by Alexey Stepin , Yaroslav Lyssenko, Anton Shilov
11/04/2009 | 03:01 PM
September 23, 2009, AMD’s graphics department announced their new-generation architecture called Evergreen and introduced two top-performance graphics cards based on the new RV870 GPU codenamed Cypress. This chip is the first in the history of consumer 3D graphics hardware to support DirectX 11. Using progressive 40nm tech process AMD’s engineers not only created a monster GPU incorporating 1600 ALUs and over 2 billion transistors but also made it economically justifiable – and even rather inexpensive! To remind you, the recommended price of the senior model of the Radeon HD 5800 series is only $399 while it delivers as much performance as the dual-processor Radeon HD 4870 X2 which had been originally targeted at the premium sector and priced at over $500. Our tests indicate that the single-processor newcomer is indeed capable of challenging the most advanced and luxurious dual-processor solutions of the previous generation.
On the other hand, it is the flagship product that defines the armada but the main battle is fought with smaller ships. That is, the most expensive and advanced solutions are meant to indicate the technological level of the developer but it is more affordable products that bring in the largest profit. People at AMD are surely aware of that even better than us and pay much attention to developing and promoting $200 and cheaper graphics cards that would be affordable for most gamers. We can recall the Radeon HD 4850 as an example: priced at $199, that card beat the more expensive GeForce 9800 GTX and got to be extremely popular. Then there appeared the Radeon HD 4770 model that AMD used to polish off its 40nm tech process. The $200 performance was made affordable at only $100.
And now it’s time for a new generation of affordable graphics cards and AMD follows its strategy again. Just when the Cypress was announced, we already knew about a simpler and cheaper GPU codenamed Juniper and scheduled for October 13.
In fact, ATI’s strategy has not changed since the RV770. First, the company rolls out a high-performance core. Then it begins to promote dual-processor solutions based on the core into the sector of premium solutions for enthusiasts. And after that, it develops a cheaper GPU or GPUs for the mainstream market. In the first quarter of the next year there will appear even cheaper chips, codenamed Redwood and Cedar, while the Juniper was indeed released on October 13. This chip is going to reinforce AMD’s position in the sector of gaming graphics cards and HTPC-oriented solutions priced at $100-160.
As you may have guessed, we will be testing Juniper-based graphics cards today. We will see if the Radeon HD 5700 series can carry on the glorious tradition of the Radeon HD 4850 and HD 4770 and reach a new performance and functionality milestone in the sector of affordable graphics cards.
Following the Radeon HD 4700, the Radeon HD 5700 series is meant to push the performance of affordable graphics cards to a whole new level. At the moment of the announcement AMD introduced two Juniper-based products: Radeon HD 5770 and Radeon HD 5750. The recommended prices of these two models are $159 and $100, respectively. Here is how the new cards compare with the previous generation:
Thus, the Juniper is comparable to the RV770/790 in terms of resources excepting the 128-bit memory interface. The similar transistor count of AMD’s new and old GPUs indicates that, too. Considering the architectural innovations and improvements, we can expect the Radeon HD 5770 to perform as well as or even somewhat better than the previous-generation solutions, except at very high resolutions together with full-screen antialiasing where its memory bandwidth may become a bottleneck. The new card is better than the Radeon HD 4770 in memory bandwidth but inferior to the Radeon HD 4890, GeForce GTX 275 and GeForce 260 Core 216.
Priced at only $100, the Radeon HD 5750 has much better specs than the Radeon HD 4770 and has every chance to get the title of the best entry-level graphics card. As users of such cards tend to run them for a year or more without upgrading, the card’s DirectX 11 support is going to come in handy.
So, the RV830 Juniper is architecturally one half of the RV870 Cypress.
It has 10 rather than 20 SIMD cores for a total of 160 universal superscalar execution processors each of which incorporates 4 universal ALUs, 1 special ALU capable of executing complex instructions, a branch control unit and a set of general-purpose registers. Of course, the GPU supports all DirectX 11/DirectCompute 11 features. Each SIMD core is serviced by four texture processors (a total of 40 TMUs).
