Biggest GPU Ever! NVIDIA GeForce RTX 3090 Unboxing

Ali Güngör

Genel Yayın Yönetmeni
Yönetici
Katılım
22 Haziran 2011
Mesajlar
53.075
Çözümler
17
Yer
İstanbul Türkiye
Daha fazla  
Cinsiyet
Erkek
Meslek
Technopat
Profil Kapağı
1523300036
Bu içeriği görüntülemek için üçüncü taraf çerezlerini yerleştirmek için izninize ihtiyacımız olacak.
Daha detaylı bilgi için, çerezler sayfamıza bakınız.

NVIDIA GeForce RTX 3090 is a BFGPU* as Jensen Huang calls it. Let's unbox it and see for ourselves.

How big it really is? Let's compare the size to RTX 3080 and understand it better. The 3090 is the card that took place of NVIDIA TITAN in the previous generations.

If you look it as a gaming card, it is the most expensive with 1500 USD price tag. But if you look it as a creator GPU, as an alternative to Quadro cards, then it is the cheapest GPU you can get for that kind of work. It is the card to make games that run on 3080 and 3070 cards. So we expect many developer studios to get 3090. So theoretically 3080 is only 16 percent behind of 3090 in gaming, but big memory boost and other features make this card valuable for creators. We will test it mostly for productivity with 3D, texture and creation applications along with NVIDIA Studio drivers.

As you will see down, it's FP16, FP32 and FP64 performance along with other features that make the difference in some software. Unlike most gaming cards, it features advanced floating point calculation capabilities. Of course it's the software that counts, because not all applications take advantage of these features.

GPU Core: GA102-300-A1 Ampere Architecture, 8 nm Samsung manufacturing process is used. 28,300 million transistors on a die size of 628 mm²
Shading Units: 10496
TMUs: 328
ROPs: 112
SM Count: 82
Tensor Cores: 328
RT Cores: 82
L1 Cache: 128 KB (per SM)
L2 Cache: 6 MB
Pixel Rate: 189.8 GPixel/s
Texture Rate: 556.0 GTexel/s
FP16 (half) performance: 35.58 TFLOPS (1:1)
FP32 (float) performance: 35.58 TFLOPS
FP64 (double) performance: 556.0 GFLOPS (1:64)

24 GB GDDR6X memory with a bus of 384 bit, along with advanced color compression algorithms for a total memory bandwidth of 936.2 GB/s.
Base Clock: 1395 MHz
Boost Clock: 1695 MHz
Memory Clock: 1219 MHz, 19.5 Gbps effective

It is a triple slot design card, 313 mm - 12.3 inches to 138 mm - 5.4 inches with a three slot width. It has 3 DisplayPort and one HDMI port. Special 12 pin connector and power cable adapter. It has a TDP of 350 Watts and requires a high quality PSU. 750 Watt is recommended. Also take note that this GPU and other 3000 series cards use PCIe 4.0 connection.

You can use them on a PCIe 3.0 motherboard - CPU combination but there may be slight performance difference between PCIe 3.0 and PCIe 4.0. First test made between these connections show there is some but negliable difference for this gen of GPU's but further testing is required to make sure.

*BFGPU: Big Ferocious GPU is the explanation in the NVIDIA site, but anyone who played Doom would beg to differ.

Possible FAQ: Why not test it instead of unboxing it? Well we recieved the card with only hours to spare to the NDA date, this was the only possible content we could publish with available time. As you are watching this, we are testing it.

Turkish version:
 
Turkish version:
Bu içeriği görüntülemek için üçüncü taraf çerezlerini yerleştirmek için izninize ihtiyacımız olacak.
Daha detaylı bilgi için, çerezler sayfamıza bakınız.
 
Nice content. I didn't know that you have a global YouTube Channel. I only knew the English Section on Sosyal.
@Ali Güngör I added Turkish subtitles. It will be available to use if you review and approve it. By the way, did you say "Micron" at 0:38? I assumed that way. I believe you can change it if you like.
That was my first YouTube translation. I don't know the whole system yet.
 
Son düzenleme:

Geri
Yukarı