[UPDATE] AMD hosting "capsaicin" live webcast at GDC on March 14th-could announce Fury X2 and demo Polaris

[Ashaira]

On 3/7/2016 at 8:33 PM, Thunderjolt said:

Oooooooh boy, If you thought the R9 Fury was a hot card just wait till you see our new "capsaicin" branded ones. You'll be able to cook dinner on it

afraid the gtx 480 beat amd to it.

[bobhays]

On 3/6/2016 at 3:09 PM, LAwLz said:

I do... Almost.

Just look at the previous dual GPU cards vs new gen single GPU cards

5970 -> 6970: About a 20% performance difference (in favor of the 5970). 600 vs 370 USD.

 

6990 -> 290X: About a 20% performance difference (in favor of the 6990). 700 vs 550 USD.

 

 

On top of that you also have to factor in all the drawbacks of dual GPU cards. It's not uncommon that crossfire doesn't work in games. Microstuttering is still a thing. Cooling is far worse because of the limited space. You know, the usual dual GPU drawbacks.

 I just think, and hope, that Polaris will be such a big leap forward that anyone who buys this Fury x2 once it launches will regret it.

Should compare 6990 to 7970 and 7990 to 290x

[Valentyn]

1 hour ago, Ashaira said:

afraid the gtx 480 beat amd to it.

Never forget!

 

I ripped all these from my ancient Photobucket account

 

[LAwLz]

2 hours ago, bobhays said:

Should compare 6990 to 7970 and 7990 to 290x

Sorry. I had a brainfart and forgot that they existed.

 

6990 -> 7970 = Same performance. The 6990 pulled slightly ahead at high resolutions (like 10%).  700 dollars vs 550 dollars (quickly dropped to 500 for the GHz edition).

7990 -> 290X = About 20% slower (in favor of the 7990). 1000 dollars vs 550 dollars.

 

So yeah. I expect a single GPU Polaris card to be significantly cheaper, and perform almost as well as the Fury X2. That's where my expectations are considering previous generations.

[carolkarine]

6 minutes ago, LAwLz said:

Sorry. I had a brainfart and forgot that they existed.

 

6990 -> 7970 = Same performance. The 6990 pulled slightly ahead at high resolutions (like 10%).  700 dollars vs 550 dollars (quickly dropped to 500 for the GHz edition).

7990 -> 290X = About 20% slower (in favor of the 7990). 1000 dollars vs 550 dollars.

 

So yeah. I expect a single GPU Polaris card to be significantly cheaper, and perform almost as well as the Fury X2. That's where my expectations are considering previous generations.

I mean, that's probably gonna be the next gen flagship. It could be as much as nine months later than the first launch. We should also consider than they'll save themselves space to increase performance on the same node in future generations by just increasing die size. I don't expect anything bigger than 350mm^2 this first gen on 14nm

[Prysin]

On 07/03/2016 at 0:09 AM, LAwLz said:

Microstuttering is still a thing. Cooling is far worse because of the limited space. You know, the usual dual GPU drawbacks.

 

Depends on the solution.

 

295x2 doesnt produce noticable micro stutters. I suspect that is because they used a PLX chip between the GPUs, giving it its own PCIe 8x Gen2? bus.

(atleast not noticeable on my 60Hz panel, i can see there is frametime spikes in RivaTuner, but i cannot visually perceive it even when i know when, where and what to look for).

 

TitanZ, GTX 590, 690, HD 5990, 6990, 7990, R9 390x2 (Powercolor made a dual 390 card) does NOT use PLX chips, they use PCB based crossfire/SLI traces. Yes even the 390 GPU can use CF bridges, but the normal 390 PCBs doesnt have the connectors or traces. But the GPU itself still has the pins/connections.

Naturally, a CF/SLI bridge wont come remotely near the bandwidth of PCIe Gen2 x8...

 

On 08/03/2016 at 1:20 AM, LAwLz said:

We don't know how well that will work yet. I would not be holding my breath that it will work flawlessly.

It already works, its been part of the LiquidVR SDK since it was released in Beta. It has been used in practice since the first Occulus demo which used dual 290Xs if my memory serves me right. I cannot vouch for how the latency and consistency of frame delivery is, obviously, but the setup has been working well enough to use it for tech demos with the press being present.