The raster back-end subsystem is cut in half, too. There are 16 RBEs and 2 memory controllers, so the memory bus width is reduced from 256 to 128 bits. The size of L2 caches is the same at 128KB per controller (for a total of 256KB). 8KB L1 caches have also been left intact. The local and global data shares, 32 and 64KB respectively, are the same size as in the RV870, too. Again, the RV830 can be considered one half of the RV870 but technically, without considering the new features and architectural improvements, one such half is almost as good as the whole RV790.
The Juniper has inherited the enhanced connectivity its elder brother. Both models of the new series can service up to three monitors simultaneously. This has little practical value for gamers who play modern games as the Radeon HD 5700 won’t be able to deliver a high frame rate at such a huge resolution, but you may try some older titles with a 3-monitor configuration. Of course, this feature can also be beneficial for designers, artists and other people who want to have a larger desktop.
Now, let’s have a closer look at the new hardware.
The Radeon HD 5770 is very much alike to its elder brother. It has a square black casing with red strip but the card is substantially shorter at 20.5 centimeters against the Radeon HD 5870’s 27 centimeters. Owners of small system cases should appreciate this compactness.
The reverse side of the card looks more exciting than that of the Radeon HD 5870 as it caries half the memory chips and the controllers of the GPU and memory voltage regulators. The cooling system can be easily dismantled by unfastening four screws on the X-shaped back-plate. The other screws fasten the heatsink of the memory chips on the face side of the PCB. The face side looks like this:
Being compact, the PCB is populated heavily. There is little free room left. The power circuit follows a 3+3 design with 3 power phases in both the GPU and memory voltage regulators. One of the memory power phases is located near the CrossFire connectors. Looks redundant for a Juniper-based solution, but there is nothing bad in having more stable power.
The GPU voltage regulator seems to be managed by an L6788A controller from STMicroelectronics. A pair of uP7701 seem responsible for power supply of memory chips.
The card has one 6-pin PCIe 1.0 connector for additional power. This should be quite enough for an RV830 even at extremely high frequencies.
The Radeon HD 5770 is equipped with GDDR5 memory in 1Gb, 1.5V chips from Hynix (H5GQ1H24AFR-T2C). The T2C suffix denotes a rated frequency of 1250 (5000) MHz. The card’s actual memory frequency is lower at 1200 (4800) MHz. The RV830 has a 128-bit memory bus, so its peak memory bandwidth is 76.8GBps which is much lower than the Radeon HD 4890’s memory bandwidth of 124.8GBps. Theoretically, this may prove to be a bottleneck when running resource-consuming games at high resolutions with full-screen antialiasing.
The die of the RV830 Juniper has a rectangular shape. The marking is still unclear for the uninitiated except for the manufacture date. This sample is dated the 34th week of 2009, i.e. the end of August. The die lacks a protective frame.
As we’ve already said, the Juniper is similar to the RV770/790 in its specs. It too has 800 ALUs grouped into 160 universal shader processors. It also has 40 texture processors and 16 RBEs. The core frequency of the Radeon HD 5770 is 850MHz.
Like its elder brother, the Radeon HD 5770 supports up to three monitors simultaneously, offering two DVI-I ports, one HDMI connector and one DisplayPort. Of course, it also supports CrossFireX configurations and has two appropriate onboard headers.
The cooling system of the Radeon HD 5770 is a smaller copy of the Radeon HD 5870’s cooler. It consists of a copper base, connected with a heat pipe to an aluminum heatsink, and a casing with blower.
You can see the trace of traditional dark-gray thermal grease on the cooler’s base. A sealed end of the heat pipe can be seen in the bottom right. Below you can see the grooves for the air to go sideways from the mounting plate. The photo shows the model of the blower. It is a well-known Delta BFB1012 that is used in many products from ATI an Nvidia.
The heatsink is rather small and the flat heat pipe serves to distribute the heat uniformly in the base rather than to carry it on to the farthest edge of the heatsink.
