Review Index:
Feedback

AMD Radeon HD 7970 3GB Graphics Card Review - Tahiti at 28nm

Author:
Manufacturer: AMD

The First 28nm GPU Architecture

It is going to be an exciting 2012. Both AMD and NVIDIA are going to be bringing gamers entirely new GPU architectures, Intel has Ivy Bridge up its sleeve and the CPU side of AMD is looking forward to the introduction of the Piledriver lineup. Today though we end 2011 with the official introduction of the AMD Southern Islands GPU design, a completely new architecture from the ground up that engineers have been working on for more than three years.

This GPU will be the first on several fronts: the first 28nm part, the first cards with support for PCI Express 3.0 and the first to officially support DirectX 11.1 coming with Windows 8. Southern Islands is broken up into three different families starting with Tahiti at the high-end, Pitcairn for sweet spot gaming and Cape Verde for budget discrete options. The Radeon HD 7970 card that is launching today with availability in early January is going to be the top-end single GPU option, based on Tahiti.

Let's see what 4.31 billion transistors buys you in today's market.  I have embedded a very short video review here as well for your perusal but of course, you should continue down a bit further for the entire, in-depth review of the Radeon HD 7970 GPU.

Southern Islands - Starting with Tahiti

Before we get into benchmark results we need to get a better understanding of this completely new GPU design that was first divulged in June at the AMD Fusion Developer Summit. At that time, our own lovely and talented Josh Walrath wrote up a great preview of the architecture that remains accurate and pertinent for today's release. We will include some of Josh's analysis here and interject with anything new that we have learned from AMD about the Southern Islands architecture.

When NVIDIA introduced the G80, they took a pretty radical approach to GPU design. Instead of going with previous VLIW architectures which would support operations such as Vec4+Scalar, they went with a completely scalar architecture. This allowed a combination of flexibility of operation types, ease of scheduling, and a high utilization of compute units. AMD has taken a somewhat similar, but still unique approach to their new architecture.

View Full Size

Continue reading our review of the AMD Radeon HD 7970 3GB graphics card and Southern Islands architecture!!

Instead of going with a purely scalar setup like NVIDIA, they opted for a vector + scalar solution. The new architecture revolves around the Compute Unit, which contains all of the functional units. The CU can almost be viewed as a fully independent processor. The unit features its own L1 cache, branch and MSG unit, control and decode unit, instruction fetch arbitration functionality, and the scalar and vector units.

View Full Size

The vector units are the primary workers in the CU when it comes to crunching numbers. Each unit contains four cores, and allows for four “wavefronts” to be processed at any one time. Because AMD stepped away from the VLIW5/4 architectures, and have gone with a vector+scalar setup, we expect to see a high utilization of each unit as compared to the old. We also expect scheduling to be much easier and efficient, which will again improve performance and efficiency. The scalar unit will actually be responsible for all of the pointer ops as well as branching code. This particular setup harkens back to the Cray supercomputers of the 1980s. The combination of scalar and vector processors was very intuitive for the workloads back then, and that follows onto the workloads of today that AMD looks to address.

View Full Size

The combination of these processors and the overall design of each CU gives it the properties of different types of units. It is a MIMD (multiple instructions multiple data) in that it can address four threads per cycle per vector, from different apps. It acts as a SIMD (single instruction multiple data) much like the previous generation of GPUs. Finally it has SMT (symmetric multi-threading) in that all four vector cores can be working on different instructions, and there are 40 waves active in each CU at any one time. Furthermore, as mentioned in the slide, it supports multiple asynchronous and independent command streams. Essentially the unit is able to work on all kinds of workloads at once, no matter what the source of the data or instructions are.

View Full Size

A full Tahiti GPU will feature 32 of the Compute Units, each made up of 64 stream processors. That brings the entire GPU to a total of 2048 SPs, a 33% jump over the Cayman architecture that was built around 1536 processors.

Memory and Caches

One area that AMD did detail extensively was the changes in the internal cache, as well as their push for fully virtualized memory. Each CU has its own L1 cache divided into data, instruction, and load/store. The GPU then has shared L2 cache which is fully coherent. Each L1 cache has a 64 bit interface with the L2, and once this scales in terms of both CU count and GPU clockspeed, we can expect to see multiple terabytes per second of bandwidth between the caches. The L1 caches and texture caches are now read/write, as compared to the read only units in previous architectures. This is a big nod not only to efficiency and performance, but also the type of caches needed for some serious compute type workloads.

