by Alexey Stepin , Yaroslav Lyssenko, Anton Shilov
03/18/2010 | 10:47 AM
Still being unable to respond to the massive onslaught of AMD’s graphics department in the sector of gaming graphics cards, Nvidia decided to make a flanking maneuver and announced a few inexpensive products based on its new 40nm GT215, GT216 and GT218 GPUs. These cards are called GeForce GT 240, GeForce GT 220 and GeForce 210, respectively, and we covered them in our articles Nvidia GeForce 210 and GeForce GT 220 Review: Revenge of the Low-End? and Palit GT 240 Sonic and Gainward GeForce GT240 1024MB GDDR5 GS: Worthy Replacement for GeForce 9600 GT. The new cards proved to be quite good as they had finally acquired DirectX 10.1 support and a fourth-generation PureVideo HD engine that offers full-featured hardware decoding for both H.264 and VC-1 (with the earlier versions of Nvidia’s video engine, bit stream processing and entropy decoding for VC-1 had been the CPU’s responsibility). Coupled with the comprehensive support for audio-over-HDMI, this was a real breakthrough for Nvidia, although the developer still had forgotten to implement Protected Audio Path. As the result, Dolby True HD and DTS HD Master Audio formats can still be output via HDMI only by AMD’s Radeon HD 4000/5000 series.
Anyway, the GeForce 210, GeForce GT 220 and GeForce GT 240 are really good solutions for their class, especially as the GeForce GT 220 can be used for some gaming at low resolutions. The GeForce GT 240, in the GDDR5 version (Nvidia’s GPUs can utilize this high-speed memory type at last), is not much inferior to its successful predecessor GeForce 9600 GT. Thanks to its enhanced multimedia capabilities, the GT 240 is a good choice for an advanced HTPC that may be used for casual gaming.
AMD, on its part, did not want to stop its progress after releasing successful Juniper (RV840) based products like Radeon HD 5770 and Radeon HD 5750. January 14, the company announced a new addition to the Radeon HD 5000 series in the way of the Redwood (RV830) GPU which had indeed been scheduled for a Q1 release.
The senior Redwood-based model, Radeon HD 5670, was already covered in our Almost a Gaming Card: ATI Radeon HD 5670 512 MB Review. We found it to be better than the GeForce GT 240 GDDR5 but not quite fast in modern games, especially DirectX 11 ones. However, the Redwood-based series includes at least two more models, Radeon HD 5570 and Radeon HD 5550, which were announced in early February, and we did not cover the multimedia capabilities of the new card in terms of decoding HD video and audio formats. Those capabilities can make the Radeon HD 5000 series highly appealing for HTPC users.
Besides that, on the 4th of February AMD added one more GPU to its line-up. It is codenamed Cedar (RV810) and is the simplest of the new GPUs from AMD. It is supposed to be used on cheapest graphics cards and complete the Radeon HD 5000 series in the low-end sector, making the Radeon HD 4000 series products based on the RV710 core obsolete. We have not yet tested any Cedar-based solutions before, but have the chance of doing that now.
Although we covered the Radeon HD 5670 in an in an earlier review, we want to give you a fuller view of Redwood- and Cedar-based graphics cards. Here is the summary table of their specifications:
Click to enlarge
It is clear that the RV830 core codenamed Redwood is in fact one half of the RV840 (Juniper) core which is used in the inexpensive gaming series ATI Radeon HD 5700. It has half the ALUs, texture-mapping units and RBEs except for the memory subsystem which still has a 128-bit memory bus and allows using different types of memory including GDDR5 and GDDR3/DDR3. The peak memory bandwidth of the more advanced version of the card is 64 GBps which is just slightly lower than that of the Radeon HD 5770. This should not be a bottleneck in the Radeon HD 5670 design. The card’s recommended price is $99, so it falls into the same category as Nvidia’s GeForce GT 24 and everything depends on how effective the drivers and the task dispatcher of shader processors are. In other words, if the Radeon HD 5670 makes use of all its ALUs, it will undoubtedly beat the GeForce GT 240. Otherwise, Nvidia’s solution is going to gain the upper hand, particularly as it has a slightly higher fill rate.
As a multimedia card, the Radeon HD 5670 is theoretically beyond competition because, as opposed to its opponent, it offers full-featured HDMI 1.3a support including bitstreaming of multichannel high-definition audio formats. Nvidia’s products do not support the Protected Audio Path which allows transferring Dolby True HD and DTS HD/DTS HD Master Audio via HDMI. It must also be noted that, like the Cypress and Juniper, the Redwood has two 1080p decoders which are necessary to display stereo 3D content in MPEG4-MVC format (Blu-ray 3D).
One step lower in the hierarchy there is the Radeon HD 5570. Its model number may be somewhat confusing as it only differs from the Radeon HD 5670 with its clock rates and graphics memory type and could be called 5650. It is much inferior to its senior cousin in one parameter only which is the peak memory bandwidth, but this parameter is only important for playing at high resolutions but neither of these cards can be viewed as purely gaming solutions. Considering its price, the 5570 competes with the GeForce GT 240 GDDR3/DDR3 while the GeForce GT 220 has no chance to stand against it. There exists a cut-down version of Radeon HD 5570 marketed under the name of Radeon HD 5550. The latter has 320 ALUs and 16 TMUs, its core frequency is reduced to 550 MHz, and its memory is slow DDR2. We will leave the Radeon HD 5550 out of this review because we haven’t got any sample of it yet. Moreover, this product is hardly interesting for consumers. It is as helpless as its junior cousin in games whereas the latter is just as good at processing high-definition video and audio (with one exception to be mentioned below) but costs less. We guess the Radeon HD 5550 will find its place in ready-made computers.
