(Tested) ASUS GeForce GTX 590 Dual-GPU Graphics Card Review (***Updated with HD6970 CF***)

ASUS GeForce GTX 590 Dual-GPU Graphics Card


1 – ASUS GTX 590 Overview

The GeForce GTX 590, launched several months ago (March 2011), is a graphics card powered by two GF100 GPUs. But in order to reduce the power consumption, both GF110 come with a lower GPU clock speed compared to the GeForce GTX 580: 607MHz for the GTX 590 against 772MHz for the reference GTX 580 board.

ASUS’s GTX 590 is available with GPU core clock speed slightly overclocked with an amazing +5MHz: 612MHz. The rest of card follows NVIDIA’s reference board specifications: 3GB of GDDR5 memory (1.5GB per GPU) clocked at 1710MHz DDR speed (or 855MHz real speed), the same two-slot height VGA cooler, and two 8-pin power connectors. For more information, check out ASUS’s GTX 590 homepage.

2 – ASUS GTX 590 Gallery

ASUS GeForce GTX 590 Dual-GPU Graphics Card

ASUS GeForce GTX 590 Dual-GPU Graphics Card

ASUS GeForce GTX 590 Dual-GPU Graphics Card

ASUS GeForce GTX 590 Dual-GPU Graphics Card

ASUS GeForce GTX 590 Dual-GPU Graphics Card

3 – ASUS GTX 590 GPU Caps Viewer Details

The following screenshots come from the versions 1.14.0 and 1.13.0 of GPU Caps Viewer.

ASUS GeForce GTX 590 Dual-GPU Graphics Card, GPU Caps Viewer details

ASUS GeForce GTX 590 Dual-GPU Graphics Card, GPU Caps Viewer OpenCL details

ASUS GeForce GTX 590 Dual-GPU Graphics Card, GPU Caps Viewer CUDA details

4 – ASUS GTX 590 GPU Benchmarks

Testbed:

The testbed:
– CPU: Core i7 960 @ 3.2GHz
– RAM: 4GB DDR3 Corsair Dominator
– Motherboard: GIGABYTE X58-A UD5
– Windows 7 64-bit
– Graphics drivers: R280.26 WHQL
– PSU: Thermaltake Thoughpower Grand 850W

Thermaltake Thoughpower Grand 850W

FurMark 1.9.1 (OpenGL 2)

Settings: Preset 1080 (1920×1080 fullscreen, no AA, no postFX, 60sec, Burn-in mode).

3517 points (58 FPS) – GeForce GTX 480 SLI
3457 points (57 FPS) – SAPPHIRE Radeon HD 6970 CrossFire
2463 points (41 FPS) – ASUS ROG MATRIX GTX 580 (GPU@926MHz)
2335 points (38 FPS) – MSI GeForce GTX 460 Cyclone SLI
1769 points (29 FPS) – GeForce GTX 480
1618 points (26 FPS) – ASUS GeForce GTX 590 (GPU@617MHz)
1183 points (19 FPS) – MSI GeForce GTX 460 Cyclone

TessMark 0.3.0 (OpenGL 4)

Settings: 1920×1080 fullscreen, no AA, no postFX, 60sec, tessellation level: moderate (8X).

60766 points (1012 FPS) – SAPPHIRE Radeon HD 6970 CrossFire (GPU@880MHz)
53863 points (898 FPS) – ASUS ROG MATRIX GTX 580 (GPU@926MHz)
52188 points (872 FPS) – ASUS ENGTX580
45723 points (763 FPS) – MSI GeForce GTX 460 Cyclone SLI
44090 points (735 FPS) – SAPPHIRE Radeon HD 6970
38147 points (636 FPS) – ASUS GeForce GTX 590 (GPU@617MHz)
30512 points (508 FPS) – MSI GeForce GTX 460 Cyclone

Settings: 1920×1080 fullscreen, no AA, no postFX, 60sec, tessellation level: normal (16X).

36019 points (601 FPS) – ASUS ROG MATRIX GTX 580 (GPU@926MHz)
35890 points (598 FPS) – SAPPHIRE Radeon HD 6970 CrossFire (GPU@880MHz)
32666 points (545 FPS) – ASUS ENGTX580
30179 points (503 FPS) – MSI GeForce GTX 460 Cyclone SLI
25275 points (421 FPS) – ASUS GeForce GTX 590 (GPU@617MHz)
19398 points (323 FPS) – SAPPHIRE Radeon HD 6970

ShaderToyMark 0.2.0 (OpenGL 2 – Pixel shader)

Settings: 1920×1080 fullscreen, no AA, 60sec.

