by Sergey Lepilov
12/20/2009 | 02:55 PM
Saving some money is a perfectly human desire. Millionaires, housewives, computer users and, of course, overclockers want to save a little bit of their cash just like everybody else. In fact, the process of overclocking itself is the example of how you can get a higher performance for less money. As you have already guessed, this review is about saving, but I want to show you a way of saving some money other than by overclocking your components.
Let’s do some math first. As you know from our graphics card reviews, AMD’s line-up of Cypress- and Juniper-based products currently consists of two pairs of cards: the Cypress-based Radeon HD 5870 and HD 5850 and the Juniper-based Radeon HD 5770 and HD 5750. The Radeon HD 5870 comes at a recommended price of $379-399 whereas one Radeon HD 5770 should cost $159 according to AMD’s marketing folks. Easy to calculate, two HD 5770 are going to be cheaper than a single HD 5870. The same goes for two HD 5750 cards (2 x $129) in comparison with a single HD 5850 ($279-299). Obviously, you can save quite a bit of money in both cases by purchasing two junior cards instead of one senior model.
The question is if pairs of Radeon HD 5770 and 5750 cards in CrossFireX mode are going to be faster than a Radeon HD 5870 and HD 5850, respectively? Won’t the narrow 128-bit memory bus be a bottleneck? What pitfalls should be expected by a user who wants to build a graphics subsystem out of two junior graphics card models? I will try to give you the answers in this review.
The graphics cards are placed in the table in the order of ascending recommended price.
Click to enlarge
The two cards released under the XFX brand come in nearly identical boxes. The difference is in the model names. The XFX Radeon HD 5770 also has BattleForge stickers on the face side of the box.
The back sides of the boxes are identical, showing information about the key features of the products and technologies implemented in the new Juniper GPUs.
The XFX Radeon HD 5770 is accompanied with the following accessories: a cable to connect additional power, a CrossFireX bridge, a DVI → D-Sub adapter, a Do Not Disturb door plaque, two installation guides, a CD with drivers, and a coupon for a free download of BattleForge.
The XFX Radeon HD 5750 has the same accessories, except for the CrossFireX bridge, BattleForge coupon and DVI → D-Sub adapter which are missing.
The cards both follow the reference design, but the stickers on the coolers indicate that they sell under the XFX brand.
The Radeon HD 5750 is only 182 millimeters long (measuring from the mounting bracket to the edge of the cooler) while the HD 5770 is 38 millimeters longer. This is an advantage in comparison with the senior Radeon HD 58x0 series products. There are no decorative covers on the reverse side of the PCBs, which is good for cooling.
Each card has a PCIe 2.1 connector, two DVI-I ports, one HDMI connector and a DisplayPort. Some of the hot air from the card’s cooler is exhausted through the slits in the mounting bracket.
With the coolers removed, the cards look like this:
The memory chips are placed on both sides of each card. There are two CrossFireX connectors and one 6-pin power plug on each PCB. The power systems differ dramatically:
These were discussed in detail in our previous report on the Radeon HD 5750 and 5770 cards.
Both GPUs, which measure a mere 137 sq. millimeters and have unusually tall dies, were manufactured on Taiwan and differ with their markings and dates of manufacture.
As you can see, there is no protective frame around the graphics cores, so you should be most careful when removing or installing coolers on them. A Radeon HD 5750 GPU incorporates 720 universal shader processors, 36 texture-mapping units, and 16 raster back-ends. A Radeon HD 5770 GPU has the same number of RBEs, but 800 shader processors and 40 texture-mapping units. The GPU frequencies differ considerably: the HD 5750 clocks its GPU at 700MHz while the HD 5770, at 850MHz.
Both cards are equipped with the same GDDR5 memory chips manufactured by Hynix (marked as H5GQ1H24AFR-T2C, a rated frequency of 5000MHz).
The Radeon HD 5750 has a memory frequency of 4600MHz and the Radeon HD 5770, 4800MHz. We have some overclocking margin here, although the simplified PCB of the Radeon HD 5750 may affect an overclocking attempt. The memory bus is 128 bit wide. The total amount of graphics memory is 1024MB.
The cards’ specs are as follows:
You may have already noticed that the Radeon HD 5750 and HD 5770 have completely different coolers. The former card has a simple black aluminum heatsink with a 9-blade 80mm impeller. The Radeon HD 5770’s cooler consists of a heatsink, a blower, a separate aluminum plate for memory chips, and a plastic casing.
