by Alexey Stepin , Yaroslav Lyssenko, Anton Shilov
12/28/2008 | 12:23 PM
The year of 2008 is nearly over, and we can say it quite confidently now that it has been a very lucky year for AMD’s graphics department ATI. This year has been crucial for the company. It is in 2008 that ATI finally recovered from a long crisis that had been provoked by management errors and the lack of competitive products, and changed the market situation in its favor. The release of the RV770 GPU proved that ATI was not only capable of rivaling Nvidia but could create solutions head above the opponent’s. We don’t need much evidence to prove our point. Suffice it to say that the ATI Radeon HD 4850 with a recommended price of $199 could deliver the same performance as the Nvidia GeForce 9800 GTX that had originally appeared at a recommended price of $349!
Of course, Nvidia won’t give up its market position without a good fight. But ATI is still counterattacking on each front, steadily and methodically. First of all, they drove Nvidia out of the premium-class sector. Having two RV770 cores, the ATI Radeon HD 4870 X2 left no chance to the Nvidia GeForce GTX 280 even though at a high price in every sense of the word: the card has a tremendous level of power consumption. The recent release of the RV730 GPU and appropriate cards has also been a success. It makes ATI not only competitive but superior in the sector of entry-level gaming cards priced at less than $100. The excellent results of the Radeon HD 4670 are indicative of that superiority.
Quite expectedly, ATI’s next strike should have occurred in the sector of even less expensive discrete graphics cards. Although such cards are not meant for serious gaming, they can offer the user such capabilities that cannot be provided by integrated graphics cores, i.e. full-featured HD video decoding, audio-over-HDMI, support for multi-monitor configurations, DisplayPort, and more or less acceptable performance in games, at least at low resolutions. Integrated graphics cores are generally very slow in 3D games even though they have been improving in this respect.
Unlike Nvidia, ATI did not use a cut-down version of the high-category core (RV730) for this purpose. Instead, the company has released a separate processor called RV710 that is meant to deliver a new level of performance and capabilities in the below-$60 sector. Thus, the Radeon HD 4000 family has been complemented with Radeon HD 4500 and 4300 series. The former of them will be the subject of this review.
The ATI Radeon HD 4500/4300 series is based on the 55nm RV710 core. This core is simpler than the RV730 in many respects as we will describe below after we compare the new chip’s specs with those of Nvidia’s products.
Click to enlarge
Click to enlarge
The RV710 is the simplest and smallest entry-level GPU in the table. It incorporates a mere 242 million transistors and measures 73 square mm (40% smaller than the 55nm G96 core from Nvidia). However, the new GPU boasts impressive computing potential. It has as many as 80 ALUs, which is 2.5 times as much as in the full-featured G96. Each five ALUs are grouped into one execution module and each eight modules are combined into a SIMD array serviced by four texture processors. Thus, the RV710 incorporates two SIMD arrays and eight texture processors, which is impressive in comparison with the GeForce 9400 GT as well as with the more expensive GeForce 9500 GT, the market opponent to the Radeon HD 4670. Now let’s check out the architectural features of the new core.
As opposed to the RV770/730, the RV710 does not have separate texture and vertex caches. Its rasterization subsystem has been simplified relative to the RV770/730. The raster back-ends are twice slower at blending operations whereas the hardware MSAA resolve is limited to 4x as opposed to 8x of the senior GPU models. This simplification is reasonable because it helped reduce the amount of transistors in the RV710 without cutting down the overall functionality of the core. After all, this GPU is not meant for playing at high resolutions with high-level antialiasing.
The memory subsystem is represented by one memory controller servicing two 32-bit channels. Thus, the total width of the external memory bus is 64 bits. The use of rather fast DDR3 memory helps achieve the same memory bandwidth as that of the GeForce 9400 GT on a simpler PCB because there is no need to wire a 128-bit memory bus. The simpler PCB lowers the manufacturing cost of the product. So, we won’t be surprised to see the developer cut the price of the Radeon HD 4550 to make it even more appealing in comparison with Nvidia’s products from the same category.