The most low-end solution in the ATI Radeon HD series, the Radeon HD 5450 is based on a different GPU developed by AMD especially for cheap products. The core is codenamed Cedar (RV810) and is very compact. Developing the most simple and economical GPU, AMD sacrificed nearly all other parameters, excepting the video engine block (partially) and the audio processor. Like Nvidia’s GT218, the RV810 has 16 unified execution processors but, considering the superscalar Radeon HD architecture, each of these processors incorporates 5 ALUs. Theoretically, this GPU can be five times as fast as its opponent if its task dispatcher should provide all the 80 ALUs with work for that. The Cedar will also have no problems decoding even the most sophisticated HD video formats. Otherwise, it is a blatantly low-end solution that is hardly superior to the Radeon HD 4550 in sheer processing power: both have 8 TMUs and 4 RBEs and their fill rates are similar. The memory bus is 64 bits wide with both these cards and the new card may even come with slow DDR2 memory instead of faster DDR3. It is in video decoding and audio processing capabilities that the Radeon HD 5450 is head above its predecessor as well as above the GeForce 210. The only thing is cannot do is to work with Blu-ray 3D since this format requires simultaneous decoding of two video streams, one for each eye, whereas the Cedar has only one HD video decoding unit. Anyway, this GPU may be a true winner in the low-end category: economical, cold, noiseless but strong enough to cope with any modern video and audio format. In other words, it can make a perfect graphics card for a modern HTPC.
We will discuss three pairs of inexpensive Radeon HD 5000 series cards in this review, each pair consisting of AMD’s reference card and an original product from AMD’s partner.
We will take a look at their design and see how good they are in practical applications.
These are the most advanced products based on AMD’s Redwood processor (RV830) which we covered in more detail in our Almost a Gaming Card: ATI Radeon HD 5670 512 MB Review. As you can guess from their names, the two cards differ in the amount of local graphics memory, yet the difference goes even further than that. The former solution is a copy of the reference card whereas Gigabyte’s engineers took a creative approach and developed their Radeon HD 5670 almost from scratch.
The GV-R567OC-1GI comes in a rather small box made from thick glossy cardboard. The packaging is almost the same as the packaging of Gigabyte’s GV-N220OC-1GI we have tested earlier. The contents of the box are up to the product’s price category. Besides the graphics card itself, you won’t find anything in there, save for a user manual and a disc with drivers. The user manual sets a record of giving detailed and informative instructions in as many as 25 languages!
The cards differ visually even if you don’t remove their coolers.
The GV-R567OC seems to have a more potent cooler thanks to the massive round heatsink that resembles the heatsinks of Intel’s boxed CPU coolers, and the large 80mm fan. Alas, it takes two slots and may not be compatible with some barebone and HTPC cases whereas AMD’s reference card can easily fit into any such computer.
The two versions of Radeon HD 5670 look obviously different when you take their coolers off.
They are only similar in some general traits just because it is hard to find an unusual way of placing GPU and memory chips on the PCB (even though some makers do show originality in this respect, too). The GV-R567OC has CrossFire connectors whereas AMD’s reference card has none, which is a sure indication that Gigabyte positions its product as a gaming card. Although we have already found out that you cannot expect anything special from the Radeon HD 5670 in games, two such cards working in tandem can be quite fast. Unfortunately, we cannot check this out because we only have one sample of the GV-R567OC and reference Radeon HD 5670 cards come without CrossFireX slots and cannot be used in full-featured hardware-based multi-GPU configurations even though they do support CrossFireX, exchanging data via PCI Express.
The cards both have their GPU power circuitry at about the same location. In each case it is a two-phase regulator with two power MOSFETs in each phase. The respective controllers can be found on the reverse side of each card.
Both controllers are from uPI Semiconductor, but Gigabyte’s card uses an uP6209AQ whereas the reference card from AMD employs an uP6201BQ.
The single-phase memory voltage regulators are located differently. The GV-R567OC has it on the reverse side of its PCB whereas the reference card, on the face side.
A pair of uP6101 and uP771 chips is used on both versions of Radeon HD 5670 for controlling the memory power supply. Of course, the economical Redwood core (RV830) makes additional power connectors unnecessary even for the most advanced Redwood-based cards like these two. The reference version and the Gigabyte GV-R567OC-1GI are both content with what they can get from the power section of the PCI Express slot.
The cards have identical GPUs manufactured on different dates: the 44th week of 2009 for the reference card and 48th week of 2009 for the Gigabyte GV-R567OC.
As you can see, the GPUs have the same configuration: 400 ALUs grouped into 80 unified shader processors which are themselves united into 5 computing SIMD modules. Each SIMD module is serviced by four texture-mapping units, so the total number of TMUs is 20. The latest version of GPU-Z erroneously reports that the Redwood has 16 raster back-ends although it has only 8. Notwithstanding the OC suffix in the Gigabyte card’s model name, the factory overclocking amounts to a mere 10 MHz: the GPU frequency is increased from 775 to 785 MHz. This can hardly affect the card’s performance in games. The memory frequency of each card is 1000 (4000) MHz, delivering a peak bandwidth of 64 GBps with 128-bit memory access. The number is quite astonishing for such entry-level products. Not so long ago, only such expensive monsters as Radeon X1950 XTX could boast such a high memory bandwidth.
