PowerColor Announces its Radeon R9 Fury Graphics Card

[btarunr]

TUL Corporation, a leading and innovative manufacturer of AMD graphic cards since 1997, has proudly announced a new graphics card that has the capability of the highest gaming power in the most advanced GPU ever created for PC gaming. The PowerColor R9 Fury 4GB HBM is the world's first graphics card with AMD-pioneered High Bandwidth Memory (HBM) integrated on-chip delivering 60% more memory bandwidth over GDDR5 powering performance and bleeding-edge technology.

HBM is a new type of memory chip with low power consumption, ultra-wide communication lanes, and a revolutionary new stacked configuration. HBM's vertical stacking and fast information transfer open the door for truly exciting performance in innovative form factors, not to mention that GPU applications are just the start - look for HBM's superior power efficiency and space savings to spark industry-wide innovation. PowerColor R9 Fury 4GB HBM easily handles the most graphically intense games today due to its advanced and innovative GPU to ever create for the beloved PC gaming community. This power comes from AMD-pioneered High Bandwidth Memory (HBM) integrated on-chip that delivers more than 3 times the bandwidth per watt over GDDR5 along with a 4096-bit memory interface for incredible new advances in power and efficiency which makes the most innovative total solution GPU available today.



Enjoy the View. Redefine Reality.
AMD's most advanced and innovative GPU to date, the AMD Radeon R9 Fury GPU delivers the highest memory bandwidth ever for unparalleled 4K and VR gaming. Push your settings to the max with your ideal screen resolution and configuration, and get ready to redefine reality with amped support for AMD LiquidVR, DirectX© 12, Vulkan APIs, AMD CorssFire, and FreeSync technologies.

You can connect displays using 3 Display Port and 1 HDMI connector and select your operational configuration with the two-position BIOS toggle switch. PowerColor R9 Fury 4GB HBM adopts the most efficient air cooling systems that has 3 of 90mm fans, 6 of 8 mm heat pipes and a black NI plating clip. This card has it all, from great designs to even greater performances, owners of the PowerColor R9 Fury 4GB HBM will be impressed at what the Fury can do and will be gaming at the highest settings for years to come.

For more information, visit the product page.

View at TechPowerUp Main Site

 

[btarunr]

Likely a rebadged Sapphire Tri-X. TUL must have initially decided not to spend a cent on Fury non-X, then realised it effed up. So now it struck a deal with PCPartner to supply this card. Notice how it lacks any "PCS/PCS+" badging.

 

[GhostRyder]

Likely a rebadged Sapphire Tri-X. TUL must have initially decided not to spend a cent on Fury non-X, then realised it effed up. So now it struck a deal with PCPartner to supply this card. Notice how it lacks any "PCS/PCS+" badging.

Hmm well the Tri-X Nitro cooler looks a bit different than that but it does look very similar. I am kinda sad with PowerColor recently and their coolers, they made those awesome Devil 13 fans and coolers yet have not released one of these high end chips with that setup.

 

[Casecutter]

Are we sure these are basically a Sapphire rebadge?

TUL (PowerColor) and PcPartners (Sapphire) and it’s been said that Sapphire is the company that is AMD contracted to place the interposer on the reference PCB's for FuryX. Perhaps PcParters made some exclusive with AMD, that they don’t have to supply the boards as standalone, but they can as complete package so pulling in loins-share of profit. So TUL can’t get the short reference board from PcPartners, but can as complete card for the time being. This while perhaps TUL is working on their own PCB and cooler, but for whatever reason can’t be to market for several months, so PowerColor sees it as a way to get something moving.

I can’t figure a reason why TUL couldn’t have a PCB ready, was AMD holding engineering info back to get a design going. I’d think the PCB is fairly uncomplicated, surfacing mount the interposer package to a PCB isn’t tough, (though new tooling might be an issue) the power section and spec’s aren't any real obstacle, and the output connections to the DVI/D-P/HDMI straight-forward. It's almost like AMD held-off providing information, other than to Sapphire and Asus as they’re the only ones in the market. Was it AMD either didn’t trust other partners enough to even provide ambiguous pin-out and power section spec’s to get them working on their own PCB’s. Or, was AMD not in a position to make the "right deal" with PcPartners (Sapphire), and had to precluded other AIB's from opening day, as where’s XFX, Gigabyte, MSI, HIS.

