AMD Computex 2016 - Polaris and Zen Revealed

[Energycore]

4 minutes ago, zMeul said:

judging by Pascal relation to Maxwell v2, a GTX1060 will be in the same perf range

Do you think we're likely to see a $250-300 priced, GP104 based 1060 Ti?

[Mira Yurizaki]

I'm still holding my breath on how Zen will actually perform. Certainly for multithreaded tasks it will eat Intel alive, but single threaded performance is still king in the realm of games. And I don't think multithreaded processing for games is really going to take off for anything other than compiling display lists.

 

And well, the daily utilities aren't going to perform any better either. They run just fine on a mobile part.

[Kobz360]

2 minutes ago, Energycore said:

Alright, these are the two relevant ashes benchmarks I could dig out from the RadeonDemo profile. As you can see they used a 5930K.

 

Crazy 1440p, GTX 1080

 

Crazy 1440p, Crossfire RX 480 x2

Yeah I saw these too and im still curious. I have a feeling they did all of this on another profile because not only is the 1080's FPS overstated in the live stream, but so is their own (which is pointless). There's also a lot more 1080 benches than 480 ones and the timing between all of them is all over the place but who knows.. 

[Megazero]

3 minutes ago, Energycore said:

Alright, these are the two relevant ashes benchmarks I could dig out from the RadeonDemo profile. As you can see they used a 5930K.

 

Crazy 1440p, GTX 1080

 

Crazy 1440p, Crossfire RX 480 x2

I could be wrong here, but I try to search for other 1080 benchmark on Ashes with the same i7 5930k, and there's only 2 other result, both have 10-20fps higher than the AMD's result, you can check here:

http://www.ashesofthesingularity.com/metaverse#/ladders/benchmark/overall/Crazy_1440p?viewType=myself&filters={"gpu":"NVIDIA GeForce GTX 1080","cpu":"Intel(R) Core(TM) i7-5930K CPU @ 3.50GHz","api":"DirectX 12"}

 

[Energycore]

1 minute ago, Kobz360 said:

Yeah I saw these too and im still curious. I have a feeling they did all of this on another profile because not only is the 1080's FPS overstated in the live stream, but so is their own (which is pointless). There's also a lot more 1080 benches than 480 ones and the timing between all of them is all over the place but who knows.. 

I'm gonna choose to disregard what they showed in the presentation and instead use these as a real data point.

[Energycore]

2 minutes ago, Megazero said:

I could be wrong here, but I try to search for other 1080 benchmark on Ashes with the same i7 5930k, and there's only 2 other result, both have 10-20fps higher than the AMD's result, you can check here:

http://www.ashesofthesingularity.com/metaverse#/ladders/benchmark/overall/Crazy_1440p?viewType=myself&filters={"gpu":"NVIDIA GeForce GTX 1080","cpu":"Intel(R) Core(TM) i7-5930K CPU @ 3.50GHz","api":"DirectX 12"}

 

EDIT: Actually, the link you posted shows people with 1080s and 5930Ks getting similar performance both in score and fps. Am I missing something?

[Megazero]

Just now, Energycore said:

Another thing they could have done is grab the clock offset in MSI Afterburner and tell it to give -400MHz to the 1080 clock or something like that

Haha, or they could cook a BBQ on their R9 290 and blow the hot air toward the case of the 1080 while sealing that case off :3 Who knows~

[Energycore]

1 minute ago, Megazero said:

Haha, or they could cook a BBQ on their R9 290 and blow the hot air toward the case of the 1080 while sealing that case off :3 Who knows~

Edited my other post

[zMeul]

8 minutes ago, Energycore said:

Do you think we're likely to see a $250-300 priced, GP104 based 1060 Ti?

TI? I don't think it's a Ti at all

and yes, sub 300$ price point

[Energycore]

Just now, zMeul said:

TI? I don't think it's a Ti at all

and yes, sub 300$ price point

But it's going to be GP106 then, right?

[Megazero]

1 minute ago, Energycore said:

Edited my other post

This is what show up if you click on those 2:

[Energycore]

1 minute ago, Megazero said:

This is what show up if you click on those 2:

Yup, I got the same. Those 2 get 49.something and RadeonDemo's run gets 51.7.

 

Not 10-20 fps apart. CPU framerate varies more so, but I don't know what that is.

[zMeul]

3 minutes ago, Energycore said:

But it's going to be GP106 then, right?

I don't exactly recall the rumors, but it could be a 3rd cut version of GP 104 

[Energycore]

Just now, zMeul said:

I don't exactly recall the rumors, but it could be a 3rd cut version of GP 104 

I hope so because even if it was 66% GP104 it would be pretty beast. 66% GP104 should be about 980 / Fury performance.

[Megazero]

2 minutes ago, Energycore said:

Yup, I got the same. Those 2 get 49.something and RadeonDemo's run gets 51.7.

 

Not 10-20 fps apart. CPU framerate varies more so, but I don't know what that is.

Nvm, I was looking at the CPU framerate LOL. But then... the Framerate in Blue color is the GPU right? if so then the 58 fps for the 1080 is from the normal batch, yeah, but where do the 65 fps of the CF come from?