AMD’s reference card is equipped with only four GDDR5 memory chips from Hynix (H5GQ1H24MFR-T0C), 1 Gbit each. The T0C suffix indicates a rated frequency of 1000 (4000) MHz. It must be mentioned that the reference design does not provide for the installation of additional memory chips on the reverse side of the PCB. So, with 1-gigabit chips, the maximum amount of graphics memory on board the Radeon HD 5670 is going to be only 512 MB. Yes, Hynix has already developed 2Gb GDDR5 chips but they are too expensive and high-frequency to be installed on entry-level products.
The Gigabyte GV-R567OC is equipped with Samsung’s K4G10325FE-HC05 memory chips that have the same capacity and rated frequency but Gigabyte’s PCB design allows installing eight memory chips, four on each side of the PCB. As a result, the Gigabyte GV-R567OC comes with 1024 megabytes of memory, which would be quite enough even for the most advanced gaming solutions, let alone the Radeon HD 5670. As we’ve said above, the memory frequency of each card is 1000 (4000) MHz.
The ATI Radeon HD 5670 and Gigabyte GV-R567OC-1GI have different interfaces. Besides a CrossFireX connector, the latter has a somewhat conservative selection of interfaces: one HDMI, one DVI-I and one 15-pin D-Sub. AMD’s reference card is equipped with a DisplayPort instead of the latter. Although DisplayPort is far less popular than HDMI or DVI, we don’t think that D-Sub is really necessary at all. If you want an analog connection, you can use an adapter for DVI-I. The only use we can think of for D-Sub is the Eyefinity feature: it is easier to find a monitor with a D-Sub input than with a DisplayPort. The inexpensive Radeon HD 5670 is unlikely to be used with more than one display, however. If installed into an all-purpose HTPC, it may be connected to a computer monitor (for work) via DVI and to an LCD panel (for movies) via HDMI.
By the way, Gigabyte could not help promoting its product with the mention of a gold-plated HDMI connector, but our practice suggests that such connectors are neither better nor worse than ordinary ones.
Of course, both cards are fully compliant with the HDMI 1.3a specifications. Thanks to the Protected Audio Path support, they can work with multichannel HD audio formats including Dolby True HD and DTS HD/DTS HD Master Audio. AMD’s graphics cards still have no rivals in this respect.
Gigabyte and AMD have taken opposite approaches to cooling their cards. Gigabyte installed a simple aluminum heatsink with a large 80mm fan whereas AMD tried to make the card compact by using a more intricate cooler design.
As you can see, there are no secrets in the Gigabyte card’s cooler but AMD’s cooler has a heat pipe pressed into its base. AMD’s cooler has a small heatsink made of an aluminum sheet shaped accordion-like. Additionally, there are a few “needles” at the top of the base. A tiny and rather noisy blower completes the whole arrangement. We don’t think this cooler has any advantages other than its compactness. Even the dedicated cooling of the memory chips via green elastic thermal pads is no benefit because the chips of the Gigabyte GV-R567OC card are cooled by the air flowing through the heatsink fins.
Gigabyte’s product looks overall better than the reference Radeon HD 5670 but its dual-slot cooler may provoke some installation-related problems. The single-slot version of Radeon HD 5670 is compatible with more system cases.
The Radeon HD 5570, sometimes referred to as Redwood PRO, belongs to the family of Redwood-based solutions, too. This model is also represented by two versions in this review – a reference version and a version from Sapphire – but they don’t differ much. Sapphire’s card is packaged into a rather large and nice-looking box designed in red colors and embellished with a picture of the official mascot of the Radeon series called Ruby.
The type of the graphics memory is indicated correctly. You can also learn from the text on the box that a low-profile mounting bracket is included into the box. A free version of the SimHD IM plug-in is included, too. “IM” stands for Internet Messaging and the plug-in itself can scale up streaming video to near-HD level. The technology may be useful, but the quality of an Internet video call depends more on the video codec, Internet bandwidth and latencies rather than on any post-processing of the image itself. The accessories are not rich but passable for a below-$85 product: discs with drivers and SimHD IM, a Sapphire sticker, a couple of low-profile mounting brackets, one of which is designed for a D-Sub connector, and a user manual. There is no DVI-I à HDMI adapter in the kit which may be a problem as we’ll explain shortly.
The two Radeon HD 5570 models we’ve got are both based on the reference PCB design. They are compact low-profile cards equipped with tiny coolers.
With such dimensions, the Radeon HD 5570 should not provoke any installation-related problems if your system case is designed for a discrete graphics card at all. You may need a half-height mounting bracket like the one included with the Sapphire HD 5570 DP. We guess that all retail versions of Radeon HD 5570 will come with such brackets.
The cards look unusual with the coolers removed. The GPU is turned around by 45 degrees, which is not often to be seen. This must be due to the specifics of wiring a low-profile PCB. This also must be the reason why the PCBs are so densely populated with small components. The power circuit is very simple and uses single-phase GPU and voltage regulators managed by uP6101 chips.
With the component density so high, it would be a daunting task to volt-mod the Radeon HD 5570 but some overclockers might do that just for fun.
The GPU chip on the reference Radeon HD 5570 was manufactured on the 38th week of 2009. The Sapphire card’s GPU was made on the 47th week of 2009. As usual, you can extract no other useful info from the markings on ATI Radeon HD series GPUs.
