Nvidia shook things up with the performance of their first Ada-based card, the RTX 4090, but for a lot of people, that level of performance isn’t needed or is out of their price range. Things start to get interesting as we start to see the rest of the product stand fill out over time. Nvidia announced two RTX 4080 models alongside the RTX 4090 but later “unlaunched” the 12GB model leaving just the RTX 4080 16GB as the one 4080 model and today is the day. The embargo’s lifted and we can finally check out how the RTX 4080 performs ahead of it launching in retailers tomorrow morning. So today I’m going to check out the RTX 4080 Founders Edition and see how it compares with the RTX 4090 as well as last generation's cards. Let’s dive in!
Product Name: Nvidia RTX 4080 Founders Edition
Review Sample Provided by: Nvidia
Written by: Wes Compton
Amazon Affiliate Link: HERE
What's the RTX 4080 All About
I dove into some of the details of the Ada architecture for the RTX 4090 launch and I don’t think we need to get into those details. Be sure to check out our RTX 4090 Founders Edition review for more information there. But I did want to at least take a look at what sets the RTX 4080 apart from the RTX 4090 that we previously took a look at. I also put the specification listing together with the RTX 3080 and RTX 3080 Ti that the RTX 4080 is replacing. From the 4090 to the 4080 the big change is the RTX 4080 is not running on the same AD102 GPU that the RTX 4090 was on, it is on the AD103 which takes the transistor count from 76.3 Billion to 45.9 billion but that is still well above the 28.3 Billion of both of the RTX 3080 cards. This drops the graphics processing cluster count from 11 down to 7 and with that, the texture processing clusters are going from 64 to 38 and streaming multiprocessors from 128 to 78. That same change can be seen in the CUDA cores which the RTX 4080 has 9728 which is nearly half of the 16384 of the RTX 4090. Things get interesting when we compare those same numbers with the 3080 Ti which has more TPCs, SMs, and CUDA cores. It even has more Tensor and RT cores but the RTX 4080 has the newer 4th gen Tensor cores and 3rd Gen RT cores.
Where the RTX 4080 stands out from the 3080 and 3080 Ti is its GPU boost clock speed which can boost up to 2505 MHz, nearly the same as the RTX 4090 compared to 1665 MHz on the 2080 Ti. The memory clock speed is cranked up as well at 1400 MHz which is higher than the RTX 4090 even. But the big downside is the RTX 4080 has a memory interface that is 256-bit whereas the 3080 was 320-bit and the other two were 384-bit. The higher clock speed gives a higher memory data rate at least. Even with the higher memory clock speed, you can see how the interface limits things in the total memory bandwidth, even with 16 GB of memory. The RTX 4080 also has a huge jump in L2 cache like the 4090 compared to the older cards.
It is built on the same TSMC 4 nm Custom NVIDIA Process that the RTX 4090 was built on which helps and the TGP or total graphics power of the card is 320 watts whereas the 4090 was 450 watts. This puts it in line with the RTX 3080 in power and lower than the RTX 3080 Ti. Even so, it does come with a three 8-pin dongle where those two cards both came with dual 8-pin adapters.
|
Specifications |
RTX 3080 |
RTX 3080 Ti |
RTX 4080 |
RTX 4090 |
|
Graphics Processing Clusters |
6 |
7 |
7 |
11 |
|
Texture Processing Clusters |
34 |
40 |
38 |
64 |
|
Streaming Multiprocessors |
68 |
80 |
76 |
128 |
|
CUDA Cores |
8704 |
10240 |
9728 |
16384 |
|
Tensor Cores |
272 (3rd Gen) |
320 (3rd Gen) |
304 (4th Gen) |
512 (4th Gen) |
|
RT Cores |
68 (2nd Gen) |
80 (2nd Gen) |
76 (3rd Gen) |
128 (3rd Gen) |
|
Texture Units |
272 |
320 |
304 |
512 |
|
ROPs |
96 |
112 |
112 |
176 |
|
Boost Clock |
1710 MHz |
1665 MHz |
2505 MHz |
2520 MHz |
|
Memory Clock |
1188 MHz |
1188 MHz |
1400 MHz |
1313 MHz |
|
Memory Data Rate |
19 Gbps |
19 Gbps |
22.4 Gbps |
21 Gbps |
|
L2 Cache Size |
5 MB |
6 MB |
64 MB |
72MB |
|
Total Video Memory |
10 GB GDDR6X |
12 GB GDDR6X |
16 GB GDDR6X |
24 GB GDDR6X |
|
Memory Interface |
320-Bit |
384-bit |
256-bit |
384-bit |
|
Total Memory Bandwidth |
760.3 GB/s |
912 GB/s |
716.8 GB/sec |
1008 GB/s |
|
Texture Rate (Bilinear) |
465.1 GigaTexels/second |
532.8 GigaTexels/second |
761.5 GigaTexels/second |
1290.2 GigaTexels/second |
|
Fabrication Process |
Samsung 8 nm 8N NVIDIA Custom Process |
Samsung 8 nm 8N NVIDIA Custom Process |
TSMC 4 nm NVIDIA Custom Process |
TSMC 4 nm NVIDIA Custom Process |
|
Transistor Count |
28.3 Billion |
28.3 Billion |
45.9 billion |
76.3 Billion |
|
Connectors |
3 x DisplayPort 1 x HDMI |
3 x DisplayPort 1 x HDMI |
3 x DisplayPort 1 x HDMI |
3 x DisplayPort 1 x HDMI |
|
Form Factor |
Two Slots |
Two Slots |
Triple Slot |
Triple Slot |
|
Power Connectors |
1x16-pin (Dongle to 2x 8-Pins) |
1x16-pin (Dongle to 2x 8-Pins) |
1x16-pin (Dongle to 3x 8-Pins) |
1x16-pin (Dongle to 4x 8-Pins) |
|
Minimum Power Supply |
750 Watts |
750 Watts |
750 Watts |
850 Watts |
|
Total Graphics Power (TGP) |
320 Watts |
350 Watts |
320 Watts |
450 Watts |
|
Maximum GPU Temperature |
93° C |
93° C |
90° C |
90° C |
|
PCI Express Interface |
Gen 4 |
Gen 4 |
Gen 4 |
Gen 4 |
|
Launch MSRP |
$699 |
$1199 |
$1199 |
$1599 |
Before getting into testing I did also run GPUz to double-check that our clock speeds match up with the specifications and it does have the 2505 MHz boost clock speed. This also documents the BIOS revision of our card and the driver I used for testing which is the 526.72 driver that Nvidia provided to press before the launch.