The memory chips on the face side of the PCB are cooled with a metallic plate with low ribbing that has elastic thermal pads. Something like that was used in the cooling system of the Radeon HD 4770. As we’ve said already, the GDDR5 chips on the reverse side of the PCB are not cooled at all. This cooler doesn’t look like a record-breaking solution, but should cope with cooling a Radeon HD 5770 considering the latter’s low power consumption.
The Radeon HD 5750 is even shorter than the Radeon HD 5770 at only 18.5 centimeters, which is a record for a graphics card with that much power. The cooling system is much simpler and employs an ordinary axial fan.
Like on the senior model of the series, half of the memory chips are on the reverse side of the PCB. Dismantling the cooling system is easy, too. You only have to unfasten four screws on the X-shaped back-plate. The face side of the PCB looks like this:
The card is compact indeed and is populated as densely as the Radeon HD 5700. The 3-phase GPU voltage regulator is controlled by an uP6201 chip from uPI Semiconductor. A single-phase memory power circuit is based on an uP6101 chip from the same maker.
Like in the senior model of the series, one 6-pin PCIe 1.0 connector is used to connect external power.
Like on the Radeon HD 5770, Hynix H5GQ1H24AFR-T2C chips of GDDR5 memory are used. Again, they have a capacity of 1Gb, a voltage of .5V and a rated frequency of 1250 (5000) MHz. The memory frequency is just slightly lower than that of the senior model – 1150 (4600) MHz. This provides a memory bandwidth of 73.6GBps. That’s an unprecedented level for a $100 solution.
This sample of the RV830 is manufactured on the 34th week of this year, too. However, it is configured differently on the Radeon HD 5750. One of the SIMD cores is turned off. Thus, the graphics card has 144 shader processors out of 160 (720 out of 800 ALUs) and 36 out of 40 texture-mapping units. The number of RBEs has remained the same at 16. The GPU frequency is only 700MHz but the Radeon HD 5750 is no weakling. It should set a new performance record in the $100 category!
Like the faster Radeon HD 5800 series cards, this model has two dual-link DVI-I ports, one HDMI port and a DisplayPort. It supports CrossFireX, too.
The reference cooling system of the Radeon HD 5750 is very simple and resembles Thermaltake Orb series solutions, differing only in size. It is based on a round aluminum heatsink cooled by an 80mm axial fan. The fan uses a 2-pin connection.
There is a layer of dark-gray thermal grease on the cooler’s base. A piece of protective film with a cutout against the GPU die can also be seen. The cooler is fastened with four poles and a spring-loaded back-plate. The egg-shaped plastic cap may also serve a practical purpose by driving some air along the PCB to cool the power transistors of the voltage regulator.
Considering the reduced frequency and cut-down functionality of the Radeon HD 5750 graphics core, the above-described design looks sufficient to cool the junior model of the new series of mainstream graphics cards from AMD.
We measured the power consumption of the new RV830-based solutions on a special testbed configured like follows:
Following our standard methodology, we used the first SM3.0/HDR test from 3DMark06 to create 3D load. We ran it in a loop at 1600x1200 with forced 4x FSAA and 16x AF. Additionally we loaded the cards by running the OpenGL-based FurMark. The Peak 2D load was created by means of the 2D Transparent Windows test from PCMark05. We’ve got the following results:
The Radeon HD 5700 series shows excellent results even making allowances for our not-quite-new measurement method. The senior model consumes no more than 62W under 3DMark06. None of the previous-generation solutions comparable to the Radeon HD 5770 in terms of performance can beat this. The numbers are somewhat higher in the OpenGL FurMark test, but still acceptable: 81W for the senior model of the new series and 60W for the junior model.
The cooling systems of the Radeon HD 5770 and HD 5750 behave like that:
The senior model has a rather high GPU temperature due to the high core frequency coupled with the heatsink. Still, we found our Radeon HD 5770 perfectly stable even under a long run of FurMark. The GPU frequency of the junior model is 150MHz lower and the core itself does not work in its full configuration. 60°C is an excellent result for such a complex solution.