Compared with the Radeon HD 5670, the clock rate is reduced from 775 to 650 MHz, but this is still the same Redwood (RV380) with 400 ALUs, 20 TMUs and 8 RBEs. However, there are internal differences explaining why this chip is called “Redwood PRO” when installed on the Radeon HD 5570 rather than “XT” as on the senior model of the series. In this case, the memory controller is set up for DDR3 whereas the senior model’s controller is set for GDDR5. AMD’s card is an engineering sample while Sapphire’s is an almost exact copy of it, so their memory frequency is 900 (1800) MHz in full compliance with the official specs. With a 128-bit memory bus, the peak memory bandwidth is 28.8 GBps.
There are eight K4W1G1646E-HC11 memory chips from Samsung at the back of each Radeon HD 5570 card, four on each side of the PCB. Although they are categorized as gDDR3 on the manufacturer’s official website, you should not confuse them with GDDR3 as the popular GPU-Z tool does. These are actually two different types of memory. GDDR3 is based on DDR2 technologies, has a higher voltage than DDR3 (1.7 to 2.1 V compared to 1.5 V) and has internal termination. The chips on board the described AMD and Sapphire cards have a voltage of 1.5 V and a rated frequency of 900 (1800) MHz. They have a capacity of 1 Gb for a total of 1024 megabytes. Although seemingly redundant, this amount of graphics memory has already become standard even for cheap graphics cards.
The Radeon HD 5570 has a typical selection of interfaces for an HTPC-oriented solution: one DVI-I, one HDMI and one D-Sub (the smaller mounting bracket does not allow using the D-Sub). The card fully supports HDMI 1.3a. The Sapphire HD 5570 DP has a DisplayPort instead of the HDMI, which may present a problem if you are going to build a three-monitor configuration. And you will also need a DVI-I à HDMI adapter to connect the card to most modern TV-sets (the adapter is not included into the kit). Sapphire also offers a Radeon HD 5570 with a more traditional selection of interfaces, though.
The reference card’s cooling system is simple and consists of a copper base with a copper heatsink soldered to it. The heatsink is made of slim interlocked fins. The Sapphire HD 5570 DP has an all-aluminum heatsink of about the same size. There are slits beneath the fan: we don’t think the PCB of the Radeon HD 5570 needs such cooling but it will be indeed cooled better here.
The heatsink is cooled by tiny 7-blade fans. Considering the reduced GPU frequency, this should be enough. We only have some apprehensions concerning the noise these fans may produce.
Compared with the Radeon HD 5670, the Radeon HD 5570 seems a better option for HTPC users. An HTPC is not expected to be fast in 3D applications and is likely to be assembled in a compact low-profile system case for which the half-height Radeon HD 5570 is going to suit perfectly. These solutions are absolutely identical in terms of software and hardware capabilities for processing HD audio and video content, at least theoretically. As we know, some manufacturers often deliberately limit the functionality of their inexpensive products to make customers prefer the more expensive ones.
Directly opposite to the first pair of graphics cards in this review, the last pair represents the most inexpensive solutions based on the simplest GPU in the Radeon HD 5000 series, the Cedar (RV810). The two cards mentioned in the headline to this section are a little different. The former, like some other cards in this review, is a copy of the reference design whereas Sapphire’s card comes into retail in this pretty packaging:
The box is surprisingly large for a junior Radeon HD 5000 series model. It is as cute as the box of the Sapphire HD 5570 1GB DDR3 DP but differs with its color scheme. Besides the mentioned disc with ArcSoft SimHD, the kit includes a user manual, a disc with drivers, a Sapphire sticker, and two low-profile mounting brackets one of which is designed for a D-Sub connector. No DVI-I à HDMI adapter here although the card has a DisplayPort instead of an HDMI.
The AMD and Sapphire cards have the same PCB design but differ with the color of the PCBs and the cooler’s heatsinks.
Both are very compact and will easily fit into any system case although the tall heatsink of the reference card may provoke some installation-related problems. Besides, each cooler is passive, so the system case has to be properly ventilated. Without normal ventilation, even the economical Cedar (RV810) may overheat.
With the coolers removed, we can see some differences between the two cards:
Besides the different configuration of the I/O interfaces, we can note that some smaller elements installed on the AMD card are missing on the Sapphire but the latter’s GPU is protected with a plastic frame.
The power circuits of both cards are very simple and based on uP6101 controllers.
The GPUs installed on the AMD and Sapphire cards differ greatly in age: 34th and 50th weeks of 2009, respectively. It means that AMD has had working samples of the Cedar (RV810) for a long time already.
As we’ve said already, this is the weakest GPU in the Radeon HD 5000 series incorporating only 80 ALUs, i.e. 16 unified shader processors. Considering the superscalar architecture of the ATI Radeon HD, the RV810 is comparable to Nvidia’s GeForce 210, which has 16 scalar execution modules, in terms of computing resources. The Cedar (RV810) also has 8 TMUs and 4 RBEs. The GPUs of both cards are clocked at 650 MHz while the memory frequencies differ: 900 (1800) MHz on the AMD card and 800 (1600) MHz on the Sapphire HD 5450 DP. With 64-bit access, the peak memory bandwidth is 14.4 and 12.8 GBps, respectively. Both numbers are ridiculously low for today and we are already quite sure that the Radeon HD 5450 won’t be any good in modern games.
Each card is equipped with four gDDR3 memory chips from Samsung (K4W1G1646E, 1 Gb) that should not be confused with GDDR3. Sapphire decided to save a little and installed HC12-suffixed chips rated for a frequency of 800 (1600) MHz whereas AMD’s card comes with HC11 chips rated for 900 (1800) MHz. As you know from the previous paragraph, the memory chips are clocked at their respective rated frequencies on both cards. Each card’s total amount of memory is 512 megabytes.
