The Geforce 10 Series is a series of graphics chips from Nvidia and a successor to the Geforce 900 series. All graphics processors in this series are based on the Pascal architecture and support the shader model 5.0 (SM 5.0) to DirectX 12.1.
GeForce GTX 1080 Ti, Titan X and Titan Xp
On Aug. 2, 2016 Nvidia introduced the successor of the Titanium X GTX Titan X ago. Since both maps are very similar to the names (they differ only by omitting the “Geforce GTX” suffix, which is curiously still present on the map itself), this caused confusion. While the previous-generation Titan X still used the GM200 GPU, the new version uses the GP102-400-A1 graphics processor.
The GP102 GPU is the graphics processor of the Pascal generation for the enthusiast sector. It is 471 mm² in size, consists of around 12 billion transistors and is manufactured at TSMC in the 16-nm FinFET manufacturing process. From a technical point of view, the GP102 is a 50% larger version of the GP104 with 96 raster, 3840 shader and 240 texture units. In previous generations, Nvidia used a graphics processor for the high-end / enthusiast sector and the professional market, which is why GPUs like the GK110 have many HPC features that are unnecessary for 3D applications. Already in the last generation Nvidia with the GM200 created a pure Gamerchip for the enthusiast market, due to the production-technical limitations of the older 28 nm manufacturing technology. In the Pascal generation, the new 16nm production cut these limitations, but Nvidia still managed to develop two different graphics processors for the professional and enthusiast market (GP100 and GP102). Although both processors have the same number of shader, texture, and raster units, they differed massively. For example, the GP100 uses a different cluster design that is different from all other Pascal GPUs, has many HPC features, and has an HBM2 memory interface . To develop two different graphics processors for the professional and enthusiast market (GP100 and GP102). Although both processors have the same number of shader, texture, and raster units, they differed massively. For example, the GP100 uses a different cluster design that is different from all other Pascal GPUs, has many HPC features, and has an HBM2 memory interface . To develop two different graphics processors for the professional and enthusiast market (GP100 and GP102). Although both processors have the same number of shader, texture, and raster units, they differed massively. For example, the GP100 uses a different cluster design that is different from all other Pascal GPUs, has many HPC features, and has an HBM2 memory interface .
In the case of Titan X , two clusters are deactivated on the GP102-400-A1, which still provides 3584 shader and 224 texture units (the full version is initially only available on the Quadro P6000 ). Furthermore, the Titan X has a 12 GB GDDR5X video memory, which is connected via a 384 Bit memory interface. This gives the Titan X a 25% higher performance under full-HD, compared to the GeForce GTX 1080, which increases to about 31% under 4K. According to earlier Titan models, the Titan X once again took the position as the fastest single-GPU graphics card on the market, fueled by the lack of AMD competing products. The fact that Nvidia did not sell the Titan X over board partners was a source of criticism, But only through their own onlineshop for a list price of 1299 €. After the Titan X has become obsolete by the Geforce GTX 1080 Ti, Nvidia replaced it with the Titan Xp . This uses the GP102-450-A1 in full scale and with slightly increased clock rates.
On March 9, 2017, Nvidia introduces the GeForce GTX 1080 Ti . This uses the GP102-350-K1-A1 graphics processor, which comes with two disabling shader clusters (SMX). Thus, the GTX 1080 Ti has over 3584 shader and 224 texture units, just like the Titan X before. In contrast to this, the video memory has also been truncated: the Geforce GTX 1080 Ti still has 11GB GDDR5X VRAM on a 352 bit memory interface. In combination with slightly higher clock rates, the GeForce GTX 1080 Ti achieves a slightly better performance compared to the Titan X, despite the smaller video memory. Nvidia set the list price for the GeForce GTX 1080 Ti to 699 US- $ and Waived a price surcharge for the Founders Edition.
Geforce GTX 1070 and 1080
Nvidia introduced the first graphics card of the Geforce 10 series, the Geforce GTX 1080, on May 6, 2016 and announced the sales start for May 27th. The GeForce GTX 1070 followed on June 10, 2016. Both cards use the GP104 graphics processor. A special feature of the cards was that Nvidia offered them in a so-called founders edition. However, this was only the reference design, which offered Nvidia under another name with extended support duration and increased list price and distributed through its own online shop. Especially the last circumstance, the increased list price (379 to 449 US- $ with the Geforce GTX 1070) against the designs of the boardpartners caused considerable confusion and criticism in the specialized press. The criticism of the founders edition was also by the, Against the design of the boardpartners, increased noise development: on the Geforce GTX 1080 reaches this, despite Vapor Chamber cooling, up to 4.2 Sone under full load.
The GP104 GPU represents the Pascal generation graphics processor for the high-end sector. It is 314 mm² in size, consists of 7.2 billion transistors and is manufactured at TSMC in the 16-nm FinFET manufacturing process. It has 2560 shader and 160 texture units organized in 20 clusters. In addition, the GP104 has 4 raster engines, 64 raster units on 8 partitions, as well as a 2048 kb L2 cache. Compared to the direct predecessor, the GM204, the GP104 now supports HDMI 2.0b (including HDCP 2.2 ), as well as DisplayPort 1.4 (although this was not yet finalized at release time).
