Testing Methodology
KFA2 GeForce RTX 3080 SG |
|---|
 |
GPU Comparisons |
||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
GPU |
Memory |
Miscellaneous |
||||||||
Cores |
Base (MHz) |
Boost (MHz) |
Type |
Size (MB) |
Clock (MHz) |
Bus (bits) |
Power Config |
Driver |
||
| Gigabyte GeForce RTX 3090 Gaming OC | 10,496 |
1,395 |
1,755 |
GDDR6X |
24,576 |
19,500 |
384 |
8+8 |
456.38 |
|
| MSI GeForce RTX 3080 Gaming X Trio | 8,704 |
1,440 |
1,815 |
GDDR6X |
10,240 |
19,000 |
320 |
8+8+8 |
456.16 |
|
| KFA2 GeForce RTX 3080 SG | 8,704 |
1,440 |
1,710 |
GDDR6X |
10,240 |
19,000 |
320 |
8+8 |
456.16 |
|
| Nvidia GeForce RTX 3080 Founders Edition | 8,704 |
1,440 |
1,710 |
GDDR6X |
10,240 |
19,000 |
320 |
12 |
456.16 |
|
| Nvidia GeForce RTX 2080 Ti Founders Edition | 4,352 |
1,350 |
1,635 |
GDDR6 |
11,264 |
14,000 |
352 |
8+8 |
456.16 |
|
| Nvidia GeForce RTX 2080 Super Founders Edition | 3,072 |
1,650 |
1,815 |
GDDR6 |
8,192 |
15,500 |
256 |
6+8 |
456.16 |
|
| Nvidia GeForce RTX 2080 Founders Edition | 2,944 |
1,515 |
1,800 |
GDDR6 |
8,192 |
14,000 |
256 |
6+8 |
456.16 |
|
| Nvidia GeForce RTX 2070 Super Founders Edition | 2,560 |
1,605 |
1,770 |
GDDR6 |
8,192 |
14,000 |
256 |
6+8 |
456.16 |
|
| Nvidia GeForce GTX 1080 Ti Founders Edition | 3,584 |
1,481 |
1,582 |
GDDR5X |
8,192 |
11,000 |
352 |
6+8 |
456.16 |
|
| AMD Radeon RX 5700 XT | 2,560 |
1,605 |
1,905 |
GDDR6 |
8,192 |
14,000 |
256 |
6+8 |
20.8.3 |
|
| AMD Radeon RX 5700 | 2,304 |
1,465 |
1,725 |
GDDR6 |
8,192 |
14,000 |
256 |
6+8 |
20.8.3 |
|
HEXUS 2020 Test Platform |
||
|---|---|---|
| Component | Product Page | |
| Processor | AMD Ryzen 9 3950X | amd.com |
| Motherboard | Asus ROG Crosshair VIII Formula | asus.com |
| CPU Cooler | Corsair Hydro Series H150i Pro RGB | corsair.com |
| Memory | G.Skill Trident Z Neo DDR4-3200 (2x16GB) | gskill.com |
| Storage | 2TB Corsair MP600 | corsair.com |
| Power Supply | be quiet! Straight Power 11 Platinum 1,000W | bequiet.com |
| Network Card | Asus XG-C100C 10GBase-T PCIe Adapter | asus.com |
| Chassis | Fractal Design Define 7 Clear Tempered Glass | fractal-design.com |
| Monitor | Philips Momentum 436M6VBPAB | philips.co.uk |
| Keyboard | Corsair K70 RGB MK.2 Low Profile Rapidfire | corsair.com |
| Mouse | Corsair Ironclaw RGB | corsair.com |
| Operating System | Microsoft Windows 10 Pro | microsoft.com |
Benchmark Suite |
||
|---|---|---|
| Synthetic Benchmarks | ||
| 3DMark | Time Spy, Time Spy Extreme, Time Spy Extreme Stress Test | |
| Gaming Benchmarks and Settings | ||
| Borderlands 3 | DX12, TAA, Badass Quality | |
| Civilization VI: Gathering Storm | DX11, 4xMSAA, Ultra Quality | |
| Far Cry: New Dawn | DX11, TAA, HD Textures Enabled, HDR10, Ultra Quality | |
| Forza Horizon 4 | DX12, 8xMSAA, Maximum Quality | |
| Gears 5 | DX12, TAA, Ultra Textures Enabled, Ultra Quality | |
| Shadow of the Tomb Raider | DX12, TAA, Highest Quality | |
| General Benchmarks | ||
| Power Consumption | To emulate real-world usage scenarios, we record system-wide mains power draw when idle and while playing Gears 5 | |
| Temperature | To emulate real-world usage scenarios, we record GPU core temperature when idle and while playing Gears 5 | |
| Noise | A PCE-318 meter is used to record noise levels when idle and while playing Gears 5 | |
Notes
KFA2 offered us a second BIOS that lifts the card's power budget from 320W to 340W and fixes a situation where setting a manual fan speed results in incorrect values. That said, the BIOS doesn't change the card's core frequencies so we ran with the original for the numbers you see on the following pages. It is unclear whether the second BIOS is available on the small smattering of retail cards in the wild.
We observed no crashes during testing with the 456.16 driver. Switching over to an Intel Core i9-10900K platform powered by a Corsair HX1000 PSU exhibited rock-solid stability, so our sample card is not blighted by the ongoing power-delivery issues reported by some users. Further, we ran a subset of numbers with 456.55, the latest driver that is supposed to address said issues, and found the benchmark numbers to be within an acceptable standard deviation from one driver to the next.