 

[Casecutter]

Can't find a view of the Tri-X that is similar to the picture of the P-C version above, but compare the opening in the cooler fins just about where the PCI-E slot ends. The position and height/width of the P-C cooler appear different.
Hopefully this goes about debunking such conspiracy theories (especially mine above).

 

[the54thvoid]

select your operational configuration with the two-position BIOS toggle switch

Wish it said what it did! Core clocks etc?

 

[btarunr]

Wish it said what it did! Core clocks etc?

Both identical BIOSes.

 

[HumanSmoke]

I can’t figure a reason why TUL couldn’t have a PCB ready, was AMD holding engineering info back to get a design going.

If that was the case, then PowerColor aren't alone. How does Asus get the Strix out, but MSI and Gigabyte get left out in the cold with their cards? Seen any HIS IceQ Fury's? XFX DD's? ColorFire? VisionTek?

I think the answer is pretty obvious. Remember when Asus and EVGA had sole distribution in the U.S. for the Titan Z ? I am betting that Fiji die production is constrained. Limited number of initial reviews with more arriving when press review boards are sent on to other sites, limited Crossfire reviews, very limited user (OcUK has a owners thread for every high-end card, the Fury X's ownership list is pitiful in comparison to any others) reviews/postings. In spite of this, the card is out of stock in a lot of markets. Kind of screams low production.

What I can't figure out is how Asus got the jump on Gigabyte, MSI, and PowerColor.

EDIT: Seems the conclusion I reached has some basis in fact

 

[INSTG8R]

Both identical BIOSes.

Well at least on the Sapphire they are different. The 2nd one on the BIOS increases the Power and Temp limit.

Yeah we get that OC'ing Fury is pretty much a pipe dream O tend to believe the new Interposer is a rather delicate thing connecting it all together that too much heat or voltage could litterally separate it all. I await the first one who figures out to overvolt and overclock a Fury only to find it's come undone.
At least Sapphire made an attempt. ASUS went with a massive 12 phase VRM and you can't do anything with it. They didn't even attempt to OC it or offer a dual BIOS.

I personally have no interest in overclocking Fury nor any GFX card of mine for many years. You guys all read Benchmarks right? We end up with 2-5FPS gains you would never even see unless you monitoring it.
My 7970 is "OC'd" I just used the 2nd BIOS Sapphire put on the card with the higher clocks.

Oh and I am actually planning to buy a Fury just not sure whether to go Sapphire or ASUS yet.

 

[AsRock]

At least it has a nice clean cut shroud on it, most are just dumb these days.

 

[Relayer]

Likely a rebadged Sapphire Tri-X. TUL must have initially decided not to spend a cent on Fury non-X, then realised it effed up. So now it struck a deal with PCPartner to supply this card. Notice how it lacks any "PCS/PCS+" badging.

It's obviously not a tri X. It doesn't even resemble one except the reference board. I can't believe you would say that?

 

[Casecutter]

If that was the case, then PowerColor aren't alone. How does Asus get the Strix out, but MSI and Gigabyte get left out in the cold with their cards? Seen any HIS IceQ Fury's? XFX DD's? ColorFire? VisionTek?

Yea, I think I made that point.

and had to precluded other AIB's from opening day, as where’s XFX, Gigabyte, MSI, HIS...

I am betting that Fiji die production is constrained.

Is it a question of TSMC not delivering? The understanding HBM as of late is supposedly not constrained. There was word that delivery of interposer as it's said to push the reticle limit of the UMC 65nm process, but that was an discussion back like May. Then there's mounting all that stuff to the interposer by Amkor, which even though it is not new, not sure if it's been on such a grand scale.

So agreed AMD is having production constricted, but which of like 4 parts/processes is effecting it's now harder to put a finger on where that lies. But this is a great pipe-cleaner to sort out all the teething pains ahead of the 2016 Arctic Islands and their coup of HBM2 on various R9 variants.

