Tuesday, December 4th 2012

Probable Radeon HD 8000 Series GPU Specifications Surface

AMD's Radeon HD 8000 "Sea Islands" family of GPUs may arrive some time in Q2-2013, but they won't arrive without a specifications overhaul. While NVIDIA is counting on performance enhancements to come out of higher clock speeds on existing silicon, while maintaining current (or lower) power-draw, with its "Enhanced Kepler" family of GPUs that will include the "GK110" juggernaut, AMD is counting on physically bigger chips with more components. AMD could step up transistor counts of its chips by as much as 20 percent, on existing 28 nm process.

According to a report, the biggest chip from AMD's fold could pack 5.1 billion transistors, amounting to 2560 stream processors, and an updated raster processing engine, with 48 ROPs ("Tahiti" has 32). While the Radeon HD 8970 maxes out the chip's capabilities, the HD 8950 could feature 256 fewer stream processors, at 2304. It could also go light on the clock speeds. AMD's performance-segment chip, codenamed "Sun" could see a similar stream processor increase to 1792, with the Radeon HD 8870 maxing it out, and HD 8850 featuring 1536. It maintains the memory bus width and raster engine layout of its predecessor.

AMD's mainstream chip, codenamed "Oland," which succeeds "Cape Verde," could address the problem of memory bandwidth, which is giving competing NVIDIA GPUs an upper hand in its segment. It could feature 896 stream processors, and a wider 192-bit GDDR5 memory interface. The Radeon HD 8770 maxes its specifications out, while the Radeon HD 8750 features a slimmer 128-bit memory interface, and 768 stream processors. The authenticity of this information can't be vouched for, but presents the only way in which AMD can end up with a competitive GPU lineup against NVIDIA, over 2013.Source: Hardware.info
Add your own comment

28 Comments on Probable Radeon HD 8000 Series GPU Specifications Surface

#1
happita
by: SonDa5
Sounds like a myth to me.


Looking at the hardware at release the HD79XX series cards had alot more performance potential then the GK104 cards that NVIDIA released. Looking at the hardware the HD79XX cards are superior to GTX 6XX cards. I believe the only reason NVIDIA chose to increase the price of their inferior cards is for profit.


Right now AMD HD79XX cards are better than ever with mature drivers. Good drivers with superior hardware is why AMD HD79XX cards are better than NVIDIA GTX6XX cards.

No matter how good NVIDIA drivers get their hardware will alway cap ultimate performance while AMD has overhead to grow because of superior hardware.
You sound like a real....oh heck, I hate even saying it...


by: m1dg3t
Picked my 5870 for just over $500, IIRC it was $540 all said and done. Picked my 7950 for $350. Double the performance and more for $150+ less. Wasn't a hard choice for me :o
I got my 5850 like 6 months after launch. I also bought a few Sapphire Toxic 5870s just to re-sell and make a small profit. But I got the Toxic's for $429 at launch and turned them around for $470.

The point is that the price points don't match up anymore based on how each card is tiered. However, prices have gone down quite a bit on the 7k series. AMD knew it was overpriced.
Posted on Reply
#2
badtaylorx
by: hardcore_gamer
GK104 was originally designed to be used in mainstream gaming cards. When 7970 didn't perform as nvidia speculated, they saw an opportunity to cut some costs. Since GK104 was good enough to beat 7970, nvidia released it as GTX 680 and delayed the real 680 (GK110).

Gaming is not the only market of these flagship GPUs. Computing also brings in significant revenue. Looking at these specs , I don't think 8970 can beat GK110 in computing.
not quite true....nv knew the gk110 was not going to make it in time from the get-go....the 110 was a "swing for the fences" chip that if they NAILED, would have DOMINATED!!! the 104 was an "overbuilt" for the mid section that served as a more than capable back up.....
Posted on Reply
#3
ThunderStorm
by: hardcore_gamer
GK104 was originally designed to be used in mainstream gaming cards. When 7970 didn't perform as nvidia speculated, they saw an opportunity to cut some costs. Since GK104 was good enough to beat 7970, nvidia released it as GTX 680 and delayed the real 680 (GK110).

Gaming is not the only market of these flagship GPUs. Computing also brings in significant revenue. Looking at these specs , I don't think 8970 can beat GK110 in computing.
Here's my thought on this:

GTX 680 is not supposed-to-be-mid-range-card like you said. The Gk104 chip is not a mid-range chip neither. What if I put a loads of transistors intended for computational purposes into GK104, would you still consider it as a mid range gpu ? I doubt highly doubt that.

GK 104 is not a wise choice when computing is evident because it is made for highly effective gaming performance. Those transistors packed in GK 104 are somewhat 90% useless for compute, just look at GK 104's DP and SP.

Now back to my assumption, the new and improved GK 104 with awesome computational power will probably consist of shit loads of transistors, more than Tahiti XT's 4.13b transistors for sure. But that's not all, the number of transistors can reach up to 5b or even more. Now that is a classic nvidia's style, crush your enemy with brute force imo.

Oh by the way, hi guys, i'm new here so go easy on me.
Posted on Reply