Testing methodology
GPU Comparisons |
|||||||
|---|---|---|---|---|---|---|---|
| Graphics Card | Peak GPU Clock (MHz) |
Stream Processors |
Peak Shader Clock (MHz) |
Memory Clock (MHz) |
Memory Bus (bits) |
Graphics Driver | |
| AMD Radeon R9 290X 4GB | 1,000 | 2,816 | 1,000 | 5,000 | 512 | Catalyst 13.11 v8 | |
| AMD Radeon R9 290 4GB | 947 | 2,560 | 947 | 5,000 | 512 | Catalyst 13.11 v8 | |
| AMD Radeon R9 280X 3GB | 1,000 | 2,048 | 1,000 | 6,000 | 384 | Catalyst 13.11 v1 | |
| AMD Radeon HD 7970 GE 3GB | 1,050 | 2,048 | 1,050 | 6,000 | 384 | Catalyst 13.11 v1 | |
| AMD Radeon HD 7950 3GB | 800 | 1,792 | 800 | 5,000 | 384 | Catalyst 13.11 v1 | |
| AMD Radeon R9 270X 2GB | 1,050 | 1,280 | 1,050 | 5,600 | 256 | Catalyst 13.11 v1 | |
| AMD Radeon HD 7870 2GB | 1,000 | 1,280 | 1,000 | 4,800 | 256 | Catalyst 13.11 v1 | |
| AMD Radeon R7 260X 2GB | 1,100 | 896 | 1,100 | 6,400 | 128 | Catalyst 13.11 v1 | |
| Nvidia GeForce GTX Titan 6GB | 876 | 2,688 | 875 | 6,008 | 384 | GeForce 331.40 | |
| Nvidia GeForce GTX 780 3GB | 900 | 2,304 | 900 | 6,008 | 384 | GeForce 331.40 | |
| Nvidia GeForce GTX 770 2GB | 1,085 | 1,536 | 1,085 | 7,012 | 256 | GeForce 331.40 | |
| Nvidia GeForce GTX 760 2GB | 1,033 | 1,152 | 1,033 | 6,008 | 256 | GeForce 331.40 | |
| Nvidia GeForce GTX 670 2GB | 980 | 1,344 | 980 | 6,008 | 256 | GeForce 331.40 | |
| Nvidia GeForce GTX 660 2GB | 1,032 | 960 | 1,032 | 6,008 | 192 | GeForce 331.40 | |
| Nvidia GeForce GTX 650 Ti Boost 2GB | 1,032 | 768 | 1,032 | 6,008 | 192 | GeForce 331.40 | |
HEXUS Graphics Test Bench |
|
|---|---|
| Processor | Intel Core i7-4770K (3.50GHz, 8MB cache, quad-core) |
| CPU Cooler | Intel reference E97378-001 |
| Motherboard | Gigabyte GA-Z87X-UD3H |
| Memory | 16GB G.Skill RipJaws (2x8GB) DDR3 @ 1,600MHz |
| Power Supply | Corsair AX760i |
| Storage Device | Crucial M500 240GB SSD |
| Chassis | Corsair Graphite Series 600T |
| Monitor | Philips Brilliance 272P (2,560x1,440) |
| Operating system | Windows 8 64-bit |
HEXUS High-End Benchmark Suite |
||
|---|---|---|
| GPU Benchmarks | Mode and Resolutions | Quality Settings |
| 3DMark | DX11 at 1,920x1,080 and 2,560x1,440 | Fire Strike and Fire Strike Extreme |
| BioShock Infinite | DX11 at 1,920x1,080 and 2,560x1,440 | Ultra + DOF |
| Crysis 3 | DX11 at 1,920x1,080 and 2,560x1,440 | 4xMSAA, High Preset |
| Far Cry 3 | DX11 at 1,920x1,080 and 2,560x1,440 | 2xMSAA, Ultra Preset |
| GRID 2 | DX11 at 1,920x1,080 and 2,560x1,440 | 4xMSAA, Ultra Preset |
| Just Cause 2 | DX10 at 1,920x1,080 and 2,560x1,440 | 8xMSAA, High Preset |
| Total War: Rome II | DX11 at 1,920x1,080 and 2,560x1,440 | 4xMSAA, High Preset |
| General Benchmarks | Description | |
| Power Consumption | To emulate real-world usage scenarios, we record mains power draw both when idle and while playing Far Cry 3 | |
| Temperature | To emulate real-world usage scenarios, we record GPU core temperature both when idle and while playing Far Cry 3 | |
| Noise | A PCE-318 meter is used to record noise levels when idle and while playing Far Cry 3 | |
Notes
AMD pulled a last-minute stunt by pushing back the launch of this card due to the arrival of a performance driver referred to as Catalyst 13.11 beta 8. Though we were content to run with the results already obtained on Catalyst 13.11 beta 5 - the Radeon R9 290X's launch driver - we have retested both the R9 290 and R9 290X on the latest version. We have also tested the R9 290X in both standard and uber modes.
So why the fuss about Catalyst 13.11 beta 8 for the R9 290 in particular? Well, AMD says: '...given that the R9 290 has so much untapped performance headroom, this new driver will allow it to beat GTX780, and achieve higher framerates than the Titan when running 4K resolutions'. That's a bombast statement indeed.
Untapped frequency headroom is another way of saying that AMD can boost the clocks to increase performance. Note that the GPU is specified to run at a peak 947MHz under optimum conditions. Here's Far Cry 3 run at 2,560x1,440 with the per-second frequency logged over a 10-minute period.
What you're looking at is the core frequency when using the Catalyst 13.11 beta 5 launch driver and the 'new-and-improved' Catalyst 13.11 beta 8, released last Thursday evening. The GPU is actually switching frequencies far more rapidly than we can graph, but what's clear is that the R9 290 sample is being locked to 947MHz using the newer driver but allowed to modulate far more with the older version.
The impact of such a move is most keenly felt when the game becomes GPU bound, usually at a 2,560x1,440 resolution on this card. This is why AMD says the new driver provides Titan-beating performance at 4K. By locking the frequency to the maximum AMD forces the R9 290 into peak mode whenever gaming. One side effect of such a move is that the fan has to work harder to keep the GPU below 95°C, and you'll see the results in the noise section.
Put simply, AMD has artificially increased the speed of the Radeon R9 290, especially at higher resolutions, by using a driver that imitates the uber mode of the R9 290X. It is therefore interesting to note that this second-rung card runs at a higher in-game frequency than a standard R9 290X.