7870 XT Review – Not long ago, we sought out to find the best video card for around $200-250. We compared the GeForce 660 to the Radeon 7870. Nvidia’s original target for the 660 was the Radeon 7850, but due to price cuts, this was the comparison that had to be made. In the end, the 7870’s superior horsepower allowed it to become our overall recommendation in the price range.
However, some readers pointed out that there is another contender in the same price range. The confusingly named Radeon 7870 XT is frequently (but not always) available, and despite having a nearly identical name, the graphics card’s core is vastly different from the plain Radeon 7870.
What is Radeon 7870 XT?
Seemingly made available only to a few select vendors such as Sapphire, the 7870 XT does not use the same Pictairn core as the Radeon 7870. Instead, it uses a stripped down version of the Tahiti core, dubbed Tahiti LE. This is the core that is used on AMD’s 7950 and above series of cards, which normally sell for at least $330 – nearly $100 more than the normal 7870.
So how do they compare on paper? Have a look:
|Clock rate (MHz)|
|Fill Rate (G/s)|
|GFLOPS (Single-precision)||Memory Bus (bits)||Mem Bandwidth (GB/s)||TDP (W)
(Radeon 7870 XT)
Notes: "Config Core" = Unified Shaders : Texture mapping unit : Render Output unit
GeForce architecture cannot be directly compared - reference only
As you can see, Tahiti LE is a detuned version of the Tahiti core used in the 7950 and above. Any 7950 core that is fabricated that cannot run at full spec gets binned and sold to Sapphire as Tahiti LE.
Shader count had been dropped from 1792 to 1536, and texture mapping units have dropped from 112 to 96. The memory bus width is also lowered to 256 bit from 384, so it suffers in the memory bandwidth department as well. AMD were able to increase the clock rate in boost mode, all the way up to 975 MHz from 800 MHz on the 7950. Thich helps move things along, but with less hardware on the chip, it is overall slower than the 7950 by quite a bit.
On the other hand, the core improves upon the Pictairn Pro core used in the 7870 due to the presence of more shaders and texture units, and higher memory speed. It does use more power than Pictairn, but on paper at least, it should be faster in most aspects. That’s what we’re here to find out!
Sapphire Radeon 7870 XT With Boost
If you were to search for a Radeon 7870 XT to buy for yourself, pretty much the only model you’ll come up with is the Sapphire version we’re reviewing today. You may come across one from PowerColor, and I have seen some from the European brand “Club3D” listed, but 9 times out of 10, you’ll come across the Sapphire version, in North America and Asia at least.
At a quite long 275mm (10.8″) in length, you can tell that the heritage of the Sapphire 7870 XT indeed lies in the 7900 series, and not the 7800 series it gets its name from. The cooler is more than adequate, with four copper heat pipes to distribute the heat to the fins, to be cooled by the fans.
The 7870 XT is carries a TDP of 185W, and is powered by a pair of 6-pin PCI-E connectors. With no need for 8-pin connectors, it can be used with older-yet-still-functioning power supplies.
Supported video ports are: DVI, HDMI, and a pair of mini-DisplayPort headers.
Sapphire goes light on the accessories, including a pair of Molex > PCI-E adapters, a DVI > VGA adapter, and an adapter that converts one of the mini-DisplayPorts to full DisplayPort. You do get a copy of Sapphire’s Trixx software, but I think most people will stick to more familiar software.
However, it is still fully supported by AMD’s excellent Never Settle Promo, which gives you a key for some very good games (depending on which region). In North America, you’ll get BioShock Infinite (which we’re giving away in this review!) and Tomb Raider.
Our Testing Methods – Frame Times
If you have been reading our video card reviews in the last year, you’ll already know that we moved away from FPS as the end-all indicator of gaming performance. As soon as Scott Wasson at Tech Report let the cat out of the bag that consistent frame time is more important, we made the switch.
If you weren’t aware then, you probably are now; last week, Nvidia provided the 4 biggest hardware sites with “FCAT” hardware that helps record frame times at the monitor level. The reason for this is because the tool we use to record frame times – FRAPS – records its numbers at the beginning of each frame, rather than the end. Because game engines perform quite a few functions after this point, further issues or improvements may occur throughout the pipeline.
Since Scott was the only one who was using FRAPS all along, he was the one who took the time to directly compare his FRAPS results to the new Nvidia FCAT results. In the end, FCAT indeed shows ‘tighter’ lines in single GPU results, so it looks like game engines and drivers do improve on frame times after FRAPS takes its recording. In single GPU testing though, the results are comparable and sometimes identical, such as in the case of Unreal engine games. In dual GPU results, FCAT shows that some games do NOT handle Crossfire very well, and end up with a very sloppy experience despite the high framerate. FRAPS doesn’t show this well though, so is not nearly as useful for dual GPU testing.
Therefore, we will continue to use FRAPS for single GPU testing. I think that even with the lack of Nvidia’s FCAT tools, FRAPS is still far superior to frame rate results.
I should also mention here that I have noticed some confusion with regards to the “Framerate over time” results that some sites use. This is not to be confused with “Frametime” or “Frame Time” despite the similar phrasing. Framerate over time still takes an average frame rate each second, dropping a lot of data that is useful. For instance with Skyrim running at 80 fps, a 1 minute benchmark will result in about 5000 frames. Using frame time, the chart will therefore show 5000 points of data, while a “framerate over time” chart will show 60 points of data. The latter will show where slowdowns occur through the run, but the former will show that and exactly how efficiently frames are being delivered.
Radeon 7870 XT Game Tests
|CPU||Intel Core i7 3770K (Review)|
|Motherboard||Gigabyte Z77X-UD5H Motherboard|
|GPU Drivers||AMD: 13.1, 13.2b7
|Memory||8GB Kingston HyperX Genesis @ 9-9-9-27 (Review)|
|OS||Windows 8 Professional x64 RTM|
|Test Notes||CPU Speed Locked at 100%|
We wanted to include some of the most modern games available from various genres, while at the same time testing as many different engines as possible. It is not as useful to test the same engine over and over again, since results will likely be the same. Here’s the list of games we uses for testing the Radeon 7870 XT:
|Assassin's Creed 3||3rd Person Action||1.03||AnvilNext|
|Batman: Arkham City||3rd Person Action||GOTY||Unreal Engine 3|
|Battlefield 3||First Person Shooter||1.04||Frostbite 2|
|Crysis 3||First Person Shooter||188.8.131.520||CryEngine 3|
|Far Cry 3||First Person Shooter||1.04||Dunia Engine 2|
|Skyrim||First Person RPG||184.108.40.206||Creation Engine|
|Sleeping Dogs||3rd Person Action||1.8||Sleeping Dogs|
|Witcher 2||3rd Person RPG||3.3||RED Engine|
Radeon 7870XT GPGPU Tests
Another growing concern of the video card market is General Purpose GPU computing. This refers to using the GPU to perform operations other than gaming, and can include video encoding, 3d rendering, physics calculation, protein folding, and more. We will be considering three areas that a casual user would be most interested in – Video encoding, 3D rendering, and distributed computing (including BitCoin mining and [email protected]).
We have a lot to cover, so let’s get right to it!
Bioshock Infinite Giveaway
We are giving away a Steam key for Bioshock Infinite to one lucky reader of this review! To enter, all you have to do is “like” this article using the Facebook button at the bottom of this page (beside “Share This”).
You can get an extra entry by liking us on Facebook as well! All current Facebook followers will automatically be entered, as well as those who “liked” our two previous articles before midnight April 3, 2013 Japan Standard Time.
Winners will be contacted via Facebook. Contest ends next week!