The reference point for our noise measurement tests is 43dBA which is the level of ambient noise in our test lab as measured at a distance of 1 meter from the testbed with a passively cooled graphics card inside. When we installed the tested graphics cards, we got the following results:
Just as expected, there is nothing extraordinary about the reference cooling systems of the Radeon HD 5770 and HD 5750. The cards are quiet, even though not silent, especially in 3D mode when their fans accelerate. The noise spectrum is quite agreeable to the ear, and this soft hiss does not irritate. People who want complete silence will anyway install a passive cooler while others, unless they are into extreme overclocking with extreme methods, will be quite satisfied with the reference coolers.
As opposed to the Radeon HD 5870, the Radeon HD 5700 series overclocks well. We overclocked our sample of Radeon HD 5770 to 940MHz GPU and 1445 (5780) MHz memory. The Radeon HD 5750 was overclocked to 870MHz core and 1430 (5720) MHz memory. We did this without any special methods, so we expect the Radeon HD 5700 series to get highly popular among serious overclockers. Unfortunately, we could not fix the result with GPU-Z screenshots because the current version of GPU-Z cannot report increased frequencies when the card is overclocked with Catalyst Control Center. This may be due to the PowerPlay technology implementation in the new ATI Radeon HD series.
We are going to investigate the performance of Radeon HD 5700 graphics cards using the following universal testbed:
The graphics card drivers were configured in the following way:
ATI Catalyst:
Nvidia GeForce:
The list of benchmarks includes the following gaming titles and synthetic tests:
First-Person 3D Shooters
Third-Person 3D Shooters
RPG
Simulators
Strategies
Semi-synthetic benchmarks
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. We made a few exceptions for selected games if that was necessary. We are going to specifically dwell on each exception like that later on in our article.
Besides Radeon HD 5770 and Radeon HD 5750 we have also included the following graphics accelerators to participate in our test session:
We ran our tests in the following resolutions: 1280x1024, 1680x1050, 1920x1200 and 2560x1600. Everywhere, where it was possible we added MSAA 4x antialiasing to the standard anisotropic filtering 16x. We enabled antialiasing from the game’s menu. If this was not possible, we forced them using the appropriate driver settings of ATI Catalyst and Nvidia GeForce drivers.
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. In the latter case we ran the test three times and took the average of the three for the performance charts.
Starting from version 1.3 the game allows to record and reproduce a demo. Unfortunately, this method does not report the bottom frame rate.
As expected, the senior model of the new series takes a good start, but starting from 1680x1050 it falls behind the Radeon HD 4890 due to lower memory bandwidth. This gap can only be bridged by overclocking. The new card’s results are high enough for comfortable play even at 2560x1600. We can also note that it is ahead of the GeForce GTX 275 at resolutions up to 2560x1600. That’s just incredible for a graphics card with a 128-bit memory interface, even though the memory is GDDR5.
The junior model is less confident but beats its previous-generation predecessor and delivers a good speed at 2560x1600. Overclocking makes it as fast as the Radeon HD 5770. Gamers with a limited budget should appreciate this.
The Radeon HD 5870 is the only solution with acceptable performance at 1280x1024. The new RV830 Juniper based models are behind the GeForce GTX 275 at every resolution save for 2560x1600 where they somehow go ahead notwithstanding the 128-bit memory bus. The Radeon HD 4890 is obviously faster than the Radeon HD 5700 and the gap cannot be eliminated even through overclocking.
We disabled the integrated frame rate limiter in the game console for the sake of comparing the cards. The game’s built-in benchmarking options do not provide information about the bottom speed, so there is no such info in the diagrams.
The game using MegaTexture technology, the graphics card’s memory bandwidth is crucial for its performance here. That’s why the Radeon HD 5770 cannot compete with the Radeon HD 4890 or GeForce GTX 275 even when overclocked. The junior model, Radeon HD 5750, is quite competitive to the Radeon HD 4770 and even better. Besides, the new cards are fast enough for comfortable play even at 2560x1600, especially as the game has an in-built frame rate limiter. Overclocking helps easily speed the junior model up to the senior model’s level.
