AMD Radeon HD 8000 Series: complete discussion

Toggle sidebar Toggle sidebar
bloc97

bloc97

Distinguished
Sep 12, 2010
1,030
0
19,460
----
AMD Sea Island “Tenerife GPU” Leaked – Features Enhanced GCN Architecture, Over 4.50 TFlops Raw Processing Power
---
tenerife-635x357.png

As you can see from the leaked slide, Tenerife would be an Enthusiast level product replacing the current flagship HD7970 GPU so we can assume that it would be called HD8970. It is said to feature 2304 Stream Processors, 128TMU’s and 48 ROP’s which would result in over 4.50 TFLOPS of processing power and over 20% better compute performance compared to HD7970. Display outputs include an HDMI and 2 x MiniDisplay ports, DVI is not listed however the card is suggested to support a maximum of 6 displays.
---
AMD Announces Sea Islands GPUs Coming in 2013
---
amd_2012-2013_client_and_graphics_roadmap.jpg

The slide showed that the Sea Islands GPU architecture will replace the Southern Islands architecture that just came out in November 2011 for discrete graphics sometime in 2013. Southern Islands is the core architecture used on AMD Radeon HD 7970 and Radeon HD 7950 GPUs.
---
AMD Sea Islands next year with Kaveri - and Kabini?
---

This looks very promising for AMD. We are anticipating a performance improvement of at least 20% without changing architecture.
The Radeon HD 8970 is also expected to have Over 4.50 TFlops of Raw Processing Power.

We don't have enough information to tell when those cards will be released, but AMD is planning to release them in around Q2 2013.
 
---
AMD 28nm Sea Islands official successor to Southern Islands
---
At the moment, we have no information on AMD's specific GPU codenames for the 28nm Sea Islands family, other than the fact that the name "Sea Islands refers to the chain of islands on the Atlantic Coast of the United States. Of course, for those speculators out there, a quick Wikipedia search will get you a list of over 100 islands found on the Atlantic Coast. It's anybody's guess as to which codenames AMD will choose to classify its Sea Islands GPUs. Nevertheless, this is the closest roadmap we could find:
sea_islands_map.jpg
 
isn't it still early to discuss about HD8k series? btw i'm interested on the HPC side of thing with amd GCN. we all know that GCN was made to be more GPGPU friendly. will amd able to secure something like super computer contract like nvidia did with their tesla?

We are anticipating a performance improvement of at least 20% without changing architecture.
Click to expand...

nvidia did that with GTX580
 
Nothing is too early 😉.

This GCN "v2" should be quite a bit better than its predecessor. It is always faster GPGPU computing power since more and more games uses GPGPU instead of Shaders.

I didn't understand very well this part "will amd able to secure something like super computer contract like nvidia did with their tesla?".
Can you clarify that?
 
well pretty much all super computer that utilizing gpu out there were using nvidia tesla gpu. even china super computer that once using amd gpu has been upgraded with nvidia tesla (fermi based). so i wonder if amd able to regain back some market share lost on the HPC sector with GCN
 
Yes, AMD could do that, but they are currently focusing on Graphics, and their profits had always been in the Graphics sector.
They don't have enough money to invest for a other branch of graphic card that may not work as expected.

nVidia can afford a 1M$ lost, AMD can't. 🙁
 
i remember reading some of the first rumors about the HD7000s that mentionned AMD was going to use XDR memory on them. The crazy bandwidth & PCIe 3.0 seemed to offer significant performance increase, and with GCN on top of that, it all looked very promising. Do you guys think it may have been reported to the HD8000s or it was rly just the rumor mill?

Very excited nontheless with the HD7000s, perf/watt and OC headroom is excellent. The complete arch overhaul turned out good. GCN 2.0.... i't'll keep me waiting for a year.
 
XDR isn't faster itself compared to GDDR5, but consumes far less power.

It can be used to create ultra-low wattage cards or overclock the Memory with Higher Clocks to have higher bandwidth/watt.

