GeForce RTX 2080 Ti remained the flagship graphics card in NVIDIA’s portfolio for a long time, although the top specification also meant a prohibitive price of over 5000 zlotys. The most powerful variation of Turing was therefore a product far beyond the reach of the average player. However, a moment of reshuffling is coming, as the champions of the previous generation give way to models positioned much lower. According to NVIDIA’s assurances, GeForce RTX 3070 was to offer GeForce RTX performance 2080 Ti, being sold for more than twice as much money. Today, on the example of GeForce RTX 3070 Founders Edition I will check whether the promises have been fulfilled, after all, the performance of GeForce RTX 2080 Ti priced at 2359 zlotys seemed a very optimistic scenario.

Author: Sebastian Oktaba

GeForce RTX 3000 is the second generation of NVIDIA graphics cards with hardware units for ray tracing acceleration and deep learning, using the new Ampere architecture, being a development version of the Turing powering the GeForce RTX family 2000. The two-year construction period was well spent – engineers have solidly improved graphics processing clusters (GPC), including increasing the cache capacity, doubling the number of shading units in SM (Streaming Multiprocessors) blocks while increasing the performance of RT and Tensor cores. A significant change is also the transition to a lower Samsung lithography, which made it possible to improve the energy efficiency factor, reduce the size of the core and increase the clock speeds. Top models received GDDR6X memory with high frequency and wide buses, the bandwidth of which is equal to the HBM2 modules. NVIDIA Ampere also introduces a chain of interesting solutions and technologies, including Reflex, RTX IO, Broadcast and HDM 2.1 connector to display 8K HDR image when refreshing 60 Hz. More strictly technical information can be found on the next page, dedicated to the most important changes introduced in the new architecture.

NVIDIA GeForce RTX graphics card 3070 was to offer GeForce RTX performance 2080 Ti, but included in the price of GeForce RTX 2070 / COOL. Will it really be that good? After all, it is a product that is more than twice cheaper.

“src =” http://www.purepc.pl/image/artykul/2020 / 10 / 26 _geforce_rtx_graphic_test _ 3070 _ vs_geforce_rtx _ 2080 _ ti_2.png “>

NVIDIA GeForce RTX 3070 is based on the GA graphics chip 104 – 300 – A1, which is the weakest iterations of the Ampere architecture so far using Samsung’s 8 nm lithography, which has more than 17 billions of transistors on the surface less than 400 mm². NVIDIA boldly announces The hero of this review as a substitute for NVIDIA GeForce RTX 2080 Ti available for much less money that’s why it will be compared with the top Turing. NVIDIA GeForce RTX 3070 has more stream processors (5888 vs 4352), while rendering units (64 vs 88) and texturing (184 vs 272) received much less. The situation is similar with specialized RT units (46 vs 68) and Tensors (184 vs 544). The dry technical specification clearly indicates that even despite the progress of the architecture and the changes made (including doubling the CUDA units for each SM), the brutal computing power of NVIDIA GeForce RTX 2080 Ti should be greater. However, NVIDIA GeForce RTX 3070 has a double L2 cache (128 KB for a single SM) and extended L3 memory (5.5 vs 4 MB). However, the graphics memory of 8 GB GDDR6 was severely trimmed on 256 – a bit bus, the bandwidth of which is also almost half that of the NVIDIA GeForce RT 2080 Ti.

	RTX 3090	RTX 3080	RTX 3070	RTX 2080 Ti
Architecture	Ampere	Ampere	Ampere	Turing
Graphics Layout	GA 102	GA 102	GA 104	HERE102
Lithography	8 nm	8 nm	8 nm	12 nm
Core size	628 mm²	628 mm²	392 mm²	754 mm²
Transistors	28 billion	28 billion	17 billion	19 billion
SP Units	10496	8704	5888	4352
TMUs	328	272	184	272
ROP units	112	96	64	88
RT Units	82	68	46	68
Tensor Units	328	272	184	544
Base clock	1395 MHz	1440 MHz	1500 MHz	1350 MHz
Boost clock	1650 MHz	1710 MHz	1730 MHz	1500 MHz
Memory clock	19500 MHz	19000 MHz	14000 MHz	14000 MHz
Memory amount	24 GB	10 GB	8 GB	11 GB
Memory type	GDDR6X	GDDR6X	GDDR6	GDDR6
Memory bus	384 – bit	320 – bit	256 – bit	352 – bit
Bandwidth	937 GB / s	760 GB / s	448 GB / s	616 GB / s
Interface	PCI-E 4.0 x 16	PCI-E 4.0 x 16	PCI-E 4.0 x 16	PCI-E 3.0 x 16
TGP Factor	350 W	320 W	220 W	250 W
Starting price	1499 USD	699 USD	448 USD	999 USD
Release date	September 2020	September 2020	October 2020	September 2018

NVIDIA GeForce RTX Base Timing 3070 Foundes Edition set to 1500 MHz, although thanks to the GPU Boost technology the actual clock is automatically adjusted to at least 1800 MHz (target value depends on temperature, load, power limit, etc.). The graphics memory works with the standard frequency 14000 MHz. NVIDIA GeForce RTX 3070 uses PCI-Express 4.0, of course, although the new interface has a marginal impact on the results in addition, being available only on the AM4 platform paired with Ryzen processors 3000. The manufacturer gives the NVIDIA GeForce RTX 3070 TPD factor 220 W, admittedly lower than NVIDIA GeForce RTX 2080 Ti, but relatively high considering lithography and a smaller number of ROPs, TMUs, memory controllers, etc. The key factor is, however, a completely different factor – the price. NVDIA retained