AMD Lists The Radeon RX 490 Flagship – Polaris based Dual GPU Graphics Card For 4K Ready Gaming

[Bouzoo]

14 minutes ago, zMeul said:

it's quite possible it will be a dual-GPU and aimed specifically at the same market Radeon Pro Duo was - not gaming

but AMD is screwing themselves with this card, the engineering alone that went into it

 

out of curiosity, how many went into a store and bough a Pro Duo?! I'd suspect not that many - because that card doesn't really makes sense

and neither a dual Polaris10 card

Pro Duo came out too late imo, and I honestly don't know a single person that bought one, irl or online. It's like the card never came out in the first place. 

Remind me, do dual GPU cards perform better than 2 equivalent crossfire cards due to chips being on the same PCB, or is it the same story? They need the same crossfire profiles ya?

[Khvarrioiren]

26 minutes ago, Trixanity said:

 

We have a bunch of companies who will get business almost no matter what. That list includes Nvidia, Intel, Apple and Samsung. There are others - many others - but those are definitely the big four in the business.

I can attest to this. Many, many of my client builds have Nvidia GPU's in them simply because the client told me to put "an Nvidia GPU " in it. No nomenclature or anything. Simply n Nvidia GPU. Some of them have some knowledge about the GPU-side of things, and still go with Nvidia even if the offerings from AMD's side are way superior in the price bracket just because hey, Nvidia. Not like I'm complaining, that let's me keep fatter margins.

[Khvarrioiren]

A screw up, please delete.

[zMeul]

20 minutes ago, Bouzoo said:

do dual GPU cards perform better than 2 equivalent crossfire cards due to chips being on the same PCB, or is it the same story? They need the same crossfire profiles ya?

dual GPU cards perform about the same as single GPU cards in CFX

and yes, they need the same profiles

[Prysin]

1 hour ago, zMeul said:

it's quite possible it will be a dual-GPU and aimed specifically at the same market Radeon Pro Duo was - not gaming

but AMD is screwing themselves with this card, the engineering alone that went into it

 

out of curiosity, how many went into a store and bough a Pro Duo?! I'd suspect not that many - because that card doesn't really makes sense

and neither a dual Polaris10 card

i'd be shocked if i SAW a Pro Duo in a store... its waaay to expensive of a niche product to be worth having in store. It would only take up shelf space.

[Khvarrioiren]

26 minutes ago, Prysin said:

i'd be shocked if i SAW a Pro Duo in a store... its waaay to expensive of a niche product to be worth having in store. It would only take up shelf space.

I've made a client build with one. Poor sod had to wait for nearly 2 months and pay 1900 € for it.

[Dabombinable]

4 minutes ago, Khvarrioiren said:

I've made a client build with one. Poor sod had to wait for nearly 2 months and pay 1900 € for it.

Ouch. 3x FuryX would have been cheaper I guess?

[Khvarrioiren]

1 minute ago, Dabombinable said:

Ouch. 3x FuryX would have been cheaper I guess?

About 200-300€ more. I'm kinda mad that he ordered the build before I started doing the client builds on my own. Would've made some nice bank instead of the shitty 16€/hr.

[Dabombinable]

2 minutes ago, Khvarrioiren said:

About 200-300€ more. I'm kinda mad that he ordered the build before I started doing the client builds on my own. Would've made some nice bank instead of the shitty 16€/hr.

I think that the Radeon Pro Duo is just a massive ripoff.

[Khvarrioiren]

1 minute ago, Dabombinable said:

I think that the Radeon Pro Duo is just a massive ripoff.

Pretty much.

[SteveGrabowski0]

6 hours ago, Humbug said:

polaris 10 was never supposed to be high end or fury X tier. That much is crystal clear from the resources it has and the die size.

 

as to whether AMD also planned rebrands or dual GPUs to fill the void until vega I have no idea.

But Polaris 10 was supposed to be crazy efficient, which it is definitely not.

[Prysin]

36 minutes ago, Dabombinable said:

Ouch. 3x FuryX would have been cheaper I guess?

nop

Also, Radeon Pro Duo has Firepro drivers AND gaming drivers. Fury X does not. There is no "Firepro" version of Fury X that allow you to make a custom bios and flash it.

Unlike the 290X which has the W9100 "Firepro" equivalent.

[Prysin]

35 minutes ago, SteveGrabowski0 said:

But Polaris 10 was supposed to be crazy efficient, which it is definitely not.

is it not?

 

ok, let us compare the performance numbers shall we?

 

R9 290

= or -10% behind RX 480

TDP 290w

 

R9 390

= or -5% behind RX 480

TDP 275w (due to fine tuning and improvements in the 28nm process and binning)

 

RX 480

TDP 150w

 

Meaning it is about 1.8-1.9x more efficient, for the same performance, as the cards it is replacing.

