unnamed file 4681

 

unnamed file 4680

 

1 – Overview

First GeForce RTX 30 have been announced and launched last September. The GeForce RTX 3060 Ti, the latest member of the RTX 30 family has been launched two weeks ago. And today, it’s the first time I really test and benchmark an Ampere-based graphics card. Let’s see if the Ampere-based graphics cards are really faster compared to previous generation.

In this article I will review the GIGABYTE GeForce RTX 3060 Ti GAMING OC 8G and compare it to this GeForce RTX 2080. GIGABYTE’s RTX 3060 Ti is based on the Ampere GA104 GPU (4864 cores, 80 ROPs, 152 texture units, 38 RT cores for a 200W-TGP) and comes with a factory-overclocked GPU: 1740MHz (reference clock speed: 1670MHz).

 

2 – Gallery

The bundle: the box, the graphics card and a paper install guide (honestly, this guide is a bit useless and some trees can be saved if this kind of doc is no longer added in the bundle).

The GPU cooler is GIGABYTE’s WINDFORCE 3X, with alternate spinning fans:

The WINDFORCE 3X cooling system features 3x 80mm unique blade fans, alternate spinning, 3 composite copper heat pipes direct touch GPU, 3D active fan and Screen cooling, which together provide high efficiency heat dissipation.

What I can say is that the card is correctly cooled and the cooling is done quietly. The card is passive is the GPU temperature is less than 42°C, which is most of the time the case at idle.

unnamed file 4681

unnamed file 4682

unnamed file 4683

unnamed file 4684

This RTX 3060 Ti has a nice backplate that brings additional mechanical protection:

unnamed file 4685

unnamed file 4686

The graphics card is powered by an additional 8-pin power connector (total possible power draw: 150 + 75 = 225W). GIGABYTE kept the regular 8-pin connector in place of the new 12-pin connector available on the RTX 3060 Ti Founders Edition.

unnamed file 4687

Two HDMI 2.1 and two DisplayPort 1.4a connectors are present:

unnamed file 4688

Last photo: the size.

unnamed file 4689

3 – GPU Information

The usual GPU Caps Viewer, GPU Shark and GPU-Z screenshots.

unnamed file 3648

unnamed file 3649

unnamed file 3650

unnamed file 3651

4 – GPU Benchmarks

 

Test System
CPU Intel Core i7-8700K @ 3.7GHz
Motherboard GIGABYTE Z390 Gaming X
Memory 16GB DDR4 Corsair Vengeance LED @ 3000MHz
PSU Corsair AX860i
Storage Samsung 840 Pro 256GB SSD
OS Windows 10 v2009 64-bit

 

4.1 – 3DMark FireStrike (1920×1080)

22185 (graphics:29130, physics:17988, combined:9100)
GIGABYTE GeForce RTX 3060 Ti Gaming OC 8G – GeForce 460.79
22166 (graphics:28314, physics:18072, combined:9686)
ASUS ROG Strix GeForce RTX 2080 OC 8GB – GeForce 460.89
19146 (graphics:22604, physics:18585, combined:9109)
ASUS ROG Strix GeForce GTX 1080 OC 8GB – GeForce 416.94
19118 (graphics:22150, physics:18587, combined:9639)
MSI GeForce RTX 2070 Ventus 8GB – GeForce 417.01
16504 (graphics:19574, physics:18586, combined:29143)
MSI Radeon RX Vega 56 AIR BOOST – Adrenalin 18.11.1
16414 (graphics:18561, physics:18629, combined:8025)
EVGA GeForce GTX 1070 FTW – GeForce 416.94
11708 (graphics:13426, physics:18640, combined:4651)
MSI Radeon R9 290X – Adrenalin 18.11.1
11538 (graphics:13079, physics:18534, combined:4711)
EVGA GeForce GTX 1060 SC 6GB – GeForce 416.94
11075 (graphics:12839, physics:18597, combined:4201)
MSI Radeon RX 470 8GB – Adrenalin 18.11.1
10292 (graphics:11368, physics:18595, combined:4325)
MSI GeForce GTX 970 – GeForce 416.94
7899 (graphics:8616, physics:18629, combined:3175)
MSI Radeon HD 7970 – Adrenalin 18.11.1