Each is equipped with one DVI-I and one D-Sub connector. The latter is connected via a flexible cable and can be removed and unplugged if you install the card into a half-height system case. The Sapphire HD 5450 DP kit includes a D-Sub mounting bracket which you can install into a neighboring slot if you need this interface. The third connector of AMD’s reference card is HDMI while the Sapphire Radeon HD 5450 DP has a DisplayPort there. Perhaps some users will appreciate this configuration, but we guess the very positioning of the Radeon HD 5450 makes HDMI a far more demanded option. The lack of native HDMI on board the Sapphire card won’t be an insurmountable problem, however. You can use a DVI-I à HDMI adapter (not included into the kit). Like all ATI Radeon HD 5000 series products, the Cedar (RV810) based cards fully support HDMI 1.3a but, as opposed to the more advanced models, cannot decode 3D video formats.
The reference Radeon HD 5450 and the Sapphire HD 5450 DP both use passive coolers.
On the Sapphire card, it is a simple black U-shaped heatsink with low fins that partially bends over to the reverse side of the PCB and thus fully fits into the single-slot form-factor. AMD’s card has a rather tall heatsink that makes the card occupy two slots. Both versions are acceptable but the tall heatsink may be compatible with fewer system cases. And we want to remind you once again that any passive cooling system requires that the system case be properly ventilated. Otherwise, the graphics card may overheat. Take care of that if you’ve chosen one of such cards because of its noiselessness.
Summing it up, the Radeon HD 5450 specs make it clear that it is not meant for modern games. The best it can do in terms of gaming is a Quake III Arena match or Sims 3.
Now that’s the end of the descriptive part of the review. Let’s move on to technicalities. First, let’s check out the electric, thermal and acoustic characteristics of the products we have described.
In our previous review of the Radeon HD 5670 we already tried to find out how energy-efficient RV830 “Redwood” was. However, we did talk about the power consumption of a complete system rather than the electrical characteristics of an individual card. Besides, we have only tested one RV830 model, while RV810 “Cedar” wasn’t included. We are going to give you this information now by testing our graphics cards on a testbed configured as follows:
The new testbed for measuring electric characteristics of graphics cards uses a card designed by one of our engineer and described in his article PC Power Consumption: How Many Watts Do We Need?. This device provides the most extensive information about the electrical characteristics of the graphics cards in various test modes. As usual, to create appropriate workload on the graphics accelerators in different modes we used the following benchmarks:
Except for the maximum load simulation with OCCT, we measured power consumption in each mode for 60 seconds. We limit the run time of OCCT: GPU to 10 seconds to avoid overloading the graphics card's power circuitry.
Here are the obtained results:
Beyond a doubt, the Radeon HD 5670 is highly economical and superior to the GeForce GT 240 in this respect, especially when playing high-definition video. Its peak power draw in the video playback mode is no higher than 16 watts whereas Nvidia’s solution needs about 25 watts. There is nothing interesting about the distribution of load among the different power lines: the +3.3V line always has the same load, its current varying within 0.6 to 0.7 amperes. The current on the +12V line grows up under higher loads, reaching a peak value of 2.6 amperes in the OCCT:GPU test. Gigabyte’s Radeon HD 5670 behaves the same, which is no wonder as the GPU frequencies differ by a mere 10 MHz.
The Radeon HD 5570 has a considerably lower GPU frequency and is equipped with slower DDR3 memory. As a result, its peak power draw is below 20 watts even in games and below 25 watts in the OCCT: GPU test. The card needs about 13 watts when playing HD video, just like the GeForce GT 220. The latter is somewhat more economical in idle mode.
The junior model of the Radeon HD 5000 series is very economical, too, but hardly differs from the GeForce 210 in this respect. Both cards have low power requirements in every mode, the difference between them being negligible. We shouldn’t forget about the Cedar’s potentially higher computing resources but that hardly matters for HTPCs whereas its 80 ALUs cannot save the day in modern games.
So, compared to Nvidia’s 40nm entry-level solutions, AMD’s Redwood (RV830) and Cedar (RV810) based products look really good. The senior model Radeon HD 5670 is especially appealing as it is more economical than the GeForce GT 240. The Radeon HD 5570 and 5450 are roughly comparable to their opponents GeForce GT 220 and GeForce 210 in terms of power consumption. Choosing a power supply for any of these Radeon HD 5000 series models won’t be difficult because a 250W and higher unit will do just fine. Of course, the PSU must be high quality. You shouldn’t save by buying cheap products from dubious brands like Codegen, for example.
And here is what the graphics cards show in terms of temperature:
* - passive cooling
The Gigabyte GV-R567OC-1GI is best here thanks to its large 80mm fan. The GPU is no hotter than 55°C under load, which is very good. The rest of the cards are hotter, especially the Sapphire HD 5450 512MB DDR3 which has the smallest heatsink. We can recall that its predecessor ATI Radeon HD 4550 suffered from the same problem whereas the tall heatsink helps the reference Radeon HD 5450. The more advanced card from Sapphire is very hot under load, too. The results are generally just as we might expect, and we want to repeat it once again that a graphics card with passive cooler you may choose for your noiseless HTPC needs a properly ventilated system case. Otherwise, its GPU, memory or voltage regulators may overheat with unpleasant consequences.
