(GPU) NVIDIA GeForce GTX 1080 Ti

(Profile updated as of 30th October 2019)

Here is my GPU profile for the GeForce GTX 1080 Ti, NVIDIA's pre-emptive strike against AMD's Vega-based cards, launched in 2017. This card has achieved 'legendary' status as it still is one of the most powerful video cards available 2 years later.

(click for full images).

(Picture 1) The silicon die of GP102-350, of the G TX 1080 Ti graphics card. Note the single missing DRAM solder point, which would be wired to the disable memory controller on the silicon die. This image is from TechPowerUp's review of the Founder Edition GTX 1080 Ti.

(Picture 2) The architectural block-diagram for GP102-350. Two Streaming Multi-processors, a 32-bit GDDR5X PHY (And its associated DRAM), a ROP partition (8 pixel/clock) and 256 KB of L2$ have been disabled.

(Picture 3) Actual silicon die-shot of the GP102 die, I have done my best to highlight an example of the blocks on the die that could be laser disabled to allow this silicon to be binned as a GTX 1080 Ti. The image credit for the original Die Shot goes to Fritzchen Fritz, do check out his Flickr page.

(GPU) NVIDIA GeForce GTX 1080 Ti
(GPU) NVIDIA GeForce GTX 1080 Ti
(GPU) NVIDIA GeForce GTX 1080 Ti

Graphics Card Information

Graphics Card: NVIDIA GeForce GTX 1080 Ti

Graphics Card Manufacturer: NVIDIA

Graphics Card Release Date: March 10, 2017

Graphics Card MSRP: $699 USD

Graphics Processor Codename: GP102-350

Graphics Processor Manufacturer: NVIDIA

Graphics Processor Implementation: Cut die

Graphics Interface: PCI-E 16x Gen3

Architecture: Pascal (Client)

Lithography Process: TSMC 16nm FinFET

Approximate die size: 471mm²

SashleyCat's GPU die Size Rating: large

Approximate Transistor Count: 12,000 Million

Approximate Transistor Density: 25.4 Million / Square Milimetre




GPU Features

Double-speed FP16 Shading: No (1:64 FP32)

Asynchronous Compute Capability: Limited (Fast Context Switching) *

DirectX Hardware Support: DX12.1 (FL 12_1)

Dedicated DXR Acelleration on chip: No

Variable-rate Shading: No

Adv. Geometry shading: No

Adv. Geometry shading (Programmable/DX12 Mesh Shaders): No

AI/ML Acceleration: No

Advanced Memory Management: No

Integer and Float Shader Co-execution: No

Tile-based Renderer: Yes


GPU Computing Resources

GPU Substructures: 6 Graphics Processing Clusters, 28 Texture Processing Clusters (30 Full Chip)

Graphics Cores: 28 Streaming Multi-processors (30 Full Chip)

Graphics Cores per Substructure: 1 per TPC, 4 x GPC with 5, 2 x GPC with 4*

Total Stream Processors (ALU/Shaders): 3584 (3840 Full Chip)

Stream Processors per Graphics Core: 128

Graphics Core SIMD Structure: 4 x 32

Total Special Execution Units: 896 Special Function Units (960 Full Chip), 112 FP64 CUDA Cores (120 Full Chip), 28 FP16x2 CUDA Cores (30 Full Chip), 896 Load/Store Units (960 Full Chip))

Special Execution Units per Graphics Core: 32 Special Function Units, 4 FP64 CUDA Cores, 1 FP16x2 CUDA Core, 32 Load/Store Units

Total Texturing Units: 224 (240 Full Chip)

Texturing Units per Graphics Core: 8

Pixel Pipelines (ROPs): 88 (11x ROP Partitions with 8 Pixels per clock) (96 (12x 8 ROP) full chip))

Level 2 shared on-chip cache: 2816 KB (3072 KB Full Chip)

Geometry/Tessellation Processors: 28 (30 Full Chip)

Raster Engines: 6


GPU Memory Subsystem

Graphics Memory Type: GDDR5X

Graphics Memory Standard Capacity: 11,264 MB

Graphics Memory Composition: 11 x 1024 MB GDDR5X SDRAM Chips

Graphics Memory Access Granularity: 32-bit (4 bytes)

Graphics Memory Standard Clock Speed / Data Rate: 1375 MHz / 11,000 MHz

Graphics Memory Full Interface Width: 352-bit (44 bytes per clock) (384-bit / 48 bytes per clock Full Chip)

