AMD Radeon HD 7000 Series Single-GPU Graphics Card Price-Points Surface

[Casecutter]

Oh another thing... these prices also probably reflect that TSMC's was to increase their price for wafers manufactured using the 28nm process, so there’s that to contend along with their lackluster yields.

 

[Boomstick777]

he is actually completely correct in what he said, maybe you shouldnt run your mouth when you dont know anything.

I think we got our wires crossed? I thought he was saying the GTX 580 3GB only has 1.5GB actual ram. He is talking about the GTX 590? Which although it says 3GB on the box, only actually uses 1.5GB per GPU becausce of it's SLI architecture.

Oops sorry dude I beg thee of forgiveness



Hd 7970 looks like a decent card, can't wait for Nvidia's line up aswell

 

[Super XP]

It seems the new HD 7900's are going in the right direction.

According to BSN, the HD 7970 GPU will be made up from 32 Compute Units for a total of 2048 cores that operate at a 1GHz clock.

These will be linked via a 384-bit wide memory bus to 3GB of GDDR5 VRAM working in a quad data-rate mode at 1.37GHz (5.5GHz effective) in order to provide a whopping 264GB/s of memory bandwidth.

1) Truly functional Virtual Memory coming to GPU with AMD Radeon HD 7900 Series
2) AMD Compute Unit: Completely redesigned compute core features dedicated L1 and L2 memory, as well as shared L1 for MIMD functionality
3) With Tahiti packing 32 Compute Units in a maximum configuration, a 32 CU GPU with 2048 processing cores features almost 5MB of on-die memory: 512KB L1 Data cache, 384KB Shared L1 cache and 2MB of LDS and 2MB of L2 Cache. This is a record amount of cache for the GPUs so far, and you can expect this trend to continue.



The new GCN architecture brings numerous innovations to GPU architecture, out of which we see x86 virtual memory as perhaps one of the most important ones. While the GPU manufacturers have promised functional virtual memory for ages, this is the first time we're seeing a working implementation. This is not a marketing gimmick, IOMMU is a fully functional GPU feature, supporting page faults, over allocating and even accepting 64-bit x86 memory pointers for 100% compatibility with 64-bit CPUs. Virtual memory is going to be the large part of next-gen Fusion APUs (2013) and FireStream GPGPU cards (2012), and we can only commend the effort made in making this possible.

LINK:
http://www.brightsideofnews.com/news/2011/11/30/radeon-hd-7000-revealed-amd-to-mix-gcn-with-vliw4--vliw5-architectures.aspx

 

[dj-electric]

If these specs are true...

If the HD7970 will really based on 2048SP at 1Ghz and a 384BIT 3GB GDDR5 memory, it will probably smack the GTX580 like a 9800XT smacking an FX5950ULTRA

 

[Delta6326]

yummy I could use a 7970

 

[dir_d]

Isnt there some law about new architectures and how they dont perform well. Im being really skeptical of this new card, i dont think it will perform as well as it should.

 

[Super XP]

The only issue I see with Crossfire/SLI is it ends up only using one cards memory. Why on earth won't these companies fix this

 

[dj-electric]

Fix? if u wanna be able to render a certain scene you have to put it on both of the cards to flexibly render, they kinda have no choice you know....
It HAS to mirrorize the same VRAM to be able to render the same frame

 

[nt300]

Fix? if u wanna be able to render a certain scene you have to put it on both of the cards to flexibly render, they kinda have no choice you know....
It HAS to mirrorize the same VRAM to be able to render the same frame

I find this interesting. Can't they make the Crossfire act as a single dual gpu HD 6990? You don't run into this problem by putting it on both of the cards to flexibly render. With technology today they must be a fix somewhere or coming out soon.

 

[RejZoR]

Yeah, one would expect that by now, they'd figure it out how to render scenes using 2 cards without the need to mirror data on both.

 

[Delta6326]

I find this interesting. Can't they make the Crossfire act as a single dual gpu HD 6990? You don't run into this problem by putting it on both of the cards to flexibly render. With technology today they must be a fix somewhere or coming out soon.

I wish because that would be awesome but the 6990 doesn't even run like that. The 6990 shares the 4GB, each GPU is backed by 2 GB of GDDR5 memory over a 256-bit wide memory interface each, totaling 4GB of memory. It's just 2 GPUs on one PCB.