173 points (28 FPS) – SAPPHIRE Radeon HD 6970 CrossFire (GPU@880MHz)
109 points (18 FPS) – MSI GeForce GTX 460 Cyclone SLI (AFR1 mode)
88 points (14 FPS) – SAPPHIRE Radeon HD 6970 (GPU@880MHz)
81 points (13 FPS) – ASUS GeForce GTX 590 (GPU@617MHz)

Unigine Heaven 2.0 (Direct3D 11)

Settings: 1920×1080 fullscreen, AA: 4X MSAA, 16X anisotropic filtering, tessellation: normal, shaders: high.

73.6 FPS, Scores: 1854 – ASUS GeForce GTX 590 (GPU@617MHz)
65.7 FPS, Scores: 1656 – SAPPHIRE Radeon HD 6970 CrossFire (GPU@880MHz)
56.4 FPS, Scores: 1421 – ASUS ROG MATRIX GTX 580 (GPU@926MHz)
50.2 FPS, Scores: 1264 – MSI GeForce GTX 460 Cyclone SLI (AFR1 mode)
42.9 FPS, Scores: 1081 – SAPPHIRE Radeon HD 6970

3DMark 11 (Direct3D 11)

Settings: Extreme mode (1920×1080).

X3502 – SAPPHIRE Radeon HD 6970 CrossFire (GPU@880MHz)
X3057 – ASUS GeForce GTX 590 (GPU@617MHz)
X2432 – ASUS ROG MATRIX GTX 580 (GPU@926MHz)
X2151 – MSI GeForce GTX 460 Cyclone SLI (AFR1 mode)
X1812 – SAPPHIRE Radeon HD 6970

DiRT 2 (Direct3D 11)

Settings: 1920×1080, 4X MSAA.

160 FPS – ASUS ROG MATRIX GTX 580 (GPU@926MHz)
155 FPS – ASUS GeForce GTX 590 (GPU@617MHz)
151 FPS – EVGA GeForce GTX 580 SC
130 FPS – MSI GeForce GTX 460 Cyclone SLI
129 FPS – SAPPHIRE Radeon HD 6970 CrossFire (GPU@880MHz)
94 FPS – SAPPHIRE Radeon HD 6970

MAFIA 2 (Direct3D 9)

Settings: 1920×1080, no MSAA, 16x anisotropic filtering, CPU PhysX.

116 FPS – ASUS GeForce GTX 590 (GPU@617MHz)
116 FPS – ASUS ROG MATRIX GTX 580 (GPU@926MHz)
106 FPS – EVGA GeForce GTX 580 SC
103 FPS – MSI GeForce GTX 460 Cyclone SLI
89 FPS – SAPPHIRE Radeon HD 6970
80 FPS – SAPPHIRE Radeon HD 6970 CrossFire (GPU@880MHz)

AvP: Aliens vs Predator (Direct3D 11)

Settings: 1920×1080, Texture Quality: 2, Shadow Quality: 3, Anisotropic Filtering: 16, SSAO: ON, Vertical Sync: OFF.

130.8 FPS – SAPPHIRE Radeon HD 6970 CrossFire (GPU@880MHz)
103.7 FPS – ASUS GeForce GTX 590 (GPU@617MHz)
79.4 FPS – ASUS ROG MATRIX GTX 580 (GPU@926MHz)
71.6 FPS – EVGA GeForce GTX 580 SC
71.1 FPS – MSI GeForce GTX 460 Cyclone SLI
64.3 FPS – SAPPHIRE Radeon HD 6970

5 – ASUS GTX 590 GPU Overclocking

To overclock the GTX 590, I used MSI Afterburner because it allowed to overclock both GPUs while I don’t know why, ASUS’s GPU Tweak could only overclock one GPU. For this OC test, I used 3DMark 11 (Extreme mode) as well as Unigine Heaven 2 (D3D 11) to validate the GPU overclocking. The overclocking is stable is both tests are okay.

