Introduction

The RADEON HD 5670 from AMD represents a very nice addition to the 5xxx series, features DirectX11 support and does not need an extra power connector to function. The card was aimed from the beginning against the GT 240 from Nvidia, let's find out if it stands a chance!



At first I would like to thank AMD for making this review possible.



About AMD:

“ Advanced Micro Devices, Inc. (AMD) (NYSE: AMD) is an American multinational semiconductor company based in Sunnyvale, California, that develops computer processors and related technologies for commercial and consumer markets. Its main products include microprocessors, motherboard chipsets, embedded processors and graphics processors for servers, workstations and personal computers, and processor technologies for handheld devices, digital television, automobiles, game consoles, and other embedded systems applications.

AMD is the second-largest global supplier of microprocessors based on the x86 architecture and also one of the largest supplier of graphics processing units. It also owns 8.6% of Spansion, a supplier of non-volatile flash memory.[2] In 2009, AMD ranked ninth among semiconductor manufacturers in terms of revenue.”

Product Features

• Expand your visual real estate across up to three displays and get lost in the action with ATI Eyefinity Technology
  
• Using ATI Stream technology, accelerate demanding applications and do more than ever with your PC
  
• Enjoy rich, realistic visuals with explosive performance
  
• Consumes less than 75 Watts under full load, eliminating the need for an external power connection and making installation a snap
  
• ATI CrossFireX™ technology dual mode support
  
• Ultra-high bandwidth GDDR5 memory

Product Specifications

The RADEON HD 5600 series take part from the budget segment of DirectX11 video cards offering from AMD. Nvidia does also have some cards that fit in the budget series, on 40nm, the GeForce GT 240; unfortunately, these do not support DirectX11, but only 10.1.

The affordable 5670 video card we are going to test next shares most of the features with the higher performance video cards from the 5000 series (including Eyefinity, meaning that we can connect up to 3 monitors on a single video card [if the video card version features at least one DisplayPort]).



The card is very surprising in the power consumption department, meaning that we do not have to use an extra PCI Express power connector, all is provided by the PCI Express 2.0 slot, which does have a 75W power limitation. This makes it a very good choice for HTPC configurations too.

Redwood block diagram



The Redwood architecture does have 5 SIMD units, each containing 80 ALUs and tied to one texture unit. After doing the math, we get a total of 400 stream processors (or ALUs), compared to the RADEON HD 5770 which has a double number of processors; the ROPs have also been reduced from 16 to 8. The 5670 does share, though, the same memory interface width with the 5770: 128bit.

The latest GPUs from AMD fully support DirectX11 instructions which include GPGPU (DirectCompute 11), tessellation and improved multi-threading; they also come with Shader Model 5, better shadows and HDR texture compression.

Tesselation, as described on the Unigine website, is a “scalable technology aimed for automatic subdivision of polygons into smaller and finer pieces, so that developers can gain a more detailed look of their games almost free of charge in terms of performance. Thanks to this procedure, the elaboration of the rendered image finally approaches the boundary of veridical visual perception: the virtual reality is vivified at your fingertips delivering engaging gaming experience.”

Here is a modeled house inside the Unigine Heaven benchmark, without and with the tessellation feature enabled:

The multi-threaded rendering is similar to the techniques applied for the current CPUs. If a shader or an instruction has to be queued up, the process creates a delay. The current GPUs can now process data completely threaded, which bring a better overall performance.

The DirectCompute feature allows access to the GPU for stream computing; it shares a range of computational interfaces with its competitors: OpenCL and CUDA.

Eyefinity is an advanced multiple-display technology from AMD which enables a single GPU to support up to six independent display outputs simultaneously. The six high-resolution displays can be operated simultaneously and independently, configured in various combinations of landscape and portrait orientations.

We can group multiple monitors into a large integrated display surface, enabling windowed and full screen 3D applications, images and video to span across multiple displays as one desktop workspace. ATI Eyefinity supports Duplicated Mode operation (PC desktop cloned on multiple displays) and Extended Mode (PC desktop extended across multiple displays).