The HD 7970 itself is nearly twice as fast compared to the HD 6970.
 
http://www.cpu-world.com/news_2012/2012061601_A_First_Look_at_AMD_s_Radeon_HD8000_Series.html
AMD's latest beta graphics drivers revealed a few new device IDs, including four Kaveri and five Kabini graphics solutions for the next generation APUs. There are also a number of IDs for what looks like the next generation of Sea Islands GPUs, code-named Mars (9 models), Oland (3 models) and Venus (4 models).
According to predictions at 3DCenter (Google translation), Venus will be used for high-end GPUs, and it is expected to have around 5.1 billion transistors in a 410mm2 die, using 28nm technology. There could be between 2 and 4 raster engines with 2560 cores, approximately 160 TMUs and 32 or 48 ROPs, and there is likely to be a 384-bit memory interface. The most powerful single-GPU card could be the Radeon HD 8970, with a 30-40% performance boost over current generation of cards.
Oland is expected to be the name given to graphics cards filling the performance sector. With an anticipated 3.2 billion transistors on a 230mm2 die, this is another 28nm product. It is likely to have 2 raster engines, 96 TMUs and 32 ROPs. The full die will likely have 1536 cores and a 256-bit memory interface. The most powerful Oland based card will be the Radeon HD 8870, with a performance increase of around 20% compared to it's current generation of counterparts.
Finally, the mainstream cards will use the Mars core, the last 28nm GPU for the 2013 generation of graphics cards, with an expected 1.7 - 2.0 billion transistors in a 135 - 160mm2 die. There will be 1 raster engine, 40-48 TMUs and 16 ROPs. The core count for the full die is around 768 or 896, with a 128-bit or 192-bit memory interface. The best solution provided by this core will be the Radeon HD 8770, with an expected performance boost of 35% over current generation of mainstream GPUs.
It should be noted that most of the information in this article comes from predictions made by 3DCenter.org, and should not be considered as final information at this stage. We will update this as soon as more information is available.
Click to expand...
 
http://www.legitreviews.com/news/13381/

These are the new devices ID in the recent beta driver.
It seems that their new codenames are Mars, Oland and Venus.

These cards should be accompanied with a 20-30% boost over their older counterparts...

AMD1305.1 = "KAVERI1 DESKTOP (1305)"
AMD1304.1 = "KAVERI1 MOBILE (1304)"
AMD1307.1 = "KAVERI2 DESKTOP (1307)"
AMD1306.1 = "KAVERI2 MOBILE (1306)"
AMD9833.1 = "KB 12W 2C (9833)"
AMD9834.1 = "2C 5W KB (9834)"
AMD9831.1 = "KB 4C 17W (9831)"
AMD9832.1 = "KB 17W 4C (N-1) (9832)"
AMD9830.1 = "KB 4C 25W (9830)"
AMD6600.1 = "MARS (6600)"
AMD6601.1 = "MARS (6601)"
AMD6602.1 = "MARS (6602)"
AMD6602.1 = "MARS (6603)"
AMD6602.1 = "MARS (6606)"
AMD6602.1 = "MARS (6607)"
AMD6600.1 = "MARS (6620)"
AMD6601.1 = "MARS (6621)"
AMD6602.1 = "MARS (6623)"
AMD6610.1 = "OLAND (6610)"
AMD6611.1 = "OLAND (6611)"
AMD6631.1 = "OLAND (6631)"
AMD682B.1 = "VENUS LE"
AMD6823.4 = "VENUS PRO"
AMD6821.1 = "VENUS XT"
AMD6820.2 = "VENUS XTX"
Click to expand...
 
wowowow....
HD 8000 series.. I'm waiting and saving money for 8870..
I hope the new HD 8000 series will consume less power and better performance


 


Very very nice, if true. I am looking to upgrade my 6870 to 8870, and it looks like release pricing won't be jacked up this time around. As for release date, last time AMD "released" 7000 series cards, it was about a month before stores actually had them on shelves. Hopefully this time release and sales will kick off at the same time.
 