 

4.2 – 3DMark FireStrike Ultra (4K / 3840×2160)

7314 (graphics:7292, physics:18008, combined:3917)
GIGABYTE GeForce RTX 3060 Ti Gaming OC 8G – GeForce 460.79
6805 (graphics:6713, physics:18125, combined:3711)
ASUS ROG Strix GeForce RTX 2080 OC 8GB – GeForce 460.89
5517 (graphics:5446, physics:18595, combined:2819)
ASUS ROG Strix GeForce GTX 1080 OC 8GB – GeForce 416.94
5242 (graphics:5113, physics:18601, combined:2777)
MSI GeForce RTX 2070 Ventus 8GB – GeForce 417.01
4634 (graphics:4654, physics:18592, combined:2148)
MSI Radeon RX Vega 56 AIR BOOST – Adrenalin 18.11.1
4593 (graphics:4512, physics:18631, combined:2304)
EVGA GeForce GTX 1070 FTW – GeForce 416.94
3117 (graphics:3013, physics:18669, combined:4711)
EVGA GeForce GTX 1060 SC 6GB – GeForce 416.94
2869 (graphics:2830, physics:18652, combined:1325)
MSI Radeon RX 470 8GB – Adrenalin 18.11.1
2681 (graphics:2594, physics:18622, combined:1321)
MSI GeForce GTX 970 – GeForce 416.94

 

4.3 – 3DMark TimeSpy (Direct3D 12, 2560×1440)

10077 (graphics:10837, CPU:7213)
GIGABYTE GeForce RTX 3060 Ti Gaming OC 8G – GeForce 460.79
9714 (graphics:10415, cpu:7248)
ASUS ROG Strix GeForce RTX 2080 OC 8GB – GeForce 460.89
8252 (graphics:8461, cpu:7239)
MSI GeForce RTX 2070 Ventus 8GB – GeForce 417.01
7524 (graphics:7562, cpu:7316)
ASUS ROG Strix GeForce GTX 1080 OC 8GB – GeForce 416.94
6284 (graphics:6144, cpu:7218)
MSI Radeon RX Vega 56 AIR BOOST – Adrenalin 18.11.1
6284 (graphics:6131, cpu:7324)
EVGA GeForce GTX 1070 FTW – GeForce 416.94
4524 (graphics:4244, cpu:7239)
EVGA GeForce GTX 1060 SC 6GB – GeForce 416.94
4259 (graphics:3966, cpu:7338)
MSI Radeon R9 290X – Adrenalin 18.11.1
3957 (graphics:3660, cpu:7330)
MSI Radeon RX 470 8GB – Adrenalin 18.11.1
3814 (graphics:3515, cpu:7368)
MSI GeForce GTX 970 – GeForce 416.94

 

4.4 – 3DMark Port Royal (Direct3D 12 + Raytracing)

6916
GIGABYTE GeForce RTX 3060 Ti Gaming OC 8G – GeForce 460.79
6863
ASUS ROG Strix GeForce RTX 2080 OC 8GB – GeForce 460.89

 

4.5 – Unigine Superposition – 1080p Medium – Direct3D

21154 (avg FPS: 175)
GIGABYTE GeForce RTX 3060 Ti Gaming OC 8G – GeForce 460.79
20903 (avg FPS: 156)
ASUS ROG Strix GeForce RTX 2080 OC 8GB – GeForce 460.89
16419 (avg FPS: 122.8)
ASUS ROG Strix GeForce GTX 1080 OC 8GB – GeForce 416.94
16001 (avg FPS: 119.7)
MSI GeForce RTX 2070 Ventus 8GB – GeForce 417.01
13039 (avg FPS: 97.5)
EVGA GeForce GTX 1070 FTW – GeForce 416.94
12078 (avg FPS: 90.3)
MSI Radeon RX Vega 56 AIR BOOST – Adrenalin 18.11.1
9082 (avg FPS: 67.9)
EVGA GeForce GTX 1060 SC 6GB – GeForce 416.94
8290 (avg FPS: 62)
MSI Radeon R9 290X – Adrenalin 18.11.1
7469 (avg FPS: 55.9)
MSI GeForce GTX 970 – GeForce 416.94
7127 (avg FPS: 53.3)
MSI Radeon RX 470 8GB – Adrenalin 18.11.1
5332 (avg FPS: 39.9)
MSI Radeon HD 7970 – Adrenalin 18.11.1