The Radeon HD 5770 is somewhat slower than the Radeon HD 4890 throughout the test and cannot maintain a comfortable bottom speed at 1920x1200. Still, this is an excellent result considering the new product’s price whereas more demanding gamers can buy an RV870 Cypress based solution. Overclocking helps boost the new card’s bottom speed up to comfortable level at 1920x1200.
The Radeon HD 5750 is not far behind the senior model despite the substantial difference in clock rates and one disabled SIMD core. Graphics memory bandwidth seems to be the limiting factor for both solutions. The Radeon HD 5750 can be used to play at 1680x1050 at maximum graphics quality settings and with turned-on full-screen antialiasing. When overclocked, it delivers the same performance as the senior model of the series.
The game runs on the Source engine and has an integrated benchmark, but the latter does not report the bottom speed information.
At the two lower resolutions the Radeon HD 5770 delivers the same performance as the GeForce GTX 275. Starting from 1920x1200, its 128-bit memory bus affects the results. Every resolution is playable, though. The same goes for the Radeon HD 5750 which looks much better than the Radeon HD 4770, just as you can expect by comparing their specs. Overclocking produces the same effect as in the previous game. Note only that the limiting effect of the 128-bit memory bus cannot be made up for by ordinary overclocking at 2560x1600.
To achieve a playable speed in this game we disabled FSAA and such resource-consuming options as Sun rays, Wet surfaces and Volumetric Smoke. We use the Enhanced full dynamic lighting (DX10) mode for our test and additionally enable the DirectX 10.1 mode for the ATI cards.
The improved architecture helps the Radeon HD 5770 beat the Radeon HD 4890 at every resolution save for 2560x1600. It is also better than the GeForce GTX 275 everywhere, including 2560x1600. The Radeon HD 5750 is farther behind the senior model here than in the previous games because S.T.A.L.K.E.R.: Clear Sky makes use of complex special effects and a resource-consuming rendering technique that depends on the GPU frequency. This is indicated by the results of the overclocked cards: the Radeon HD 5770 overclocked to a GPU frequency of 940MHz leaves behind both the Radeon HD 4890 and GeForce GTX 275. The junior model of the series is almost as fast as the Radeon HD 4890 when overclocked.
Even at the relaxed settings the Radeon HD 5770 is limited to 1680x1050 whereas the Radeon HD 5750 can only deliver a comfortable frame rate at 1280x1024.
The junior model of the ATI Radeon HD 5700 series performs successfully, leaving its predecessor behind at high resolutions and being somewhat slower than it at low resolutions. The senior Radeon HD 5700 series model is no match to the Radeon HD 4890, but this is not a problem at all. The average frame rate is over 60fps at 2560x1600, which is more than enough even for first-person shooters that require much more accurate aiming.
Overclocking gives both cards a considerable boost, but has little practical value. On the other hand, overclocking can give you the performance of a Radeon HD 5770 for the price of a Radeon HD 5750.
The junior Radeon HD 5700 model is preferable again. It is somewhat faster than the Radeon HD 4770, delivering a comfortable frame rate at 2560x1600. The faster Radeon HD 5870 cannot match the Radeon HD 4890 or GeForce GTX 275 although its performance is good enough for its price. You can have higher results by overclocking the card, yet the 128-bit memory bus acts as a bottleneck, The Radeon HD 5770 is slower than the Radeon HD 4890 and GeForce GTX 275 even when overclocked.
We see the same picture in Fallout 3, the original game with all the add-ons. The ATI Radeon HD 5770 is inferior to the cards with wider memory bus but allows playing comfortably at 2560x1600. The bottom speed of the Radeon HD 5750 is barely comfortable at that resolution although is higher than that of the ATI Radeon HD 4770. You can get some reserve of speed by overclocking the junior model. And we don’t actually think that too many people will be using Radeon HD 5750 cards together with 30-inch monitors.
