you believe the benches as well as the aots Computex demo - is materially inferior in dx12.

One of the following must be true:

-The leaks are wrong

-P10 has such extraordinary command processor benefits that its gap over gtx 980 is greater in dx11 vs dx12

-amd purposely downplayed p10 performance at Computex

-there were driver issues in all of the aots benches we've seen thusfar

Take your pick. I have no idea.

 

[boozzer]

That's fine, but the results imply that p10 is gimped in dx12 vs 11.

Previously we saw amd dx12 vs dx11 was superior vs nvidia, and that nvidia basically performed the same in both. P10 -if you believe the benches as well as the aots Computex demo - is materially inferior in dx12.

One of the following must be true:

-The leaks are wrong

-P10 has such extraordinary command processor benefits that its gap over gtx 980 is greater in dx11 vs dx12

-amd purposely downplayed p10 performance at Computex

-there were driver issues in all of the aots benches we've seen thusfar

Take your pick. I have no idea.

just 1 more day. no need for more guesses or speculation.

 

[JDG1980]

Unless amd downplayed rx480 perf in aots, the numbers don't add up. In aots, we're clearly seeing dx12 perf below 390x, and closer to 390, but slightly above 980, and now in dx11 we're seeing it way above 390x and way above 980.

28nm GCN products had bottlenecks that prevented full resource utilization in DX11. That was most notable with the Fiji cards, but it existed across the board to some extent. AotS in DX12 was one of the very few games that could push 28nm GCN cards near their theoretical limits.

If Polaris fixes the bottlenecks that plagued previous GCN generations, then it wouldn't be surprising to see the gap close between DX11 and DX12. That would mean even if the raw horsepower (and DX12 performance in AotS) was similar to Hawaii, it would do better in most real-world DX11 games and benchmarks.

 

[renderstate]

Weren't they already miles ahead of nvidia in terms of VR performance? Nvidia's new VR features just made their GPU's not suck at VR which wasn't an AMD problem in the first place.

In the universe where I live AMD had hardly any advantage over NVIDIA before Pascal, but if you have numbers to corroborate your statement I an curious too see them.

 

[sontin]

There doesnt exist any hardware bottlenecks. AMD hasnt cared to fix their DX11 driver. Look at Tomb Raider: No difference with DX12. Or use the both Vulkan games Talos and Dota 2.

 

[renderstate]

I dont get the VR point. Does this mean VR games dont get better? Or AMD wont release anything new for the next two years?!

IMO it means that for $200 you get to pass the min requirements for performance as specified by Oculus and Valve.

 

[geoxile]

What do you think of AMD's new marketing campaign Comrades ?

They really missed the mark, but I think at worst it'll probably be harmless.

 

[Armsdealer]

28nm GCN products had bottlenecks that prevented full resource utilization in DX11. That was most notable with the Fiji cards, but it existed across the board to some extent. AotS in DX12 was one of the very few games that could push 28nm GCN cards near their theoretical limits.

If Polaris fixes the bottlenecks that plagued previous GCN generations, then it wouldn't be surprising to see the gap close between DX11 and DX12. That would mean even if the raw horsepower (and DX12 performance in AotS) was similar to Hawaii, it would do better in most real-world DX11 games and benchmarks.

...not just close, but flip to the other direction.

Dx11: p10 >> 980 (firestrike)
Dx12 p10 >= 980 (aots)

Whereas other amd cards were exactly the opposite.

Get my point?

 

[BlitzWulf]

There doesnt exist any hardware bottlenecks. AMD hasnt cared to fix their DX11 driver. Look at Tomb Raider: No difference with DX12. Or use the both Vulkan games Talos and Dota 2.

Please don't use ROTR's awful and unfinished DX12 mode as a yardstick for anything, it's intellectually dishonest at best.
Both your other examples have also been said to be works in progress by the developer's. Talo's loses perf for all cards in Vulkan mode not sure about Dota

 

[boozzer]

Please don't use ROTR's awful and unfinished DX12 mode as a yardstick for anything, it's intellectually dishonest at best.
Both your other examples have also been said to be works in progress by the developer's. Talo's loses perf for all cards in Vulkan mode not sure about Dota

why do you think he brought it up

 

[coercitiv]

It wasn't my intent to upset anyone with that joke my apologies

It's not upsetting (I defined my reaction as visceral to portray subjectivity on this matter), but I see no value in it, their message can be easily perverted. Then again, I doubt it will do any harm considering it's limited exposure. Maybe their target audience will relate better.

 

[tviceman]

just 1 more day. no need for more guesses or speculation.

Wait, 1 more day? Until reviews? I thought the NDA was June 29?!?!

 

[boozzer]

Wait, 1 more day? Until reviews? I thought the NDA was June 29?!?!

E3 2016 was what I read on this very forum.

 

[frozentundra123456]

Well it is like saying: I will not believe that that comet is gonna destroy my planet even if it is in front of my eyes. It has no credibility, because it could destroy jupiter, sun or any other bigger planet, than mine.

Dafaq? He may have no credibility for you. But ignoring what is displayed on recorded video is...

Reasonable??? You tube videos are the last source I expect for reliable benchmarks.

Seriously, looks like a very good price/performance card, but I am withholding judgement until we get benchmarks from reliable testing sites, not just some guy on the internet.

