News
 

Bookmark and Share

(15) 

Advanced Micro Devices confirmed on Tuesday at its AMD Fusion Development Summit (AFDS) that it would release a new breed of graphics processing units (GPUs) later this year. Apparently, AMD's code-named Southern Island graphics processing units are on-track to be launched in 2011.

"In a couple of days you are going to hear about our exciting new graphics architecture that will be coming out later this year and will be utilized by our future APUs," said Rick Bergman, senior vice president and general manager of AMD products group.

During his keynote at the AFDS, Eric Demers, AMD corporate vice president and chief technology officer at graphics division of the company, will recap the evolution of GPUs in the recent years, including the latest VLIW5 and VLIW4 core architectures and instruction sets. But more importantly Mr. Demers will present an overview of the next generation of AMD cores under development, which will propel forward new capabilities and continue the GPU evolution.

While it is not completely clear what kind of peculiarities of next-generation graphics processors will be revealed at the summit, it is highly likely that going forward both AMD and its arch-rival Nvidia will attempt to make GPUs more programmable and better suitable not only for graphics processing, but also for complex computing.

Later during this year AMD is projected to release its first graphics processing units made using 28nm process technology code-named Southern Islands. The SI family is projected to support DirectX 11 functionality along with architectural improvements and some new features. Theoretically, the company may start revealing architectural peculiarities about the chips behind closed doors in order to let software designers to prepare their software for new chips.

Tags: ATI, AMD, Southern Islands, 28nm, Radeon

Discussion

Comments currently: 15
Discussion started: 06/15/11 10:52:28 PM
Latest comment: 07/05/11 01:00:31 AM
Expand all threads | Collapse all threads

[1-3]

1. 
Wow, its about time they moved to a 28nm core/design
40nm came out on.. what, Apr 2009! (with the HD4770) so two and a bit years later they move onto 28nm. Here's hoping the move brings with it some serious performance!
4 1 [Posted by: vid_ghost  | Date: 06/15/11 10:52:28 PM]
Reply
- collapse thread

 
normaly fermi and so on would be on 32 nm but it was scrapped by tsmc so they had to wait a lot longer.
1 0 [Posted by: massau  | Date: 06/16/11 02:30:40 AM]
Reply
 
Nope, Fermi was made for 40nm

Kepler and Southern Islands will be duking it out @ 28nm HKMG

Kepler is just a die shrink of Fermi, I believe
Southern Islands might be VLIW4 or Wide-band SIMD(ACE)
1 1 [Posted by: seronx  | Date: 06/16/11 12:36:19 PM]
Reply
 
Kepler is not a die-shrink of Fermi. It likely will be a very heavily revised Fermi architecutre (i.e., Fermi II if you will).

Kepler's will be based on a 28nm fabrication process, and is expected to deliver an estimated 3 to 4 times the performance per watt compared to Fermi.

http://www.guru3d.com/new...-fermi--due-2011-already/

http://semiaccurate.com/2...l-barely-beat-moores-law/

http://www.vizworld.com/2...dias-kepler-maxwell-gpus/
1 1 [Posted by: BestJinjo  | Date: 06/16/11 01:41:31 PM]
Reply
 
I know it's mainly going to be at a high clock and some tweaks here and there to make it faster

But mainly its going to just be Fermi @ a better fab process

Where as the 7000 could be VLIW4 or ACE(More likely)

Fermi GF110 512@772MHz= 790.55 DP GFlops
Cayman XT 384@880MHz = 675 DP GFlops

Keplar GF1x0 Unknown@Unknown = 3162.2 DP GFlops@Best
Southern Islands (NextGen GPU Architecture) Unknown@Unknown = 2700 DP GFlops

If it was actually based on VLIW4

Next Gen Architecture it should also have the 4x this architecture is 15-20% better than fermi(gonna pull it out of a hat 850 DP GFlops if placed on 40nm then to 28nm 3400 DP GFlops)
0 0 [Posted by: seronx  | Date: 06/16/11 03:09:09 PM]
Reply
 
Your DP calculations for Fermi are way off btw.

GTX580
Single Precision = 1581 Gflops
http://techreport.com/articles.x/19934

Double Precision is 1/8th of SP for consumer Fermi = 197 Gflops
http://techreport.com/articles.x/18332/5

But either way, you shouldn't derive Gaming performance by comparing Gflops across 2 different architectures. DP performance only matters for GPGPU tasks such as MilkyWay@Home, BitCoin mining, Collatz Conjecture, etc.

Otherwise, GTX580 is faster than the HD6970 despite a massive deficit in DP performance.

