AMD Computex 2016 - Polaris and Zen Revealed

[Brianyu]

@Clanscorpia Does anybody know his real name?  Does he keep it private? i know him irl. like we are bff's. 

[LabRat]

we'll find out in July what the real side by side numbers are.... real world that is.

[Japi]

http://videocardz.com/amd/radeon-rx-400/radeon-rx-490

[Valentyn]

10 hours ago, Briggsy said:

Yeah, on second thought I hope that's not an overclocked 480, or its weaker than a 390. Maybe Drivers aren't up to snuff for the 480 yet.

 

Here's my 390 score, overclocked to 1150/1650 which is really modest and doable by all 390s. My CPU is somewhat overclocked, but that shouldn't make a huge difference on the graphic score.

 

http://www.3dmark.com/3dm11/11299972

 

Did another run with stock clocks. Very close to the 480 score a few posts back. I really hope it was a stock 480.

http://www.3dmark.com/3dm11/11300012

 

 

The interesting thing found at Overclocker.co.uk was that the 390 somehow scores better than even 390X, 290X and even the Fury's when it has even the slightest of OCs.

There's something odd with the 390 in that benchmark, also the 480 isn't OC'd, and neither is the CPU there; which is a 4770K in an H87M motherboard near useless chipset and mobo for overclocking.

Also the benchmark seems to love a good CPU OC, each 150Mhz adding around 200 points to the GPU Score.

Also worth nothing your 390 is also beating a lot of GTX 980's there.

Here is the thread at OcUK, and the last 2 pages are covered in folks with various AMD cards from 290-s Fury's running at stock and OC to compare.

A stock 290X seems to get a GPU score of ~16200
 

https://forums.overclockers.co.uk/showthread.php?t=18711396&page=224

 

So stock for stock it seems the 480 is in Fury/Nano performance in 3DMark 11. Although synthetics are odd, as shown by your 390 spanking even a Fury that's had it's extra shaders unlocked.

I do wonder that proper drivers will also do for the new Polaris cards, AMD hasn't been good at getting great performance drivers out on launch the past 1.5 years.

[patrickjp93]

14 hours ago, mariushm said:

If anyone still has that 51% figure in their heads... you guys are understanding it wrong.

 

Read this:  https://www.reddit.com/r/Amd/comments/4m692q/concerning_the_aots_image_quality_controversy/

 

2x Radeon RX 480 - 62.5 fps | Single Batch GPU Util: 51% | Med Batch GPU Util: 71.9 | Heavy Batch GPU Util: 92.3%

GTX 1080 – 58.7 fps | Single Batch GPU Util: 98.7%| Med Batch GPU Util: 97.9% | Heavy Batch GPU Util: 98.7%

 

Reddit user asks :

 

AMD Rep answered:

 

So basically with a "robust" game like AOTS they achieve 1.83x out of a maximum 2x ... with arguably still some unrefined drivers, with room for improvement. The 51% number was cherry-picked.

 

Also note the other comment about nVidia cards showing better performance due to a bug in how they render:

 

 

 

 

That comment from AMD wrt the 1080 not rendering correctly is bullshit and AOTS developers even said so. You can even tell with how the 480 handles shadows to lower quality levels than the 1080. When independent reviews come out, AMD will get burned for this like it always does. AMD needs to stop hiring habitual liars for PR reps.

[Briggsy]

3 hours ago, Valentyn said:

The interesting thing found at Overclocker.co.uk was that the 390 somehow scores better than even 390X, 290X and even the Fury's when it has even the slightest of OCs.

There's something odd with the 390 in that benchmark, also the 480 isn't OC'd, and neither is the CPU there; which is a 4770K in an H87M motherboard near useless chipset and mobo for overclocking.

Also the benchmark seems to love a good CPU OC, each 150Mhz adding around 200 points to the GPU Score.

Also worth nothing your 390 is also beating a lot of GTX 980's there.

Here is the thread at OcUK, and the last 2 pages are covered in folks with various AMD cards from 290-s Fury's running at stock and OC to compare.

A stock 290X seems to get a GPU score of ~16200
 

https://forums.overclockers.co.uk/showthread.php?t=18711396&page=224

 

So stock for stock it seems the 480 is in Fury/Nano performance in 3DMark 11. Although synthetics are odd, as shown by your 390 spanking even a Fury that's had it's extra shaders unlocked.