View Full Size

The next level of memory support is that of full virtualization of memory with the CPU. Previous generations of products were limited to what memory and cache were onboard each video card. This posed some limitations on not just content in graphics, but were also problematic in compute type scenarios. Large data sets proved to be troublesome, and required a memory virtualization system which was separate from the CPUs virtual memory. By adopting x86-64 virtual memory support on the GPU, this gets rid of a lot of the problems in previous cards. The GPU shares the virtual memory space, which improves data handling and locality, as well as gracefully surviving unhappy things like page faults and oversubscriptions. This again is aimed at helping to improve the programming model. With virtual memory, the GPU’s state is not hidden, and it should also allow for fast context switches as well as context switch pre-emption. State changes and context switches can be quite costly, so when working in an environment that features both graphics based and compute workloads, the added features described above should make things go a whole lot smoother, as well as be significantly faster, thereby limiting the amount of downtime per CU.

View Full Size

It also opens up some new advantages to traditional graphics. “Megatextures” which will not fit on a card’s frame buffer can be stored in virtual memory. While not as fast as onboard, it is still far faster than loading up the texture from the hard drive. This should allow for more seamless worlds. I’m sure John Carmack is quite excited about this technology.

Obviously the 32 CU on Tahiti make up the majority of the architecture but there are several other keys to look at.  Just as we saw with Cayman, Southern Islands will offer dual geometry engines for improved scalability as well as eight render back-ends with 32 ROP engines

View Full Size

The memory interface gets a nice boost moving from the 256-bit GDDR5 interface on Cayman to a 384-bit interface capable of 264 GB/sec of bandwidth.  There are six individual 64-bit dual-channel memory controllers that create an unbalanced render back-end ratio of 4:6. 

View Full Size

And of course, the architecture rounds out with the controllers for the PCI Express 3.0 interface capable of 8 GigaTransfers / second (essentially double that of PCIe 2.0), Eyefinity display controllers, UVD engine, CrossFire compositor, etc.  All of this adds up to an amazing 4.31 billion transistors on a 28nm process technology inside a 365mm2 die

You might be wondering why we don't have the typical die shot of the lovely new 28nm Tahiti GPU.  In truth, I have no answer for you, other than we asked several times and were told in each instance that they didn't have one.  While this did arouse some suspicion from us as the the design of Tahiti, AMD assured us there were no tricks up its sleeve, the die was built with 36 CUs with 4 disabled, for example.

December 22, 2011 | 07:29 AM - Posted by Ryan Shrout

Sure man, we'd love to have you write a guest story about the upgrade experience. I am sure quite a few readers will be in the same boat.

December 22, 2011 | 05:56 AM - Posted by Richard (not verified)

I personally think that graphics cards are way overpriced. I will keep my 570 for a good while I suppose. I have a hard time spending more than 350$ especially when you see how games for PC are having problems to run well. Major issues with Battlefield 3, even in the menus, like it is corrupted, image and sound. The beta ran fine which is weird. Will try one more install and if not fixed will have to throw it away, 60$ badly invested. Other games run just fine. Wrote to the company, no reply. By the way, game prices are also way too high in my opinion. Saw Rage for 20$ free delivery at newegg.com, same game at newegg.ca = 60$ +++ 10$ delivery. That is not right. Waiting for a newegg response...they will not sell to a Canadian on their US site of course.

December 22, 2011 | 07:30 AM - Posted by Ryan Shrout

Not sure about your BF3 issues, sorry to hear.

But yes, I tend to agree - if you have a GTX 570/580 or a Radeon HD 6900 card the difference in performance doesn't warrant the price right now.

December 23, 2011 | 05:36 AM - Posted by Richard (not verified)

Good news, the new drivers seem to have fixed Battlefield.
I gave it a try last night and I got stucked on it for a few hours. Good game by the way, it looks really good on my system. I get a few weird horizontal lines here and there but only in the menu section, strange. Once in the game it runs fine. I play it at 1080p and it is smooth. Happy!!!!!
Thanks for your excellent reviews by the way.

December 23, 2011 | 06:46 AM - Posted by Ryan Shrout

Thanks for reading them Richard!