Dark-gray thermal grease is used as a thermal interface for both GPUs. The Radeon HD 5750’s cooler uses a 2-pin connection but notwithstanding this fact the fan speed is adjusted by the card automatically depending on the GPU temperature. The Radeon HD 5770 has a blower from AVC with a 4-pin connection and PWM-based speed adjustment.
The blower is driving the air through a small aluminum thin-ribbed heatsink. The ribs are placed 2 millimeters apart and soldered to the copper base which is encircled by a copper heat pipe. The pipe’s job is to distribute the heat uniformly in the heatsink base.
I tested the efficiency of these coolers in a closed system case. The room temperature was 24°C. The cards were loaded by FurMark 1.7.0 running in stability check mode at 1920x1200. The frequencies and temperatures were monitored withMSI Afterburner 1.4.1.
First let’s see how effective the coolers are in automatic fan management mode.
Curiously enough, the GPUs clocked at considerably different frequencies prove to have the same temperature under load, which is the consequence of the cards’ using different coolers. The Radeon HD 5770’s cooler is expectedly more effective although its blower works at 49% speed as opposed to the HD 5750’s 45% speed.
Then, I tested the cards at the full speed of their coolers.
The Radeon HD 5770 is more effective than the Radeon HD 5750 in this case, too. The HD 5770 is noisier than the HD 5750, but I will discuss the noise factor later.
I hadn’t enough time to test one pair of cards, so I only tried to overclock one Radeon HD 5770 and one HD 5750 (with their default coolers and in a closed system case). As a GPU stability check, I ran OCCT’s GPU test and two scenes from 3DMark 2006. I also made sure the graphics memory was stable by running two 3DMark 2006 scenes with 8x full-screen antialiasing and three heavy 3D games. Increasing the GPU voltage to 1.2V, I managed to overclock my XFX Radeon HD 5750 to 855/1425MHz (805MHz at the default voltage) and XFX Radeon HD 5770 to 960/1420MHz (910MHz at the default voltage).
The temperatures of the graphics cards did not change much after overclocking but the coolers were working at higher speeds than at the default frequencies.
I had six alternative cooling systems at hand but could not install any of them on the Radeon HD 5750 and 5770. The distance from the center of the GPU die to the aluminum screen of the DVI outputs is a mere 60 millimeters on both cards, so the coolers’ heatsinks just pressed against that screen and could not be installed. Considering that the GPU is so close to the graphics cards’ outputs, it is going to be hard to select an alternative cooler. By the way, the diagonal distance between the cooler’s mounting holes in the PCB is 61 millimeters.
Finally, I can show you a photo of two Radeon HD 5750 cards installed inside the system case and connected in CrossFireX mode:
As always, let me start with a few words about the hardware that we used during this test session. All graphics cards were benchmarked in a closed system case with the following configuration:
To minimize the processor influence on the performance of the tested graphics cards in certain operational modes and gaming applications I overclocked our 45 nm quad-core CPU with the multiplier set at 21x and “Load-Line Calibration” enabled to 4.1GHz. The processor Vcore was increased to 1.3825V in the mainboard BIOS:
The system memory worked at 1.56 GHz frequency with 7-7-7-14_1T timings and 1.62V voltage:
All other parameters available in the mainboard BIOS and connected with CPU or memory overclocking remained unchanged (set to Auto).
We are going to use reference ATI Radeon HD 5850 and HD 5870 working at their nominal frequencies:
Besides that, we will also include the results for HIS Radeon HD 4770 512 MB working at the nominal frequencies as well as ZOTAC GeForce GTX 260 AMP2! Edition 896 MB brought down to nominal frequencies:
We added the first graphics card to be able to compare it against the success – Radeon HD 5770. As for GeForce GTX 260, we also included it to compare against Radeon HD 5770, because they are currently retailing at almost the same price. All graphics cards were tested at their nominal frequencies. As for the results of Radeon HD 5770 and HD 5750 overclocking, you can check them out in one of our earlier articles.
Now let’s move on to software and benchmarking tools that we used. The tests started on November 30th and were performed under the new Windows 7 Ultimate RTM x64 operating system with the following drivers:
The graphics cards were tested in two resolutions: 1280x1024 and1920x1200. We decided to eliminate the tests in 1680x1050, because according to our previous reviews, the difference between the results in this resolution and in 1920x1200 is about 6-8%, while you need to perform 33% more tests. Our monitor didn’t support resolutions above 1920x1200, but it is a minor issue, because very few gamers use higher screen resolutions anyway and the tested graphics cards cannot provide sufficient performance to ensure comfortable gaming experience in 2560x1600 resolution.