 

[kraatus77]

In the universe where I live AMD had hardly any advantage over NVIDIA before Pascal, but if you have numbers to corroborate your statement I an curious too see them.

is that why amd has 80+ % market of vr? because if we believe usual suspects nvidia's pc gpu market lead is because performance.

so if we go by same logic, they have better performance in vr compared to nvidia. :thumbsup: ( atw better preemption and something about that 10ms response time, that's why developers recommend amd over nvidia for vr)

regardless it's a mere 200$ gpu, bringing vr spec to lower prices points. it does that successfully so i don't see anything wrong here.

 

[sirmo]

That's fine, but the results imply that p10 is gimped in dx12 vs 11. The aots (dx12) benchmarks are implying way way worse performance vs both amd AND nvidia cards. Than these leaks.

Previously we saw amd dx12 vs dx11 was superior vs nvidia, and that nvidia basically performed the same in both. P10 -if you believe the benches as well as the aots Computex demo - is materially inferior in dx12.

One of the following must be true:

-The leaks are wrong

-P10 has such extraordinary command processor benefits that its gap over gtx 980 is greater in dx11 vs dx12

-amd purposely downplayed p10 performance at Computex

-there were driver issues in all of the aots benches we've seen thusfar

Take your pick. I have no idea.

Or it could be just what I said it was. New architecture responds differently to different benchmarks.

 

[VulgarDisplay]

In the universe where I live AMD had hardly any advantage over NVIDIA before Pascal, but if you have numbers to corroborate your statement I an curious too see them.

I asked a question I didn't make a statement. Do you have any numbers to corroborate your constant anti amd posting? It's tiresome.

Sent from my XT1575 using Tapatalk

 

[Armsdealer]

Or it could be just what I said it was. New architecture responds differently to different benchmarks.

If all these leaks are correct, and you're correct, then amd picked the worst game possible to demonstrate Polaris 10 at Computex. They should have picked a dx11 game.

 

[sirmo]

If all these leaks are correct, and you're correct, then amd picked the worst game possible to demonstrate Polaris 10 at Computex. They should have picked a dx11 game.

They are obviously hiding performance, so I would say it's probably the best game possible for that.

 

[Armsdealer]

They are obviously hiding performance, so I would say it's probably the best game possible for that.

That's a perfectly reasonable conclusion.

I'd like to believe that you're right because the idea of having 3515 fs ultra points at stockclocks is absolutely mindboggling (imagine a 15-25% oc!). but I'm still skeptical about these leaks...

 

[Despoiler]

In the universe where I live AMD had hardly any advantage over NVIDIA before Pascal, but if you have numbers to corroborate your statement I an curious too see them.

Oh yah that thing called preemption that Pascal just finally added? Please dude....

There doesnt exist any hardware bottlenecks. AMD hasnt cared to fix their DX11 driver. Look at Tomb Raider: No difference with DX12. Or use the both Vulkan games Talos and Dota 2.

It's not the driver. DX11, being largely reliant on a single thread, simply cannot feed GCN fast enough. Do you think that if AMD just magically came up with better drivers for DX12 and somehow wouldn't put them in play for DX11? It's the API. DX11 is pure garbage.

Tomb Raider is an example that no one uses. Even the devs said not to use DX12 in their own game.

 

[BlitzWulf]

They are obviously hiding performance, so I would say it's probably the best game possible for that.

Exactly, they were able to raise doubts about the value of their competitors products with a product in a dramatically different price segment,all while revealing almost nothing about their offerings performance.

Very well played in my opinion,almost hard to believe it's AMD .
They have been very focused and executing well since Lisa and Raja have fully taken the reigns.

 

[topmounter]

That's fine, but the results imply that p10 is gimped in dx12 vs 11. The aots (dx12) benchmarks are implying way way worse performance vs both amd AND nvidia cards. Than these leaks.

Previously we saw amd dx12 vs dx11 was superior vs nvidia, and that nvidia basically performed the same in both. P10 -if you believe the benches as well as the aots Computex demo - is materially inferior in dx12.

One of the following must be true:

-The leaks are wrong

-P10 has such extraordinary command processor benefits that its gap over gtx 980 is greater in dx11 vs dx12

-amd purposely downplayed p10 performance at Computex

-there were driver issues in all of the aots benches we've seen thusfar

Take your pick. I have no idea.

Considering Computex was at least one month prior to GA, if nothing else, then I'd expect AMD to keep something in their back pocket so they can respond to 1060 rumors and/or announcements if needed. They did announce the "up to" $300 price target without any specifics beyond the $229 8GB RX 480.

 

[Armsdealer]

Considering Computex was at least one month prior to GA, if nothing else, then I'd expect AMD to keep something in their back pocket so they can respond to 1060 rumors and/or announcements if needed.

I'm starting to buy into it myself. Looking back at that leak video, a guy with a cat as his desktop wallpaper and a pink phone case cover doesn't seem like he gets his kicks by posting fake benchmarks to mislead people on purpose. This is such a huge win for amd if it's real...

 

[3DVagabond]

I had a visceral reaction, but then again I'm from Eastern Europe: that red revolution was actually about 1% of the population having full control over the rest, exactly the opposite of what they were claiming.

I don't like it, even as a joke.

This is where red being AMD's color is a problem.