The Geforce GTX 1080 was introduced by Nvidia on May 27, 2016, with the GP104-400-A1 used in the full version and has an 8GB GDDR5X video memory. With Ultra-HD, the card has a 64% to 73%, the Geforce GTX 980 Ti a 30% to 37%, and the AMD Radeon R9 Fury X a performance of 26% to 33% higher than its predecessor, the Geforce GTX 980 , In this way, the GeForce GTX 1080 soon established itself as the fastest single GPU card on the market until Nvidia introduced Titan X. The high performance combined with the once again improved energy efficiency led to corresponding positive evaluations in the specialist press. This, however, was again offset by an increased list price of 599 and 699 US $ in the “Founder Edition” (with the Geforce GTX 980, the list price was still 549 US- $), fueled by the lack of competition from AMD. Only with the launch of the GeForce GTX 1080 Ti Nvidia officially lowered the list price to 499 US- $. On 28 February 2017, Nvidia introduced the GP104-410-A1 GPU on the Geforce GTX 1080, which supports the GDDR5X video memory up to 11 Gbps.
The GeForce GTX 1070 followed the GTX 1080 on the 10th June 2016 on the market. The GP104-200-A1 graphics processor used is, however, partially trimmed: so 5 of the 20 shader clusters are deactivated, that is, a complete GPC unit is not available. In contrast to the direct predecessor, the Geforce GTX 970 , Nvidia renounced a pruning of the video memory, however instead of the GDDR5X the conventional GDDR5 memory is used. In this configuration, the GeForce GTX 1070 reached 82 to 86% of the performance of the GTX 1080. Compared to the top models of the predecessor generation, the Geforce GTX 980 Ti and the Radeon R9 Fury X , it achieves a performance gain of 3 to 9% under Ultra-HD.
Geforce GTX 1060
The GP106 GPU is the graphics processor of the Pascal generation for the performance sector. It is 200 mm², consists of 4.4 billion transistors and is also produced at TSMC in the 16-nm FinFET manufacturing process. Technically, the GP106 is a smooth halved GP104: it has 1280 shader and 80 textures organized in 10 clusters. In addition, the GP106 has 2 raster engines, 48 raster units on 6 partitions, as well as a 2048 kb L2 cache. In contrast to the graphics processors of previous generations from the performance sector, the GP106 does not support SLI .
On July 19, 2016, Nvidia introduced the GeForce GTX 1060 , which utilizes the GP106-400-A1 in full-scale design and has over 6 GB of VRAM. In contrast to the other previously presented graphics cards of the series, stood with the Radeon RX 480 a direct competitor of AMD . Compared to this, the GeForce GTX 1060 under full HD has a 6% higher performance, compared to its direct predecessor, the GeForce GTX 960 (4 GB), a roughly 75%. Since the difference in performance was roughly the same as the price difference (€ 269 to € 279), both cards were correspondingly positively valued for their price / performance ratio in the trade press. A pre-assumed disadvantage of the Geforce GTX 1060 compared to the Radeon RX 480 was the smaller video memory (6 to 8 GB). In the practice test, however, due to the improved storage management and the improved color compression, no disadvantage could be found. In contrast, the Radeon RX 480 could have a higher performance under DirectX 12 and volcano , which gives them a higher degree of future security. Nvidia came up with the criticism when it released a 3 GB version of the Geforce GTX 1060 at the beginning of August. In this case, a significant storage limitation could be detected. The performance also had a negative effect on the performance, That the GP106-300-A1 is not used in the full version, but a shader cluster is deactivated, which is not directly recognizable for the customer; Both models can only be distinguished from the memory size. In analogy to the Geforce GTX 1080, Nvidia introduced the GP106-410-A1 GPU on the Geforce GTX 1060 on 28th February 2017, which supports the GDDR5 video memory up to 9 Gbps.
Geforce GTX 1050 and 1050 Ti
On October 25th, 2016, Nvidia introduced graphics cards Geforce GTX 1050 and Geforce GTX 1050 Ti to their entry-level models using the GP107 graphics processor and the features of the previously released Nvidia Geforce 10 Series Pascal graphics cards for performance And high-end market. Nvidia thus aims at the price segments below 200 euros, correspondingly the GeForce GTX 1050 for 125 euros on the market come. A reference model like the Founders Edition of the Geforce GTX 1080 and Geforce GTX 1070 will not give it for the Geforce GTX 1050. With the GeForce GTX 1050 and the variant GTX 1050 Ti planned for the low-end entry-level market, Nvidia is positioning itself explicitly between AMD’s Radeon RX 470 and Radeon RX 460. Compared to the top models of the Nvidia-Geforce-10 series, the Geforce GTX 1050 with the GP107 has clear cuts: on the smaller die base area of 132 mm² (compare GP104 / GTX 1080: 314 mm²) the GP107 significantly less Raster, shader and texture units. The GTX 1060 as a low-critical memory size (3 GB GDDR5) was again reduced to 2 GB GDDR5 – apart from the variant Geforce GTX 1050 Ti, which is equipped with 4 GB GDDR5 even more lavishly equipped than the Geforce GTX 1060. The equipment Does not meet all the current games at the time of the release and achieves satisfactory results, even in the overclocked state, only when games are clearly reduced in the detailed level. Performance tests were limited to the widespread Full HD resolution with 1,920 × 1,080 pixels.