The tests were performed in two image quality modes: “High Quality” without any image quality enhancements and “HQ+ AF16x+AA4/8x” with enabled 16x anisotropic filtering and 4x full screen anti-aliasing (or 8x FSAA if the average framerate was high enough for comfortable gaming experience). We enabled anisotropic filtering and full-screen anti-aliasing from the game settings or configuration files. If the corresponding options were missing, we changed these settings in the Control Panel. Vertical sync was always off in driver control panels.
All games were updated with the latest patches available at the time of tests. So, the complete list of test applications included two popular semi-synthetic benchmarking suites and 16 games of various genres, including such new titles as Borderlands and Left 4 Dead 2. Here is the complete list of tests used with the settings (all games listed in their release order):
Here I’d like to add that if the game allowed recording the minimal fps readings, they were also added to the charts. We ran each game test or benchmark twice and took the best result for the diagrams, but only if the difference between them didn’t exceed 1%. If it did exceed 1%, we ran the tests at least one more time to achieve repeatability of results.
Now let’s talk about the actual benchmarks.
The ATI Radeon 5870 1GB and HD 5850 1GB cards are marked purple in the diagrams. The new Radeon HD 5770 1GB and 5750 1GB and their CrossFireX configurations are marked red. The GeForce GTX 260 896MB is traditionally green while the old Radeon HD 4770 512MB is blue. Let’s start with synthetic benchmarks.
The CrossFireX configurations are both faster than the top-end single-chip cards in 3DMark 2006. The two junior Radeon HD 5750 are not only ahead of the single Radeon HD 5850 but stop very short of the Radeon HD 5870! This is an impressive result considering the difference in price. As for the single cards, the Radeon HD 5770 is competing with the GeForce GTX 260 and is expectedly superior to the Radeon HD 4770.
We see a different picture in the newer and more demanding 3DMark Vantage. The CrossFireX tandems, although delivering impressive scalability over the single cards, are no match to the Radeon HD 5870 and HD 5850. The GeForce GTX 260 is a little better than the Radeon HD 5770. The latter routs the Radeon HD 4770.
According to the heavenly demo from Unigine, the CrossFireX configurations are faster than their single-GPU opponents Radeon HD 5870 and HD 5850. The two junior Radeon HD 5750 cards are as fast together as the single Radeon HD 5870. The high results of the GeForce GTX 260 and Radeon HD 4770 (in ordinary mode) are due to the lack of DirectX 11 support whereas the cards based on AMD’s new GPUs all work in DirectX 11 mode.
And again the CrossFireX tandems built out of Radeon HD 5770 and 5750 cards show their best, delivering higher average frame rates than the Radeon HD 5870 and HD 5850, respectively, in every test mode. Moreover, the HD 5750 pair proves to be ahead of the Radeon HD 5870 when FSAA and anisotropic filtering are turned off. The overall picture is not so bright, though. At the high-quality settings the cards with 128-bit memory bus suffer a performance hit, resulting in a rather low bottom speed. The same 128-bit bus is the reason why the Radeon HD 5770 falls behind the GeForce GTX 260 which copes better with the high-quality test modes. The Radeon HD 4770 is expectedly last.
The highly resource-consuming Crysis agrees with World in Conflict in that the CrossFireX configurations of the junior Radeons scale up well and compete successfully with the single top-end solutions, but both of them fall behind the Radeon HD 5870 and 5850 in terms of bottom speed when the memory bandwidth requirements grow up. The single Radeon HD 5770 is competitive to the GeForce GTX 260 and 29-46% ahead of the Radeon HD 4770.
Unfortunately, Unreal Tournament 3 does not allow recording the bottom frame rate. But when it comes to average speed, the CrossFireX configurations are still ahead of the Radeon HD 5870 and HD 5850. It is only in the hardest test mode (if we can say hardest with respect to the UT3 engine) that the competing solutions are equals. Nvidia’s card still comes out the winner in the race between the GeForce GTX 260 and Radeon HD 5770. The latter is about three times as fast as the Radeon HD 4770, though.
There is again no data about the cards’ bottom frame rate. As for the average speeds, the Radeon HD 5750 tandem easily competes with the Radeon HD 5870!
