Talk about improved versions of the top-end GeForce 30 series graphics cards (namely the RTX 3070 and RTX 3080 ) began almost immediately after the original devices hit the market. A common practice for chip makers is that in the early days of the new GPU architecture, fully functional crystals are reserved for professional solutions, and consumer video cards are equipped with significantly cut down graphics processors, which in the future can be easily upgraded (not to mention the variability in the amount of RAM). if prompted by a changed competitive environment.
However, NVIDIA’s gaming products, which belong to the highest echelon of price and performance, are currently not under strong pressure from the “red” competitors. AMD introduced an alternative to all GeForce 30-series models, with the exception of the RTX 3090, but if you throw the clear advantage of Ampere silicon in game ray tracing on the scales (as well as the ability to use DLSS in order to increase the frame rate at any time at the cost of a minor loss in image quality), NVIDIA accelerators initially turned out to be in a more advantageous position in comparison with similar Radeon ones, when the composition of both lines was settled. Nevertheless, NVIDIA decided that the right moment for the update had come. We previously learned about the specs of the GeForce RTX 3070 Ti and RTX 3080 Ti, along with suggested retail prices, and they fully support the RTX 3080 Ti’s mission:
Speaking about prices, we have already made a mistake more than once, assuming that a long-term shortage of semiconductor crystals, for one reason or another, will bypass the novelty. NVIDIA has taken further measures in order to artificially limit Ti performance in cryptocurrency mining operations, but whether it will be possible to ensure wide availability of video cards this time, readers will find out on their own in the coming days. In the meantime, we’ll figure out what the GeForce RTX 3080 Ti is from a technical point of view and, most importantly, how great is the increase in performance compared to the first edition of the RTX 3080. The next in line is the RTX 3070 Ti, but the second NVIDIA novelty will be delayed for another week, debut is scheduled for June 9th.
The configuration of the original GeForce RTX 3080 left a good headroom for future spec revisions. While the fully functional GA102 crystal contains 84 streaming multiprocessors (SM), the RTX 3080 has only 68 of them at its disposal. In the Ti modification, the budget of active SMs was increased to 80, which equals an additional 1,536 FP32-compatible shader ALUs, 48 texture mapping units and 16 ROPs. It turns out that only two active streaming multiprocessors (or 256 CUDA-cores of standard precision) separate from the older model of the series, RTX 3090, GeForce RTX 3080 Ti.
Of course, such significant changes in the characteristics of the GPU have left their mark on the power consumption of the device. The RTX 3080 Ti is rated to operate within the same TBP as the RTX 3090, which is a whopping 350 watts. Nevertheless, although the new product is inferior to the flagship video card in the number of active GPU components, NVIDIA had to underestimate the design clock frequencies compared to the original characteristics of the RTX 3080 and RTX 3090: from 1,710 and 1,695 MHz, respectively, to 1,665 MHz. Boost Clock in NVIDIA device specifications is a conservative estimate, and nevertheless measurements in real conditions will confirm that some down-clocking does take place. It looks like there is a selection between the RTX 3080 Ti and the RTX 3090 for the quality of the GA102 crystals, which is reflected in the higher specific power consumption of the chips intended for the RTX 3080 Ti, and, accordingly, reduced clock frequencies within the same TBP. Nevertheless, if we focus on the passport specifications, the final increase in theoretical throughput compared to the RTX 3080 in floating point operations in standard precision was about 15%, which potentially translates into an equivalent change in the game frame rate under maximum load – in high resolution or tracing conditions. Rays.
In order to give the enhanced graphics processor an adequate communication channel with the external memory array, two more 32-bit controllers were put into operation. Thus, the GeForce RTX 3080 Ti has a full 384-bit VRAM bus, like the RTX 3090, but instead of expensive GDDR6X chips designed for a speed of 21 GB / s, it uses chips with a nominal value of 19 GB / s. It is equally important that along with the increased memory bandwidth of the RTX 3080 Ti, one more gigabyte of local memory got. Thus, the paradoxical situation was resolved when the junior model of the series, RTX 3060, has more VRAM (12 GB) than all the older ones (with the exception of RTX 3090). As for the latter, with such a slight difference in GPU configuration between the RTX 3080 Ti and RTX 3090, the only practical advantage of the flagship is the doubled set of GDDR6X chips (24 GB).
And finally, we will answer the most exciting question: “How much?” The GeForce RTX 3080 Ti was priced at $ 1,199 (or 116,990 in rubles), which is $ 500 more than the recommended cost of the RTX 3080 and exactly matches the cost of the RTX 2080 Ti Founders Edtion. The price of the new item, on the one hand, confirms the success of the previous version, by and large and does not need to be updated to compete with the older AMD 6000 series models. On the other hand, it seems that NVIDIA has acknowledged an abnormal market situation that has developed due to the huge demand for computer components and the popularity of cryptocurrency mining. The estimate of $ 1,199 looks more realistic than any numbers below a thousand dollars at a time when the RTX 3080 is already difficult to find below $ 2,500. In addition, it is worth paying attention to the changed vocabulary of the chip maker:
Sales of the GeForce RTX 3080 Ti will begin tomorrow, in all the variety of partner versions of the accelerator. However, the basic level of performance of the new items, as usual, will be demonstrated by the proprietary NVIDIA graphics card. When it comes to the original RTX 3080, the Founders EditionThey did their job perfectly well: the compact cooler of unusual design dissipates the previously unprecedented high heat flux of 320 W surprisingly efficiently, while maintaining a low noise level. Given that the RTX 3080 Ti’s rated power is on par with the RTX 3090 (and it also overclocks!), we expected to see the same huge cooling system. However, NVIDIA did not consider it necessary to change anything in the physical design of the RTX 3080, aside from minor cosmetic details such as the polished bevels of the aluminum edging.
The RTX 3080 Ti’s cooling system is still powered by two 85mm fans. A fan, mounted on the front of the card, draws heated air out of the computer through wide openings in the mounting plate. The second heatsink is installed on the opposite side of the PCB and rotates in the opposite direction, and the PCB itself was reduced in size in order to provide blowing through the entire tail of the heatsink.