Most likely HD7000 will use the VLIW4 design, and only next generation HD8000 will adopt the more advanced architecture.
1 1 [Posted by: BestJinjo  | Date: 06/16/11 09:08:39 PM]
Reply
 
HD7000 has a 99.8% chance being the GCN Architecture

http://www.anandtech.com/...md-architects-for-compute

I'm not talking about consumer or enterprise, I am talking about MAXs

Fermi (800 DP)
SP = (1/2)DP
VLIW4 (600 DP)
SP = (1/4)DP


Kepler (2000 GFlops DP)
SP = (1/2)DP
GCN (2000 GFlops DP)
SP = (1/2)DP

GCN once it hits PCI-e 3.0 CPU L2 cache will be paralleled to GPUs L2

Bulldozer FX has 1MB per core(512KB Max will be used for GPU Synchronization)

F@H by Windows 8 might use C++ AMP which will allow the CPU+GPU Parallelism

Both Kepler and GCN will be able to do parallelism with the FX CPU

Kepler/Fermi being the better as they are using the mature architecture

With Both halves releasing this year by 2013 AMD will have the Haswell competitor it needs(2011 enthusiast Haswell competition, 2013 mainstream/low-end competiton)

Enhanced Bulldozer + Graphic Core Next by 2013(Mainstream)
Next Gen Bulldozer + Next Gen GCN by 2013(Enthusiast)

2011 Enthusiast(Scorpius Platform: FX + VLIW4) = 2012 Mainstream
2012 Enthusiast(Corona Platform: FX v2 + GCN) = 2013 Mainstream
1 1 [Posted by: seronx  | Date: 06/17/11 11:13:43 PM]
Reply
 
Fermi's DP is not equal to 1/2 of SP. It's 1/8th.

Did you even read the article you posted when you made claims that HD7000 is 99.8% GCN?

"Certainly it looks like it will be 2013 before GCN has a chance to become a top-to-bottom architecture"
0 0 [Posted by: BestJinjo  | Date: 06/19/11 02:59:31 PM]
Reply

2. 
Once again who cares anymore, Your current gen card is all that is needed as game developers just tone pc version down to a shitty console port. 1 version for all platforms so they can save money.
3 0 [Posted by: ozegamer  | Date: 06/17/11 04:43:31 AM]
Reply
- collapse thread

 
Maybe not right now, but games will eventually get more graphically intensive (BF3). Then we would need more powerful hardware.

Personally, I never buy the latest and greatest and wait until the end of a generation to upgrade. So in 12 months from now a $250 GTX580 or faster performance sounds good to me. For that to happen we need next generation mid-range cards (and therefore a next generation of GPUs).
0 1 [Posted by: BestJinjo  | Date: 06/17/11 09:51:45 AM]
Reply

3. 
ATI/Radeon is an 'evil' neoliberal company, that basically tells support seeking customers to piss off, games look worse in favor of performance, unlike NVIDIA, they fire new workers on the fourth day because their shoes are not classy, their executives are disgusting egoistic personalities and they cobble up smaller bourgeoning tech firms by hostile company take-overs. But they have the best graphics cards, and they should give up CPUs to Intel and concentrate just on the graphics cards. Face it!
1 2 [Posted by: TeemuMilto  | Date: 06/19/11 10:26:39 AM]
Reply
- collapse thread

 
can anyone else make sense of this post ?

"games look worse in games...." mirror in a mirror...ah i'm trapped
0 0 [Posted by: ultimaone  | Date: 06/20/11 12:26:20 AM]
Reply
 
Sure I followed the post completely. You were told ATI supports stinks and is rude ( it most does and it most certainly is) Radeon cheats on IQ in favor of a frame rates (10% cheat across the board). Their executives are arrogant, they fire workers for petty crap, and they were sucked up by the crappy cpu company amd.
1 0 [Posted by: SiliconDoc  | Date: 07/05/11 01:00:31 AM]
Reply
 
show the post
0 3 [Posted by: 63jax  | Date: 06/21/11 12:57:51 AM]
Reply
 
Never!
1 1 [Posted by: TeemuMilto  | Date: 06/21/11 11:08:37 PM]
Reply

[1-3]

Add your Comment




Related news

Latest News

Monday, August 4, 2014

4:04 pm | HGST Shows-Off World’s Fastest SSD Based on PCM Memory. Phase-Change Memory Power’s World’s Fastest Solid-State Drive

Monday, July 28, 2014

6:02 pm | Microsoft’s Mobile Strategy Seem to Fail: Sales of Lumia and Surface Remain Low. Microsoft Still Cannot Make Windows a Popular Mobile Platform

12:11 pm | Intel Core i7-5960X “Haswell-E” De-Lidded: Twelve Cores and Alloy-Based Thermal Interface. Intel Core i7-5960X Uses “Haswell-EP” Die, Promises Good Overclocking Potential

Tuesday, July 22, 2014

10:40 pm | ARM Preps Second-Generation “Artemis” and “Maya” 64-Bit ARMv8-A Offerings. ARM Readies 64-Bit Cores for Non-Traditional Applications

7:38 pm | AMD Vows to Introduce 20nm Products Next Year. AMD’s 20nm APUs, GPUs and Embedded Chips to Arrive in 2015

4:08 am | Microsoft to Unify All Windows Operating Systems for Client PCs. One Windows OS will Power PCs, Tablets and Smartphones