December 22, 2011 | 06:33 AM - Posted by Shadow (not verified)

This card is not worth the $600 price tag, clearly overpriced in my opinion. And here I thought AMD was the best price to performace PC hardware company, but I guess if you want the first 28nm graphic card you have to pay up. I hope Nvidia comes out with their 600 series soon so the price wars can start on these new graphic cards, but I will be holding onto my EVGA Classified 590 for awhile though. I can play BF3, WOW, and Diablo 3 beta all at high/ultra -1920/1080 settings with no lag or hicups perfectly fine.

December 22, 2011 | 07:31 AM - Posted by Ryan Shrout

True, this is not a replacement for a GTX 590 at all.

December 22, 2011 | 08:57 AM - Posted by Imperfectlink

Looks like a solid showing from AMD. As long as the drivers match the drapes it should do pretty well.

We have to remember that TSMC is having a rough time meeting demand at the 28nm fabrication process so price will reflect that. Signs on the intarwebs point to Kepler lagging behind so why not take advantage of the lead? Bleeding edge has a price and right now it's $550. No doubt AMD will cut the price when yields are good, competition arrives and they can afford to.

December 22, 2011 | 09:51 AM - Posted by Swoosh (not verified)

@ Shadow

Not so fast, this is just a review and AMD is still in the
full authority for the "The last minute say" as far as retail pricing is concerned. For all we know and for what
AMD is known for is that the way they priced their video card products always makes it hard for nVidia to cope up
and compete, this is not just for the video card performance alone BUT for the price to performance ratio
which nVidia are loosing years past since the release of AMD's 5900 series video cards up to now and they may never
regain major market shares unless nVidia would drop their high end video card prices less than to what AMD's pricing
on counterpart models, WHICH will never happen.

Im not saying nVidia video cards are not good but, what AMD
did and currently doing (as we all know) it seems its a one sided race to price to performance ratio and chances are even if nVidia were to release their up coming "not sure
when" 6000 series video cards the same day as AMD's 7900 series video card launch, their prices would be more
expensive and doesnt favor the taste of many gamers now. Another thing that makes nVidia suffer from low sales is the issue of hypocrisy and bad publicity and getting a lot of bad karma on their unfair doings.

AMD for sure will try to make their prices reasonable. Talking about latest technology and hi performance offerings particularly the 7970 model, It is like releasing their 6900 series video cards where gamers have no qualms or issues since day one.

- Very bad for nvidia and its biased bloggers and reviewers.

December 22, 2011 | 10:42 AM - Posted by Anonymous (not verified)

How can I tell if my motherboard supports PCI Express 3.0? It's about 1.5 years old.

Thanks,
Kevin

December 22, 2011 | 10:59 AM - Posted by Ryan Shrout

It doesn't require a PCIe 3.0 motherboard, PCIe 1.0 and 2.0 are backwards compatible still.

December 22, 2011 | 11:43 AM - Posted by Mechromancer (not verified)

We will finally have an AMD GPU worthy for FOLDING@HOME!!!!!!!!!!!!!!!!!!!!!!!!!

I've waited since the 2900XT for this day. Maybe Stanford will get around to making a good client for it in less than a year this time >:P

December 22, 2011 | 07:23 PM - Posted by nabokovfan87

Heh, and I thought waiting with the 4850 was a long time. Patience is a virtue, well done!

December 23, 2011 | 06:46 AM - Posted by Ryan Shrout

Wow, yah, you are going to see a HUGE performance difference here.

December 22, 2011 | 11:49 AM - Posted by Luzer (not verified)

I have two 5870's in Crossfire connected to three Dell 24"(something like 6080x1200 with bezel compensation on). If I sell both of my cards for ~$300 is it worth the $250 to get a 7970 or should I pick up another 5870 ($150) for triple crossfire or even two 6970's ($300 each)? I can’t really play BF3 with crossfire enabled and using all three displays, av. FPS is around 20 on high, but I can play on a single display, 1900x1200, on high, av. FPS 55, with crossfire.

In the end I probably wait for the 7970 to get down to $375 and put my 5870's in an old pc and use it as a space heater.

December 22, 2011 | 11:57 AM - Posted by JohnE (not verified)

I want to purchase the 6970 MSI Lightning today. I'll be happy with this card as it overclocks to 1ghz stable. When do you think the "older" cards will drop in price? I'm afraid i'll have to wait a month, I WANT MY TOYS NOW! (jk)

December 22, 2011 | 02:36 PM - Posted by Wolvenmoon (not verified)

Will hold on to my 9800GTX+ until prices come down or performance across the board doubles again.

