AMD Radeon HD 9970 Hawaii Detailed, Volcanic Islands GPUs Set for Late 2013 (chip hell leak/rumour)
Sainatarajan
Wise Old Owl
Re: AMD Radeon HD 9970 Hawaii Detailed, Volcanic Islands GPUs Set for Late 2013 (chip hell leak/rumo
wth , 9xxx series!!! Will there wont be 8xxx series???
wth , 9xxx series!!! Will there wont be 8xxx series???
Re: AMD Radeon HD 9970 Hawaii Detailed, Volcanic Islands GPUs Set for Late 2013 (chip hell leak/rumo
7 series rebranded for OEMs
7 series rebranded for OEMs
Re: AMD Radeon HD 9970 Hawaii Detailed, Volcanic Islands GPUs Set for Late 2013 (chip hell leak/rumr
As Extreme Gamer has said, OEM only.
I'd guess the pricing will be at around the Titan. Nvidia really needs to up their game if this is true.
@Bold: You trying to start a flame war here, mate?
As Extreme Gamer has said, OEM only.
Hope it beats the crap out of nvidia and is priced right
I'd guess the pricing will be at around the Titan. Nvidia really needs to up their game if this is true.
@Bold: You trying to start a flame war here, mate?
Re: AMD Radeon HD 9970 Hawaii Diled, Volcanic Islands GPUs Set for Late 2013 (chip hell leak/rumour)
I am stating my opinion.If it turns out into a flame war so be it
Seriously want AMD to up their game and better Nvidia, which will be a boon for the consumers as the pricing will come down to a more sane level.
The volcanic island gpu's from AMD, If and when they come out will be a direct competitor to Maxwell, titan won't even be close to them in performance.
I am stating my opinion.If it turns out into a flame war so be it
Seriously want AMD to up their game and better Nvidia, which will be a boon for the consumers as the pricing will come down to a more sane level.
The volcanic island gpu's from AMD, If and when they come out will be a direct competitor to Maxwell, titan won't even be close to them in performance.
vickybat
I am the night...I am...
Re: AMD Radeon HD 9970 Hawaii Detailed, Volcanic Islands GPUs Set for Late 2013 (chip hell leak/rumo
This looks more like a competitor to titan or titan ultra than maxwell. From the specs itself, it seems more powerful than titan.
Architecturally, i doubt this to be GCN 2.0 ( i can be wrong though).
Maxwell is a different beast though and will be completely different from fermi and kepler architecture.
Such a wide bus also comes with its fair share of troubles unless amd made it more efficient. Even this gen, GCN was no match for kepler's memory management unit.
That explains logically how a gtx 680 with a 256 bit bus and 2gb vram could keep up with a 7970 with 384 bit bus and 3gb vram even at resolutions higher than 1440p.
This looks more like a competitor to titan or titan ultra than maxwell. From the specs itself, it seems more powerful than titan.
Architecturally, i doubt this to be GCN 2.0 ( i can be wrong though).
Maxwell is a different beast though and will be completely different from fermi and kepler architecture.
Such a wide bus also comes with its fair share of troubles unless amd made it more efficient. Even this gen, GCN was no match for kepler's memory management unit.
That explains logically how a gtx 680 with a 256 bit bus and 2gb vram could keep up with a 7970 with 384 bit bus and 3gb vram even at resolutions higher than 1440p.
Re: AMD Radeon HD 9970 Hawaii Diled, Volcanic Islands GPUs Set for Late 2013 (chip hell leak/rumour)
TSMC to install 20nm fab equipment ahead of schedule, says report
This does lend some more credence to the leak
I am stating my opinion.If it turns out into a flame war so be it
Seriously want AMD to up their game and better Nvidia, which will be a boon for the consumers as the pricing will come down to a more sane level.
The volcanic island gpu's from AMD, If and when they come out will be a direct competitor to Maxwell, titan won't even be close to them in performance.