The multimedia subsection of the core has been left intact. It incorporates the industry-leading UVD2 video-processor that offers full support for decoding of both H.264 and VC-1 formats. Nvidia’s solutions are still deficient in this respect.
The new GPU also features an integrated audio core which can output eight-channel audio over HDMI, also in the high-definition formats Dolby TrueHD and DTS-HD. ATI is still superior here because even Nvidia’s flagship solutions do not offer such advanced audio capabilities, being limited to simple translation of S/PDIF audio into HDMI.
Overall, the Radeon HD 4550 looks like a bargain for its recommended price of $45-55, especially in comparison with its opponents that are not only more expensive but also have weaker technical capabilities, particularly in terms of HD video decoding and HDMI audio support. The new card should find its place in advanced HTPCs that are not intended for intensive gaming. The Radeon HD 4350 comes at a recommended price of only $39, which makes it an appealing buy, too. However, it has slow memory and can hardly be much faster than integrated graphics solutions in 3D games.
ATI developed two different PCB designs for the new entry-level card: a full-size and a low-profile PCB. And we were lucky to get both samples. We will describe the full-size version first as it is the more interesting one due to its design specifics.
Notwithstanding its full-size form-factor, the card is very compact. Most of the PCB is covered with a large needle-shaped heatsink that cools both the GPU and the memory chips.
The power circuit is simple, consisting of two single-phase regulators. The GPU is serviced by the more advanced regulator incorporating two power transistors. The other regulator has only one transistor and is responsible for the memory chips. Both regulators can be reinforced with additional transistors, but it can hardly be necessary as the power consumption of the RV710 is unlikely to exceed even 30 watts. Perhaps this PCB design was developed with ATI’s next-generation entry-level GPUs in mind – they are going to be faster and require more power. Of course, the card lacks an additional power connector. The PCI Express x16 slot can provide the card with all the power it may want.
When we removed the heatsink, we saw that the PCB of this version of Radeon HD 4550 has a cut-out at the back. It must be made to save on textolite. This saving is substantial when it comes to a $45-45 product. There are four Samsung K4W1G1646D-EC12 memory chips at the back part of the PCB, two chips on either side of it. According to the official specifications, these are DDR3 chips with a capacity of 1Gb (64Mb x 16) and a voltage of 1.8V.
The EC12 suffix means a rated frequency of 800 (1600) MHz and this is indeed the frequency the chips are clocked at on this card. The four 1Gb chips provide a total of 512 megabytes of memory. The total width of the memory bus is 64 bits, ensuring a memory bandwidth of 12.8GBps. That’s quite high for such a cheap solution and more than enough for processing HD video content. As for gaming, the Radeon HD 4550 is not positioned as a gaming graphics card, but it may be fast enough at low resolutions, especially in comparison with Nvidia’s opponents. Strangely enough, the memory chips have not thermal pads and are not actually cooled in any way.
The GPU die is small thanks to 55nm tech process and some architectural improvements over the RV730. The marking indicates that this sample of the chip was manufactured on the 30th week of 2008. The rest of the numbers make no sense for an ordinary user. The die packaging lacks a protective frame and does not actually need one. The heatsink is secured properly: its halves are fastened together with nine screws and there is no possibility of misalignment that might damage the chip. A plastic material that looks like chewing gum is used as the thermal interface here. Its efficiency is low but sufficient for a chip that dissipates 20 watts of heats only.
As we already noted, the RV710 boasts advanced specifications for its class. It incorporates 80 ALUs, eight texture processors and four raster back-ends. The GPU clock rate is 600MHz. The peak specified power draw is 20W.
This version of Radeon HD 4550 is equipped with a full set of modern digital interfaces including DVI-I, HDMI and DisplayPort and does not support the obsolete analog interfaces Composite, S-Video and YPbPr. It doesn’t support CrossFire either – joining two Radeon HD 4550 into one graphics subsystem wouldn’t make sense. With these interfaces on board, the card seems to be optimal for HTPCs that allow to install a full-size graphics card (it can be an ATX-like case, for example the Thermaltake Bach, or a low-profile case with an adapter that turns the PCI Express x16 slot by 90 degrees, for example the Cooler Master Media 260).