We also measured the amount of noise produced by the cards with active coolers. The two Radeon HD 5450 versions have no fans and do not increase the noise produced by our testbed at a distance of 1 meter, which is 37 dBA.
The Gigabyte GV-R567OC-1GI is louder than the reference Radeon HD 5670 according to our noise-level meter, but the later is actually more irritating because of its small and high-speed fan. Gigabyte’s cooler, on the contrary, merges into the noise produced by the other system components. The reference Radeon HD 5570 is even more irritating but its cooling efficiency is higher than that of the Sapphire HD 5570 1GB DDR3 DP. The latter tries to keep the fan speed low, reluctantly increasing it under OCCT: GPU. As a result, the GPU temperature is as high as 80°C but the card is virtually silent, despite the active cooler. We wouldn’t recommend installing it into a poorly ventilated system case, though, due to the high GPU temperature. So, the nonstandard cooling solutions are more comfortable for the ear although sometimes at the expense of cooling performance as is the case with the Sapphire HD 5570 1GB DDR3 DP. We guess that most junior models of AMD’s Redwood and Cedar based cards will be equipped with nonstandard cooling solutions.
Now we are ready to move on to practical tests. By the way, overclocking the Redwood and Cedar based cards makes little sense. They still won’t be fast in games and the higher frequencies will be perfectly useless for multimedia applications. We did try to overclock the Radeon HD 5670 in an earlier review and you can refer to it to see the result. Right now we will check out the multimedia capabilities of the new entry-level cards in the Radeon HD line-up.
The ATI Radeon HD 4000 series has been the best choice for HTPCs throughout 2008 and 2009. And the ATI Radeon HD 5000 is expected to be just as successful thanks to their heritage as well as to a few innovations including:
So again, the new cards can make a good choice for an HTPC. But how exactly good are they? We will check this out right now using the newest toolset for video playback quality measurements called HQV Benchmark 2.0.
We are going to investigate the decoding performance and playback quality of our today’s testing participants on the following platform:
The following graphics cards and integrated graphics processors took part in our tests:
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 suite requirements, the noise suppression and detail levels for Nvidia GeForce and ATI Radeon HD graphics cards were set higher, but not to the maximums, in order to avoid artifacts in case of extremely high level of detail combined with aggressive noise suppression.
Since the owners of high-end sound systems will be extremely interested in the results of lossless threads playback, we also included DTS-HD Master Audio and Dolby Digital TrueHD (where available) in order to increase the CPU load in all played movie fragments.
Keeping in mind that all tests are run under Windows 7/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.
To estimate the CPU utilization during full-HD video playback (1920x1080) and full-HD video with enabled “picture-in-picture” (PiP) or Bonus View (according to Blu-ray disc Association classification) feature, we used the following movies:
Like in a number of our previous articles, we decided to give up tests with free online media content (MKV rips of well-known shows), because its decoding is currently not supported by existing commercial software and the low bit-rates of the rips can be played perfectly by the system CPU.
The HQV benchmarks from Silicon Optix used to be our basic tools for video playback quality measurements but in October 2008 this manufacturer of high-quality video processors began to sell out its assets. The Reon processor technologies and the HQV benchmarks were bought by IDT. The new owner released new versions of HQV tests in 2010.
HQV Benchmark 2.0 for Blu-ray and DVD is a sophisticated toolset that allows checking out the various capabilities of a video processor. The new tests are similar in content and functionality but optimized for Blu-ray/DVD players as well as TV-sets. As the consequence, a number of specific features prove to be not implemented in modern GPUs which cannot reach the maximum of 210 points. We are going to dedicate a special article to the methods of testing image quality with HQV Benchmark 2.0 for Blu-ray and DVD.
Considering that some video processing capabilities depend on the graphics card driver and that the test suite is new, you should not view the HQV results as the ultimate truth, especially as there can be differences of opinion concerning the recommended evaluation procedures.
As you can see, none of the tested graphics cards can boast to have scored the maximum number of points. AMD’s Redwood (ATI Radeon HD 5500/5600) and Nvidia’s GeForce GT 240 score more points than the others. This indicates the high computing requirements of HQV Benchmark 2.0 because these GPUs are not superior due to some functionality enhancements. They just can process visual effects faster than the cheaper GPUs.
The tested GPUs seem to have problems reproducing video from different sources, with simultaneous upscaling and noise reduction (this especially concerns AMD’s GPUs), color correction, etc. Perhaps AMD and Nvidia need to have some time to implement the driver features necessary to improve the video processing quality up to the level of specialized video processors installed in expensive players and TV-sets.
HQV Benchmark 2.0 Blu-ray is exactly like the DVD version except for minor differences in the content and the different resolution. As a result, the GPUs don’t have to upscale DVD to full-HD format and the final results are somewhat higher.
The new GPUs from AMD look better than Nvidia’s but the gap is rather small and can be negated by driver updates.
Like the DVD version of the benchmark, the Blu-ray version reveals problems the modern GPUs have reproducing video from different sources, with simultaneous upscaling and noise reduction, color correction, etc.
The diagrams suggest that AMD’s new cards don’t have any problems reproducing HD video encoded in VC-1 format.
Save for the GeForce 210, each of the tested graphics cards shows a low usage of CPU resources when decoding VC-1 streams, so we don’t want to single out leaders or losers here.
When the H.264/MPEG4-AVC format is used, the overall picture is the same as with VC-1: AMD’s new GPUs offload the CPU well while decoding video and are roughly as good as their opponents, excepting the low-performance GeForce 210.