The new mainstream Radeons show their best in NFS: Shift. The senior model is but slightly slower than the Radeon HD 4890 and even goes ahead at 2560x1600 whereas the junior model shows good performance at resolutions up to 1920x1200 inclusive. When overclocked, the Radeon HD 5750 allows playing comfortably at 2560x1600 thanks to the significant growth of the bottom speed.
We use the in-game benchmarking tools that do not allow to measure the bottom frame rate. We also enable DirectX 10.1 support for ATI’s solutions.
Oddly enough, the gap between the Radeon HD 5770 and Radeon HD 4890 is shrinking as we switch to higher resolutions. This must be the effect of the new, improved architecture. The Radeon HD 5750 beats the Radeon HD 4770 again, which is not surprising already. The game’s engine is still incompatible with AMD’s Adaptive AA algorithms, so you may want to turn this option off on your Radeon card in order to have higher speed.
We enable DirectX 11 support for the Radeon HD 5x00 series.
The new cards’ DirectX 11 support brings considerable benefits here. The Radeon HD 5770 is at first equal to the GeForce GTX 275 and then leaves the latter far behind. The narrow memory bus prevents it from beating the Radeon HD 4890, though. By the way, the Radeon HD 5750 also beats Nvidia’s solution at 1920x1200. Alas, both cards, just like the top-end Radeon HD 5870, cannot offer a high bottom speed in this game. Overclocking does not help here although makes the senior model as fast as the Radeon HD 4890 and the junior model as fast as the senior one.
The senior model of the ATI Radeon HD series can only beat the Radeon HD 4770 but is no rival for the GeForce GTX 275 or Radeon HD 4890. That’s normal considering the new cards’ 128-bit memory bus. In fact, their performance is excellent for their price category! Moreover, the overclocked Radeon HD 5770 allows playing comfortably enough at 1280x1024 even though its bottom speed might be higher.
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 across all display resolutions to provide a full picture.
The narrow memory bus shows up in synthetic benchmarks, too. The Radeon HD 4890 scores more points. Although the new card can outperform the older solution when overclocked, this is not its standard operation mode. The performance of the junior Radeon HD 5700 series model is just as expected, too.
The Radeon HD 5770 behaves like in Tom Clancy’s H.A.W.X., getting closer to the Radeon HD 4890 at higher resolutions although we might expect the opposite theoretically. The Radeon HD 5750 shows a somewhat chaotic behavior, its lead over the opponent changing in no particular order. For example, it enjoys the largest lead at 1680x1050 and 2560x1600 in the first test. The new cards do not show anything extraordinary here.
Considering its extremely low idle power consumption and modest power requirements under load, the ATI Radeon HD 5770 seems to be an ideal choice for HTPCs. In this section we will compare the ATI Radeon HD 5770 against its rivals in terms of high-definition video playback in the Microsoft Windows 7 environment. (In this section of this article we will try to compare ATI Radeon HD 5770 against the rivals in terms of high-definition video playback in Microsoft Windows 7 operating system).
We are going to investigate the decoding performance and playback quality of our today’s testing participants on the following platform:
We also used the following additional testbed:
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 HD suite requirements, the noise suppression and detail levels for Nvidia GeForce graphics cards were set to the maximums.
Unfortunately, Windows 7 doesn’t support HD DVD playback using CyberLink PowerDVD player. That is why we used an additional testbed with Windows Vista OS and traditional HQV HD test to estimate the HD playback quality. Moreover, since absolutely all graphics accelerators participating in our today’s test session demonstrate extremely low results in the HQV test, we also estimated the DVD playback quality in Windows Vista. As a result, you can see the results obtained at different time, on different computer systems and with different drivers on the diagrams. Although in this case you can argue about the fairness of such comparison, since the HQV HD results may vary depending on the driver version, we believe that by providing the HQV HD data we show the maximum potential of the participating GPUs very well.
Since the owners of high-end sound systems will be extremely interested in the results of loseless 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, because its decoding is currently not supported by existing commercial software.
