So Nvidia has launched its flagship GTX 1080 a few days ago at DreamHack, but the company failed to reveal any official information regarding the card or its several variants. However, specifications and stock performance numbers of the card have already been leaked, and all we wanted to know now was the number of different versions that the GTX 1080 has. Well, thanks to a report from VR-World.com, we have got some exciting information for all Nvidia fans.
There are Four Variants of the GTX 1080
The new Nvidia Geforce GTX 1080 has got a total of four variants, including a liquid cooled one that can reach a clock rate of up to 2.50 GHz! There’s a Founders Edition, which is a slightly modified version of the Reference Edition. Then there’s an air cooled Plain AIB version and an air cooled Custom AIB version. Lastly, there’s a water cooled Custom AIB version.
Talking about the Founder’s Edition, Nvidia gave a new name to the Reference Edition along with a few bells and whistles. The Founder’s Edition gets an improved power circuitry, a better design, and a BIOS that does not mind overclocking and allows the clock to attain the 2.0 GHz mark with air cooling. On stock, the card clocks 1.66 Ghz, and it reaches 1.73 GHz with the turbo mode on. Do note that the card has a single 8-pin connector, so overclocking overhead won’t happen as the card can only squeeze out up to 225W from PSU and PCI-e together.
Next up is the plan AIB version, which costs $100 lesser than the Founder’s Edition and comes with lesser gimmicks. The familiar reference cooler isn’t there anymore and you can choose a custom cooler of your choice of the particular AIB. You can get this variant if you want to get more out of your money and are okay with losing some overclocking.
The custom AIB variants can get clock rates of 2.1 GHz and above, and graphics cards manufacturers like Gigabyte and MSI are already trying to work on a method that can push more power to the GTX 1080. But what’s really worth talking about is the liquid cooled custom AIB variant that’s going to get a better power supply system and can get clock rates of up to 2.5 GHz, which will be a huge improvement over this card’s predecessor. The card clocked at that rate can easily handle 4K/60fps.
| GTX 980 Ti | GTX 980 | GTX 1080 | GTX 1070 | GTX 1060 | TESLA P100 (GP100) | |
| GPU | GM200 | GM204 | GP104 | GP104 | GP106 | GP100 |
| Process Node | 28nm | 28nm | 16nm FinFET | 16nm FinFET | 16nm FinFET | 16nm FinFET |
| Transistors | 8 Billion | 5.2 Billion | TBA | TBA | TBA | 15.3 Billion |
| CUDA Cores | 2816 CUDA Cores | 2048 CUDA Cores | ~2560 CUDA Cores | ~2048 CUDA Cores | ~1280 CUDA Cores | 3840 CUDA Cores |
| VRAM | 6 GB GDDR5 | 4 GB GDDR5 | 8 GB GDDR5X | 8 GB GDDR5 | 4 GB GDDR5 | 16GB HBM2 |
| Memory Bus | 384-bit | 256-bit | 256-bit | 256-bit | 128-bit | 4096-bit |
| Memory Speed | 7Gbps | 7Gbps | 10-12Gbps | 8Gbps | 8Gbps | 1.4Gbps |
| Bandwidth | 336GB/s | 224GB/s | 384-320GB/s | 256GB/s | 128GB/s | 720GB/s |
| TDP | 250W | 165W | TBA | TBA | TBA | 300W |
| Launch Date | May 2015 | September 2014 | June 2016 | June 2016 | Autumn 2016 | 2017 |
GTX 1080 Beats the P100 By 20%!
P100 has got 3584 CUDA Cores and is itself based on GP100. The P100 turbo clocks at 1480 MHz. It offers a total peak single precision performance of 10.6 TFLOPs. The Pascal architecture is the same here, so we can compare the P100 to the GTX that is based on the GP104.
The GTX 1080 gets 2560 CUDA cores, but do note that it can reach the 2.1 GHz mark on air cooling and 2.5 GHz on liquid cooling. The GTX 1090 can make a single precision performance of 10.7 TFLOPs at 2.1 Ghz. And when running at 2.5 Ghz, it will make 12.8 TFLOPs, thus defeating the P100 badly. That’s around 20% faster and is a great thing to know for those who have been waiting for the GTX 1080.
Also, remember that the GTX 1080 has got GDDR5X memory and is made with 16nm FinFET+ process, which tells us that the card will surely break the barriers. It will at least get you 2.0 GHz+ speeds directly out of the box.