Judging by the diagram, AMD’s Redwood and Cedar solutions find it harder than their opponents to decode MPEG2 HD, but the difference is 4% can hardly matter much.
AMD’s Redwood and Cedar processors have very advanced video playback capabilities. Save for the lack of Blu-ray 3D support in the junior chip, these GPUs, together with the more expensive models of the Radeon HD 5000 series, seem to be an ideal choice for HTPCs. They have a hardware video decoder and an integrated audio controller. They can output audio in 7.1 format over HDMI 1.3a as well as transfer audio via Protected Audio Path to decode on an external device. They also support a variety of video formats including Adobe Flash 10.1 and offer a wide selection of interfaces (depends on the specific model).
Our tests have shown that the new GPUs can effectively cope with HD video decoding. They can also process two video streams simultaneously, which saves you the cost and trouble of purchasing a top-end CPU.
However, the tests from the IDT HQV Benchmark 2.0 suite suggest that none of modern GPUs can show ideal image quality in all situations whereas special-purpose video chips, e.g. from IDT itself, can do that. AMD and Nvidia will have to do a lot of work on their drivers and, perhaps, even on their hardware to reach the score of 210 points in the HQV tests.
We are going to investigate the performance of AMD Redwood and Cedar based graphics cards using the following universal testbed:
The graphics card drivers were configured in the following way:
Since the junior models of ATI Radeon HD graphics cards based on RV830, not to mention RV810, are definitely not the primary gaming choices we decided to use a lite version of our benchmarking suite. It included the following gaming titles and tests:
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 in any way, because the ordinary user doesn’t have to know how to do it.
Since the systems equipped with Radeon HD 5670/5570/5450 and similar solutions often serve as HTPCs and are connected to a plasma or LCD screen, we decided to run our gaming tests in a standard resolution for devices of this sort: 1360x768 (720p) and 1920x1080 (1080p). We forced anisotropic filtering 16x in each case, while MSAA 4x was only used in the least resource hungry games such as Left 4 Dead 2 and Resident Evil 5
Performance was measured with the games’ own tools and the original demos were recorded if possible. We measured not only the average speed, but also the minimum speed of the cards where possible. Otherwise, the performance was measured manually with Fraps utility version 3.0.2. In the latter case we ran the test three times and took the average of the three for the performance charts.
Each of the Radeon HD 5670 cards delivers an excellent speed in both 720p and 1080p mode. Their average performance is roughly similar to that of the GeForce GT 240 GDDR5 but the latter has a higher bottom speed, even though this has no effect on the gamer’s comfort. So, these cards allow playing Modern Warfare 2 on a big screen but we must note that even the senior Redwood (RV830) based card cannot beat the oldie Radeon HD 4770. They are not much faster than the Radeon HD 4670, either.
The Radeon HD 5570 is far from impressive. Equipped with slow memory and working at a reduced core frequency, it cannot deliver a playable frame rate even in the 720p mode although its average frame rate is good. As for the Radeon HD 5450, you can start modern games on it but you cannot play: the frame rate of 15-25 fps makes the process far from enjoyable. Thus, the Cedar is only limited to decoding high-definition video.
The in-game benchmarking tools are more accurate than Fraps but cannot report the bottom frame rate.
We don’t have any bottom speed info, but the overall picture is the same as in the previous game: the Radeon HD 5670 copes with both HD-ready and full-HD modes, the tiny factory overclocking of the Gigabyte card giving it a tiny advantage. Take note that the GeForce GT 240 GDDR5 falls behind AMD’s solutions at 1920x1080.
There are differences, though. The Radeon HD 5570 feels confident in this game. Although somewhat slower than the old Radeon HD 4670, it allows playing with enough of comfort. The Radeon HD 5450 is poor, again. The gameplay is jerky even at 1360x768 and the first-person shooter is reduced to a kind of a turn-based strategy as soon as there appear enemies in the scene.
The Radeon HD 5670 cards turn in good results for their class although both are inferior to the Radeon HD 4770. The gameplay is smooth at both 1360x768 and 1920x1080. The Radeon HD 5570 is almost as good as the Radeon HD 5570 and slightly inferior to the GeForce GT 240 GDDR5, but we’d recommend to turn FSAA off for the 1080p mode because the game begins to lag noticeably in some scenes.
This won’t do for the cards based on the junior GPU in the Radeon HD 5000 series. You have to lower the level of detail, losing some of the game’s visual impact. So, if you are going to play more or less seriously on your HTPC, you may want to choose a more advanced graphics card.
The game has such low system requirements that you can enjoy it fully even on a Radeon HD 5450 in HD-ready mode, i.e. at 1360x768. As for the faster cards, the Radeon HD 5670 is almost as fast as the Radeon HD 4770 and beats the GeForce GT 240 GDDR5 at 1920x1080. The most compact Radeon HD 5570 model competes with the Radeon HD 4670 and outperforms the GeForce GT 220 at 1360x768.
As we’ve shown in an earlier report, Mass Effect 2 is not a very demanding application if you don’t turn on FSAA. The inexpensive Radeon HD 5000 series cards agree with us: you can even play the game in the 1080p mode on the Radeon HD 5570 whereas the Radeon HD 5670 can give you a nice reserve of speed at that resolution.
You can also try to use the Radeon HD 5450 at 1360x768 but there can be occasional slowdowns in gameplay. Perhaps it is better to lower the game’s settings. As you can learn from our Mass Effect 2 review, the visuals won’t suffer much whereas the frame rate may grow up considerably.