---
AMD HD 8000 specs leak, point to major performance boost.
---
Last week, spec sheets on AMD’s Radeon HD 8000 series leaked out to the press. The chart shows the two upcoming graphics cards that would theoretically replace AMD’s HD 7870 and 7850. The new HD 8000 family, codenamed Sea Islands, will be AMD’s first major refresh of its GCN (Graphics Core Next) architecture that debuted last year. The two cards we have information on are based on the Oland GPU, and would replace the Pitcairn family. Like the 7870/7850, the new GPU is built on a 28nm process at TSMC.

With leaks like this, there’s always the question of whether or not the data is valid. We’ve gone through and crunched some numbers, and figured out a few metrics that the original table doesn’t show. Here’s the original first, without any additional data.
Videocardzchart-640x275.png

These improvements align nearly perfectly with what we’d expect a next-gen architecture to deliver in the same price bracket. Memory bandwidth bumps upwards thanks to the use of faster GDDR5 RAM, pixel fillrates are up slightly, texture fillrates improve significantly. Since we know how many Render Output Units (ROPs) and Texture Mapping Units (TMUs) are in the 7870, we can extrapolate what the 8870 is packing in those departments.
Click to expand...

8870vs7950.png

We’ve added data from the HD 7950 Boost Edition and included current prices for Radeon cards as seen on Newegg. Columns labeled in orange reflect our extrapolations on the new core. The math fits, perfectly, in every example. The Radeon HD 8870 is built like a Radeon 7950.

Note that I said “built like,” not “equivalent to.” The current Radeon HD 7950 is a cut down Radeon 7970; AMD and Nvidia both improve yields by selling cards with certain function blocks disabled. This allows them to recoup a return on products that aren’t quite perfect, rather than tossing them on the slag heap. As a result, the 7950′s die is the same size as the 7970′s, despite the disparity in core count.

This data suggests that AMD felt the 7950 presented an ideal performance balance for a midrange card and opted to scale the chip downwards with a few changes made to decrease die size. The 384-bit memory bus is gone, though the faster GDDR5 memory compensates well for this. On the whole, the 8870 should be faster than the 7950 it replaces thanks to a higher core clock.

There are other reasons why this shift makes sense. Transistor densities across AMD and Nvidia are typically quite similar when comparing them at the same node.
Click to expand...

GPU-Transistor.png

Oland’s 3.4B and 270mm sq die size are right where we’d expect them to be.

Just because the Radeon 8870 is specced like a 7950 doesn’t mean it’ll perform like one. AMD will have tweaked the architecture to improve performance and power efficiency; the lower TDP is proof enough of that. This data suggests that AMD is pursuing an evolutionary strategy with the HD 8K GPU. When the company built the HD 6000-era Barts and Turks chips, they were a continuation and enhancement of the highly successful HD 5000 core. Cayman, the chip at the heart of the HD 6900 family, was the new architecture.

AMD could be planning to open a major new front with the high-end HD 8900 parts, but we’ll have to wait and see. GCN remains an excellent architecture, and AMD has enough on its plate trying to deal with bringing new SoCs to market — the company may have chosen a single GPU strategy to divert engineering resources where they’re needed most.
Click to expand...
 
^Very very impressive if it turns out to be true. I can't wait to go get next get AMD card. lol Last year we had release date by mid october, but now things have shifted I guess.
 
Are these prices accurate though....?

You would think that it would be the opposite and be much more expensive, considering it will be an unchallenged new generations of cards until the GTX700 series is released
 
I cant imagine the Nvidia 700 series having better card than what we are looking at here i heard they are looking at 5-15% boost in performance but also focusing on everything else like power consumption, bit interface, and initial pricing. cant remember where i read that though... sorry.
 
the same can be said to AMD next gen card. AMD claims to increase the performance up to 30% but nothing is confirm until we can see the actual product. Charlie at SA said that based on his 'internal' info you're looking at 15% performance increase at best.

i think that's why amd did not releasing their HD8k series at the end of this year to further tweak their design. the latest spec listing as the flagship will have bigger die with increase to SP count (2500+). with 7970 (2000 SP) amd already have 250w TDP if they going to make it into 2500 SP then just imagine about the power consumption. and it is not easy to trim down the power consumption because the HD8000 series will still based on 28nm node.
 
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom