| GPU | Launch price | Spring MSRP | Summer MSRP |
|---|---|---|---|
| Radeon HD 7970 | $549 | $479 | $429 |
| Radeon HD 7950 | $449 | $399 | $349 |
| Radeon HD 7870 | $349 | $349 | $299 |
| Radeon HD 7850 | $249 | $249 | ~$239 |
| Radeon HD 7770 | $159 | $139 | ~$119 |
| Radeon HD 7750 | $109 | $109 | ~$99 |
With both AMD and Nvidia trying hard to push its graphics technologies beyond PC gaming, including the use of embedded graphics in motherboards and processors, that has also meant that the discrete GPU market on the low-end segment is shrinking and thus mid-range offerings are taking a new level of precedence.
Holding the performance crown can be a big deal, and AMD did so uncontested for quite some time with the Radeon HD 5970. As you may recall, the GeForce GTX 400 series was simply too hot to stick a pair of GPUs on a single PCB and therefore Nvidia was never able to respond with a dual-GPU offering of their own. Nvidia has overcome all those issues now, having also released their latest generation architecture with multiple GeForce 500 series products on offer today.
The GeForce GTX 580 is able to match the dual-GPU Radeon HD 5970 and even beat it in a number of areas. With the arrival of the Radeon HD 6990, AMD is hoping to take back the performance crown which they will likely hold until the GeForce GTX 590 arrives.
Under the hood, the Radeon HD 6990 is essentially a pair of underclocked Radeon HD 6970 GPUs on top of a massive PCB. This is the same of what we found with the Radeon HD 5970 which featured a pair of HD 5870 GPUs operating at slightly lower frequencies.
When designing the older Radeon HD 5970, AMD worked hard to keep within the PCI Express specification which calls for a maximum TDP of 300 watts. At 294 watts the HD 5970 just scrapped in and was still the most power hungry graphics card of its time. However, the new Radeon HD 6990 is much closer in specification to the HD 6970 than the HD 5970 was to the HD 5870. As a result the card's TPD rating has increased to a staggering 375 watts.
Perhaps for some that won't be the biggest shocker. The new Radeon 6990 is set to sell for $699, which translates in a $100 premium versus the HD 5970's price at launch. At this price point gamers can alternatively purchase a pair of Radeon HD 6970 cards and still save a few bucks while scoring a little extra performance. With that in mind, let’s check out the almighty Radeon HD 6990 in better detail...
 |
If the $350-380 price tag is too hard to justify, the Radeon HD 6870 and GeForce GTX 460 are more affordable alternatives in the $200-240 range, though they are quite a bit slower.
The previous generation Radeon HD 5870 offers a fraction more performance at around $280. Meanwhile, Nvidia is still offering the (not very efficient but speedy) GeForce GTX 470 for around $249, which makes the Radeon HD 5870 a tough sell these days.
Therefore, between the GeForce GTX 470 at $250 and the GeForce GTX 570 at $350 there is a $100 price gap with no real contenders -- at least until AMD launched their new $299 Radeon HD 6950 graphics card. At this price point, the Radeon HD 6950 has no direct competition, but we'll be keeping a close eye on how it compares to the GTX 470 and 570 and whether it makes more sense to scale down or up to get better value.
Much of the information covered here was already reviewed in our Radeon HD 6970 review earlier this month, so feel free to skip this page if you read that already.
The high-end "Cayman" architecture used by the Radeon HD 6900 series is a little different from Barts, which debuted with the Radeon HD 6870 and HD 6850 graphics cards.
Barts used the same VLIW5 configuration as the Radeon HD 5000 series, which features SIMD units with 4 simple and 1 complex stream processing unit.
The Radeon HD 6900 series, however, adopts a VLIW4 configuration, which features stream processing units arranged in groups of four along with general purpose registers. Although the four have equal capabilities, two (occupying 3 and 4 issue slots) are assigned with some special functions. AMD claims that VLIW4 configuration gives similar computational power as VLIW5, with 10% reduction in die area.
The Cayman GPUs feature a greater level of parallelization when compared to the Evergreen/Cypress architecture used by the Radeon HD 5800 series. Barts GPUs sit somewhere in the middle, as they were a step up from Cypress in assigning individual dispatch processors for each of the two SIMD Engine blocks.
The Cayman GPUs take things a step further with dual GPEs (Graphics Processing Engines) and assigning each to an SIMD engine block. What this means is that the Cayman architecture* now features two physical tessellation units whereas Barts only featured a single tessellation unit with improved efficiency when compared to the Radeon HD 5800 series as claimed by AMD.
What this means for the Radeon HD 6900 series is that tessellation performance can be improved quite significantly. In fact, AMD is claiming up to 3x the performance of the Radeon HD 5870 in these scenarios which, as you might remember, were the ones Nvidia liked to point out when claiming Fermi was architecturally superior than previous-generation Radeons.
In addition, the Cayman architecture also features reworked render backends consisting of 128 Z/Stencil ROPs, and 32 color ROPs, with up to 2x faster 16-bit integer operations and 2-4x faster 32-bit floating-point operations.
Note: For sake of simplicity we have referred to Cayman and Barts codenames that identify the high-end and mid-range desktop products in AMD's latest 6000-series GPU family (codenamed "Northern Islands" as a whole).
Just as its predecessor, the Radeon HD 6950 GPU has been fabricated using the 40nm process, yet AMD has squeezed in 486 million more transistors resulting in a die increase of 16%.
The GPU core is clocked at 800MHz, 9% lower than the Radeon HD 6970, while the GDDR5 memory also operates 9% slower at 1250MHz. Pairing that frequency with a 256-bit wide memory bus gives the Radeon HD 6950 a theoretical bandwidth of 160GB/s which, as you may have guessed, is 9% less than the HD 6970.
The Radeon HD 6950 also differs from the more expensive HD 6970 in its core configuration. Whereas the Radeon HD 6970 featured 1536 SPUs, 96 TAUs and 32 ROPs, the cut down HD 6950 packs 1408 SPUs, 88 TAUs, and the same 32 ROPs (8% less SPUs and TAUs).
Cooling the "Cayman Pro" GPU is a fairly large aluminum vapor chamber heatsink made up of 39 fins measuring 13.5cm long, 6.5cm wide, and 2.5cm tall. The vapor chamber design was first implemented in the Radeon HD 5970 and has recently also been adopted by Nvidia with their GeForce GTX 580 and GTX 570 graphics cards. Cooling the heatsink is a 75x20mm blower fan that draws air in from the case and pushes it out through the back.
For the most part, this fan operates very quietly, helped by the impressively low 20 watt idle consumption of the Radeon HD 6950. When gaming the fan will spin up as the card can consume up to 200 watts under load. This is ~6% more than the Radeon HD 5870, but even with the increased thermal stress, the fan didn't scream at intolerable levels.
The heatsink and fan have been enclosed within a custom-built housing that conceals the entire graphics card, a common practice for AMD when designing their most elite cards. We actually like this setup, as it helps protect the product very well. Nvidia has also done this in the past with their most prized graphics cards, such as the GeForce GTX 295 (dual GPU), though they have not used this design for their most recent flagship product, the GeForce GTX 580.
In order to feed the card enough power, AMD includes dual 6-pin PCI Express power connectors. This is the same setup used on the Radeon HD 5870, HD 6870 and GeForce GTX 580, GTX 570 graphics cards, so it's a configuration usually employed by high-end graphics cards.
Naturally, the Radeon HD 6950 supports Crossfire, so it has a pair of connectors to bridge two or more cards together. The only other connectors are on the I/O panel. Our AMD reference sample featured two dual DL-DVI connectors, a single HDMI 1.4a port, and two mini-DisplayPort 1.2 sockets.
It's worth noting that all Radeon HD 6950 graphics cards can support a max resolution of 2560x1600 on up to three monitors. With a multi-stream hub using the mini-DisplayPort 1.2 sockets, the card can power up to six monitors.
Using the new Futuremark 3Dmark 11 benchmark, the Radeon HD 6950 was 20% faster than the GeForce GTX 470 when running the extreme test. Should this be the case when running actual games this could be very good news for AMD.
The same test also reported that the Radeon HD 6950 was ~5% slower than the GeForce GTX 570 and 12% slower than the HD 6970 while it was just 2% faster than the HD 5870.
| Radeon HD 6850 | Radeon HD 5850 | GeForce GTX 470 | Radeon HD 6870 | Radeon HD 5870 | Radeon HD 6950 | GeForce GTX 570 | GeForce GTX 480 | Radeon HD 6970 | GeForce GTX 580 | Radeon HD 5970 | |
| Shader units | 960 | 1440 | 448 | 1120 | 1600 | 1408 | 480 | 480 | 1536 | 512 | 2x 1600 |
| ROPs | 32 | 32 | 40 | 32 | 32 | 32 | 40 | 48 | 32 | 48 | 2x 32 |
| GPU | Barts | Cypress | GF100 | Barts | Cypress | Cayman | GF110 | GF100 | Cayman | GF110 | 2x Cypress |
| Transistors | 1700M | 2154M | 3200M | 1700M | 2154M | 2640M | 3000M | 3200M | 2640M | 3000M | 2x 2154M |
| Memory Size | 1024 MB | 1024 MB | 1280 MB | 1024 MB | 1024 MB | 2048 MB | 1280 MB | 1536 MB | 2048 MB | 1536 MB | 2x 1024 MB |
| Memory Bus Width | 256 bit | 256 bit | 320 bit | 256 bit | 256 bit | 256 bit | 320 bit | 384 bit | 256 bit | 384 bit | 2x 256 bit |
| Core Clock | 775 MHz | 725 MHz | 607 MHz | 900 MHz | 850 MHz | 800 MHz | 732 MHz | 700 MHz | 880 MHz | 772 MHz | 725 MHz |
| Memory Clock | 1000 MHz | 1000 MHz | 837 MHz | 1050 MHz | 1200 MHz | 1250 MHz | 950 MHz | 924 MHz | 1375 MHz | 1002 MHz | 1000 MHz |
| Price | $180 | $260 | $260 | $240 | $360 | $300 | $330 | $450 | $370 | $500 | $580 |
