Meet the Radeon HD 7970

Now that we’ve had a chance to discuss the features and the architecture of GCN and Tahiti, we can finally get to the  end result: the card. AMD’s first product in the Southern Islands family is the Radeon HD 7970, continuing AMD’s tradition of launching their fastest single-GPU first.

As we’ve already covered in our discussion on GCN/Tahiti’s architecture, Tahiti shares a lot of physical similarities with Cayman, and so then does the 7970 with the 6970. With the 7970 AMD has targeted a very similar power profile as the 6970, and while AMD has not published the typical board power of the 7970 we know the PowerTune limit is 250W, the same as with the 6970. As a result the 7970 is at least superficially designed to work in the same environments/constraints as the 6970.

With that said, AMD has not sat idle when it comes to the design of the card – this isn’t just a Tahiti GPU put in a 6970 shell. Livery changes aside it’s clear that the 7970 is a distinct card just from looking at it. AMD’s ill-fated boxy design for their cards is gone; the removable plastic shroud is now once again a rounded design similar to the 5800 series, and this time AMD takes it a step further by slightly rounding off the rear of the card for airflow purposes. Furthermore the shroud is now made of a very hard, very shiny plastic, versus the soft plastic used in past cards.

But the bigger change is on the back of the card, where AMD has completely done away with the backplate. First used in the 5800, backplates help to protect the card from users (and users from the sharp bits of the card), but the tradeoff was that the backplate occupied 2.7mm of space. What’s the significance of 2.7mm? When you’re trying to put these cards adjactent to each other for CrossFire, it’s everything as we have found out.

The boxy design coupled with the backplate meant that the 6900 series used virtually every last millimeter of space they were allowed under the PCIe specification; the cards were so wide that when adjacent it was easy for a card to shift and block the airflow of the neighboring card. The backplate contributed to this problem by consuming 2.7mm of space that could otherwise be used to channel airflow, and as a result it’s gone. AMD’s design doesn’t have the overt wedge that NVIDIA’s does to allow airflow, but it should be enough to keep the cards well enough separated to allow them to breathe when they’re closely together for CrossFire.

Overall the card is 10.5” going by the PCB, but AMD has a metal reinforcement ring/plate running along the entire card that sticks out the rear. After accounting for this plate the total length of the card is just shy of 11”, making the card roughly half an inch longer than the 6970 and 5870. The difference is not huge, but it will make the 7970 ever so slightly harder to fit than the 6970 in space-constrained cases.

Moving on, while AMD has made some changes to the shrouding to improve cooling, they haven’t stopped there. The blower has also been tweaked slightly compared to what we’ve seen on the 6970. The 7970’s blower is a bit larger (~75mm) and the fins are slightly larger to make use of that space. Overall this should improve the amount of air moved at speeds similar to the blower on the 6970, though AMD didn’t provide any numbers.

Meanwhile the heatsink is very similar to the 6970’s. As with the 6970 an aluminum heatsink sits on top of a vapor chamber cooler that draws heat from the GPU and other components towards the heatsink. Other than being a bit larger than the 6970 the biggest difference is that AMD is now using the same higher performance phase-change TIM that they used on the 6990, which also means that AMD is highly recommending that the 7970 not be disassembled as the TIM won’t operate nearly as well once it’s been separated. Furthermore as we found out the specific TIM AMD is using is screen printed onto the GPU, so reapplying a new TIM in the same manner is virtually impossible.

 

Finally, it’s once we move towards the front that we see the biggest change in the name of cooling: AMD has once again moved back to a full slot exhaust vent. As you may recall, starting with the 5800 series AMD moved to a half slot vent configuration so that they could use the other half of the second slot to fit a second DVI port along with their DisplayPort and HDMI ports. The half slot vent did not prove to be a huge problem for the 5800 or 6900 series but it still placed some limits on AMD’s ability to cool their cards and made the process a bit noisier. As the second DVI port has become redundant (more on that later), AMD has opted to get rid of it and go back to using the whole slot for cooling. One way or another though this was probably necessary – looking at our data the 7970 is a bit more power hungry than the 6970 even if the specifications are similar, and as a result AMD needs better cooling to keep parity with the 6970.

Moving on, tweakers will be happy to see that the dual BIOS feature first introduced on the 6900 series is back. The 7970 will feature the same dual BIOS configuration, with a locked factory BIOS (2), and a rewritable BIOS (1) for other uses. As with the 6900 series this is primarily to allow failsafe BIOS flashing, but the implications for GPU unlocking lower tier cards are clear. In fact we’re surprised that AMD included the switch given how rampant 6950 unlocking was, as while it was good PR it must have been bad for their 6970 sales.

Next to the BIOS switch we will find the PCIe power sockets, which given the 250W PowerTune limit of the card mean we’re looking at the same 6+8pin configuration as the 6970. Enthusiasts who caught on to the fact that AMD had to shave some PCIe sockets on the 6900 series should note that the sockets are untouched on the 7970, as the blower now sits above the PCIe sockets. Elsewhere at the front end of the card we’ll find the two CrossFire connectors, and as always when it comes to their high-end cards AMD is supporting up to 3-way CF with the 7970.

Back to the front of the card we can see AMD’s new Southern Islands port configuration. As you may recall from the 6000 series, with the 6000 series AMD moved from being able to drive 2 dual-link DVI ports (2 sets of paired TMDS transmitters) to being able to drive 1 dual-link DVI port + 1 single-link DVI port, as they removed the 4th TMDS transmitter. Furthermore as AMD has only been able to drive 2 TMDS-type ports at once, the 2nd DVI port was largely redundant as everything it could do the HDMI port could do with a mechanical adaptor.

So for Southern Islands AMD has taken this to its logical conclusion and cut out the 2nd DVI port entirely. There is now a single DL-DVI port, along with an HDMI port and 2 miniDP ports all along a single slot. The 7970 still has the internal logic to drive the same monitor configurations as the 6970, but anyone using the SL-DVI port will now be fed by the HDMI port. In order to make this transition easier on buyers, AMD will be requiring that partners ship both an HDMI to SL-DVI adaptor and an active miniDP to SL-DVI adaptor with their 7970s, so 7970 users will be able to drive up to 3 DVI monitors out of the box, which is actually better than what the 6970 could do. Of course we expect that this will be a limited time offer; once AMD’s partners start putting together cheaper cards later in the 7970’s life, the mDP to SL-DVI adaptor will be the first thing to go.

On that note, for anyone who is curious about idle clockspeeds and power consumption with multiple monitors, it has not changed relative to the 6970. When using a TMDS-type monitor along with any other monitor, AMD has to raise their idle clockspeeds from 350MHz core and 600Mhz memory to 350MHz core and the full 5.5GHz speed for memory, with the power penalty for that being around 30W. Matched timing monitors used exclusively over DisplayPort will continue to be the only way to be able to use multiple monitors without incurring an idle penalty.

Next on the docket we wanted to quickly touch on the subject of RAM. As with their past cards AMD has outfitted the 7970 with RAM rated beyond their memory speed requirements, in this case the 7970 is outfitted with 6GHz modules even though it only needs to operate at 5.5GHz. We haven’t been able to take the card apart so we haven’t seen whose modules AMD is using, but we strongly suspect they’re the same 2Gb Hynix modules the reference 6970 used.

With the move to a 384bit bus AMD has increased the chip count from 8 to 12, and the total RAM size from 2GB to 3GB. As games are only now starting to effectively use more than 1GB of RAM this should offer plenty of headroom for future games, above and beyond even their existing 2GB cards.

At the same time with the move to a 384bit bus this has raised the question of where AMD goes from here. It’s well published that GDDR5 is an intricate memory technology to work with, with the memory bus being the hardest part. To run a 384bit bus at 5.5GHz is quite the accomplishment (NVIDIA didn’t get nearly that high with Fermi), but it also means that we’re near the end of the road for GDDR5. While GDDR5 is rated for up to 7GHz in the real world buses will never be able to scale quite that well and the required memory voltages are on the high side. Meanwhile a 512bit GDDR5 bus is possible but would be even more expensive and difficult than a 384bit bus, and it’s safe to say that’s not the preferred route.

So what comes after GDDR5? At this point AMD tells us that they’re looking at a few different things, but of course it’s far too early to discuss anything in detail. The JEDEC has not passed anything such as a GDDR6 standard, though we expect whatever technology AMD will eventually use will come out of the JEDEC. But if nothing else at this point it’s safe to assume that by the time we’re on 20nm GPUs we won’t be using GDDR5.

Finally, while we haven’t had a chance to tinker with overclocking due to our limited time with the 7970, but AMD is telling us that the 7970 is going to be good overclocker. Most cards should be able to hit 1GHz or higher on the core clock and 6GHz or higher on the memory clock with little effort even with the reference cooler, but the tradeoff will of course be in power consumption and noise, as you’ll need to increase the PowerTune limits to make those overclocks matter.

Drivers & ISV Relations The Test
Comments Locked

292 Comments

View All Comments

  • Zingam - Thursday, December 22, 2011 - link

    And at the time when it is available in D3D. AMD's implementation won't be compatible... :D That's sounds familiar. So will have to wait for another generation to get the things right.
  • Ryan Smith - Thursday, December 22, 2011 - link

    As for your question about FP64, it's worth noting that of the FP64 rates AMD listed for GCN, "0" was not explicitly an option. It's quite possible that anything using GCN will have at a minimum 1/16th FP64.
  • Sind - Thursday, December 22, 2011 - link

    Excellent review thanks Ryan. Looking forward to see what the 7950 performance and pricing will end up. Also to see what nv has up their sleeves. Although I can't shake the feeling amd is holding back.
  • chizow - Thursday, December 22, 2011 - link

    Another great article, I really enjoyed all the state-of-the-industry commentary more than the actual benchmarks and performance numbers.

