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

[IllogicalGlory]

So 38 fps in 4k with 4xmsaa seems VERY promising for a $199-229 card.

Someone testing out overclocking perhaps?

That result isn't nearly as impressive as it seems because this is Ashes; MSAA and resolution increases don't work the normal way (I observed a 1 FPS decrease going from 4X to 8X, for example, at 1080p), and the details are turned down to high or something, which does affect FPS significantly. Unfortunately, I can't test it on my R9 290 as VSR doesn't work well enough but I imagine I'd be right there with it.

 

[IllogicalGlory]

It's getting thrashed by an R9 390(X) at 4K high, which is slightly below the settings tested on that Polaris chip, but...

http://www.ashesofthesingularity.co...-details/b807149c-9536-434a-b21b-af3a408515e8

http://www.ashesofthesingularity.co...-details/62fc254a-6f1d-4f4c-bfd0-b4195a0f3af6

 

[Bacon1]

It's getting thrashed by an R9 390(X) at 4K high, which is slightly below the settings tested on that Polaris chip, but...

http://www.ashesofthesingularity.co...-details/b807149c-9536-434a-b21b-af3a408515e8

http://www.ashesofthesingularity.co...-details/62fc254a-6f1d-4f4c-bfd0-b4195a0f3af6

What are those supposed to show?

Different client versions (1.12.19928.0 vs 1.2.18933.0) and 2x MSAA vs 4xMSAA @ 4k.

 

[IllogicalGlory]

Minor updates and MSAA is basically free.

This is at most an overclocked 290X using the same version and it's winning by plenty:

http://www.ashesofthesingularity.co...-details/f0b14c37-b2f6-4b82-84f2-35971f25edcf

 

[renderstate]

Most of the shading work in AOTS happens in texture space. When polygons are rendered on the screen a super simple shader with a single texture lookup is used to fill the screen. since the texture space resolution is decoupled from screen space resolution one can increase the latter without increasing the former, which results in performance much less dependent upon screen resolution. On the other hand going to very high res is not going to buy you any extra image quality on the interior of polygons but you'll get less jagged edges.

 

[Blitzvogel]

That could be very possible. Been awhile since a lineup of Radeons came out that could overclock well.

Is there any data on pipeline length of current GCN Radeons? Almost seems a case of short pipelines for AMD, hence "moar cores"/stream processors, versus Nvidia with a fewer faster ones that can accept such high clocks. However, I wonder how efficient Vega will be versus the GTX 1080.

Nvidia seems to have that secret sawce these days with power efficiency. Maybe Nvidia just has altogether better dynamic power switching technology than AMD?

 

[VulgarDisplay]

Minor updates and MSAA is basically free.

This is at most an overclocked 290X using the same version and it's winning by plenty:

http://www.ashesofthesingularity.co...-details/f0b14c37-b2f6-4b82-84f2-35971f25edcf

CPU overclock have any effect in this game?

 

[gamervivek]

That result isn't nearly as impressive as it seems because this is Ashes; MSAA and resolution increases don't work the normal way (I observed a 1 FPS decrease going from 4X to 8X, for example, at 1080p), and the details are turned down to high or something, which does affect FPS significantly. Unfortunately, I can't test it on my R9 290 as VSR doesn't work well enough but I imagine I'd be right there with it.

I remember previous AMD cards doing much better than nvidia at 4x to 8x AA performance delta than 2x to 4x. Perhaps it's the same now.

Besides, it's only ~10% slower than this Fury series card at 1440p extreme with similar hardware.(from another AMD tester and only the game version is different, but the results were the same for a previous game version)

http://www.ashesofthesingularity.co...-details/10ff69fc-86dc-4fed-b03f-82a7ae10f2f0

It's only slightly slower than this 390 series card but which is rendering at 1080p extreme.

http://www.ashesofthesingularity.co...-details/c9c7cc8a-6263-452c-b252-8f2db12a894f

And a Fury series card is 27% faster than the above 390S result at 1080p extreme and 18% faster than its 1440p result.

http://www.ashesofthesingularity.co...-details/857b2b5a-bde9-46e8-9b63-6c32fa0722b1

I think it's pretty likely the above scores are at stock settings and somewhat likely that the cards are full chip versions. So quite likely that it's in the 390X-Fury range of performance.