[Energycore]

1 minute ago, Megazero said:

Nvm, I was looking at the CPU framerate LOL. But then... the Framerate in Blue color is the GPU right? if so then the 58 fps for the 1080 is from the normal batch, yeah, but where do the 65 fps of the CF come from?

Probably ran the benchmark multiple times and this one time it posted 62 (I recall it being 62.something, not 65) and compared that to this one run of the 1080.

 

I'd do it. If I wanted to show off my card in benchmarks, I'd run 5 and use the best result. And overclock the bananas out of it. To be fair we don't know whether the 480s are overclocked.

[Curufinwe_wins]

5 minutes ago, Energycore said:

Probably ran the benchmark multiple times and this one time it posted 62 (I recall it being 62.something, not 65) and compared that to this one run of the 1080.

 

I'd do it. If I wanted to show off my card in benchmarks, I'd run 5 and use the best result. And overclock the bananas out of it. To be fair we don't know whether the 480s are overclocked.

I mean guys lets not be disingenuous... IF the 480 offers around Hawaii perf in AotS, then to be blatantly honest, you SHOULD expect two 480s to beat the 1080 (which was compared so many times to be about on par with dual 980s) with good scaling.

 

Now how they do in other games, we will see.

[Megazero]

6 minutes ago, Energycore said:

Probably ran the benchmark multiple times and this one time it posted 62 (I recall it being 62.something, not 65) and compared that to this one run of the 1080.

 

I'd do it. If I wanted to show off my card in benchmarks, I'd run 5 and use the best result. And overclock the bananas out of it. To be fair we don't know whether the 480s are overclocked.

Hmm...

 

On a side note, if they want to only show us a CF benchmark, then let's use it to determine how powerful the GPU is. 2 R9 390 in CF with a 5930k have the score of 5400-5900 and the framerate average of all batch is 55-61fps, 4-10fps higher than the 51fps of the 2 480 LOL. Or as they say 62fps then it's about the same as 2 390,

[Energycore]

Just now, Curufinwe_wins said:

I mean guys lets not be disingenuous... IF the 480 offers around Hawaii perf in AotS, then to be blatantly honest, you SHOULD expect two 480s to beat the 1080 with good scaling.

 

Now how they do in other games, we will see.

I agree. I'm not not excited about the card, but the seemingly different benchmarks was a real flop from AMD. Considering they actually have bench in the archive that makes their cards look good anyway. But no, rather than say "Our 2 cards legit perform at 98% of a 1080" they had to go with the skewed results.

At least looking at the bench archive from the ashes site puts my mind at rest. And speaking of rest!

[Energycore]

2 minutes ago, Megazero said:

Hmm...

 

On a side note, if they want to only show us a CF benchmark, then let's use it to determine how powerful the GPU is. 2 R9 390 in CF with a 5930k have the score of 5400-5900 and the framerate average of all batch is 55-61fps, 4-10fps higher than the 51fps of the 2 480 LOL. Or as they say 62fps then it's about the same as 2 390,

It's like they purposefully didn't show us actual single-card performance so we'd stress about it. Damn you AMD.

[Spectrez]

Very weird to see so little from AMD.

 

All we can hope for at this point is some good benchmarks from reviewers  

[Kobz360]

Even though it's unlikely to perform as well as we'd all like it, (in terms of being close to a 1070 and whatnot) $199 is a still pretty insane price when compared to the GPUs its supposedly supposed to perform around. 

 

I think many of you were expecting an insane high power GPU following what NVIDIA delivered us, but in reality we all knew it wasn't going to be overly spectacular. I'm just waiting to see some other benchmarks, as well as witness what the AIB partners can deliver us in terms of overclocking.. Overall, I'd say what AMD has shown us has been very good. But of course, that's just my opinion 

[StainlessSR]

3 hours ago, zMeul said:

regarding the RX480 vs GTX1080 Ashes so called "benchmark" - you don't even need to look closely to see AMD tweaked the settings to work in their favor; what's up with the ground quality Rajah?!?!

snow. the left frames have more(and deeper) snow cover making the ground "softer".  As a benchmark tool for comparisons having a difference like that is not a good thing. As a benchmark tool to see how well your pc may run that game it wouldn't matter.

[Carclis]

32 minutes ago, StainlessSR said:

snow. the left frames have more(and deeper) snow cover making the ground "softer".  As a benchmark tool for comparisons having a difference like that is not a good thing. As a benchmark tool to see how well your pc may run that game it wouldn't matter.

I'd be interested to hear from someone who does own Ashes whether that is true or AMD was actually running on lower settings. I don't feel like they did themselves any favours by demonstrating crossfire as opposed to a single 480 though.

[zMeul]

34 minutes ago, StainlessSR said:

snow. the left frames have more(and deeper) snow cover making the ground "softer".  As a benchmark tool for comparisons having a difference like that is not a good thing. As a benchmark tool to see how well your pc may run that game it wouldn't matter.

there are a lot of terrain patches that aren't covered in snow and they still don't look as good as the GTX1080 run