    One thing I may have missed was any coverage at all of GCN. Usually you guys have all those block diagrams and arrows explaining the changes in architecture. I know you or Anand did a write-up on GCN awhile ago, but I may have missed the link to it in this article. Or maybe put a quick recap in there with a link to the full write-up.

    But with GCN, I guess we can close the book on AMD's past Vec5/VLIW4 archs as compute failures? For years ATI/AMD and their supporters have insisted it was the better compute architecture, and now we're on the 3rd major arch change since unified shaders, while Nvidia has remained remarkably consistent with their simple SP approach. I think the most striking aspect of this consistency is that you can run any CUDA or GPU accelerated apps on GPUs as old as G80, while you even noted you can't even run some of the most popular compute apps on 7970 because of arch-specific customizations.

    I also really enjoyed the ISV and driver/support commentary. It sounds like AMD is finally serious about "getting in the game" or whatever they're branding it nowadays, but I have seen them ramp up their efforts with their logo program. I think one important thing for them to focus on is to get into more *quality* games rather than just focusing on getting their logo program into more games. Still, as long as both Nvidia and AMD are working to further the compatibility of their cards without pushing too many vendor-specific features, I think that's a win overall for gamers.

    A few other minor things:

    1) I believe Nvidia will soon be countering MLAA with a driver-enabled version of their FXAA. While FXAA is available to both AMD and Nvidia if implemented in-game, providing it driver-side will be a pretty big win for Nvidia given how much better performance and quality it offers over AMD's MLAA.

    2) When referring to active DP adapter, shouldn't it be DL-DVI? In your blurb it said SL-DVI. Its interesting they went this route with the outputs, but providing the active adapter was definitely a smart move. Also, is there any reason GPU mfgs don't just add additional TMDS transmitters to overcome the 4x limitation? Or is it just a cost issue?

    3) The HDMI discussion is a bit fuzzy. HDMI 1.4b specs were just finalized, but haven't been released. Any idea whether or not SI or Kepler will support 1.4b? Biggest concern here is for 120Hz 1080p 3D support.

    Again, thoroughly enjoyed reading the article, great job as usual!
  • Ryan Smith - Thursday, December 22, 2011 - link

    Thanks for the kind words.

    Quick answers:

    2) No, it's an active SL-DVI adapter. DL-DVI adapters exist, but are much more expensive and more cumbersome to use because they require an additional power source (usually USB).

    As for why you don't see video cards that support more than 2 TMDS-type displays, it's both an engineering and a cost issue. On the engineering side each TMDS source (and thus each supported TMDS display) requires its own clock generator, whereas DisplayPort only requires 1 common clock generator. On the cost side those clock generators cost money to implement, but using TMDS also requires paying royalties to Silicon Image. The royalty is on the order of cents, but AMD and NVIDIA would still rather not pay it.

    3) SI will support 1080P 120Hz frame packed S3D.
  • ericore - Thursday, December 22, 2011 - link

    Core Next: It appears AMD is playing catchup to Nvidia's Cuda, but to an extent that halves the potential performance metrics; I see no other reason why they could not have achieved at varying 25-50% improvement in FPS. That is going to cost them, not just for marginally better performance 5-25%, but they are price matching GTX 580 which means less sales though I suppose people who buy 500$ + GPUs buy them no matter what. Though in this case, they may wait to see what Nvidia has to offer.

    Other New AMD GPUs: Will be releasing in February and April are based on the current architecture, but with two critical differences; smaller node + low power based silicon VS the norm performance based silicon. We will see very similar performance metrics, but the table completely flips around: we will see them, cheaper, much more power efficient and therefore very quiet GPUs; I am excited though I would hate to buy this and see Nvidia deliver where AMD failed.

    Thanks Anand, always a pleasure reading your articles.
  • Angrybird - Thursday, December 22, 2011 - link

    any hint on 7950? this card should go head to head with gtx580 when it release. good job for AMD, great review for Ryan!
  • ericore - Thursday, December 22, 2011 - link

    I should add with over 4 billion transistors, they've added more than 35% more transistors but only squeeze 5-25% improvement; unacceptable. That is a complete fail in that context relative to advancement in gaming. Too much catchup with Nvidia.
  • Finally - Thursday, December 22, 2011 - link

    ...that saying? It goes like this:
    If you don't show up for a race, you lose by default.
    Your favourite company lost, so their fanboys may become green of envydia :)

    Besides that - I'd never shell out more than 150€ for a petty GPU, so neither company's product would have appealed to me...
  • piroroadkill - Thursday, December 22, 2011 - link

    Wait, catchup? In my eyes, they were already winning. 6950 with dual BIOS, unlock it to 6970.. unbelievable value.. profit??

    Already has a larger framebuffer than the GTX580, so...

Log in

Don't have an account? Sign up now