Editing GeForce (section)

=== {{Anchor|NVLINK|PASCAL}} GeForce 10 series ===
{{Main|GeForce 10 series}}
In March 2014, Nvidia announced that the successor to Maxwell would be the [[Pascal (microarchitecture)|Pascal microarchitecture]]; announced on May 6, 2016, and were released several weeks later on May 27 and June 10, respectively. Architectural improvements include the following:<ref name="nvidia-blog-20140325">{{cite web | last = Gupta | first = Sumit | url = http://blogs.nvidia.com/blog/2014/03/25/gpu-roadmap-pascal/ | title = NVIDIA Updates GPU Roadmap; Announces Pascal | publisher = Blogs.nvidia.com | date = 2014-03-21 | access-date = 2014-03-25 | url-status = live | archive-url = http://archive.wikiwix.com/cache/20140325074350/http://blogs.nvidia.com/blog/2014/03/25/gpu-roadmap-pascal/ | archive-date = March 25, 2014 | df = mdy-all }}</ref><ref>{{cite web | url = http://devblogs.nvidia.com/parallelforall/ | title = Parallel Forall | publisher = Devblogs.nvidia.com | work = NVIDIA Developer Zone | access-date = 2014-03-25 | url-status = dead | archive-url = https://web.archive.org/web/20140326025738/http://devblogs.nvidia.com/parallelforall/ | archive-date = March 26, 2014 | df = mdy-all }}</ref>
* In Pascal, an SM (streaming multiprocessor) consists of 128 CUDA cores. Kepler packed 192, Fermi 32 and Tesla only 8 CUDA cores into an SM; the GP100 SM is partitioned into two processing blocks, each having 32 single-precision CUDA Cores, an instruction buffer, a warp scheduler, 2 texture mapping units and 2 dispatch units.
* [[GDDR5 SDRAM#GDDR5X|GDDR5X]]{{snd}}New memory standard supporting 10&nbsp;Gbit/s data rates and an updated memory controller. Only the Nvidia Titan X (and Titan Xp), GTX 1080, GTX 1080 Ti, and GTX 1060 (6&nbsp;GB Version) support GDDR5X. The GTX 1070 Ti, GTX 1070, GTX 1060 (3&nbsp;GB version), GTX 1050 Ti, and GTX 1050 use GDDR5.<ref>{{cite web|url=http://www.geforce.com/hardware/10series|title=GEFORCE GTX 10 SERIES|website=www.geforce.com|access-date=April 24, 2018|url-status=live|archive-url=https://web.archive.org/web/20161128103547/http://www.geforce.com/hardware/10series|archive-date=November 28, 2016|df=mdy-all}}</ref>
* Unified memory{{snd}}A memory architecture, where the CPU and GPU can access both main system memory and memory on the graphics card with the help of a technology called "Page Migration Engine".
* [[NVLink]]{{snd}}A high-bandwidth bus between the CPU and GPU, and between multiple GPUs. Allows much higher transfer speeds than those achievable by using PCI Express; estimated to provide between 80 and 200&nbsp;GB/s.<ref>{{cite web | url = https://devblogs.nvidia.com/parallelforall/inside-pascal/ | title = nside Pascal: NVIDIA's Newest Computing Platform | date = 2016-04-05 | url-status = live | archive-url = https://web.archive.org/web/20170507110037/https://devblogs.nvidia.com/parallelforall/inside-pascal/ | archive-date = May 7, 2017 | df = mdy-all }}</ref><ref>{{cite web | url = http://devblogs.nvidia.com/parallelforall/nvlink-pascal-stacked-memory-feeding-appetite-big-data/ | title = NVLink, Pascal and Stacked Memory: Feeding the Appetite for Big Data | date = 2014-03-25 | access-date = 2014-07-07 | author = Denis Foley | website = nvidia.com | url-status = live | archive-url = https://web.archive.org/web/20140720130522/http://devblogs.nvidia.com/parallelforall/nvlink-pascal-stacked-memory-feeding-appetite-big-data/ | archive-date = July 20, 2014 | df = mdy-all }}</ref>
* 16-bit ([[Half-precision floating-point format|FP16]]) floating-point operations can be executed at twice the rate of 32-bit floating-point operations ("single precision")<ref>{{cite web | title = NVIDIA's Next-Gen Pascal GPU Architecture to Provide 10X Speedup for Deep Learning Apps | url = http://blogs.nvidia.com/blog/2015/03/17/pascal/ | website = The Official NVIDIA Blog | access-date = 23 March 2015 | url-status = live | archive-url = https://web.archive.org/web/20150402135434/http://blogs.nvidia.com/blog/2015/03/17/pascal/ | archive-date = April 2, 2015 | df = mdy-all }}</ref> and 64-bit floating-point operations ("double precision") executed at half the rate of 32-bit floating point operations (Maxwell 1/32 rate).<ref>{{cite news | last1 = Smith | first1 = Ryan | date = 2015-03-17 | title = The NVIDIA GeForce GTX Titan X Review | url = http://www.anandtech.com/show/9059/the-nvidia-geforce-gtx-titan-x-review/2 | newspaper = [[AnandTech]] | page = 2 | access-date = 2016-04-22 | quote = ...puny native FP64 rate of just 1/32 | url-status = live | archive-url = https://web.archive.org/web/20160505122454/http://www.anandtech.com/show/9059/the-nvidia-geforce-gtx-titan-x-review/2 | archive-date = May 5, 2016 | df = mdy-all }}</ref>
* More advanced process node, TSMC 16mm instead of the older TSMC [[32 nm process|28 nm]]