TSMC to install 20nm fab equipment ahead of schedule, says report
This does lend some more credence to the leak
sukesh1090
Adam young
Re: AMD Radeon HD 9970 Hawaii Detailed, Volcanic Islands GPUs Set for Late 2013 (chip hell leak/rumo
^^
that could be for "APPLE".
^^
that could be for "APPLE".
Re: AMD Radeon HD 9970 Hawaii Detailed, Volcanic Islands GPUs Set for Late 2013 (chip hell leak/rumo
Nice(Though most of that was already mentioned in the posted link/apparent from the diagram). Here's to hoping that it is true.
here's some spec details though not confirmed
AMD Radeon HD 9970 | techPowerUp GPU Database
Nice(Though most of that was already mentioned in the posted link/apparent from the diagram). Here's to hoping that it is true.
Re: AMD Radeon HD 9970 Hawaii Detailed, Volcanic Islands GPUs Set for Late 2013 (chip hell leak/rumo
The codename Volcanic Islands is almost confirm as it has been talked off in an interview by AMD officials some days ago,still there is no official word on specifications and the manufacturing process it will use. Rage3D.com : The Future of AMD: Jim Keller and Chekib Akrout Interviewed [ Looking forward ]
Here is more on the image by chiphell :Unknown AMD SoC Features 8 Bulldozer Modules, 1024 Stream Processors, 512-Bit Memory Bus
Doesn't look like a standalone GPU. AMD-P?It is a power saving technology used in some AMD opteron processors.
There are some serial processing units in the image..GPU to have serial processing units as in CPU?maybe
Maybe a fake picture or something else(APU based Opteron?).Only time will tell.
Anyway unless something gets more concrete about it we can't make any conclusion from it,till then it(chiphell image) is just a speculation.
As far as 1440p or higher resolution performance is concerned single hd7970GHz has the upper hand over single gtx680(both 2GB and 4GB versions).
Frame Rating: High End GPUs Benchmarked at 4K Resolutions | PC Perspective
Frame Rating Dissected: Full Details on Capture-based Graphics Performance Testing | PC Perspective
The codename Volcanic Islands is almost confirm as it has been talked off in an interview by AMD officials some days ago,still there is no official word on specifications and the manufacturing process it will use. Rage3D.com : The Future of AMD: Jim Keller and Chekib Akrout Interviewed [ Looking forward ]
Here is more on the image by chiphell :Unknown AMD SoC Features 8 Bulldozer Modules, 1024 Stream Processors, 512-Bit Memory Bus
Doesn't look like a standalone GPU. AMD-P?It is a power saving technology used in some AMD opteron processors.
There are some serial processing units in the image..GPU to have serial processing units as in CPU?maybe
Maybe a fake picture or something else(APU based Opteron?).Only time will tell.
Anyway unless something gets more concrete about it we can't make any conclusion from it,till then it(chiphell image) is just a speculation.
As far as 1440p or higher resolution performance is concerned single hd7970GHz has the upper hand over single gtx680(both 2GB and 4GB versions).
Frame Rating: High End GPUs Benchmarked at 4K Resolutions | PC Perspective
Frame Rating Dissected: Full Details on Capture-based Graphics Performance Testing | PC Perspective
vickybat
I am the night...I am...
Re: AMD Radeon HD 9970 Hawaii Detailed, Volcanic Islands GPUs Set for Late 2013 (chip hell leak/rumo
The point was not about out doing but keeping up. 680 competes with a paltry 256 bit bus. 7970 does not seem to have an advantage using a wide 384 bit bus.
Wider bus is like a wide highway for better streamlined data feed. Apart from that , you also need an efficient MMU to manage data efficiently through the bus.
The 680's are just a couple of fps lower than the GHz edition. They would still match a standard 7970 with their 256 bit bus. Read up about kepler's memory management unit. They were explained in several early reviews. Its superior to what GCN offers.
Besides, 4k is an overkill resolution for single gpu gaming. Apart from the titan, non of the cards are fast enough. Even titan falters in some.