 

Other benefits of the RX 480?

Tesselation performance has increased and is now ON PAR WITH A GTX 970

A GTX 970 is 25% faster in tesselation then a 290 or 390.

 

naturally, 25% better tesselation performance is irrelevant, because obviously isnt any games using heavy tesselation. Especially not AAA games.

 

 

Yes Steve. I agree. There seems to be NO efficiency improvements here. None at all.

[SteveGrabowski0]

4 minutes ago, Prysin said:

is it not?

 

ok, let us compare the performance numbers shall we?

 

R9 290

= or -10% behind RX 480

TDP 290w

 

R9 390

= or -5% behind RX 480

TDP 275w (due to fine tuning and improvements in the 28nm process and binning)

 

RX 480

TDP 150w

 

Meaning it is about 1.8-1.9x more efficient, for the same performance, as the cards it is replacing.

 

Other benefits of the RX 480?

Tesselation performance has increased and is now ON PAR WITH A GTX 970

A GTX 970 is 25% faster in tesselation then a 290 or 390.

 

naturally, 25% better tesselation performance is irrelevant, because obviously isnt any games using heavy tesselation. Especially not AAA games.

 

 

Yes Steve. I agree. There seems to be NO efficiency improvements here. None at all.

LOL the spin. So it's as efficient as the cutdown version of Nvidia's 28 nm GM204 offering from almost two years ago.

[Prysin]

42 minutes ago, Dabombinable said:

I think that the Radeon Pro Duo is just a massive ripoff.

for consumers. Yes

For PRO-SUMERS or enterprise. HELL NO

 

Dab, i have some homework for you.

 

Find me ANY product from AMD, Nvidia, VIA, Intel, Qualcomm, Apple, Samsung, IBM or whatever that can deliver the following:

 

16.8TFLOPs single precision performance

1/32 double precision

DX12_0 compliancy

Validated professional drivers, optimized for professional software solutions such as CAD/CAE

Pricetag: $1,500 US

 

Go on. PLEASE find me such a product.

 

Incase you do not believe me, here is the official Firepro drivers for the Pro Duo,

http://support.amd.com/en-us/kb-articles/Pages/Workstation-Graphics-Drivers.aspx#desktop

[Dabombinable]

11 minutes ago, Prysin said:

for consumers. Yes

For PRO-SUMERS or enterprise. HELL NO

 

Dab, i have some homework for you.

 

Find me ANY product from AMD, Nvidia, VIA, Intel, Qualcomm, Apple, Samsung, IBM or whatever that can deliver the following:

 

16.8TFLOPs single precision performance

1/32 double precision

DX12_0 compliancy

Validated professional drivers, optimized for professional software solutions such as CAD/CAE

Pricetag: $1,500 US

 

Go on. PLEASE find me such a product.

 

Incase you do not believe me, here is the official Firepro drivers for the Pro Duo,

http://support.amd.com/en-us/kb-articles/Pages/Workstation-Graphics-Drivers.aspx#desktop

2x FuryX can deliver that performance. And you do realise that its possible to tweak the settings of graphics cards for the same end result as professional cards.

[Prysin]

4 minutes ago, SteveGrabowski0 said:

LOL the spin. So it's as efficient as the cutdown version of Nvidia's 28 nm GM204 offering from almost two years ago.

Well Steve, what are you comparing it to?

There is no 1060 out. So there is no current gen product that you can compare it to.

 

We already know that under DX12, it is closer to a 980 then a 970. Mostly because of how shit Maxwell is at Async compute. That is but a "flaw" in the design of Maxwell.

So until 1060 comes out and has been reviewed, there is NO REAL competitor for it. The 960, 970. 980. 380, 380X, 390, 390X are all made obsolete by this ONE product.

 

But wait, i can feel something coming. Oh yes, it's your next argument.

"But who cares for DX12, it will take years before it catches on"

"Async compute is but a lame feature nobody will use"

 

And you would be wrong, on both accounts.

Why?

CONSOLES

Consoles, namely the "pisspoor" and "fagbox" has been out since 2013. They are both using APIs which is similar to DX12 in many ways, in the case of the "fagbox", it is more or less the same. Both consoles uses x86-64 architectures, and due to this, porting to PC has NEVER been easier. Async compute is an essential part in getting the most performance out of a console. Without using it, you WONT get the best out of a console. It is a requirement in order to achieve the best visuals and effects.

 

So essentially, developers has spent three YEARS "training" on how to use DX12 features. All they need to learn now is what effect corresponds to what liberary inside DX12. Meaning, it is essentially like reading a book in a different language and just translate it word for word.

