The GTX 980 Ti Poseidon Platinum from ASUS is the only card, which ships with hybrid watercooler from the factory. Therefore it’s possible to used this card either aircooled or watercooled. Apart from that there is a complex custom PCB with a 10 phase power design as well as a factory overclock. Overall we’re very curious how this card does in our review.
ASUS was amongst the first Nvidia add-in-card (AIC) partners who have decided to introduce a new custom GTX 980 Ti graphics card which is using NVIDIAs latest and greatest GM200-310-A Maxwell chip. For this graphics card, ASUS makes use of a custom PCB along with a new version of their DirectCU H2O cooler, which also supports watercooling. On the backside of the card, there is a good looking backplate. A quick glance at the specs also reveals that this card ships factory overclocked.
Browsing the specifications of this card we find 2816 CUDA cores, 176 TMUs and 96 ROPs. For comparison reasons, the GTX 980 features 2048 CUDA cores, 128 TMUs and 64 ROPs. This means that, compared to the smaller model, there are significantly more CUDA cores, TMUs and ROPs. Looking for reference clocks we find 1’141 MHz base clock and 1’229 MHz boost clock, whereas the NVIDIA reference cards run at 1000MHz/1075MHz, therefore there is a massive 14.3% factory overclock on the GPU.
As we already mentioned, ASUS decided to use a significant factory-overclock on their new GTX 980 Ti Poseidon. While the reference GTX 980 Ti is set to work at 1000 MHz for the base clock and 1075 MHz for the GPU boost clock, the GTX 980 Ti Poseidon runs at 1’141 MHz base clock and 1’229 MHz boost clock.
When it comes to memory clock speeds we see a slight, 200 MHz frequency boost, which makes the chips run at 7200 MHz. Combine this with a 384 Bit memory interface and you end up with a total of 336.5 Gigabyte of memory bandwidth.
Whereas the typical boost clock is set at 1’229 MHz, the maximum boost of 1’341 MHz was achieved quite easily and the GTX 980 Ti Poseidon held that clock most of the load time due to the capable DirectCU H2O cooler as well as good TDP target (nVidia’s Boost technology being power-based and not temperature-based on this card). At this point we also ran Furmark to see how high the card overclocks, when the GPU is under maximum load. This way we can determine a worst case scenario regarding maximum boost clock. With the card we see clocks speeds of 1151 MHz at 1.018v.
|ASUS GTX 980 Ti Poseidon Platinum||GeForce GTX 980Ti||GeForce GTX 980|
|Chip||GM200-310-A1 Maxwell||GM200-310-A1 Maxwell||GM204 Maxwell|
|Process||28 nm||28 nm||28 nm|
|Transistors||8.00 billion||8.00 billion||5.20 billion|
|GPU clock||1’114 MHz||1’000 MHz||1’126 MHz|
|GPU Boost clock||1’229 MHz||1’076 MHz||1’216 MHz|
|Memory GDDR5||6’144 MB||6’144 MB||3’072 MB|
|Memory clock||1’750 (7’000) MHz||1’750 (7’000) MHz||1’750 (7’000) MHz|
|Memory interface||384 Bit||384 Bit||256 Bit|
|Memory bandwidth||336’600 MB/s||336’600 MB/s||224’000 MB/s|
|Shader Cores||2’816||2’816||2’048 (12 SMX)|
|TDP||250 Watt||250 Watt||165 Watt|
|PCB Type||Custom||Reference Design||Reference Design|
|Cooler||DirectCU H2O||NVIDIA Reference||NVIDIA Reference|