This S.T.A.L.K.E.R. series game is known to scale up well and, like in the previous tests, the two CrossFireX configurations are no slower than the single top-end graphics cards. Take note that there are no slowdowns in this game (I do not use antialiasing because the frame rate is already too low).
CrossFireX technology works blamelessly in Far Cry 2. The 90% performance growth relative to the single cards helps the Radeon HD 5770 and 5750 CrossFireX configurations outperform the Radeon HD 5870 and HD 5850 in the low-quality test mode and equal them in the FSAA+AF mode. The bottom frame rate is high enough, so there is no discomfort with the CrossFireX tandems when playing this game. The new Radeon HD 5770 is overall slower than the GeForce GTX 260: a 128-bit memory bus just can’t beat a 448-bit one.
Call of Duty 5: World at War is one more game that does not record the bottom frame rate. The CrossFireX tandems are all right in terms of average speed, though. The GeForce GTX 260 enjoys a big advantage over the Radeon HD 5770 here. The Radeon HD 4770, on the contrary, is closer to the HD 5770 than in the previous tests.
Warhammer 40 000: Dawn of War II is one of the most CPU-dependent games in this test session. Suddenly, we see the Radeon HD 5770 and HD 5750 deliver higher performance in CrossFireX mode. I say suddenly because the Radeon HD 5870 and HD 5850 had performance drop in this mode in our earlier tests. This problem must have been solved in Catalyst 9.11 and CrossFireX technology works in this game now.
There are no definite leaders here. The CrossFireX configurations are as fast as the Radeon HD 5870 and HD 5850 in terms of average frame rate, except at 1920x1200 with the high-quality settings. The opponents are not always equal in terms of bottom speed, though.
The GeForce GTX 260 is ahead of the Radeon HD 5770 at 1920x1200 and is equal or slower than the later at 1280x1024. Thus, the Radeon HD 5770 betrays a lack of memory bandwidth again.
The CrossFireX tandems beat the single Radeon HD 5870 and HD 5850 cards in the low-quality mode and equal them in the FSAA+AF mode. The game does not allow measuring its bottom frame rate but it seems to run smoothly on either CrossFireX configuration. No lags or jerks in the integrated benchmark as well as in the game proper.
Stormrise seems to be the only game where the bottom frame rate of the CrossFireX tandems built out of entry-level cards is not lower than with the top-end single cards. These graphics subsystems are also equals in terms of average frame rate. The GeForce GTX 260 is beaten by the Radeon HD 5770 for the first time in this test session and is even overall inferior to the Radeon HD 4770.
The entry-level CrossFireX is superb in the low-quality mode! It is also good enough in the FSAA mode although we don’t know the bottom speed.
Developed for AMD’s GPUs, Call of Juarez: Bound in Blood just couldn’t say No to the CrossFireX configurations which are ahead in terms of average as well as bottom speed. The GeForce GTX 260 is surprisingly competitive to the single Radeon HD 5770 while the latter is 29-32% ahead of the Radeon HD 4770.
The CrossFireX tandems are very good in Wolfenstein, matching the single top-end graphics cards easily. The integrated benchmark cannot measure the bottom frame rate.
The CrossFireX configurations are ahead of the top-end single cards in the low-quality mode and similar to them in the FSAA+AF mode. The GeForce GTX 260 beats the Radeon HD 5770.
Although the average frame rate is good, the CrossFireX configurations built out of two HD 5770 and HD 5750 cards are worse than the single HD 5870 and HD 5850, respectively, and even occasionally than the single HD 5770 and HD 5750. Indeed, the game would run jerkily on the CrossFireX tandems. Hopefully, this is a driver issue (the game is new and AMD’s programmers are known for their sluggishness). Besides, FSAA does not work on AMD’s cards in Borderlands while I could easily turn it on in the control panel for the GeForce GTX 260.
The new zombie-killing game, just as the original, easily runs on the CrossFireX configurations which are ahead of AMD’s top-end cards. The GeForce GTX 260 is about as fast as the Radeon HD 5770. The latter is 12 to 29% ahead of the Radeon HD 4770.
The diagrams are based on the average frame rate in the games only. First, let’s see how faster the CrossFireX configurations of Radeon HD 5770 and HD 5750 cards are in comparison with the respective single cards. Here are the diagrams for the low-quality settings:
The Radeon HD 5750 boasts better scalability than the Radeon HD 5770 at 1280x1024: +79.7% against +70.2% on average across all the games. At a resolution of 1920x1200 the addition of a second card produces a bigger effect and the scalability of the two configurations is similar: +84.7% and +80.1%.