 

[IllogicalGlory]

To illustrate my point about MSAA being basically free:

(stock R9 290)

1800p High 2x MSAA 44.3 FPS
1800p High 4x MSAA 42.7 FPS
1800p High 8x MSAA 41.7 FPS

We lost 3.6% performance going from 2x to 4x, less from 4x to 8x.

And just some other results, some of which I've posted here before, as a reminder.

Stock:
1440p Extreme 39.3 FPS
1080p Crazy + 8x MSAA 34.0 FPS (settings same as Computex demo)

Overclocked (1128/1530):
1440p Extreme 46.0 FPS

As for my CPU, RTG's 5930K is more capable in this game, at least in single-GPU results.

 

[3DVagabond]

Is there any data on pipeline length of current GCN Radeons? Almost seems a case of short pipelines for AMD, hence "moar cores"/stream processors, versus Nvidia with a fewer faster ones that can accept such high clocks. However, I wonder how efficient Vega will be versus the GTX 1080.

Nvidia seems to have that secret sawce these days with power efficiency. Maybe Nvidia just has altogether better dynamic power switching technology than AMD?

I think nVidia's boost technology is the key, personally. And I also believe AMD is going wider. Everyone wants to talk clock speeds in a vacuum. But if Hawaii ran 1800MHz+ nothing would touch it. Never mind their newer stuff.

 

[Sweepr]

According to the chinese it will be in this order:

- Reference Radeon RX 480 8GB available on June 30th
- Custom Radeon RX 480 8GB available on mid-July
- Radeon RX 480 4GB available on August

http://tieba.baidu.com/p/4598447555

 

[kraatus77]

According to the chinese it will be in this order:

- Reference Radeon RX 480 8GB available on June 30th
- Custom Radeon RX 480 8GB available on mid-July
- Radeon RX 480 4GB available on August

http://tieba.baidu.com/p/4598447555

it's 29 june according to amd themselves. add 10-15 days for aib cards. just like nvidia.

 

[antihelten]

According to the chinese it will be in this order:

- Reference Radeon RX 480 8GB available on June 30th
- Custom Radeon RX 480 8GB available on mid-July
- Radeon RX 480 4GB available on August

http://tieba.baidu.com/p/4598447555

So the $200 card (the 4GB version) won't actually be available to begin with? That seems like a bit of a blunder by AMD.

 

[tential]

According to the chinese it will be in this order:

- Reference Radeon RX 480 8GB available on June 30th
- Custom Radeon RX 480 8GB available on mid-July
- Radeon RX 480 4GB available on August

http://tieba.baidu.com/p/4598447555

Sounds like an AMD thing to do. Lets see if it holds true.

 

[crisium]

Yeah seems very AMD. Just when some people (GCN 1.0 and Kepler/Fermi users especially) are really interested in a $199 upgrade, would AMD only offer $229 (or more)?

Of course Reference only at launch has been a mistake from AMD in the past. Given the low TDP I'm hopeful for reasonable acoustics, but having AIB cards at launch reviews would have been nice even if in small quantities relative to blower.

If that's true, it's possible 4GB $199 Polaris 10 was an afterthought. Maybe GP104 release did catch them off guard and forced AMD to have as low as prices as possible and they see 4GB as the only tenable way to get to $199 without eating too much of the margins, and they initially planned 8GB only.

 

[sirmo]

Yeah seems very AMD. Just when some people (GCN 1.0 and Kepler/Fermi users especially) are really interested in a $199 upgrade, would AMD only offer $229 (or more)?

Of course Reference only at launch has been a mistake from AMD in the past. Given the low TDP I'm hopeful for reasonable acoustics, but having AIB cards at launch reviews would have been nice even if in small quantities relative to blower.

If that's true, it's possible 4GB $199 Polaris 10 was an afterthought. Maybe GP104 release did catch them off guard and forced AMD to have as low as prices as possible and they see 4GB as the only tenable way to get to $199 without eating too much of the margins, and they initially planned 8GB only.

AMD talked about growing the TAM via low price way before Pascal launch.

 

[Qwertilot]

Is it definite that the 4GB one isn't slightly cut down or something? That would be one very rational reason to delay releasing it for a month.