The max stable GPU clock I found is 670MHz, without change in GPU voltage (0.913V). Here are the 3DMark 11 and Heaven 2.0 scores with a GPU core clock of 670MHz:

Unigine Heaven 2.0 (DX11)
Settings: 1920×1080 fullscreen, AA: 4X MSAA, 16X anisotropic filtering, tessellation: normal, shaders: high.

78.2 FPS, Scores: 1969 – ASUS GeForce GTX 590 (GPU@670MHz)
73.6 FPS, Scores: 1854 – ASUS GeForce GTX 590 (GPU@617MHz)

3DMark 11 (DX11)
Settings: Extreme mode (1920×1080).

X3455 – ASUS GeForce GTX 590 (GPU@670MHz)
X3057 – ASUS GeForce GTX 590 (GPU@617MHz)

And the power consumption test with FurMark? NVIDIA has totally blocked FurMark with a too aggressive OCP (OCP = Over Current Protection –check out the score with FurMark 1.9.x) because the green firm wants to hide something…

But to give you an idea of the GTX 590 power draw under approved and gentle apps, here is the total power consumption of the tesbed at idle and under Heaven (1920×1080, DX11):
– Idle: 150W
– Load: 450W

The CPU power consumption under Heaven is roughly approximated at 25W. Then we can estimate the GTX 590 power draw with a PSU efficiency factor or 0.9:
(450 – 150 – 25) * 0.9 = 247W

The TDP of the GTX 590 is 365W. With Unigine Heaven, we are far from this value. Wait a minute… I just found that one of my stress test apps is not too much capped by NV drivers, hehe… Here is the total power draw with this secret app:
– Load: 545W
Which leads to:
(545 – 150 – 25) * 0.9 = 333W

Do the GPUs run at full speed with this secret stress test app or is there a little capping? I don’t know.

6 – ASUS GTX 590 Conclusion

There are two problems with the GTX 590: correct support of SLI and power limiting system. When the application has an optimized SLI profile in NVIDIA drivers, GTX 590 GPUs are correctly exploited and you get high scores like in AvP, 3DMark11 or Unigine Heaven. But in other cases, the gain of two GF110 GPUs is not so obvious like in MAFIA 2 or in DiRT 2. The worst case is ShaderToyMark: a SLI of GTX 460 gets a higher score than the GTX 590.

But for me the biggest problem of the GeForce GTX 590 is the power limiting system. FurMark is limited with much exaggeration since the launch of the GTX 580 and with latest NV drivers and the launch of the GTX 590, we don’t know anymore if an application is capped or not by the power limiter. For example, scores in TessMark or in ShaderToyMark are meaningless: is it a problem of SLI profile or is there a power limitation?

If you know that the games you plan to play have a good support of the GTX 590 and if a high price for two GPUs is not a problem for you, then the GTX 590 is a correct choice, this card being in the top in many well profiled benchmarks 😉 Two GTX 460 / GTX 560 Ti in SLI would be, perhaps, a better idea: much less expensive for correct performance. Unless you really need a dual-GF110 card or you can’t live without a quad-SLI system (for 3D surround gaming…), robust single GPU cards (read: without hardware power limiter!) like ASUS’s Matrix GTX 580 or ASUS’s GTX 570 DirectCU II are, IMHO, good alternatives. Now for a very robust dual-GF110 card, I’m sure that ASUS’s latest product, the MARS II (a kind of dual Matrix card), will satisfy your craving for big GPUs with no concession!

Geforce GTX 590 burns @ SweClockers.com



3 thoughts on “(Tested) ASUS GeForce GTX 590 Dual-GPU Graphics Card Review (***Updated with HD6970 CF***)”

  1. SMiThaYe

    Good review. If anyone buys this monster GPU they will have 3 displays for surround gaming, that would show the true potential of this GPU. Still, I think the 480 SLI would give it a good run for its money and is still an outstanding product – in comparison the 590 look quite ordinary.

    To the author: In the Furmark bench you had the GPUs in CFX/SLI, but the next batch of benches the others are described as single GPU (bar 460), why you didn’t continue to use CFX/SLI against a premium 600 product as that’s what most people expect?

  2. JeGX Post Author

    @SMiThaYe: you’re right, I will update the benchmarks with a CrossFire of HD 6970. Will do it asap!

Comments are closed.