I do wonder that proper drivers will also do for the new Polaris cards, AMD hasn't been good at getting great performance drivers out on launch the past 1.5 years.

Yeah, will definitely need a more modern test to get a better idea.

 

This post that i think you were referencing leans towards what I was thinking about my own overclocked cpu speeds effecting graphic score. Also makes the 480 look pretty good. guess it will depend on how far the 480 can overclock (will probably need an AIB card with 8-pin to really push it).

https://forums.overclockers.co.uk/showpost.php?p=29580136&postcount=6711

[IGJoe2192]

21 hours ago, Clanscorpia said:

I mean like their buisness plan

The majority of sales and market are in the mid-low range components that most people can afford or are willing to pay for. AMD is looking to expand their market share and increase profits by releasing great price/performance products in the mid range it will make them a no brainer when it comes to budget builds. People want to get the most out of their money and it looks like AMD is lining up to do just that.

[Kobz360]

36 minutes ago, patrickjp93 said:

That comment from AMD wrt the 1080 not rendering correctly is bullshit and AOTS developers even said so. You can even tell with how the 480 handles shadows to lower quality levels than the 1080. When independent reviews come out, AMD will get burned for this like it always does. AMD needs to stop hiring habitual liars for PR reps.

Source?

[Briggsy]

4 minutes ago, Kobz360 said:

Source?

+1 

 

would like to see a source on that too.

[Valentyn]

51 minutes ago, Briggsy said:

Yeah, will definitely need a more modern test to get a better idea.

 

This post that i think you were referencing leans towards what I was thinking about my own overclocked cpu speeds effecting graphic score. Also makes the 480 look pretty good. guess it will depend on how far the 480 can overclock (will probably need an AIB card with 8-pin to really push it).

https://forums.overclockers.co.uk/showpost.php?p=29580136&postcount=6711

Yup, I've been a member of OcUK for years, and there's always some interesting testing going on there.

Will be interesting to see what the 480 can do on more modern software, with release drivers. 

 

Even so it's shaping up to be a very good card for the mainstream so far.

[Starelementpoke]

1 hour ago, patrickjp93 said:

That comment from AMD wrt the 1080 not rendering correctly is bullshit and AOTS developers even said so. You can even tell with how the 480 handles shadows to lower quality levels than the 1080. When independent reviews come out, AMD will get burned for this like it always does. AMD needs to stop hiring habitual liars for PR reps.

Like others are saying, source?

[BingoFishy]

On 6/1/2016 at 7:10 AM, zMeul said:

because you can't run the game in off-line ?!?!

 

---

 

this is supposedly AMD's run

• 2x RX480: http://ashesofthesingularity.com/metaverse#/personas/b0db0294-8cab-4399-8815-f956a670b68f/match-details/0561a980-78ce-4e24-a7c3-2749a8e33aac
• GTX1080 http://ashesofthesingularity.com/metaverse#/personas/b0db0294-8cab-4399-8815-f956a670b68f/match-details/d80e73ae-e7ed-46e9-9b73-551276ac8a16

check the Terrain Shading Samples (basically Tessellation) for each run:

• RX480 - 8mil
• GTX1080 - 16mil

AMD still has problems with tessellation in their own fucking game 

The framerate difference is greater than another set of links I saw. I think that these two weren't meant to be compared. In this, the 480s are significantly better, whereas with the other links, the 480s and the 1080 were roughly equivalent.

[christianled59]

I really hope VR starts getting better crossfire and sli support. I'd drop $400 on 2 480s in a heartbeat if it did, but I think I'm going to wait out the market.

[Valentyn]

22 minutes ago, Starelementpoke said:

Like others are saying, source?

I would also like to see his source where the AoTS devs responded.

As here you can clearly see the GTX 1080 on the right is Not fully rendering snow. There is clear visible ridges, and mountain tops that as much less snow than that of the RX480 CF on the left. 

It's clear the 480's are busy rendering what looks like a snow storm, and extra ground clutter snow as well.

[LeStringMan]

6 minutes ago, christianled59 said:

I really hope VR starts getting better crossfire and sli support. I'd drop $400 on 2 480s in a heartbeat if it did, but I think I'm going to wait out the market.

Well that's up to the developers of said games to add multi-gpu support, but Directx12 should theoretically make adding multi gpu support easier, again though its up to developers.

