MSI GeForce GTX 1070 Ti Titanium 8 GB Review – Twin Frozr VI Now In Quick Silver Colors
MSI GeForce GTX 1070 Ti Titanium2nd November, 2017
NVIDIA Embraces FinFET With Pascal Graphics Architecture
The Pascal GPUs were designed to deliver a generational leap in performance over their predecessors by embracing the latest FinFET process. The increasing demand for graphics performance in high-end gaming PCs and also the emergence of the virtual reality market have opted graphics manufacturers to focus on next generation DirectX 12 and Vulkan graphics while delivering increased performance to drive high-resolution, HDR capable displays.
Using the TSMC 16nm FinFET process, the Pascal GP104 GPU was crafted for GeForce gamers and offered three products that include the GeForce GTX 1080, GeForce GTX 1070 and today's GTX 1070 Ti. Right now, we are going to give you a run through of the entire GP104 GPU which is the heart of the GeForce GTX 1070 Ti graphics card.
NVIDIA GP104-300 GPU Dissected - A Beefier GTX 1070 That Still Lacks G5X Memory
The GeForce GTX 1070 Ti uses a GP104 SKU that falls in between the GTX 1070 and GTX 1080. It's not fully enabled, but with just one SM disabled, it's very closer to the GTX 1080 than the 1070. The NVIDIA GP104 Pascal GPU packs 7.2 Billion transistors and has 4 graphics processing clusters, all of which are enabled. This adds up to a total of 19 SM units with 128 cores each. The card is geared to power compute and memory hungry gaming applications at higher resolutions with tons of texture and geometry processing power.
The GeForce GTX 1070 Ti features 2432 CUDA Cores, 152 Texture Mapping Units and 64 ROPs. These are clocked at a base clock of 1607 MHz and boost clock of 1683 MHz which can go as high as 2 GHz with overclocking. The Pascal cards are built to overclock like crazy and custom models will further boost the clock rates when overclocked manually as NVIDIA has restricted AIBs from adding custom factory overclock BIOS's in their cards.
The Pascal FinFET architecture allows for a significant increase in clock frequency. The reference GeForce GTX 1070 Ti is clocked at 1607 MHz base and 1683 MHz boost clock. The model MSI sent us actually clocks much higher than the frequencies mentioned above due to GPU Boost 3.0.
The GeForce GTX 1070 Ti delivers 8.1 TFLOPs of compute performance in its stock configuration. All of this circuitry is housed in a 314mm2 die size which is impressive. We have all the numbers in our reviews mentioned later in our review.
The card also features 8 GB GDDR5 VRAM that run across a 256-bit bus interface. NVIDIA is still using the 8 Gbps GDDR5 chips on the GTX 1070 Ti instead of either G5X or 9 Gbps GDDR5 chips like the GTX 1060 uses now. There's still an ample amount of bandwidth available on the card at 256 GB/s. The card is powered by a single 8 pin power connector and has a rated TDP of 180W.
With the new compression and tiled caching system, the bandwidth on GTX 1070 Ti can be boosted beyond its full available limit.
In addition to better performance, NVIDIA’s GeForce GTX 1070 Ti is also packing new technology features such as compression and tiled caching boost bandwidth, tiled rendering, DirectX 12 Flex and Flow, NVIDIA Aftermath (a new tool to identify the causes of GPU crashes) and ShadowPlay Highlights which will further boost gamer experiences running GeForce hardware.
Faster clocked GDDR5 memory still offers great performance and loads of bandwidth to the card. There are tons of features enabled on the Pascal cards such as Compute Preemption, memory compression, simultaneous multi-projection which you can read about in our detailed post here. Also, NVIDIA has launched the full fat configuration of the GP102 GPU on the latest NVIDIA Titan Xp graphics card which was launched a few months back.