Sapphire Aftermarket R9 Fury Revealed

[utahraptor]

http://www.overclock3d.net/articles/gpu_displays/sapphire_aftermarket_r9_fury_revealed/1

 

[Shehriazad]

Yet another Twohandsword....


Sheesh...I really hope that with 14/16nm and some energy efficiency enhancements we finally get to have them in small again...really not a fan of such huge cards.

 

[railven]

Yet another Twohandsword....


Sheesh...I really hope that with 14/16nm and some energy efficiency enhancements we finally get to have them in small again...really not a fan of such huge cards.

You kidding me? As the nodes got smaller the heatsinks got bigger. WIth the hardships Intel was seeing, I expect them to all be Watercooled or 2.5 handswords. Haha.

 

[Stuka87]

Yet another Twohandsword....


Sheesh...I really hope that with 14/16nm and some energy efficiency enhancements we finally get to have them in small again...really not a fan of such huge cards.

Not going to happen. Any savings on the energy side will be spent on adding more transistors to gain performance.

It also gets harder and harder to cool these smaller and smaller dies.

If you want a smaller card, get a lower end card. Top tier cards will always be pushing the limits of cooling.

 

[railven]

I was thinking...

What if the volts lock is across all Fiji chips?

 

[utahraptor]

I was thinking...

What if the volts lock is across all Fiji chips?

Considering the factory overclock version is a measly 40 MHz over the base unit and still 10 MHz under the Fury X I would say that may be so.

 

[xthetenth]

I thought there were rumors Sapphire's next version of their OC utility would have volt changing functionality?

 

[Lalilulelo]

Woah, a back plate

 

[Shehriazad]

Woah, a back plate

That's probably to keep the card from ripping apart...after all the card needs to be strong enough to be used in a duel.

Jokes aside...backblates are always nice...I won't ever buy a card without backplate again if I can help it. Really gives it a nice finishing touch even if the gains on the cooling side are a little questionable.

 

[Elixer]

Since when do these cards need a crossfire cable?
I thought they were all XDMA now?

*edit n/m, reading fail.

 

[railven]

Considering the factory overclock version is a measly 40 MHz over the base unit and still 10 MHz under the Fury X I would say that may be so.

Sort of what made me think that. First glance I didn't realize it was two separate products, then I noticed the clock diff.

Woah, a back plate

I think that backplate looks nice! Outside of reusing heatsinks they probably already have, I think this card would have gotten by fine with a dual fan so they could have hacked off 1/3 of the setup.

 

[Headfoot]

Since when do these cards need a crossfire cable?
I thought they were all XDMA now?

It doesnt? Where are you seeing that?

 

[cbrunny]

I dont understand. Lower clock speed, cut down card, same TFLOPS? Something isn't right. I'm guessing typo.

 

[Kallogan]

stream processors : up to 3584

made me lol. up to ? wat ?

 

[LTC8K6]

"up to"...

Don't they know yet?

 

[LTC8K6]

Foot long video cards need to be boycotted...

It's ridiculous.

 

[beginner99]

Well due to my case it seems Fury is of the table. Either nano or X or skip. Even most 290/390 are too long...

 

[LTC8K6]

Well due to my case it seems Fury is of the table. Either nano or X or skip. Even most 290/390 are too long...

According to the AMD presentation, air cooled Fury cards are supposed to be 10 inches long, iirc.

This one just happens to be longer, I think?

 

[Rvenger]

Sapphire is not going to be the only AIB custom solution. They will have a smaller 2 fan option. This one is the Tri-X cooler model. It would be silly to call it a Tri-X with 2 fans. Sapphire will probably have a Nitro version as well sporting 2 fans.

 

[SK10H]

So Sapphire and Asus Strix are rumuored to drop the DVI since they are using AMD original PCB design. Probably will affect other vendors with these reference pcb design. There's space to put the DVI at the top slot and leave the rest as exhaust hole like the 980ti, unlike the Fury X where the waterblock may be blocking the connector.

Given the $100 difference ~ 14% ~ 512 /4096, Fury isn't much of a deal relative to Fury X even excluding the AIO watercooling.

If Gameworks, 3.5GB, record margin didn't exist, I would have vote with my wallet already.

 

[Elixer]

It doesnt? Where are you seeing that?

Under Specifications, last item.

*edit n/m, reading fail.

 

[LTC8K6]

R7 370 still requires a bridge connector. It's still a Pitcairn chip.

 

[Hitman928]

Under Specifications, last item.

That says crossfire cable not required. . .

 

[Eymar]

Haha, sucks to be a technical writer. No matter what you write, more than a few people is going to read the opposite.

 

[PPB]

The back of this card got me thinking: it is probably time to better use those bigass heatsinks and probably add another fan in the exposed area the pcb doesnt reqch anymore because of the benefit of HBM? 4 slim 80mm coolers at lower rpm prpbably behave better that 3 of them. It also lets you push/pull 1/3rd of the whole heatsink