The second reference version of Radeon HD 4550 is significantly different from the first one because it uses a half-size PCB.
This solves many problems concerning the installation into small system cases. Oddly enough, this version supports analog interfaces. The DVI-I port is universal and allows to connect both digital and analog sources, so the dedicated D-Sub connector looks redundant here. Well, this connector is attached to the PCB via a cable and won’t be used in a small system case because you need to install an appropriate mounting bracket for that. The mini-DIN port offers the user a wider choice of interfaces, so we won’t criticize the card for having it. Of course, this version of Radeon HD 4550 supports HDMI – by means of an adapter. Its specifications are identical to the full-size version except for the memory amount. The low-profile version has 256 rather than 512 megabytes of memory.
The low-profile PCB is more compact and doesn’t have the room necessary to install a large passive heatsink. Therefore it is equipped with an active cooler. The latter is simple and consists of a small aluminum heatsink with some ribbing and a modest fan. The cooler is covered with a plastic casing.
Considering the simple PCB design of RV710-based solutions, we guess there will soon be many unique versions of such graphics cards on the market.
The Radeon HD 4550 claims to require no more than 20 watts of power and to have an appropriate level of heat dissipation. We checked this out on a special testbed configured like follows:
The 3D load was created by means of the first SM3.0/HDR test from 3DMark06 running in a loop at 1600x1200 with 4x FSAA and 16x AF. The Peak 2D mode was emulated by means of the 2D Transparent Windows test from PCMark05. This test is important as it simulates the user’s working with application windows whereas Windows Vista’s Aero interface uses 3D features. We got the following results:
The results surpass our best expectations. The Radeon HD 4550 consumed less than 20 watts even under load. In idle mode it is the most economical graphics card we have ever tested in our labs, including the GeForce 7300 GS (8.7W under the same conditions).
The GPU temperature varied from 56°C in idle mode to 78°C under load. The GPU and memory clock rates are reduced to 110 and 300 (600) MHz in 2D mode. The temperature is not low because of the modest cooling system.
The full-size version of the card is absolutely silent as it uses a passive heatsink, but the low-profile version is very quiet, too. You can’t hear its sound in a working system case. Our measurements revealed no difference: the level of noise was 43dBA in both cases, the same as at distance of 1 meter from the working testbed with a passively cooled graphics card inside.
Thus, the Radeon HD 4550 boasts superb electrical, thermal and noise characteristics. Let’s now check out its capabilities in games and at decoding and playing HD video.
Besides the required interfaces and technologies such as HDCP, a multimedia graphics card must provide hardware acceleration of video decoding and improve playback quality through post-processing.
Since ATI Radeon HD 4550 is first of all targeted for home-theater personal computer systems (HTPC), we checked out not only the content playback quality but also the CPU utilization. For our tests we built the following platform:
Since ATI Radeon HD 4550 is a very inexpensive graphics card we compared it against the following discrete and integrated solutions:
Mainboards for AMD processors were tested with the similar software and hardware components with that only difference that we used an AMD Phenom X4 9550 processor. Since it is quad-core CPU, integrated AMD platforms demonstrate higher results than the competitor solutions.
We used the following tools to estimate the video playback quality in standard (SD) and high-definition (HD) resolutions:
The driver settings remained the same. However, according to the HQV HD suite requirements, the noise suppression and detail levels for Nvidia GeForce graphics cards and Intel Graphics Media Accelerator 4500 were set to the maximums.
Considering that the CPU utilization was measured under Windows Vista OS without disabling background services, the CPU utilization peaks shouldn’t be regarded as critical. It is much more important how much time it takes the CPU on average to complete the task. Note that the CPU utilization may vary. Therefore, 1-2% difference is not indicative of any advantage of a certain graphics accelerator over the competitor.
Since MPEG2 decoding and DVD playback are no longer a complicated task for GPUs, we didn’t measure the CPU utilization in this case.
To estimate the CPU utilization during full-HD video playback (1920x1080) and full-HD video with enabled “picture-in-picture” feature, we used the following movies:
Since there is also a lot of HD content available online these days, we also measured CPU utilization during playback of several free videos with the following parameters:
The HQV benchmarks from Silicon Optix are among the few available tools for evaluating playback quality, but there is one drawback about them. The tester’s perception is always subjective. Therefore you should note take these benchmarks as the final truth.
Moreover, ATI and Nvidia are constantly optimizing video settings in their drivers, so the playback quality in general and the HQV HD result in particular has improved considerably over the last year. Unfortunately, both companies take the HQV test less seriously as is indicated by the lower results.
The score of 91 out of 130 possible points is satisfactory. The Radeon ND 4550 delivers satisfactory playback quality with SD content. However, it is not as good as the other solutions, particularly the Radeon HD 3300, the integrated graphics core of the AMD 790GX chipset. ATI’s programmers must have not had enough time yet to find the optimal settings of the Avivo HD video-processor for ordinary DVDs.
High-definition video playback is getting more important because video content in standard resolution is dying out while HTPC users are likely to use a full-HD TV-set with a resolution of 1920x1080.
As you can see, the Radeon HD 4550 scores 100 out of 100 points which is higher than the opponents or integrated graphics processors can score in this test.
ATI’s Radeon HD processors traditionally have a lower CPU load when playing video encoded with the VC-1 codec. As you can see, the CPU load doesn’t grow up much even if we turn on the Picture-in-Picture mode on graphics cards that officially cannot accelerate two VC-1 streams simultaneously.
The Radeon HD 4550 has a very low CPU load in comparison with the other Radeons as well as Nvidia’s solutions.
Modern GPUs find it an easy job to decode MPEG4-AVC/H.264 streams. Few of them have an average CPU load of 25% even when decoding two streams simultaneously.
The Radeon HD 4550 offloads the CPU effectively when decoding video. Decoding a scene from The Day After Tomorrow with a bitrate of 40Mbps is not a problem for it.
We did not test the cards with MPEG2 video although some Blu-ray movies available on the market are encoded with MPEG2 HD. If you’ve got such discs, you may be interested to check out the CPU load when decoding a MPEG2 stream with a bitrate of 20-25Mbps (a free clip by NASA about launching a space shuttle).
Once the most popular codec for illegally distributed video, DivX has lost its ground by now. MPEG4-AVC is getting more popular for HD content while Xvid is used for ordinary DVD ripping. However, Divx company is making efforts to get back to the scene. Perhaps they will succeed.
But it is clear that the GPU developers do little for hardware acceleration of DivX decoding. Many graphics cards, including the Radeon HD 4550, require more than half of the CPU resources to decode DivX HD.
Considering that every modern GPU incorporates an MPEG2 decoder, no wonder that all of them deliver good results when playing MPEG2 clips or movies with a resolution of 1920x1080.
MPEG4-AVC/H.264 enjoys tremendous popularity, so it is good that the average CPU load is no higher than 3.5% when decoding a 720p stream.
The 720p clip encoded with VC-1 has a frame rate of 60fps, so playing it requires a lot of resources, particularly CPU resources. The Radeon HD 4550 copes with the job well enough. It is equal to its opponents in this test.
The WMV HD format is leaving place for VC-1, but we still publish the results of this test. The graphics cards deliver similar performance here in terms of CPU load. However, we had expected more from the Radeon HD 4550 considering its targeting at HTPCs.
For our performance tests of ATI Radeon HD 4550 we used two test platforms. One of them was built for gaming tests and featured the following configuration:
According to our testing methodology, the drivers were set up to provide the highest possible quality of texture filtering and to minimize the effect of software optimizations used by default by both: AMD/ATI and Nvidia. Also, to ensure maximum image quality, we enabled transparent texture filtering. As a result, our ATI and Nvidia driver settings looked as follows:
For our tests we used the following games and synthetic benchmarks:
First-Person 3D Shooters
Third-Person 3D Shooters
We selected the highest possible level of detail in each game using standard tools provided by the game itself from the gaming menu. The games configuration files weren’t modified manually in any way, because the user doesn’t have to know how to do it. The only exception was Enemy Territory: Quake Wars game where we disabled the built-in fps rate limitation locked at 30fps. Games supporting DirectX 10 were tested in this particular mode.
Since ATI Radeon HD 4550 belongs to the sub-$60 price range, we decided not to run any tests in resolutions over 1280x1024. Nevertheless, we enabled MSAA 4x antialiasing everywhere where it was possible, because it is no exotic function anymore and at the same time allows improving image quality significantly. We enabled antialiasing and anisotropic filtering from the game’s menu. If this was not possible, we forced them using the appropriate driver settings of ATI Catalyst and Nvidia ForceWare drivers. As we have already said, we didn’t modify the games configurations files.
Performance was measured with the games’ own tools and the original demos were recorded if possible. Otherwise, the performance was measured manually with Fraps utility version 2.9.1. In the latter case we ran the tests several times and took the average value for further analysis. We measured not only the average speed, but also the minimum speed where possible.
Alas, the Radeon HD 4550 is not strong enough to run the game at an acceptable speed even though this game doesn’t support antialiasing. The RBE subsystem proves to be a bottleneck but seldom today, but this card has a weaker RBE subsystem even in comparison with the GeForce 9500 GT, let alone the more sophisticated products.
The GPU’s relatively advanced computing section just can’t show its full potential within these limitations.
The Radeon HD 4550 is very slow here, but this is a resource-consuming game: the GeForce 9800 GT and Radeon HD 4830 are slow, too.
The Radeon HD 4550 is obviously not a gaming card even though it outperforms the GeForce 9500 GT. The somewhat more expensive Radeon HD 4670 is almost three times as fast!
The new card’s average frame rate is no higher than 10fps and it is slower by 45% than the GeForce 9500 GT, the slowest of Nvidia’s solutions. It seems like we should forget about serious gaming with Radeon HD 4550.
The frame rate is fixed at 30fps in this game as this is the rate at which the physical model is being updated at the server. Thus, this 30fps speed is the required minimum for playing the game.
Quake Wars is not a heavy application, but the Radeon HD 4550 is an outsider again. It is unable to deliver the desired 30fps whereas the GeForce 9500 GT can do that.
There is no need for any comments here. The cheapest graphics cards do not allow playing this game even at a resolution of 1280x1024 with 4x FSAA.
Like Crysis, this game is too heavy even for top-end graphics hardware, let alone such entry-level solutions as Radeon HD 4550 or GeForce 9500/9400.
Dead Space has modest system requirements and runs fast even on inexpensive graphics cards, but the Radeon HD 4550 is not the case. It only yields an average speed of 20fps, which is not enough to enjoy the gloomy atmosphere of the game fully.
The Radeon HD 4550 is over 50% slower than the GeForce 9500 GT. And while the latter can be used to play this game with occasional slowdowns due to the low bottom speed, the Radeon is no good at all as its average speed is only as high as 20fps.
The Radeon HD 4550 shows that it is no gaming card, again. Of course, you can try to achieve a higher speed by lowering the level of detail and resolution, but this would make the game less appealing visually and thus less enjoyable.
The Radeon HD 4670 and HD 4550 are both far slower than the Radeons that have a 256-bit memory bus. The 4550 model has a speed of 10fps, which is completely unplayable.
The Radeon 4550 is about as fast as the GeForce 9500 GT, neither card being able to provide a playable frame rate.
As we found out in our recent review, Fallout 3 is an optimized application, yet this optimization is not enough to ensure a comfortable speed on every graphics card. So if you want to explore the radioactive ruins of the post-nuclear Washington, you have to lower the level of detail or think about buying a Radeon HD 4830 at least.
The new Radeon finds itself in last place again. Like in most other cases, it is the cut-down architecture of the raster back-ends that prevents the Radeon HD 4550 from competing even with the GeForce 9500 GT.