We enable DirectX 11 mode for graphics cards that support it.
It is a real pleasure to drive a virtual car on a big screen. We have no doubt many HTPC users would appreciate the opportunity to play a racing sim on a large LCD panel. The Radeon HD 5670 allows using the highest-quality DirectX 11 mode but cannot deliver a high enough bottom speed even at 1360x768. The resolution of 1920x1080 is totally unplayable. The older Radeon HD 4000 series cards and the GeForce GT 220/240 cope much better with this game even though their graphics quality is inferior (they don’t support DirectX 11). Of course, you can switch into DirectX 9 mode, but you may want to purchase a Radeon HD 5750 or 5770 to enjoy Dirt 2 fully.
The game did not launch on the Radeon HD 5450 at 1920x1080. At 1360x768, this graphics card has a very low frame rate, making the game unplayable.
We use DirectX 11 and DirectX 10.1 modes for graphics cards that support them.
This beautiful real-time strategy can be no less visually stunning on a big screen than a fighting or a racing sim. Here, the Radeon HD 5670 does well at 1360x768, although its bottom frame rate, like that of the other tested cards, is low. At 1920x1080 the card’s average speed falls below 25 fps. We must confess that the DirectX 11 mode has but a very small effect on the game’s visuals but improves speed. This doesn’t help the Radeon HD 5670 much in its contest with the Radeon HD 4770.
The most inexpensive Redwood and Cedar-based cards are absolutely helpless here. Even if you don’t use FSAA, you can get no faster than 20 fps at best. The bottom speed is as low as 5 fps, effectively transforming the game into a slideshow. And you can’t do anything about that even by dropping all settings to their minimums. This seems to be the final proof that the Radeon HD 5450 is not a gaming graphics card.
We minimize the CPU’s influence by using the Extreme profile (1920x1200, 4x FSAA and anisotropic filtering). We also publish the results of the individual tests at the HD resolutions of 1360x768 and 1920x1080 without FSAA.
You can’t set any 3DMark records with a Radeon HD 5670, 5570 or 5450, that’s clear. The latter card sets a kind of an anti-record, scoring less than 500 points, though. Anyway, the most advanced of these Radeons is somewhat ahead of the GeForce GT 240 GDDR5. The compact and functional Radeon HD 5570 leaves no chance to the GeForce GT 220 and Radeon HD 4670.
It’s somewhat different in the individual tests: the Radeon HD 5670 both try to compete with the GeForce GTS 250 and even succeed at the resolution of 1360x768. The Radeon HD 5570 lacks fast memory. It is competitive to the GeForce GT 240 GDDR5 in the HD-ready mode but slows down in the full-HD mode because of that. The Radeon HD 5450 fails as a gaming graphics card again.
We think that AMD’s Redwood (RV830) and Cedar (RV810) based graphics cards are good and appropriate for their respective price categories. It does not mean they are free from any shortcomings. We can note that even the most advanced model, the Radeon HD 5670 with GDDR5 memory, is not very good for modern games. Our brief test session has shown that some games can be played in the HD-ready or even full-HD mode, but you cannot hope playing Crysis on a large screen as this wouldn’t be an easy task even for the top-end products from AMD and Nvidia. Anyway, we would dare recommend the Radeon HD 5670 as the best choice for a power-efficient HTPC that can be used for occasional gaming as this graphics card is far superior to the Nvidia GeForce GT 240 in terms of functionality. However, its full-size PCB won’t fit into compact low-profile system cases. You’ll have to buy a full-size “flat” system case which can be quite expensive. As for the specific models, the Gigabyte GV-R567OC-1GI leaves a good impression with its high-quality components and effective cooler. The cooler takes two slots, though, making this version of Radeon HD 5670 even more fastidious in its system case requirements than the reference card with reference cooler.
If you want to assemble your HTPC in a low-profile system case, you may prefer the Radeon HD 5570. It is a low-profile card that often comes with a low-profile mounting bracket included into the box. Its slow DDR3 memory and reduced GPU frequency won’t allow you to enjoy modern 3D games, though. According to our tests, the Radeon HD 5570 is only more or less good at 1360x768 or 1280x720 but generally cannot cope with the full-HD mode (1920x1080). By the way, Sapphire should be given credit for setting up the cooler of the Sapphire HD 5570 1GB DDR3 DP in such a way that the card is nearly silent whereas the reference card is not so comfortable acoustically. Sapphire’s card is hotter because of this, but a well-ventilated system case should solve this problem.
Finally, the Radeon HD 5450 presents no interest for gamers. But despite its weak 3D rendering capabilities, this card has a full-featured video processor and supports HDMI 1.3a, which may make it preferable to an integrated graphics core. The only exception is the Blu-ray 3D support. The Cedar has only one HD video decoding engine and cannot work with that upcoming format. The version from Sapphire discussed in our review is smaller than the reference card but gets hotter at work. It needs a well-ventilated system case. The reference card, on its part, is colder but may not fit into some compact enclosures due to its taller heatsink.
Our today’s tests have shown that AMD’s Radeon HD 5600, 5500 and 5400 series cards all cope nicely with HD video decoding, also with decoding two HD video streams simultaneously. Unfortunately, HQV Benchmark 2.0 suggests that none of modern GPUs can deliver ideal image quality which you can expect from special-purpose video chips installed into players and TV-sets. Hopefully, AMD will be able to improve the video playback quality of its solutions by polishing off their drivers.