Time consuming, yes, but sooner or later you learn several of the new words and wont have to look them up.

 

I think DX12, for what it is and what it offers, will have the fastest API "adoption" ever. Not just because the functions and features has been in use for years. But because AMD, Nvidia, Intel and Microsoft are heavily pushing for it. As is Chronos on the "Open" side.

DX12 is already being implemented into Unity and Unreal engines. The two biggest, most widespread engines in use, both among AAA titles and indy games.

 

 

[Prysin]

6 minutes ago, Dabombinable said:

2x FuryX can deliver that performance. And you do realise that its possible to tweak the settings of graphics cards for the same end result as professional cards.

no you cannot tweak the Fiji based cards for the same results without the professional drivers. That only worked with the Tahiti and Hawaii based cards.

Pro Duo also has ECC memory.

Fury X does not. Never had never will.

2x Fury X is also limited by space restraints due to 2x radiators. If you want to buy 2x waterblocks + single rad + hose/pipes and pump + rad, then the price will be WAY over 1500 USD.

2x Fury X will draw more power (because non-binned GPUs)

 

 

The only cards you can "tweak" into Firepro cards are pre-Fiji cards. because they had Firepro equivalents.

You could just flash the bios to a modded Firepro version. Which would allow you to install the Firepro driver stack.

Without a card using a Firepro BIOS, the Firepro driver stack will NOT install properly.

 

 

 

[Humbug]

1 hour ago, SteveGrabowski0 said:

But Polaris 10 was supposed to be crazy efficient, which it is definitely not.

Yes. It's efficient compared to previous AMD GCN cards.

But certainly not compared to Pascal. Anyway I don't think it's a big deal on desktop, because it's certainly efficient enough to not be a concern.

 

But it is a big concern for AMD in non-desktop market. So I'm interested to see if the laptop parts like Pascal 11 are more efficient, that's where efficiency really matters with all the battery life and thermal constraints. If those are only as efficient as Polaris 10 that's a fail for AMD.

[gilang01]

7 minutes ago, Prysin said:

Well Steve, what are you comparing it to?

There is no 1060 out. So there is no current gen product that you can compare it to.

 

We already know that under DX12, it is closer to a 980 then a 970. Mostly because of how shit Maxwell is at Async compute. That is but a "flaw" in the design of Maxwell.

So until 1060 comes out and has been reviewed, there is NO REAL competitor for it. The 960, 970. 980. 380, 380X, 390, 390X are all made obsolete by this ONE product.

 

But wait, i can feel something coming. Oh yes, it's your next argument.

"But who cares for DX12, it will take years before it catches on"

"Async compute is but a lame feature nobody will use"

 

And you would be wrong, on both accounts.

Why?

CONSOLES

Consoles, namely the "pisspoor" and "fagbox" has been out since 2013. They are both using APIs which is similar to DX12 in many ways, in the case of the "fagbox", it is more or less the same. Both consoles uses x86-64 architectures, and due to this, porting to PC has NEVER been easier. Async compute is an essential part in getting the most performance out of a console. Without using it, you WONT get the best out of a console. It is a requirement in order to achieve the best visuals and effects.

 

So essentially, developers has spent three YEARS "training" on how to use DX12 features. All they need to learn now is what effect corresponds to what liberary inside DX12. Meaning, it is essentially like reading a book in a different language and just translate it word for word.

Time consuming, yes, but sooner or later you learn several of the new words and wont have to look them up.

 

I think DX12, for what it is and what it offers, will have the fastest API "adoption" ever. Not just because the functions and features has been in use for years. But because AMD, Nvidia, Intel and Microsoft are heavily pushing for it. As is Chronos on the "Open" side.

DX12 is already being implemented into Unity and Unreal engines. The two biggest, most widespread engines in use, both among AAA titles and indy games.

 

 

Let me help you.

BF1 and all future DICE/EA game will support DX12 thanks to frostbite. Rise of the Tomb Raider supports dx12 and async compute. Ported Xbone games supports dx12 right out of the box. All future Total War Games supports DX12. 

DX12 is the future. 

[Dabombinable]

7 minutes ago, Prysin said:

Well Steve, what are you comparing it to?

There is no 1060 out. So there is no current gen product that you can compare it to.

 

We already know that under DX12, it is closer to a 980 then a 970. Mostly because of how shit Maxwell is at Async compute. That is but a "flaw" in the design of Maxwell.

So until 1060 comes out and has been reviewed, there is NO REAL competitor for it. The 960, 970. 980. 380, 380X, 390, 390X are all made obsolete by this ONE product.