Graphics Memory Peak Memory Bandwidth: 484 GB/s


GPU Frequency and Peak performance

Graphics Engine Clock: 1582 MHz+ *

GPU Computing Power FP16: 177,184 Million operations per second with FMA

GPU Computing Power FP32: 11,339,776‬‬ Million operations per second with FMA

GPU Computing Power FP64: 354,368 Million operations per second with FMA

GPU Texturing Rate INT8: 354,368 Million texels per second

GPU Texturing Rate FP16: 354,368 Million texels per second

GPU Pixel Rate: 139,216 Million pixels per second

GPU Primitive Rate: 9,492‬ Million triangles per second *


GPU Thermal and Power

Standard Cooling Solution: Blower cooler with vapour chamber

Typical Board Power: 250 W

Maximum Board Power: 300 W *

Maximum Allowed Junction Temperature (TJ Max): 91*C




Graphics Card description

GeForce GTX 1080 Ti launched on the 10th of March, 2019. This video card was a pre-emptive counter to AMD's impending re-entry to the high-end GPU market, with the Vega-based cards that launched in August of the same year. The Graphics card uses the 'GP102' silicon, the first time such a designation has been given to a GPU from NVIDIA. This silicon is essentially the 32-bit Compute cability and rasterisation rate of the larger, HPC-focused GP100 chip, but without the dedicated FP64 CUDA cores and expensive HBM2 memory subsystem. The GP102, like the GP104 before it, uses a different SM design to that processor, also, in which the chip is more similar to previous-generation 'Maxwell' processors in its layout.



The GeForce GTX 1080 Ti has a salvaged back-end, with a ROP partition and a block of its associated L2 cache disabled, along with the 32-bit PHY and DRAM chip that connected to it. As result, it has fewer ROPs, L2 cache, memory bus width and capacity, than the TITAN X (Pascal), a video card that is otherwise comparable in performance and design, the former due to the elevated operating clock rates on the GTX 1080 Ti.


Graphics Card approximate 3D Performance

Sashleycat gaming performance rating (2019): Great for 4K high settings 60 Hz, or 1440p maximum settings and high refresh



NVIDIA's GTX 1080 Ti provides high performance as of writing. It performs slightly behind the current-generation RTX 2080 Super, by about 10%, more or less depending on the game, but offers more video memory (11GB vs 8GB). Performance is as a result, similar to the older RTX 2080 (non Super) and Radeon VII video cards, which themselves are in the same league as the current Radeon RX 5700 XT, and RTX 2070 Super. Overall, there isn't much to seperate these cards in normal gaming performance, but the GTX 1080 Ti remains one of the most powerful video cards on the market, 2 years later, and with the 11 GB of video memory, is set to be quite ready for next-gen games with high resolution textures. Though a keen reader will notice the lack of advanced GPU features such as double-speed FP16 shading. This could hamper the GTX 1080 Ti in truly next-gen engines, but the effects of which are yet to be seen. This card has a legendary status to me similar to its predecessor, the GeForce GTX 980 Ti.



As of writing, the GTX 1080 Ti provides ample performance paired with a 4K monitor, high or maximum settings and playable FPS, although often short of 60 frames per second. Alternatively, Ultra-wide resolutions such as 3440x1440, or QHD 1440p, are dealt with easily, the latter of which this card could likely provide a high-refresh gaming experience.


Notes

Graphics Engine Clock

GPU Boost on Nvidia cards of this generation results in higher observed core clock speeds in actual games. The 1080 Ti will often run at upwards of 1700 MHz if power or thermal limits are not reached. The number here is the NVIDIA-spec boost rating. Please note all the peak performance figures are using this base spec number. Actual performance will be higher but it's impossible to exactly quantify what each card will boost to.



GPU Primitive Rate

Raw triangle output based on my understanding of the Raster Engines. PolyMorph engines attached to each TPC may have an effect on total triangles rastered.



Ayschronous Compute Capability

As far as I know Pascal GPUs are unable to execute graphics and compute tasks asychronously. However, I am listing it as "Limited" due to this architecture's ability to switch between both tasks very quickly and this allows Pascal to still do very well in frames requiring both types of workload. This is still effective for Pascal. To my knowledge Pascal would not have benefitted a lot, if at all, from true Asynchronous Compute Capability



Stream Processors per Graphics Core

If you count the very small number of Dedicated FP64 and FP16x2 Cores this would be higher but for simplicity's sake they are listed as special execution units not towards the Core's full ALU compliment (that are used in games).



Maximum Board Power

I believe GTX 1080 Ti Founder's Edition has a +20% power target. That would allow up to 300 Watts maximum board power.



Misc.

This bit is for my personal opinion on this Graphics card / Graphics processor



Sashleycat's Awesomeness Rating: Legendary.

©2020 by Sashleycat (Read the Legal stuff)