The point was not about out doing but keeping up. 680 competes with a paltry 256 bit bus. 7970 does not seem to have an advantage using a wide 384 bit bus.
Wider bus is like a wide highway for better streamlined data feed. Apart from that , you also need an efficient MMU to manage data efficiently through the bus.
The 680's are just a couple of fps lower than the GHz edition. They would still match a standard 7970 with their 256 bit bus. Read up about kepler's memory management unit. They were explained in several early reviews. Its superior to what GCN offers.
Besides, 4k is an overkill resolution for single gpu gaming. Apart from the titan, non of the cards are fast enough. Even titan falters in some.
Re: AMD Radeon HD 9970 Hawaii Detailed, Volcanic Islands GPUs Set for Late 2013 (chip hell leak/rumo
The point is,it's not that having a wider memory bus with higher bandwidth would benefit you in all situations in gaming but it does benefit in some.Sure different architectures would have different ways of dealing with memory management and though we can talk about if one has a considerably higher memory B/W but that doesn't make it outrightly better in every situation though things change with higher memory uitlisation,resolution,AA.
Wider memory bus and B/W of GPUs of same architecture can be compared better than comparing memory management of two different archtectures as they don't work identically.With kepler nvidia have more cache and higher bandwidth,full 512KB of L2 cache shared through the GPU,the cache hit-rate is optimized than their Fermi chips,SMX units together are more optimized than previous SM units for parallelism.
Even at 1080p hd7970GHz and gtx680 trade blows but at higher resolutions the former is overall better which is quite clear.
At higher resolutions mostly the gap is always in the favor of hd7970GHz.
Far cry 3 is an example,here at 1080p hd7970GHz edition was losing about 20% and tables are turned as soon as we go to 1440p or higher.
So yes higher memory specs definitely help hd7970GHz edition at higher resolutions but how efficiently,is a different question as the question how would gk104 with identical specs to gtx680 with 384bit wide bus,3GB VRAM would fair(definitely better than gtx680 but how much better?) is also a different question.At the end memory bandwidth is just one part of the equation there are other things as well.
And it is not like as if it is just gtx680 is just "couple of fps" lower than hd7970GHz.In many cases it is just much more substantial than that at 1440p or higher,sometimes like 15-20% and sometimes even more than that(dirt3,skyrim,sleeping dogs).Gtx680 can't keep up with hd7970GHz edition here at higher resolutions.
For sure at 1440p,4k and higher resolutions most would prefer a mutli-GPU set up but that doesn't change the thing that hd7970GHz is better at 1440p or higher,a step ahead of gtx680.
The point is,it's not that having a wider memory bus with higher bandwidth would benefit you in all situations in gaming but it does benefit in some.Sure different architectures would have different ways of dealing with memory management and though we can talk about if one has a considerably higher memory B/W but that doesn't make it outrightly better in every situation though things change with higher memory uitlisation,resolution,AA.
Wider memory bus and B/W of GPUs of same architecture can be compared better than comparing memory management of two different archtectures as they don't work identically.With kepler nvidia have more cache and higher bandwidth,full 512KB of L2 cache shared through the GPU,the cache hit-rate is optimized than their Fermi chips,SMX units together are more optimized than previous SM units for parallelism.
Even at 1080p hd7970GHz and gtx680 trade blows but at higher resolutions the former is overall better which is quite clear.
At higher resolutions mostly the gap is always in the favor of hd7970GHz.
Far cry 3 is an example,here at 1080p hd7970GHz edition was losing about 20% and tables are turned as soon as we go to 1440p or higher.
So yes higher memory specs definitely help hd7970GHz edition at higher resolutions but how efficiently,is a different question as the question how would gk104 with identical specs to gtx680 with 384bit wide bus,3GB VRAM would fair(definitely better than gtx680 but how much better?) is also a different question.At the end memory bandwidth is just one part of the equation there are other things as well.
