AMD to Launch New Single-GPU Card This Summer, to Take on GTX 780 Ti

[btarunr]

AMD is reportedly working on a new single-GPU graphics card SKU to compete with the likes of GeForce GTX 780 Ti, and perhaps even take a swing at the GTX TITAN Black, since it's not too far ahead of the GTX 780 Ti at single-display gaming. The new SKU will be more than just a clock-speed bump, it will leverage HBM (high-bandwidth memory), a cutting-edge new technology that relies on stacking multiple DRAM dies with dedicated memory paths into a single package, cutting down on power-draw, thermals, and PCB real-estate.

The first kind of HBMs to hit the market are 8 Gbit 4Hi, which will interface with the GPU over a 128-bit wide path, which means there will be just four memory packages on the card (since the "Hawaii" silicon features a 512-bit wide memory bus), with improvements in the area of power-draw and heat output. The memory could be clocked higher, too. Sadly, memory bandwidth is not the prime-mover in VGA performance, and AMD will have to offer higher GPU clocks for the card to stand a chance against NVIDIA's high-end single-GPU offerings. As for the name this card would bear, we get a sense of deja vu about how NVIDIA launched the single-GPU GeForce 7950 GTX alongside the dual-GPU 7950 GX2, replacing the 7900 GTX. And so, it wouldn't surprise us if AMD named it Radeon R9 295X.



View at TechPowerUp Main Site

 

[Eroticus]

Looks more like GTX Titan Black.


GTX 780 Ti vs GTX 780 / R290 / 290X it's just higher memory bandwidth (400~500mhz)

 

[Assimilator]

Hopefully AMD will also take the opportunity to upgrade the stock cooler past "barely adequate" levels, although I'm not holding my breath.

 

[Sony Xperia S]

All major news websites publish this as a very reliable piece of information but still the doubts are strong.

The source also claims that AMD will ditch its partnership with TSMC (not saying that TSMC is that good partner, after all) for GLOBALFOUNDRIES which is a very significant change. I for one think it would be for the better and I highly appreciate and welcome the idea.

 

[brunello]

Hopefully AMD will also take the opportunity to upgrade the stock cooler past "barely adequate" levels, although I'm not holding my breath.

Given how good the closed loop system work I hope they are going to use something like the cooler on the 295x2

They'll get temps in the 50~60°C region with a single gpu which could translate in a quiet card hopefully, with a good oc potential too

 

[arbiter]

Hopefully AMD will also take the opportunity to upgrade the stock cooler past "barely adequate" levels, although I'm not holding my breath.

"barely adequate" that isn't just sugar coating the issue, that is a few tons of sugar dumped on it.

With that aside, would this be out in time? 800 series which seems like could be not far off based on maxwell? AMD could just be chasing previous gen in all this.

 

[Sony Xperia S]

With that aside, would this be out in time? 800 series which seems like could be not far off based on maxwell? AMD could just be chasing previous gen in all this.

It doesn't matter so much. The most important is the pricing. AMD can move with older videocards in the proper price segments and still can be successful.

---------------------------------------------


28nm TSMC Q4 2013 : High-end GPU, VI 1.0
28nm GlobalFoundries Q4 2014 : Mid-end GPU and Low-end GPU, VI 2.0
28nm GlobalFoundries Q2 2015 : High-end GPU, VI 2.0
20nm GlobalFoundries Q4 2015 : Mid-end GPU and Low-end GPU, PI 1.0
20nm GlobalFoundries Q2 2016 : High-end GPU, PI 1.0
14nm GlobalFoundries Q4 2016 : Mid-GPU and Low-GPU, PI 2.0
14nm GlobalFoundries Q2 2017 : High-GPU, PI 2.0


R(x) 3(xx) will be using HBM 1st Gen.
R(x) 4(xx) will be using HBM 1st Gen.
R(x) 5(xx) will be using HBM 2nd Gen.


Volcanic Islands 1.0 - 28nm TSMC
Volcanic Islands 2.0 - 28nm GlobalFoundries
Pirate Islands 1.0 - 20nm GlobalFoundries
Pirate Islands 2.0 - 14nm GlobalFoundries

http://www.overclock.net/t/1488641/next-r9-r7-graphics-to-use-hbm

 

[praze]

Personally, I'd like to see the XT and XTX monikers find their way back out of the chip names and into the model names.

This could easily just slide right into the 290 single-GPU line by naming it the R9 290XT.

 

[buildzoid]

I wonder what they will do to get more performance because if there is one thing Hawaii doesn't need then it's memory speed. They've got to either add more cores or clock it 100+mhz higher to catch up to the 780Ti.
Adding cores is unlikely because that would need a new die. And a 10% clockspeed increase would mean that AMD has been with holding some really well binned chips. Which is very likely because the R9 295X2 chips pull less than 200W from the VRM output side since the memory pulls around 50W from the VRM and there is a 10-15% efficiency loss on the VRMs.
Proof that the Hawaii cores on R9 295X2 pull less than 200W each
So ram power draw = 50/0.85 = 58W
GPU power draw = 200/0.85 = 236W
236W+236W+58W = 528W
W1zz's review says that peak gaming is 500W so my numbers are a little too big.
That means that AMD has a stock pile of sub 200W Hawaii chips so this could very easily be a 1100mhz or 1150mhz Hawaii chip.
The last option is that AMD has gone to GloFlo's 28nm SOI which would allow then to clock higher at the cost of a power draw increase and a cost of production increase.

 

[Assimilator]

Given how good the closed loop system work I hope they are going to use something like the cooler on the 295x2

Not a chance, an AIO closed-loop cooler is probably an extra $50 onto the cost. For a $ 1,500 video card that's insignificant, for a $400 one it makes a huge difference.

"barely adequate" that isn't just sugar coating the issue, that is a few tons of sugar dumped on it.

Hence the air quotes.

With that aside, would this be out in time? 800 series which seems like could be not far off based on maxwell? AMD could just be chasing previous gen in all this.

A good question, particularly since Hawaii isn't faster than Kepler, despite being a generation newer.

Volcanic Islands 1.0 - 28nm TSMC
Volcanic Islands 2.0 - 28nm GlobalFoundries
Pirate Islands 1.0 - 20nm GlobalFoundries
Pirate Islands 2.0 - 14nm GlobalFoundries

Considering how badly TSMC has screwed the pooch on 20nm, switching to GloFo is probably a wise move on AMD's part. That said, GloFo doesn't have working 20nm either - but at least AMD has the option of switching, nVIDIA is pretty much tied to TSMC regardless.

 

[LeonVolcove]

Hopefully AMD will also take the opportunity to upgrade the stock cooler past "barely adequate" levels, although I'm not holding my breath.

AGreed

 

[HumanSmoke]

All major news websites publish this as a very reliable piece of information but still the doubts are strong

As they should be.

 

[Sony Xperia S]

As they should be.

That opinion is very radical.

There is probability the information is not 100% accurate but to completely disregard it for the sake of it, is not good either.

Seems pretty solid confirmation that AMD is indeed going to launch something new soon.

 

[ensabrenoir]

SPLENDID!!! As my eyes read these words, all my brain comprehended was this could lead to more price cuts.
This could lead to more price cuts. This could lead to more price cuts. This could lead to more price cuts.

 

[TheoneandonlyMrK]

All those dates look like reasonable expected tape out dates not shipping dates so I would not expect anything so soon.
Could be though , well 2015 is my next expected gpu update slot so I am in with speculating until then.

 

[Steevo]

A die respin with less memory interface and another set of cores in its place, or some extra on die memory, or lots of tweaks with that much real estate to use.

At the worst we have some extra cores, at the best we have better thermals. Either of which could mean higher performance, in the middle it could be a smaller cooler die and cost less too.

 

[TheoneandonlyMrK]

Well I think it possible they had a core or two fused off on there top asic already and delivering a dual gpu card is perfect for building up a nice collection of binned parts

 

[LAN_deRf_HA]

If this is true it's nice they're trying for a bit more efficiency rather than their usual brute force approach to catching up to nvidia.

 

[HisDivineOrder]

I'm surprised AMD didn't get a SKU with 8 GB out anyway. It seems like the smart thing to do what with the next gen and 4K displays being their bread 'n butter as a "focus" atm.

Given that, having 8 GB seems not only smart, but prudent. CF setups for 4K are currently very constrained by the 4GB's of RAM they're using.

Toss on a Watercooled+Aircooled cooler not totally unlike the one they invented for the R9 295X2 and they'd finally have a reference cooler people would get behind. Ramp up the clockspeeds.

Call it the R9 290XL.

 

[Popocatepetl]

The last option is that AMD has gone to GloFlo's 28nm SOI which would allow then to clock higher at the cost of a power draw increase and a cost of production increase.

What is this "28 nm SOI" you speak of ? Global Foundries does 32 nm SOI and 28 nm bulk sillicon.

 

[TheoneandonlyMrK]

What they Do does not exclude them from evolving their processes or node lines to be fair

 

[Hilux SSRG]

That HBD die stacked memory looks like an interesting step forward for AMD. I'm all for reducing power & heat, while allowing for larger memory overclocking.

 

[TheoneandonlyMrK]

http://electroiq.com/blog/2012/06/globalfoundries-to-produce-sts-28nm-and20nm-fd-soi/

28-20nm Fd Soi capabilities of glo fo and st micro.
And these nodes are being opened up to external customers, could make for a very interesting Vi asic for example.
This is however definitely a bit of a leap of dot joining and rumour milling on my part but Fd soi is good enough and cheap enough for mobiles so you never know.

 

[HM_Actua1]

yawn, already developed by Nvidia (pascal)

 

[GhostRyder]

Will be interesting to see what comes of this, I have little doubt the actual performance of this will be anything to shake a fist at. Though im quite curious why they would not just go ahead and wait and start a fresh generation in all honesty but its coming down to time.

Im very interested in the stacked HBD die stacked memory, higher clocked memory is going to be great for the high resolutions that are starting to become more of a norm. Im already about to grab a 4k Monitor to replace my eyefinity setup.