 

4.6 – Unigine Superposition – 4K Optimized – Direct3D

9495 (avg FPS: 71)
GIGABYTE GeForce RTX 3060 Ti Gaming OC 8G – GeForce 460.79
9241 (avg FPS: 69)
ASUS ROG Strix GeForce RTX 2080 OC 8GB – GeForce 460.89
7014 (avg FPS: 52.5)
ASUS ROG Strix GeForce GTX 1080 OC 8GB – GeForce 416.94
6936 (avg FPS: 51.9)
MSI GeForce RTX 2070 Ventus 8GB – GeForce 417.01
5639 (avg FPS: 42.2)
EVGA GeForce GTX 1070 FTW – GeForce 416.94
5060 (avg FPS: 37.8)
MSI Radeon RX Vega 56 AIR BOOST – Adrenalin 18.11.1
3886 (avg FPS: 29.1)
EVGA GeForce GTX 1060 SC 6GB – GeForce 416.94
3026 (avg FPS: 22.6)
MSI Radeon RX 470 8GB – Adrenalin 18.11.1

 

4.7 – FurMark 1.24

P1080 (1920×1080)

10530 (avg FPS: 175)
GIGABYTE GeForce RTX 3060 Ti Gaming OC 8G – GeForce 460.79
10320 (avg FPS: 172)
ASUS ROG Strix GeForce RTX 2080 OC 8GB – GeForce 460.89
9270 (avg FPS: 154)
MSI GeForce RTX 2070 Ventus 8GB – GeForce 417.01
7322 (avg FPS: 122)
ASUS ROG Strix GeForce GTX 1080 OC 8GB – GeForce 416.94
6691 (avg FPS: 112)
MSI Radeon RX Vega 56 AIR BOOST – Adrenalin 18.11.1
6201 (avg FPS: 103)
EVGA GeForce GTX 1070 FTW – GeForce 416.94
4946 (avg FPS: 83)
MSI Radeon R9 290X – Adrenalin 18.11.1
4556 (avg FPS: 76)
EVGA GeForce GTX 1060 SC 6GB – GeForce 416.94
3845 (avg FPS: 64)
MSI GeForce GTX 970 – GeForce 416.94
3367 (avg FPS: 56)
MSI Radeon RX 470 8GB – Adrenalin 18.11.1
3246 (avg FPS: 54)
MSI Radeon HD 7970 – Adrenalin 18.11.1

 

P2160 (3840×2160)

4018 (avg FPS: 67)
GIGABYTE GeForce RTX 3060 Ti Gaming OC 8G – GeForce 460.79
3819 (avg FPS: 64)
ASUS ROG Strix GeForce RTX 2080 OC 8GB – GeForce 460.89
3165 (avg FPS: 52)
MSI GeForce RTX 2070 Ventus 8GB – GeForce 417.01
2734 (avg FPS: 46)
ASUS ROG Strix GeForce GTX 1080 OC 8GB – GeForce 416.94
2261 (avg FPS: 38)
MSI Radeon RX Vega 56 AIR BOOST – Adrenalin 18.11.1
2217 (avg FPS: 37)
EVGA GeForce GTX 1070 FTW – GeForce 416.94
1600 (avg FPS: 27)
MSI Radeon R9 290X – Adrenalin 18.11.1
1556 (avg FPS: 26)
EVGA GeForce GTX 1060 SC 6GB – GeForce 416.94
1330 (avg FPS: 22)
MSI GeForce GTX 970 – GeForce 416.94
1071 (avg FPS: 18)
MSI Radeon HD 7970 – Adrenalin 18.11.1
1030 (avg FPS: 17)
MSI Radeon RX 470 8GB – Adrenalin 18.11.1

 

4.8 – Rhodium LC

Rhodium LC (LC for Liquid Carbon!) is a new GPU pixel shader benchmark made with GeeXLab and based on this shadertoy demo. You can download RhodiumLC from THIS PAGE.