And it is not like as if it is just gtx680 is just "couple of fps" lower than hd7970GHz.In many cases it is just much more substantial than that at 1440p or higher,sometimes like 15-20% and sometimes even more than that(dirt3,skyrim,sleeping dogs).Gtx680 can't keep up with hd7970GHz edition here at higher resolutions.
For sure at 1440p,4k and higher resolutions most would prefer a mutli-GPU set up but that doesn't change the thing that hd7970GHz is better at 1440p or higher,a step ahead of gtx680.
Cilus
laborare est orare
Re: AMD Radeon HD 9970 Hawaii Detailed, Volcanic Islands GPUs Set for Late 2013 (chip hell leak/rumo
In GCN architecture, AMD has designed their memory modules not for just higher bandwidth but for superior computing power too. They are designed in such a manner that programmers will find it easy to program with GPU computing based languages. In fact wider memory bus does have higher performance impact when coupled with wider memory bus. In GCN, AMD actually brought best of the both world, not just gaming performance or Compute performance.If you look at it closely, you will find it is more like a CPU buss of a many core CPU architecture where multiple cluster of memory bus can access Video Ram simultaneously and assign threads more efficiently to the different compute engines fo better efficiency and resource utilization. Also the design is suitable for X86 type memory management techniques which helps the programmers greatly to optimize the code blocks.
In Kepler design, nVidia almost completely omits the GPU computing part and focused on pre gaming performance. Their memory bus is more like fetching as much data as possible in every clock cycle to feed the CUDA cores. Now in gaming, the design is the best case as here GPU needs to perform very limited number of operations over millions of pixels and coordinates. But in case of general programming, not all the time, you have a huge chunk of data available for a single operation. Here the data needs be divided into smaller groups based on the type of instructions which are needed to be operated over them. in that case, more number of memory bus clusters (384 bit bus is divided into 6 chunks of memory controller of 64 bit width, each handling 512 MB of memory Block), can independently fetch data from their respective memory locations simultaneously and assign them to different memory compute engines. It also makes Thread scheduling more easier.
In GCN architecture, AMD has designed their memory modules not for just higher bandwidth but for superior computing power too. They are designed in such a manner that programmers will find it easy to program with GPU computing based languages. In fact wider memory bus does have higher performance impact when coupled with wider memory bus. In GCN, AMD actually brought best of the both world, not just gaming performance or Compute performance.If you look at it closely, you will find it is more like a CPU buss of a many core CPU architecture where multiple cluster of memory bus can access Video Ram simultaneously and assign threads more efficiently to the different compute engines fo better efficiency and resource utilization. Also the design is suitable for X86 type memory management techniques which helps the programmers greatly to optimize the code blocks.
In Kepler design, nVidia almost completely omits the GPU computing part and focused on pre gaming performance. Their memory bus is more like fetching as much data as possible in every clock cycle to feed the CUDA cores. Now in gaming, the design is the best case as here GPU needs to perform very limited number of operations over millions of pixels and coordinates. But in case of general programming, not all the time, you have a huge chunk of data available for a single operation. Here the data needs be divided into smaller groups based on the type of instructions which are needed to be operated over them. in that case, more number of memory bus clusters (384 bit bus is divided into 6 chunks of memory controller of 64 bit width, each handling 512 MB of memory Block), can independently fetch data from their respective memory locations simultaneously and assign them to different memory compute engines. It also makes Thread scheduling more easier.
Re: AMD Radeon HD 9970 Hawaii Detailed, Volcanic Islands GPUs Set for Late 2013 (chip hell leak/rumo
*videocardz.com/43446/amd-radeon-hd-9000-series-launching-october
Possibly coming this october?
AMD synkar nya Radeon med Battlefield*4 - Grafikkort - SweClockers.com
Coincidence?
*videocardz.com/43446/amd-radeon-hd-9000-series-launching-october
Possibly coming this october?
AMD synkar nya Radeon med Battlefield*4 - Grafikkort - SweClockers.com
Coincidence?