Either that or there just won't be much stock/production for a bit. It does look like producing these kinds of 'big'(ger) chips on these tiny processes is fairly clearly not that easy. They'll sort it out of course, but there are reasons the big chips aren't rolling out right off!

I definitely wouldn't expect tons of stock at launch, or they would launch with AIB stuff/the 4GB version etc all at once.

 

[antihelten]

If that's true, it's possible 4GB $199 Polaris 10 was an afterthought. Maybe GP104 release did catch them off guard and forced AMD to have as low as prices as possible and they see 4GB as the only tenable way to get to $199 without eating too much of the margins, and they initially planned 8GB only.

So perhaps the initial plan was $250 for the 8GB 480 and $350 for the 8GB 480X, which would be exactly the same pricing as the 7850 and 7870.

But then they realised that an 8GB 480X at $350 wouldn't be competitive with 1070 at $380 and therefore adjusted the whole stack down by about $50 and introduced 4GB to help make it possible for them margin wise.

It's possible I suppose.

 

[tonyfreak215]

The 1266 clock number that's been seen seems fairly low for what we've seen on 14/16nm process. This is purely speculation, but if we translate clocks from last generation into what we can expect for an OC on AMD with 14/16nm then Nvidia's 1950-2100mhz max overcocks, could put rx480 overclocks in the 1700-1800mhz range.

Again that's just pure speculation based off overclocks from last generation. I could see the rx480 picking up quite a lot of ground on gtx1070 at that clock speed especially if memory speeds start to limit both cards.

Even on a Hawaii GPU you could see a nice jump in performance on a $199 GPU at those clock speeds.

Didn't Nvidia tweak their architecture to get higher clocks? (similar to what they did with resister level tweaks to get good power efficiency with Maxwell)

 

[tonyfreak215]

So the $200 card (the 4GB version) won't actually be available to begin with? That seems like a bit of a blunder by AMD.

Sounds like an AMD thing to do. Lets see if it holds true.

Yeah seems very AMD. Just when some people (GCN 1.0 and Kepler/Fermi users especially) are really interested in a $199 upgrade, would AMD only offer $229 (or more)?

Of course Reference only at launch has been a mistake from AMD in the past. Given the low TDP I'm hopeful for reasonable acoustics, but having AIB cards at launch reviews would have been nice even if in small quantities relative to blower.

If that's true, it's possible 4GB $199 Polaris 10 was an afterthought. Maybe GP104 release did catch them off guard and forced AMD to have as low as prices as possible and they see 4GB as the only tenable way to get to $199 without eating too much of the margins, and they initially planned 8GB only.

Damnit AMD! Just when you have everyone all excited about your new cards, you screw up the launch.
I really hope this isn't true....I was looking forward to getting an AIB card on launch day.

 

[Yakk]

AMD talked about growing the TAM via low price way before Pascal launch.

Yup, Raja Koduri and company have been pushing the TAM line for 6+ months now. There's even a video of Raja talking about doing this almost 2 years ago. Been in the works for a while it seems.

 

[krumme]

Damnit AMD! Just when you have everyone all excited about your new cards, you screw up the launch.
I really hope this isn't true....I was looking forward to getting an AIB card on launch day.

Good grief. Give it some rest. This is the standard procedure.

 

[krumme]

Yup, Raja Koduri and company have been pushing the TAM line for 6+ months now. There's even a video of Raja talking about doing this almost 2 years ago. Been in the works for a while it seems.

Well looking at the small diesize and low price they did more than talk about it. Its been planned for several years so this bs about pascal whatever is nonsense. The decisions was made like what 1-2 years ago?

 

[Mopetar]

Well looking at the small diesize and low price they did more than talk about it. Its been planned for several years so this bs about pascal whatever is nonsense. The decisions was made like what 1-2 years ago?

But how many other things have the talked about over that time that haven't come to fruition? You can't point to a single instance of something someone said years ago and use it to suggest it was the plan all along without also looking at all of the things that didn't come true.

 

[lukart]

Is it possible amd is waiting for gddr5x availability to drop a $299 full p10?

It would be a good reason for the lack of news regarding the suspected full chip.



Sent from my XT1575 using Tapatalk

Not happening. GDDR5X is around 5-8% faster costing 20% more. Not pratical for a mainstream card. :thumbsdown: