[ad_1]
In the last five years, real upheavals took place in the video card market. Either the crypto fever, which caused a shortage of graphics accelerators, then the official increase in the price of cards with the release of the GeForce RTX 20 series. Then mining becomes popular again, which leads to another shortage and an even greater increase in prices. With the transition of «ether» to PoS and the increasing difficulty of mining other cryptocurrencies, there is hope that this whole epic has finally ended or taken a break for a long time. Of course, no one is going to reduce the cost after two years of excess profits, so graphics adapters will still cost most of the PC. But with the price increase, developers again promise incredible speed for games and creativity. For example, in the new Nvidia GeForce RTX 40 series.
GeForce RTX 40
Video cards based on the Ada Lovelace architecture were developed to provide even greater performance in ray tracing and AI-based neural graphics. The new GPUs use RT-cores of the third generation and tensor cores of the fourth generation, the speed of which has increased several times compared to the GeForce RTX 30 series. In addition, new more energy-efficient shader processors are used and the new DLSS 3 technology is supported, which also affects performance.
The new graphic cores are manufactured on the basis of the 4 nm technological process and have 76 million transistors in their composition, which is more than 2.5 times compared to Ampere, while the area of the crystal has become slightly smaller. The maximum number of stream processors has increased by 1.5 times, and the operating frequencies have increased to the level of 2200-2500 MHz. But the use of modern technology and advanced development, unfortunately, could not keep power consumption at the old level, and new video cards require more efficient cooling and are equipped with a 12VHPWR power connector, which has already managed to light up for no good reason.
Video adapter | GeForce RTX 4090 | GeForce RTX 4080 | GeForce RTX 4080 12GB* | GeForce RTX 3090 Ti | GeForce RTX 3090 | GeForce RTX 3080 Ti | GeForce RTX 3080 | GeForce RTX 3070 Ti | GeForce RTX 3070 |
---|---|---|---|---|---|---|---|---|---|
Core | AD102 | AD103 | AD104 | GA102 | GA102 | GA102 | GA102 | GA104 | GA104 |
Number of transistors, million pcs | 76300 | 45900 | 35800 | 28300 | 28300 | 28300 | 28300 | 17400 | 17400 |
Techproces, n.m | 4 | 4 | 4 | 8 | 8 | 8 | 8 | 8 | 8 |
Core area, sq. mm | 608 | 379 | 295 | 628 | 628 | 628 | 628 | 392 | 392 |
Number of CUDA stream processors | 16384 | 9728 | 7680 | 10752 | 10496 | 10240 | 8704 | 6144 | 5888 |
The number of tensor cores | 512 | 304 | 240 | 336 | 328 | 320 | 272 | 192 | 184 |
The number of RT cores | 128 | 76 | 60 | 84 | 82 | 80 | 68 | 48 | 46 |
Number of texture blocks | 512 | 304 | 240 | 336 | 328 | 320 | 272 | 192 | 184 |
Number of rendering blocks | 176 | 112 | 80 | 112 | 112 | 112 | 96 | 96 | 88 |
Base core frequency, MHz | 2235 | 2205 | 2310 | 1560 | 1395 | 1365 | 1440 | 1575 | 1500 |
Boost frequency, MHz | 2520 | 2505 | 2610 | 1860 | 1695 | 1665 | 1710 | 1770 | 1725 |
Memory bus, bit | 384 | 256 | 192 | 384 | 384 | 384 | 320 | 256 | 256 |
Memory type | GDDR6X | GDDR6X | GDDR6X | GDDR6X | GDDR6X | GDDR6X | GDDR6X | GDDR6X | GDDR6 |
Memory frequency, GHz | 21 | 23 | 21 | 21 | 19,5 | 19 | 19 | 19 | 16 |
Memory capacity, GB | 24 | 16 | 12 | 24 | 24 | 12 | 10 | 8 | 8 |
Supported version of DirectX | 12 Ultimate (12_2) | 12 Ultimate (12_2) | 12 Ultimate (12_2) | 12 Ultimate (12_2) | 12 Ultimate (12_2) | 12 Ultimate (12_2) | 12 Ultimate (12_2) | 12 Ultimate (12_2) | 12 Ultimate (12_2) |
Interface | PCI-E 4.0 | PCI-E 4.0 | PCI-E 4.0 | PCI-E 4.0 | PCI-E 4.0 | PCI-E 4.0 | PCI-E 4.0 | PCI-E 4.0 | PCI-E 4.0 |
Power, W | 450 | 320 | 285 | 450 | 350 | 350 | 320 | 290 | 220 |
Release date | October 12, 2022 | November 16, 2022 | – | March 29, 2022 | September 24, 2020 | June 3, 2021 | September 17, 2020 | June 10, 2021 | October 29, 2020 |
MSRP price, $ | 1599 | 1199 | 899 | 1999 | 1499 | 1199 | 699 | 599 | 499 |
* — cancelled
A total of three GeForce RTX models were announced on September 20: 4090, 4080 and 4080 12GB. The first based on the AD102 core is now top and has 16384 stream processors, operates at a frequency of 2235-2520 MHz and is equipped with a video buffer of 24 GB, which is capable of transferring data at a speed of 21 Gbit/s through a 384-bit bus. Power consumption is 450W and the price is $1,599, which is $100 more than the price of the GeForce RTX 3090 when it was released. Interestingly, the RTX 4090 has 1.5 times more computing units than the GeForce RTX 3090 Ti!
The second model, which will be released in a few weeks, has an AD103 core with 9,728 stream processors, which is only 10 percent more than the GeForce RTX 3080. The same goes for the other units — their number is not very impressive against the background of the older 40 series card. The chip operates at a frequency of 2205-2505 MHz and has a buffer of 16 GB with a 256-bit bus that transmits data at a speed of 23 Gbps. The power of the graphics accelerator is «only» 320 W, and the cost is almost twice as much as that of its predecessor.
The last card, namely the GeForce RTX 4080 12GB, turned out to be very strange, its characteristics are not at all suitable for its level and more in line with the hypothetical GeForce RTX 4070. Fortunately, it can be canceled at the last moment and it will not be able to mislead an untrained user. It is quite likely that we will see it as a younger model of the 40 series, but now we will consider the oldest, produced by ASUS — TUF Gaming GeForce RTX 4090 OC Edition.
ASUS TUF-RTX4090-O24G-GAMING
Producer | ASUS |
---|---|
Model and product page | TUF-RTX4090-O24G-GAMING |
Core | AD102 |
Techproces, n.m | 4 |
Number of stream processors | 16384 |
Base frequency, MHz | 2235 |
Boost Clock, MHz | 2565 (2595 OC Mode) |
Memory bus, bit | 384 |
Memory type | GDDR6X |
Memory frequency, MHz | 21000 |
Memory capacity, GB | 24 |
Supported version of DirectX | 12 Ultimate (12_2) |
Interface | PCI-E 4.0 (x16) |
Image output | 3x DisplayPort 1.4a, 2x HDMI 2.1a |
Energy consumption, Vt | 450 |
Dimensions, mm | 348,2 x 150 x 72,6 |
Weight, kg | n/a |
Average cost | n/a |
The graphics card comes in a large black box with chamfered edges on the sides, which has an image of the product and its key features.


In a separate envelope you can find documentation, a TUF certificate, a puzzle, a power adapter and a screwdriver stand (not in the photo).
The adapter has a 12VHPWR connector on one side, and four regular 8-pin connectors on the other, to which you need to connect the cables from the power supply. Near the new connector, it should be handled very carefully, not bending it close to the connector, because then there may be a bad contact, which will lead to heating or even melting of the plastic.


The card holder can change its length and turn into a screwdriver as needed. A very useful thing — assembled a PC, supported the card at the end.
And they put it there for a reason. The video card is very massive, big and heavy. Without this support, it can break the PCI Express graphics slot.
Externally, the card follows the contours of the GeForce RTX 3090 Ti, also from the TUF Gaming series, but this one is more brutal in appearance and not so fragile to the touch. It seems monolithic and you immediately feel when you take it in your hands that it is a real thing, also of industrial design. But to be honest, the rear side stands out a bit from the updated styling.
If you compare the new video card with the ROG-STRIX-RTX3090-O24G-GAMING, the latter will begin to seem like a mid-level adapter, and not a top card — that’s how big the GeForce RTX 4090 is.
Its dimensions are 348.2x150x72.6 mm, it occupies almost four slots, and this should be taken into account when choosing a case for a PC.
Looking at the case, it becomes clear why the card is so heavy — it is completely cast from metal, which adds to its brutality. The cover of the old cards was bent from aluminum and bolted to the frame.


There are three DisplayPort 1.4a outputs on the back panel, as well as two HDMI 2.1a outputs. Near it, we can also see cast ribs on the casing.
Three fans with an impeller diameter of 105 mm and a thickness of about 15 mm are used to cool this design.
From the back, we can also see the ribs on the casing and three threaded mounting holes.
Of course, there was also a 12VHPWR connector. It is located on top of the video card, almost in the center.
The cast casing does not give an excuse to give up the RGB lighting with Aura Sync support, and it found a place in the back of the card. It is managed through proprietary software.
If you need to clean the video card from dust, you will have to disassemble it, and not just remove the cover. Fortunately, the radiator is screwed to the board separately and disassembling the adapter will not affect the warranty period, but you will still have to unscrew all but four screws. The fans are mounted on a thick metal frame and can be easily removed. The back plate has a large cutout for better air flow from the outer fan, which will also provide additional circulation in the PC case.
The radiator is made with an evaporation chamber and eight heat pipes that penetrate two massive sections of aluminum nickel-plated plates. A viscous gray paste is used to transfer heat from the chip, and 2.5–3 mm thick thermal pads are used from the components.


The printed circuit board is small relative to the card itself and occupies less than 2/3 of its length.


The AD102-300-A1 chip has a protective frame and communicates via a 384-bit bus with a 24GB video buffer, drawn from 12 Micron 2MU47D8BZC chips. The number of power phases is 18 on the sides and 4 more phases on the corners of the chip. Throttles turned out to be very singing, so activating the V-Sync mode will not hurt. With it, even the crunching of the roller bearings of the old system fans will be heard.
The GPU operates at a standard frequency of 2235 MHz, but its Boost is increased to 2565 MHz, and in OC Mode it can also be overclocked to 2595 MHz. The memory operates at an effective 21 GHz (or 21 Gbps).
In games with 4K resolution and maximum graphics quality, the chip worked at a frequency of 2730-2760 MHz, and its temperature did not exceed 64 °C on an open bench at 20 degrees indoors.


In general, if you look at the peak indicators, you can see 2775 MHz on the chip. The maximum Hot Spot temperature was 74 °C and the total consumption of the card was 430 W.


ASUS TUF-RTX4090-O24G-GAMING is generally quiet, so Quite Mode was not used.
[ad_2]