[Trixanity]

45 minutes ago, Starelementpoke said:

Like others are saying, source?

 

1 hour ago, Briggsy said:

+1 

 

would like to see a source on that too.

 

1 hour ago, Kobz360 said:

Source?

Guys. It's actually true. It has been confirmed that the issue is not the rendering path. It has been discovered that the GTX 1080 runs so hot that it melts the snow in the game which explains why there is so much more snow when running the game on the RX 480.

[christianled59]

25 minutes ago, LeStringMan said:

Well that's up to the developers of said games to add multi-gpu support, but Directx12 should theoretically make adding multi gpu support easier, again though its up to developers.

Not entirely. SteamVR has issues with multi gpus. I'm currently developing for the Vive myself, and its very weird how the drivers work for it.

[LeStringMan]

6 minutes ago, christianled59 said:

Not entirely. SteamVR has issues with multi gpus. I'm currently developing for the Vive myself, and its very weird how the drivers work for it.

 Perhaps this is the reason for AMD's Crossfire API on gpuopen?

[xAcid9]

32 minutes ago, Valentyn said:

I would also like to see his source where the AoTS devs responded.

As here you can clearly see the GTX 1080 on the right is Not fully rendering snow. There is clear visible ridges, and mountain tops that as much less snow than that of the RX480 CF on the left. 

It's clear the 480's are busy rendering what looks like a snow storm, and extra ground clutter snow as well.

-snip-

Yep, even 980 Ti rendered similar to the 480 or my 290x in latest build as shown by DudeRandom84

 

Check 1:25 and compare to this.

Spoiler

 

[patrickjp93]

1 hour ago, Starelementpoke said:

Like others are saying, source?

They used the review driver instead of the release one which had been out for a week before Computex: https://m.reddit.com/r/pcmasterrace/comments/4iix1i/gtx_1080_ashes_of_the_singularity_benchmarks/

 

AMD manipulated the data, again.

[zMeul]

3 hours ago, patrickjp93 said:

That comment from AMD wrt the 1080 not rendering correctly is bullshit and AOTS developers even said so. You can even tell with how the 480 handles shadows to lower quality levels than the 1080. When independent reviews come out, AMD will get burned for this like it always does. AMD needs to stop hiring habitual liars for PR reps.

 

this is a screen from someone with a GTX1080 at UHD resolution:

Spoiler

AMD's PR department should melt for the bullcrap they spew

 

my guess is that they used different versions of the game

[christianled59]

1 hour ago, LeStringMan said:

 Perhaps this is the reason for AMD's Crossfire API on gpuopen?

Maybe. I think is most cases, it'll have to be something worked out on Valve's end. By drivers, I meant VR drivers for the headset its self. I'm unsure if the oculus has the same issues as the vive, but it'd be nice if it worked. 

 

For all I know, sli not working in vr may be intentional. Multi gpu setups are plagued with problems and its the reason why I tend to stay with a single gpu.

 

For me, the 480 might be a sign from God. I have a crossfire motherboard with a NVIDIA card and no intentions of multi gpu. Lol

 

[LeStringMan]

3 hours ago, christianled59 said:

Maybe. I think is most cases, it'll have to be something worked out on Valve's end. By drivers, I meant VR drivers for the headset its self. I'm unsure if the oculus has the same issues as the vive, but it'd be nice if it worked. 

 

For all I know, sli not working in vr may be intentional. Multi gpu setups are plagued with problems and its the reason why I tend to stay with a single gpu.

 

For me, the 480 might be a sign from God. I have a crossfire motherboard with a NVIDIA card and no intentions of multi gpu. Lol

 

Ah yea i wasn't thinking about drivers for the headset lol, i havent had any problems with crossfire honestly, but then again i only went ahead and did it because most of the games i play at the moment do support it.

[Kobz360]

8 hours ago, patrickjp93 said:

They used the review driver instead of the release one which had been out for a week before Computex: https://m.reddit.com/r/pcmasterrace/comments/4iix1i/gtx_1080_ashes_of_the_singularity_benchmarks/

 

AMD manipulated the data, again.

So you don't have a source of the AotS developers saying anything?

 

[patrickjp93]

14 hours ago, Kobz360 said:

So you don't have a source of the AotS developers saying anything?

 

I do. Look inside.