December 23, 2011 | 06:47 AM - Posted by Ryan Shrout

LOL

December 24, 2011 | 04:04 AM - Posted by rrr (not verified)

If you have a small monitor, by all means do it.

February 4, 2012 | 04:19 AM - Posted by SiliconDoc (not verified)

I'm holding on to the 9500GT because it rocks out in BF3 at 12-15 fps and that's on a lower end pentium D with 256x4 of 533ddr2 and a 40g ide drive.
Whoo hoo, we iz gamin' !
It doesn't crash, at all, either, unlike the red cards.

December 22, 2011 | 06:04 PM - Posted by Anonymous (not verified)

omg did you test those benchmarks with 0% power control at the CATALYST control center ?. You shoulda test it with +20 power setting, and then try to OC it... you might even get a higher oc since other reviewers get more then 1100mhz out of the core.

December 23, 2011 | 06:47 AM - Posted by Ryan Shrout

We will give that a shot after the holiday.

December 22, 2011 | 07:22 PM - Posted by nabokovfan87

I wonder how the 3 gb Frame buffer affects 3D gaming w/ eyefinity.

December 23, 2011 | 06:48 AM - Posted by Ryan Shrout

It might not help THAT much for 3D as the amount of memory required for 3D isn't much than standard - just processing power. The reasoning: the textures / data need for the left eye are usually a pretty close match to what is needed for the right eye.

December 26, 2011 | 05:56 PM - Posted by nabokovfan87

But doesn't doing 3d require twice as much processing then traditional graphics? For instance, the metro 2033 normal AVG FPS is double what is in 3d. I wonder if the frame buffer would allow more things to be in VRAM or something would happen to give a boost, slight if any. It doesn't seem likely from what you are saying, but it would be something interesting to see.

Especially considering how the 69xx series has had 2 and 1 gb variants and the slight hits in perf. on those cards.

December 23, 2011 | 08:12 AM - Posted by soldierguy

Hey Ryan

Thanks for the review..luv your shows. I've got a pair of 5870 2Gb's in Crossfire powering 5760x1080 and an i7930 oc to 3.8ghz. I'm thinking of a pair of these for my next build or maybe just buying a pair of these and clunking them in the existing unit. Of course money and value are an issue. And would I really see any multi player benefit from just upgrading the gpu's? I guess you'll be doing a comparison against my setup or similar and that would help for sure. But in the real world would there be any "real" improvement in multi player for me. I mostly need the performance for the hi twitch games--for me COD. I play other games too but don't need more computing power for them. Goodluck with the new fiber optics....I'm green with envy..that's where I have problems and limitations.
soldierguy

December 23, 2011 | 10:41 AM - Posted by Mark (not verified)

What I'm getting out of the 7970 is that they realized Nvidia had a superior architecture and mimicked it. Now how this will affect their drivers will be interesting. If anything this equalizes the 2 companies.

Their charging a high premium for cards because they can. Prices will go down SIGNIFICANTLY when Kepler rears its turtle head......TURTLE!!!

Pretty much all this says is that AMD was early this round but Nvidia again will take the performance crown because if they don't surpase a product released 6-9 months prior to theirs then it will not bode well for the green machine.

God I sound like a Nvidia shill....

December 23, 2011 | 11:23 AM - Posted by Davo (not verified)

The prices will go down and quite quickly. AMDs top end cards are cheaper than Nvidia's at present and this series will follow suit.

These cards have a premium at the moment (and we are only speculating at present since they arent yet on sale) but competition amongst AMD resellers will see it go down anyway.

I cant blame them for making hay while the sun shiones at the moment. This card is the top end and represents a small percentage of the overall market and they will sell like hotcakes no matter where they are priced to people who want the top end.

Im happy AMD has at least one division thats actually making great product and competing strongly.

December 24, 2011 | 11:19 AM - Posted by Ryan Shrout

I hope you are right Davo.

February 4, 2012 | 05:45 AM - Posted by SiliconDoc (not verified)

Cheaper by what ? $10 per three hundred, or 3 percent...
Yeah, cheaper by three percent.
$10 bucks is not enough to suffer through the driver issues, the crash issues, the lack of new game support issues, the no PhysX issue...
Nope.
To all those who've "never had a problem with their radeon cards" congratulations, miracles still do happen.

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

By submitting this form, you accept the Mollom privacy policy.