Like some other projects from EA, Spore has an integrated frame rate limiter, so you should compare the bottom speeds of the cards in the first place.
Spore is not a demanding application but the Radeon HD 4550 cannot cope with it right after the game’s unicellular phase. The GeForce 9500 GT feels better despite its weak computing resources.
The frame rate of 6fps is an unsatisfactory result, but the GeForce 9500 GT with its 15fps is not far better when it comes to practical play.
The Radeon HD 4550 cannot match the GeForce 9500 GT in terms of overall score or in the individual groups of tests. The gap is smaller in the SM3.0/HDR tests due to the good computing capabilities of the new Radeon.
The SM2.0 tests agree with the overall scores but the gap is smaller in the second test. And in the second SM3.0/HDR test the Radeon HD 4550 is even better than the GeForce 9500 GT. The overall results are logical enough, confirming our supposition about the misbalance of the Radeon HD 4550 architecture: the card has rather advanced computing resources but a weak texture-mapping and rasterization subsystem.
We minimize the CPU’s influence by using the Extreme profile (1920x1200, 4x FSAA and anisotropic filtering).
The Radeon HD 4550 just fails this test although 3DMark Vantage is meant for benchmarking modern graphics architectures. It cannot score even 400 points whereas the GeForce 9500 GT easily scores 900 points and more, which is a good result for an entry-level card.
The same goes for the individual tests: the Radeon HD 4550 doesn’t yield even two frames per second in them. We shouldn’t blame the developer, however. This graphics card is not actually positioned as a gaming solution.
There are both highs and lows about the Radeon HD 4550 graphics card we have tested today. Priced at $50, it offers a lot of features such as the hardware video-decoder ATI UVD2 that can decode H.264 and VC-1 formats (Nvidia’s solutions have more limited decoding capabilities). UVD2 is complemented with an integrated eight-channel HDMI audio core that supports the high-definition formats Dolby TrueHD and DTS-HD. Coupled with its very low power draw (less than 20W under full load), this makes the Radeon HD 4550 a perfect choice for a home HD content-oriented multimedia center connected to a full-HD LCD panel via HDMI.
Our tests showed that the Radeon HD 4550 ensures unrivalled quality when playing Blu-ray and HD DVD movies but it is somewhat inferior to its opponents when it comes to playing ordinary DVDs. This is a drawback since many users have accumulated huge collections of DVDs and want to enjoy them for many years. On the other hand, the playback quality depends largely on driver optimizations, so it is quite possible that the HD 4550 will play DVDs with as high quality as the other such cards from ATI. Moreover, it is important to remember that DVDs have a resolution of only 720x480 (for NTSC) and 720x576 (for PAL), so old movies do not look very pretty on large LCD panels irrespective of what video-processor you use.
The Radeon HD 4550 showed itself good at hardware decoding of high-resolution video. The CPU load was mostly within 30% with occasional peaks above 40%, which is an excellent result for any graphics card.
However, the owner of a HTPC and a full-HD panel may also want to use his system for playing games, especially as the large LCD panel ensures a deeper immersion into the gaming world than an ordinary PC monitor, but the Radeon HD 4550 is virtually unable to run modern 3D games at a high speed due to its cut-down subsystems of texture processors and raster back-ends. This card could not provide a playable frame rate in any of the games we tested it in.
Considering that we performed our tests at 1280x1024, we can suppose that the Radeon HD 4550 is going to be just as slow at the HD-ready resolution of 1366x768 and far slower at the full-HD resolution of 1920x1080.
Thus, the Radeon HD 4550 should only be installed into a HTPC if the latter is not going to run modern 3D games. It may be an inexpensive, compact and quiet system with low power consumption. If you want a universal HTPC, you should take a look at the more advanced products from ATI, e.g. the Radeon HD 4850. Having similar multimedia capabilities, this card is far faster in games than the Radeon HD 4550 and costs less than $150.
The Radeon HD 4550, on its part, is meant for small and quiet HTPCs with low gaming performance or as an alternative to an integrated graphics core if you don’t want to use the latter for some reason.