CrossFireX is even more effective at the high-quality settings.
The average performance growth is higher now: +81.4% for the HD 5750 and +76.2% for the HD 5770 at 1280x1024, and +81.2% and 81.1%, respectively, at 1920x1200.
Next, let’s compare the CrossFireX Radeon HD 5750 configuration with the Radeon HD 5850 and the CrossFireX Radeon HD 5770 configuration with the Radeon HD 5870. The single top-end cards are the point of reference in the diagrams.
In the low-quality test mode the CrossFireX tandems enjoy an advantage of 9.3-12.4% at 1280x1024 and 8.9-10% at 1920x1200. The only game where the CrossFireX tandems fail to win is Borderlands, at 1920x1200 only.
When FSAA and AF are turned on, the narrow 128-bit memory bus of the Radeon HD 5770 and HD 5750 affects their performance and the gap from the top-end single cards shrinks.
The CrossFireX configurations of entry-level cards are 5.2-7.7% ahead of the Radeon HD 5870 and HD 5850 at 1280x1024 and 3.5-5.6% ahead at 1920x1200. Besides the loss in Borderland we’ve already seen, the CrossFireX tandems now fall behind in Warhammer 40000: Dawn of War II. In the rest of the games the CrossFireX configurations are equal or better than their opponents. On the other hand, you should be aware that the diagrams are based on the average frame rate while the CrossFireX tandems are generally worse when it comes to bottom speed. This is all due to the AFR mode together with the 128-bit memory bus, which limit the graphics subsystem’s performance in high-quality modes.
The next pair of diagrams compares the GeForce GTX 260 (taken as the reference point) with the Radeon HD 5770.
As you can see, the GeForce GTX 260 wins more tests.
And the last pair of summary diagrams compares the Radeon HD 5770 1GB with the Radeon HD 4770 512MB.
The newer card is 196% and 329% faster than the Radeon HD 4770 512MB in Unreal Tournament 3 and Far Cry 2 but the diagrams are built to the same scale, so you can see only 100% there. The Radeon HD 5770 is 24-29% ahead of the Radeon HD 4770 in the low-quality mode and 35-70% ahead in the FSAA+AF mode.
I measured the power consumption of systems with different graphics cards using a special power supply. To create maximum load I launched FurMark 1.7.0 in stability check mode at 1920x1200 and Linpack x64 (LinX 0.6.3, 3500MB, 7 threads). These two programs load heavily the graphics card and CPU, respectively, so we can determine the peak power draw of the whole system.
You can see the results in the diagram:
The systems with CrossFireX configurations need more power than the systems with single-GPU top-end cards. Anyway, a 600W power supply is going to be sufficient for every tested system. A computer with a single Radeon HD 5770 or HD 5750 can do even with a 450W power supply, especially as such a computer is unlikely to have an Intel Core i7 overclocked to 4.1GHz.
Although the CrossFireX configuration built out of two Radeon HD 5750 was overall superior to the single Radeon HD 5850 and the two Radeon HD 5770 were faster than the single Radeon HD 5870, I wouldn’t say that they are preferable. The problem is that the CrossFireX tandems are often inferior in terms of bottom speed due to the AFR rendering mode and 128-bit memory bus. The latter shows up in high-quality modes and at high resolutions. So, while CrossFireX boosts the average frame rate to the level of top-end graphics cards, the bottom frame rate remains rather low. And it is the bottom speed that determines the level of playing comfort.
Considering the results of this and earlier test sessions with Radeon HD 5770 and HD 5750, including those at overclocked frequencies, purchasing two inexpensive Radeon HD 5750 cards and overclocking them to 850/5000MHz seems to be the most profitable option. At such frequencies, the cards totaling $250-260 will be faster than a single Radeon HD 5870 which currently costs $400 or more. The difference can be spent for a good mainboard and a couple of CrossFire bridges (the XFX Radeon HD 5750 comes without them, for example). So, this configuration seems appealing indeed.
The XFX Radeon HD 5770 and XFX Radeon HD 5750 cards discussed and tested in this review are good products following the reference design. The only thing I can complain about is that there is no CrossFireX bridge in the XFX Radeon HD 5750 kit, but the rest of the accessories are up to the product class. As these are one of the early Juniper-based cards available on the market, they do not have pre-overclocked frequencies. But XFX is known for its pre-overclocked products and I have no doubt the company will release such versions soon.