A Closer Look:

The card has arrived as bulk, protected inside an anti-static bag:



The top side of the black PCB reveals a black cooler, which is really small and does not have any fancy stickers on it:



The fan inside is red, like the ATI logo and is meant to suck air from the environment and blow it on the passive heatsink inside to cool the GPU along with the memory around it:



When looking from the other side, we can discover that the fan does have only two wires (because of this I could not monitor its exact RPM speed with software applications like GPU-Z) and the special designed heatsink which also covers the RAM, that is designed in such a way to cool better when air goes through its little “spikes”; I have seen that lateral case fans which blow air directly onto the card help a lot (the temperature on the GPU can be reduced with at least 2-3 degrees:



On the back of the PCB, we can see the metal frame that is used to distribute the forces evenly and the rest of the memory chips:



The Samsung K4G10325FE-HC05 memory chips are set to operate at 2000MHz and we can find on the card eight of them:



Crossfire connectors are also present on the card:



The card features one D-SUB, a HDMI and a DVI port:

Test Setup

Stefan's Test Setup
CPU	Core i7 920 @ 3.2Ghz
Cooling	Cogage True Spirit
Mainboard	Asrock X58 Extreme
Memory	G.Skill F3-12800CL8T-6GBPI PI Series
VGA	AXLE GeForce 210 512MB DDR2 Gainward Geforce GTS 250 Deep Green Gainward Geforce GT240 Inno3D Geforce GT240 Sparkle Geforce GTS 250 LP ATI Radeon HD 4550 ATI Radeon HD 4890 ATI Radeon HD 5770 ATI Radeon HD 5830 ATI Radeon HD 5450 ATI Radeon HD 5670
Other	Power Supply : Antec True Power New 750W HDD : Seagate Barracuda 320GB 7200.10 Case : Cooler Master HAF922 Case with 2 NB-Multiframe S-Series MF12-S3HS@1800RPM on the side

With the help of the GPU-Z 0.4.4 utility, we can extract lots of information regarding the video card clocks,memory type, pixel and texture fill rate and so on:



I have used the Everest Ultimate Edition utility to find out the ATI PowerPlay pre-defined clocks and some more information regarding the GPU Code name, Part Number or BIOS Version:



With the same utility, we can find in-depth information regarding the GPGPU capabilities:



To find out the IDLE and maximum temperatures, I have used the Furmark utility, with the Stability Test in “Xtreme Burning mode” and let it run for about 15 minutes. To record the temperatures, the GPU-Z utility was used, thanks to its Sensors monitoring tab. The recorded room temperature at the time of testing was 28.7 degrees Celsius:

IDLE



Full Load



In IDLE mode, the card remains very silent, but after 15 minutes with the GPU at 100% load, the cooling system starts to be heard, but all within acceptable levels.

Synthetic Benchmarks

We first converted the obtained scores/FPS for each benchmark to a percentage in comparison to the highest score (=100%), we then averaged all the benchmark scores per resolution/IQ and the results are put into these charts. So these are relative performance charts in % to the fastest card. (full performance numbers here)





In the first batch of synthetic benchmarks without AA, the RADEON HD 5670 is just above the GT 240 video cards from Nvidia, but the difference gets bigger as the resolution increases





When 4xAA is added, we get exactly the same behavior: the 5670 does better as the resolution is increased, leaving the GT 240 in the dust.

Games Benchmarks

Same as on the previous page, we first converted the obtained scores/FPS for each benchmark to a percentage in comparison to the highest score (=100%), we then averaged all the benchmark scores per resolution/IQ and the results are put into these charts. So these are relative performance charts in % to the fastest card. (full performance numbers here)





In the Games Benchmarks without AA filtering enabled, the 5670 video card is still on top of the GT 240 and builds a solid 4.8% difference between it and the 240 at the 1920x1200 resolution.





When 4xAA is added, we can see a 5.68% difference between the 5670 and the GT 240.

Conclusive Thoughts

The RADEON HD 5670 1GB from AMD succeeded to best the similarly priced Geforce GT 240 in all synthetic and games benchmarks, no matter the resolution and AA setting. Considering that this do not need an extra power connector and fits perfectly into the PCIe power limit specifications, it makes for a higher end HTPC friendly solution, if your HTPC has enough cooling power.

During the tests, the noise remained at acceptable levels even in the most stressful conditions like the Furmark Xtreme Burning mode; this is good because the card comes with a single slot cooling system, taking up very little space inside our case.

Considering that they can be found in stores for as low as €72 and do not require an expensive and big power supply, they are an interesting option for lots of us, considering that the GT 240 from Nvidia can be found in the same price range. Going Crossfire with two identical cards seems also an option to consider, without the need of extra wires, bridges or dongles.



I would like to thank again to AMD for making this product review possible.