News Headlines
- Tue, May 21
- Microsoft's Next-Generation Xbox Reveal: Liveblog and Discussion
- Xbox One officially announced, here are the facts
- EA teasing a new Need for Speed with image of a police supercar
- Check out the Call of Duty: Ghosts reveal trailer and Infinity Ward's behind-the-scenes tech video
- Xbox One games confirmed thus far, major publishers like EA, Square Enix and Ubisoft on board
New Articles
Related Articles
The GTX 680 is designed to fit into the enthusiast market and stands above all other single GPU graphics cards currently available, even the HD 7970. Of course, the Kepler series is going to have a full range of graphics cards from low to high-end, all of which will be built on the Kepler architecture.
Aesthetically, the GTX 680 actually looks extremely similar to graphics cards in the 500 series. As you can see, the body still has a predominantly black color scheme and uses a rear mounted blower style fan with a concave surface area. Even though the overall look has not been completely redesigned, the visual elements remain appealing and the board definitely looks high-end. The dimensions of the graphics card are slightly smaller than that of the HD 7870, as the GTX 680 measures roughly 11.5-inches long, while the HD 7970 is 11.75-inches.
The GTX 680 is Nvidia's first graphics card built on a 28nm fabrication process, and surprisingly the die is considerably smaller than the AMD's Tahiti GPU. As it stands, the GTX 680 has a die size of around 295mm² and packs in 3.54 billion transistors. By comparison, the Tahiti GPU is 365mm² and has roughly 4.31 billion transistors. With the smaller node, Nvidia was able clock the GTX 680 above 1GHz, as it has a base of 1006MHz and a Boost clock of 1058MHz. However, as stated before, the Boost Clock can go higher than the target frequency. Additionally, the GTX 680 includes 4 GPCs giving it a grand total of 8 Streaming Multiprocessors, 1536 CUDA cores, 32 ROPs, and 128 Texture units.
As for the memory, Nvidia has decided not to go with a larger interface, as the GTX 680 utilizes a 256-bit wide GDDR5 memory bus. However, the memory has been optimized to run at extremely high frequencies, so even with the smaller memory interface the total bandwidth still achieves a high rating of 192.26 GB/s. This is only achievable because the memory is clocked at 1500MHz (6000MHz effective), which is currently the fastest memory frequency on a graphics card. However, the total memory throughput is still considerably lower than the 264GB/s bandwidth of the HD 7970.
The back of the PCB is clean for a high-end graphics card, but there are a few interesting features to note. First off we can see the power regulation circuitry runs vertically instead of the standard horizontal positioning. The position of the VRM was actually designed to optimize the board space, and position the fan slightly off center. The back of the PCB also shows off the dual SLI connection points, meaning the GTX 680 is scalable up to 4-way SLI. The GTX 680 also runs on a Gen 3.0 PCIe interface which doubles the maximum data rate over Gen 2.0, giving the card up to 32 GB/s of bi-directional bandwidth on an x16 connector.
The GTX 680 was designed to offer best-in-class performance-per-watt and with a power rating of only 170 watts, we would say mission accomplished. Let’s again compare the GTX 680 to its closest competitor, the HD 7970. In the best case scenario the HD 7970 will drain around 220 watts of power, and has an 8+6 pin power configuration to ensure the board can continuously run stable. The HD 680 on the other hand only requires dual 6-pin connectors. This is an impressive feat for a high-end graphics cards, and according to our math the GTX 680 is just shy of being 28% more efficient than the HD 7970.
Another interesting aspect of the power design is the stacked 6-pin connectors. This design, like rearranging the power circuitry, allows Nvidia to optimize the board space which in turn reduces the size of the PCB and the total length of the graphics cards.
The video outputs on the GTX 680 have been completely reworked to support the expanded 3D Vision support. In total there are two DL-DVI ports, a single HDMI port and a full-sized DisplayPort. Out of the ports, both the DisplayPort and DVI connections can support resolutions of up to 2560x1600, while the HDMI port is capable of supporting resolutions of up to 1080p and comes with native support for all the latest HDMI 1.4a features.
Article Index |
|
~51fps in bf3 @ 2560 x 1600 w/ 4xAA & 16xAF? thats a big meatball!
and theres enough performance left to force all those other fancy features, like transparency AA.
when you got it, nvidia, you got it.
the next generations of GPUs will be very interesting. i dont think the gtx 680 will be anywhere near quite as powerful to max out next-gen console games as i would bf3 (2560x1600 & beyond, bunches of AA, big ol' AF, etc) assuming directx doesnt have any major efficiency reworks. its a fun ride.
either i wait and see what the 690 is like (and its cost!), or ill probably have to settle for eyefinity on the 7970, and sacrifice the CUDA that Folding@Home greatly benefits from.
Any news on how much bandwidth the 7970 and 680 need? As far as I'm aware PCIe 2.1 x16 hasn't yet been saturated yet.
AVP matched the 7970 with both at stock levels.
Only 8.6% faster at the highest settings in Arkham City on with both at stock clocks.
Smokes the 7970 by 18.6% at best in Battlefield.
Loses to the 7970 in Crysis by 21.6% at the highest settings.
Up to 24% faster than the 7970 in Dirt three in the middle tier settings. Around 18% in the others.
Loses by 15-25% in metro compared to the 7970. And Metro is not an AMD biased or optimized game. Until the 6xxx cards came out AMD always lost by a wide margin. The 7970 either is a better architecture for the game or currently has better drivers. I have no doubt the 680's performance could have been better with the game as it's a new architecture and could probably use some driver work.
The 680 also loses in Total War between 10 and 17%.
On average that makes the 680 about 5% slower at the highest settings.
Like I said, hardly a 7970 killer, very efficient architecture and it certainly trades blows the AMD but it's not nearly 10-15% faster across the board as stated in the article. I really have to question whether you did the math or just eyeballed it.
The overclocking of the 680 was 15.5% core and about 19% vram.
The 7970 on the other hand had a gpu OC of 21.6% (925mhz to 1125 mhz) and about 14.5% vram OC. So the overclocking is kind of 'meh' compared to it. It's a nice increase no doubt but it's nothing to write home about. The 7970 is able to go past the limits set in CCC in the 7970 review but the gtx 680 actually capped out before its software limits, whether that's due to power or physical limitations of the architecture at that voltage I don't know though.
The pricing and efficiency is what makes this a great card, not the raw performance.
Perspective is important.
to play next-gen console ports, your compy will need to be considerably more powerful to run 'em. you want to notch up the resolution, add mods and force spiffy gfx options? be ready to throw a ton more power at it.
and emulation is a whole 'nother story.
just sayin' by the time next-gen console games are ported to PCs, a gtx 680 probably wont be too relevant.
Weird how it can sometimes have negative scaling (if that is the correct usage of this somewhat technical term).
Thanks for the response on our GTX 680 review. I noticed this is the second time you have mentioned not being sure of the reliability of our scores. Can you tell use what you would like to see different in the way we approach and analyses our benchmarking results ?
Thanks