 

But wait, i can feel something coming. Oh yes, it's your next argument.

"But who cares for DX12, it will take years before it catches on"

"Async compute is but a lame feature nobody will use"

 

And you would be wrong, on both accounts.

Why?

CONSOLES

Consoles, namely the "pisspoor" and "fagbox" has been out since 2013. They are both using APIs which is similar to DX12 in many ways, in the case of the "fagbox", it is more or less the same. Both consoles uses x86-64 architectures, and due to this, porting to PC has NEVER been easier. Async compute is an essential part in getting the most performance out of a console. Without using it, you WONT get the best out of a console. It is a requirement in order to achieve the best visuals and effects.

 

So essentially, developers has spent three YEARS "training" on how to use DX12 features. All they need to learn now is what effect corresponds to what liberary inside DX12. Meaning, it is essentially like reading a book in a different language and just translate it word for word.

Time consuming, yes, but sooner or later you learn several of the new words and wont have to look them up.

 

I think DX12, for what it is and what it offers, will have the fastest API "adoption" ever. Not just because the functions and features has been in use for years. But because AMD, Nvidia, Intel and Microsoft are heavily pushing for it. As is Chronos on the "Open" side.

DX12 is already being implemented into Unity and Unreal engines. The two biggest, most widespread engines in use, both among AAA titles and indy games.

 

 

390, 390X and GTX 970 and GTX 980 made obsolete by the RX 480?

 

[SteveGrabowski0]

9 minutes ago, Prysin said:

Well Steve, what are you comparing it to?

There is no 1060 out. So there is no current gen product that you can compare it to.

 

We already know that under DX12, it is closer to a 980 then a 970. Mostly because of how shit Maxwell is at Async compute. That is but a "flaw" in the design of Maxwell.

So until 1060 comes out and has been reviewed, there is NO REAL competitor for it. The 960, 970. 980. 380, 380X, 390, 390X are all made obsolete by this ONE product.

 

But wait, i can feel something coming. Oh yes, it's your next argument.

"But who cares for DX12, it will take years before it catches on"

"Async compute is but a lame feature nobody will use"

 

And you would be wrong, on both accounts.

Why?

CONSOLES

Consoles, namely the "pisspoor" and "fagbox" has been out since 2013. They are both using APIs which is similar to DX12 in many ways, in the case of the "fagbox", it is more or less the same. Both consoles uses x86-64 architectures, and due to this, porting to PC has NEVER been easier. Async compute is an essential part in getting the most performance out of a console. Without using it, you WONT get the best out of a console. It is a requirement in order to achieve the best visuals and effects.

 

So essentially, developers has spent three YEARS "training" on how to use DX12 features. All they need to learn now is what effect corresponds to what liberary inside DX12. Meaning, it is essentially like reading a book in a different language and just translate it word for word.

Time consuming, yes, but sooner or later you learn several of the new words and wont have to look them up.

 

I think DX12, for what it is and what it offers, will have the fastest API "adoption" ever. Not just because the functions and features has been in use for years. But because AMD, Nvidia, Intel and Microsoft are heavily pushing for it. As is Chronos on the "Open" side.

DX12 is already being implemented into Unity and Unreal engines. The two biggest, most widespread engines in use, both among AAA titles and indy games.

 

 

Have fun fighting that strawman son

[Prysin]

3 minutes ago, Dabombinable said:

390, 390X and GTX 970 and GTX 980 made obsolete by the RX 480?

 

yep.

it invalidates them all based on price to performance. A metric that nobody on these forums cares about except when they feel like it.

[Prysin]

2 minutes ago, SteveGrabowski0 said:

Have fun fighting that strawman son

you could just admitt you were deliberatly shitposting and spreading misinformation, and thus be over with it.

But that would involve admitting to being wrong. Wouldn't it steve? We can't be wrong if our ego is large enough to cover our bias, can we?

 

[SteveGrabowski0]

5 minutes ago, Humbug said:

Yes. It's efficient compared to previous AMD GCN cards.

But certainly not compared to Pascal. Anyway I don't think it's a big deal on desktop, because it's certainly efficient enough to not be a concern.

 

But it is a big concern for AMD in other markets. So I'm interested to see if the laptop parts like Pascal 11 are more efficient, that's where efficiency really matters with all the battery life and thermal constraints.

I think it's a big concern because it makes it sound like they'll have no answer to GP102 or GP100. If their answer to GP104 is going to be some 300W monstrosity then where do they go from there? I think it's troubling that they're now trying to push multi-gpu. I'm fine with 250W-300W for king of the hill class performance like from the R9 290x when it came out, that was an incredible gpu, or from Fury X when it launched. But not for something that's not going to be in the highest performance bracket.