 

[HumanSmoke]

Is it a question of TSMC not delivering?

Undoubtedly, but what is the cause? TSMC has no problem delivering full-die GM 200's that are the same size as Fiji. Now check the facts that 1. Fiji is based on Tonga - a full die that took an age to appear (and even when it did, it was a low production mobile GPU), and 2, transistor density. Fiji has a density of almost 15 million per square mm. The only other commercial IC anywhere close to that size of the same density, Intel's Broadwell, is made on a full node smaller (and isn't without its own issues).

The understanding HBM as of late is supposedly not constrained. There was word that delivery of interposer as it's said to push the reticle limit of the UMC 65nm process,

Reticle limits the size of the interposer, not the yield associated with it. 65nm is a very old process, and all that is required is that the there are sufficient layers to lay down the metal traces.

Then there's mounting all that stuff to the interposer by Amkor, which even though it is not new, not sure if it's been on such a grand scale.

Microbump density is probably as great or greater than other those used in other IC's. What is different is that virtually every other IC doesn't have the power budget required for Fiji.
You can blame everyone under the sun except AMD, but at the end of the day, it is AMD that developed and contracted for the design on. Except for the HBM implementation (which would be more an architecture-to-interface issue), all three components are on mature processes. GPU on 28nm, HBM DRAM on 30nm, interposer on 65nm.
It was AMD who decided on going with their largest die in history by some margin and ally it with a complex microsolder assembly in close proximity to that die. Somehow I doubt that these factors and the subsequent lockdown of voltage control on the Fury cards are unconnected.

 

[Casecutter]

You can blame everyone under the sun except AMD.

Was I blaming anyone? If we can say anything, AMD has positioned themselves with new technology and it is their burden, but they honestly weren't in a position to play is safe either. Without going to the largest die they wouldn't be as competitive to the GM200. It is valiant to press the boundaries as give the state of technology (still 28nm), their architecture, engineering, and resource's from the path/play they had set out on long ago partnering and building on HBM. Long term how this plays we've to wait and see.

Somehow I doubt that these factors and the subsequent lockdown of voltage control on the Fury cards are unconnected.

Definitely a learning curve for going out of the gate first, though I'd like AMD to "clarify" the state the voltage control and OC'n in regards to Fury. Although, I think this is as official as we might get... Around June 18th, "AMD tells us that a 1150Mhz clock speed, a 100Mhz overclock over the 1050Mhz stock, is quite easily achievable with AMD’s OverDrive... does not support over-volting". I haven't heard anything more.

 

[HumanSmoke]

Was I blaming anyone?

You mean like:

Is it a question of TSMC not delivering?

BTW, doesn't it seem strange that TSMC can produce a (marginally) larger GPU in GM 200 for Nvidia with adequate supply of fully enabled dice - TSMC is hardly supply constrained for 28nm HP

Definitely a learning curve for going out of the gate first, though I'd like AMD to "clarify" the state the voltage control and OC'n in regards to Fury. Although, I think this is as official as we might get... Around June 18th, "AMD tells us that a 1150Mhz clock speed, a 100Mhz overclock over the 1050Mhz stock, is quite easily achievable with AMD’s OverDrive... does not support over-volting". I haven't heard anything more.

I doubt that you, or anyone else will - at least not officially. No doubt there will continue to be PR double-speak and vague hyperbole from Huddy and Macri and Co. I wouldn't hold your breath waiting for the Fury to be crowned benchmark king.

 

[Casecutter]

I doubt that you, or anyone else will - at least not officially. No doubt there will continue to be PR double-speak and vague hyperbole from Huddy and Macri and Co. I wouldn't hold your breath waiting for the Fury to be crowned benchmark king.

Nice try spinning such thinking as any of mine, that more your flamboyant hyper-bole at work. It's never been my belief or an outcome I subscribe to. As for Huddy and Macri... I've said on various occasions Huddy should be dumped. Whil Macri should never be out able to speak off the cuff again.