The HQV benchmarks from Silicon Optix are one of the few available methods of evaluating the playback quality of Blu-ray, DVD and HD DVD movies. They have one drawback, however. The tester’s perception is subjective while the notion of an ideal picture is rather vague.
We have been criticized for following the HQV HD test instructions too closely and giving out too low scores. The situation is different now: DVD video is so bad on Windows 7 with the CyberLink player that we can’t recommend using DVD.
Our experience suggests that video playback can improve or get worse both with DVDs and Blu-ray discs.
So, considering the subjective nature of this test, you should not view the HQV and HQV HD results as the ultimate truth.
Although DVD has become obsolete, not all modern GPUs can play that old format in high enough quality. As you can see, the Radeon HD 5700 series is not brilliant here: its major drawbacks are pretty low noise suppression and inability to display video with scrolling text correctly.
However, none of the tested GPUs could achieve the maximum score of 130 points. The main problems are the inability of the GPUs to smooth out the jaggies pattern and transform filmed content (24fps, progressive scan) into DVD or 1080i HDTV with the 3:2 pulldown method. Besides Radeon HD 5770, the whole lot of graphics cards proved unable to render text over video, which is totally unacceptable nowadays.
Standard-resolution video is dying out. TV channels in the United States are already broadcasting in 720p whereas HTPC owners are going to use Full-HD TV-sets with a resolution of 1920x1080. As a result, it is far more important to ensure high-quality playback of high-definition video.
As you can see, many modern GPUs boast exceptional quality of HD video playback (considering the problems with HQV HD + Windows 7 + CyberLink PowerDVD 8/9, we use data from earlier reports). In fact, nearly every mainstream and performance-mainstream GPU is very good at reproducing HD content. Unfortunately, the ATI Radeon HD 5700 series have poor results in the HD noise reduction test, which must be due to software flaws. Driver updates should resolve the issue.
Note that we are pretty liberal about the Film Resolution Loss Test - Stadium. ATI Radeon HD 5700 and other GPUs that received 10 points for that test had absolutely no moiré but had some slight flickering. Although it was barely noticeable in most cases, some graphics processors got 10 points for it. However, if you feel like fault-finding, you should subtract these 10 points from the total score because, according to the HQV HD instructions, flickering means 0 points.
ATI’s graphics cards demonstrate stably lower CPU utilization than their opponents. I can’t say that the results of ATI Radeon HD 5700 are dramatically different from the CPU utilization with previous generation graphics cards, which is in fact pretty logical considering how powerful ATI Radeon HD 4000 solutions are.
As opposed to VC-1, there are no clear leaders or losers when the cards are decoding MPEG4 AVC/H.264. The average CPU load is within 16%, which is rather low. ATI Radeon HD 5700 shows more of the low rather than high CPU utilization.
ATI and Nvidia took into account that decoding of MPEG2 HD video is an important (as there are first-generation Blu-ray discs in the market still), even though not crucial (in fact all new Blu-ray discs use more progressive MPEG4 AVC or VC-1 codecs) factor of a graphics card’s performance. As you can see, both companies have done a good job improving their HD video decoding engines. As we see, only a not very powerful ATI Radeon HD 4550 requires 21% CPU time on average to decode the scene from Alien vs. Predator movie. The new ATI Radeon HD 5700 needs even less CPU participation during MPEG-2 HD playback compared against the predecessors.
Without a doubt, the ATI Radeon HD 5700 series is a good choice for movie fans. High playback quality and HD video decoding support (also of two streams simultaneously) may them especially appealing. The only downside of the 5750 and 5770 models in the eyes of an HTPC user is that these cards are equipped with active cooling systems. If you are looking for exceptional 3D performance and DirectX 11 support, ATI Eyefinity and other innovations offered by ATI Radeon HD 5700, but you also would like to have a very quiet graphics card, then maybe you should wait until ATI’s partners, such as Gigabyte Technology for instance, will offer their own Radeon HD 5700 solutions with proprietary passive coolers onboard.