unnamed file 4690

 

P1080 (1920×1080)

4853 (avg FPS: 80)
ASUS ROG Strix GeForce RTX 2080 OC 8GB – GeForce 416.94
3905 (avg FPS: 65)
GIGABYTE GeForce RTX 3060 Ti Gaming OC 8G – GeForce 460.79
3650 (avg FPS: 60)
ASUS ROG Strix GeForce GTX 1080 OC 8GB – GeForce 416.94
3461 (avg FPS: 57)
MSI GeForce RTX 2070 Ventus 8GB – GeForce 417.01
2835 (avg FPS: 47)
EVGA GeForce GTX 1070 FTW – GeForce 416.94
2640 (avg FPS: 43)
MSI Radeon RX Vega 56 AIR BOOST – Adrenalin 18.11.1
1923 (avg FPS: 32)
EVGA GeForce GTX 1060 SC 6GB – GeForce 416.94
1484 (avg FPS: 24)
MSI GeForce GTX 970 – GeForce 416.94
1418 (avg FPS: 23)
MSI Radeon R9 290X – Adrenalin 18.11.1
1192 (avg FPS: 19)
MSI Radeon RX 470 8GB – Adrenalin 18.11.1
937 (avg FPS: 15)
MSI Radeon HD 7970 – Adrenalin 18.11.1

 

P2160 (3840×2160)

1269 (avg FPS: 21)
ASUS ROG Strix GeForce RTX 2080 OC 8GB – GeForce 416.94
995 (avg FPS: 16)
GIGABYTE GeForce RTX 3060 Ti Gaming OC 8G – GeForce 460.79
930 (avg FPS: 15)
ASUS ROG Strix GeForce GTX 1080 OC 8GB – GeForce 416.94
892 (avg FPS: 14)
MSI GeForce RTX 2070 Ventus 8GB – GeForce 417.01
735 (avg FPS: 12)
EVGA GeForce GTX 1070 FTW – GeForce 416.94
669 (avg FPS: 11)
MSI Radeon RX Vega 56 AIR BOOST – Adrenalin 18.11.1
497 (avg FPS: 8)
EVGA GeForce GTX 1060 SC 6GB – GeForce 416.94
378 (avg FPS: 6)
MSI GeForce GTX 970 – GeForce 416.94
305 (avg FPS: 5)
MSI Radeon RX 470 8GB – Adrenalin 18.11.1
302 (avg FPS: 5)
MSI Radeon R9 290X – Adrenalin 18.11.1
231 (avg FPS: 3)
MSI Radeon HD 7970 – Adrenalin 18.11.1

Interesting, in a pure math / procedural test, the RTX 3060 Ti is slower than the RTX 2080.

 

4.9 – MSI Kombustor 2019 – MSI-01 OpenGL Test

MSI Kombustor 2019 (v4.1.9) has an interesting benchmark (MSI-01) which is also a very good burn-in test. This graphics test mixes tessellation, shadow mapping, DoF, HDR and high resolution textures in a simple scene.

unnamed file 4691

 

P1080 (1920×1080)

4433 (avg FPS: 73)
ASUS ROG Strix GeForce RTX 2080 OC 8GB – GeForce 460.89
3873 (avg FPS: 64)
GIGABYTE GeForce RTX 3060 Ti Gaming OC 8G – GeForce 460.79

 

P2160 (3840×2160)

2320 (avg FPS: 38)
ASUS ROG Strix GeForce RTX 2080 OC 8GB – GeForce 460.89
2231 (avg FPS: 37)
GIGABYTE GeForce RTX 3060 Ti Gaming OC 8G – GeForce 460.79

 

4.10 – MSI Kombustor 2019 – TessMark X32 OpenGL

unnamed file 4692

 

P1080 (1920×1080)

19985 (avg FPS: 333)
ASUS ROG Strix GeForce RTX 2080 OC 8GB – GeForce 460.89
18406 (avg FPS: 306)
GIGABYTE GeForce RTX 3060 Ti Gaming OC 8G – GeForce 460.79

 

4.11 – Basemark GPU (Vulkan)

Basemark GPU is a professional benchmark to evaluate and compare graphics performance across mobile and desktop platforms and supports all major graphics APIs and operating systems.

unnamed file 4693

Version: Basemark GPU 1.2.3
Settings: official test, Vulkan, 3840×2160, bc7

 

15968 (avg FPS: 160)
ASUS TUF GAMING RX 6800 XT – Adrenalin 20.12.2
11860 (avg FPS: 119)
GIGABYTE GeForce RTX 3060 Ti Gaming OC 8G – GeForce 460.79
11389 (avg FPS: 114)
ASUS ROG Strix GeForce RTX 2080 OC 8G – GeForce 460.79
7502 (avg FPS: 75)
ASUS Strix Radeon RX 5700 8G – Adrenalin 20.12.2

 

5 – Burn-in Test

This burn-in test has been done with FurMark 1.24.1.

As usual, the first thing to do before running a stress test is to set to the max the power target limit. I used MSI Afterburner for this job. The RTX 3060 Ti Ventus has a max power limit of 104% TDP. This card is not designed for massive overclocking (and don’t forget that the card is already overclocked). So I set the power limit to 104% TDP using MSI Afterburner. If you don’t set the power limit to the max, the GPU is throttled when FurMark is running.

In previous reviews, I gave an estimation of the power consumption using the total power consumption minus the CPU power consumption and I moduled the result with the PSU efficiency factor. This is a rough estimation. Now with recent GeForce, we have several sensors dedicated to the power draw. Then it’s easier to get the power consumption of the entire graphics card as well as the power draw of the GPU chip itself. GPU-Z has all those sensors.

Idle state
– total power consumption: 50W
GPU temperature: 32°C
Board power draw: 16W
GPU chip power draw: 12W

Load state – 104% TDP
– total power consumption: 280W
GPU temperature: 61°C
Board power draw: 205W
GPU chip power draw: 164W

Let’s do another test, this time with the MSI-01 test of MSI Kombustor.

unnamed file 4694

unnamed file 4695

Load state – 104% TDP
– total power consumption: 300W
GPU temperature: 62°C
Board power draw: 210W
GPU chip power draw: 180W

The GPU temperature reached a max value of 62°C which is very nice. The reference RTX 3060 Ti has a TGP (total graphics power = total graphics card power draw) of 200W. This overclocked RTX 3060 Ti has a power draw that reaches 213W which is ok.

The nice thing is that this RTX 3060 Ti is able to run at 213W / 62°C very quietly. The VGA cooler is efficient!

6 – Conclusion

In most tests, this GeForce RTX 3060 Ti is faster than the previous generation RTX 2080. In some particular tests, like RhodiumLC (a pure pixel shader / procedural test), the RTX 3060 Ti is slower than the RTX 2080. Ok, more or less, the RTX 3060 Ti has the same level of performances than the RTX 2080. Now if we take into account the power draw, the RTX 3060 Ti is suddenly a better choice. In full load, the RTX 2080 has a power draw of 250W while the RTX 3060 Ti is limited to 200W.

The RTX 3060 Ti has 4864 CUDA cores. It’s nearly twice as many cores as the RTX 2080 (2944). Something does not work… until we consider the number of compute units: the RTX 3060 Ti has 38 compute units (128 cores per compute unit on Ampere architecture) and the RTX 2080 has 46 compute units (64 cores per compute unit on Turing architecture). From the compute unit point of view, the Ampere architecture offers more performance that’s why the RTX 3060 Ti is more or less equivalent to the RTX 2080. And in some tests (the pure pixel shader test, the tessellation test), the higher number of compute units of the RTX 2080 is an advantage.

The passive / active VGA cooler does its job very well. At idle, the RTX 3060 Ti is absolutely quiet because the fans are not spinning. Under heavy load, fans spin but remain quiet.

For graphics developers, this card is perfect because all features are present: Ampere GPU with all latest techs, OpenGL 4.6, Vulkan 1.2, Direct3D 12, OpenCL, CUDA, PhysX, realtime raytracing and serious graphics drivers. What else?

I’m currently working on Vulkan raytracing and I hope to have in the next weeks some cool / interesting demos to test all those RT cores.

unnamed file 4696

 

作者 frank

发表评论