AMD keeps on revolutionizing the graphics card market. After giving us more speed, the company now offers more functionality by introducing DirectX 11 features into the mainstream sector. The Radeon HD 5770 and Radeon HD 5750 are both highly appealing products. Compact, quiet, economical, affordable and DirectX 11 compatible, AMD’s new RV830 Juniper based graphics cards are a treat for every gamer who cannot invest into a more expensive RV870 Cypress based product.
It is the senior model of the new series, Radeon HD 5770, which is the most interesting. Let’s take a look at its results.
Of course, the Radeon HD 4770 is out of competition. The new card is 6 to 56% faster depending on the particular game. The average gap is as large as 34%. It is harder for the newcomer to compete with the Radeon HD 4890 which has a higher memory bandwidth. This can be observed even at 1280x1024 where the Radeon HD 5770 is an average 12% slower than the RV790-based solution. On the other hand, notwithstanding the 128-bit memory bus, it beats the GeForce GTX 260 Core 216 in five out of the 13 tests!
When the resolution is higher, the average advantage of the Radeon HD 5770 over the Radeon HD 4770 grows to almost 40% but the new card is still about 12% slower than the Radeon HD 4890 (but it wins in S.T.A.L.K.E.R.: Clear Sky thanks to architectural improvements). The race with the GeForce GTX 260 Core 216 produces mixed results: 5 wins and 8 losses, the loss in Tom Clancy’s H.A.W.X. being still due to the game’s incompatibility with Adaptive AA. It is only in World in Conflict: Soviet Assault that the gap is really large.
The new card stays at the same distance from the Radeon HD 4890 here, but increases its advantage over the Radeon HD 4770 to 46%. The Radeon HD 5770 improves in its fight with the GeForce GTX 260 Core 216. Notwithstanding the higher resolution and 128-bit memory bus it wins seven tests at 1920x1200! It falls far behind Nvidia’s solution only in Far Cry 2 and World in Conflict: Soviet Assault.
AMD’s new mainstream generation enjoys a 50% advantage over the previous one at 2560x1600, yet the new card cannot overtake the Radeon HD 4890. The Radeon HD 5770 wins 8 and loses 5 tests in its competition with the GeForce GTX 260 Core 216. It is hard to believe this as Nvidia’s card has a 448-bit memory bus!
The Radeon HD 5750 is inferior to the Radeon HD 5770 due to lower GPU and memory frequencies and one disabled SIMD core. The gap is small at 14-15%, though. It is only in some applications that the gap grows to 19-20%. The card delivers a comfortable frame rate in most tests, excepting the extremely high resolution of 2560x1600. This is an excellent choice for an inexpensive gaming platform or an HTPC, especially as the Radeon HD 5750 is compact, has low heat dissipation and offers the same multimedia capabilities as the more advanced members of the fifth ATI Radeon HD generation.
The Radeon HD 5700 models both boast good overclockability without any special tools or methods. In some applications overclocking makes the senior Radeon HD 5700 series card as fast as the Radeon HD 4890 whereas the junior model, when overclocked, easily overtakes its senior cousin. The average performance growth of the Radeon HD 5770 overclocked to 940MHz GPU and 1445 (5870) MHz memory frequencies is about 12%, reaching 14% in some applications.
For the Radeon HD 5750 overclocked to 870MHz GPU and 1430 (5720) MHz memory the average and maximum performance growth are 14% and 16%, respectively. This is a real gift for overclockers, especially as PC enthusiasts will surely reach better results using more advanced coolers. On the other hand, we have to note that the main bottleneck of the new series is the 128-bit memory bus which is going to limit the effect from overclocking at high resolutions.
AMD’s mainstream products have taken up the DirectX 11 banner now. We guess the new Radeon HD 5700 series is a success and virtually free from noticeable drawbacks. There is currently a hiatus in the Radeon HD 5x00 line-up at $160-199. What card is supposed to take this niche? We will surely find the answer soon. People at AMD cannot neglect such an obvious fact.
Highs:
Lows: