by Alexey Stepin , Yaroslav Lyssenko
11/28/2008 | 12:12 PM
Our recent testing of the new entry-level Radeon HD 4670 card from ATI showed that the below-$100 price category had progressed just like others. The mentioned card did very well, especially against Nvidia’s solutions falling into the same price category. Of course, the cutting down of the render back-ends subsystem and memory bus affected the new card’s performance, yet it proved to be competitive even to the more expensive GeForce 9600 GT. The Radeon HD 4670 left no chance to its direct market opponent GeForce 9500 GT.
The next step undertaken by AMD’s graphics department was the announcement of the Radeon HD 4500/4300 series. Here, in the most affordable price category, the former ATI Technologies had advanced significantly as its solutions were head above Nvidia’s counterparts, at least in decoding and output of HD video and audio. Moreover, the Radeon HD 4550 proved to have a very low power draw. Coupled with its enhanced multimedia capabilities, this made it an ideal solution for inexpensive, quiet, but advanced HTPCs.
There was only one niche left between the Radeon HD 4850 ($150-199) and Radeon HD 4670 ($79) and Nvidia filled it with its GeForce 9800 GT which was based on the time-tested G92 core and, judging by test data, was quite worth its price. However, product diversification basing on the same graphics core is a game that can be played by two, and ATI’s answer didn’t take long in coming.
It was logical for ATI to develop a new inexpensive below-$150 graphics card based on the superb RV770 core. This could help put to good use the cores that didn’t pass the frequency check and/or had some defective subunits. Thus, the release of the Radeon HD 4830 is an expected step in the evolution of the Radeon HD 4000 series and complies with ATI’s overall strategy.
Our today’s review is about this new addition to the Radeon HD 4800 family. As usual, we will report to you on the practical performance of the new card in a number of tests. We will also check out such important parameters as power consumption, heat dissipation and noise. But let’s first read through the Radeon HD 4830 specifications.
While the Radeon HD 4600 acquired a special processor called RV730, the Radeon HD 4830 doesn’t have a new chip. It uses an ordinary RV770 with some subunits disabled.
Click to enlarge
Click to enlarge
This is a justifiable approach because the manufacturing cost of the RV770 is not too high, especially in comparison with Nvidia’s G200. Cores that didn’t pass the frequency check or have defects in some subunits may also be installed on the new card. As a result, the Radeon HD 4830 differs from the Radeon HD 4850 with lower frequencies and fewer ALUs and TMUs. The difference isn’t big, though. Two out of the ten (160 ALUs) available SIMD arrays and two groups of corresponding texture processors (8 TMUs) are cut off. Thus, the new card looks preferable to the GeForce 9800 GT, being inferior to it in the frequency of the execution part of the core but far superior – nearly six times as good as its opponent – in terms of ALUs.
When it comes to the TMU subsystem, the GeForce 9800 GT formally has 56 texture processors, but they are only half efficient under real conditions. This card has only 28 TMUs effectively. So, instead of a formal advantage on Nvidia’s part, the ATI solution is actually better as it has 32 full-featured texture processors whose architecture has already proved its worth in games. The RBE and memory subsystems of the competing cards from ATI and Nvidia are roughly equal, the GeForce being somewhat better when working with the Z-buffer.
Apart from games, the Radeon HD 4830 has advanced multimedia capabilities such as hardware support for decoding of VC-1/H.264. It also has an integrated audio core with support for multi-channel HD audio formats. The GeForce 9800 GT can only transfer S/PDIF input from the sound card to HDMI and supports hardware video decoding for H.264 format only (VC-1 decoding is done together with the CPU).
So, it seems that the GeForce 9800 GT will find it hard to compete with the new card of the Radeon HD 4800 series in terms of gaming performance as well as functionality. Our gaming tests will show you if this is indeed so. But let’s first take a closer look at the card.
We’ve got a reference sample of the Radeon HD 4830 card which is technically identical to the reference Radeon HD 4850 and has the same PCB. ATI’s development strategy proves to be profitable again. Thanks to the simple and inexpensive design with a 256-bit memory bus, the PCB could be used to release a new, less expensive, Radeon HD 4800, without making it too costly to manufacture.
The new card has the same power circuit as the senior model. It was not changed, being quite simple as it was, and the peak power draw of the Radeon HD 4830 is declared to be 110W like that of the Radeon DH 4850. We’ll check this out, of course. The GPU voltage regulator is based on a two-phase uPI Semiconductor uP6201 controller with four power transistors in each phase. Another chip from uPI, UP6101, is responsible for the memory chips. The PCB carries one 6-pin PCIe 1.0 power connector with a load capacity of 75W.
Like the senior model, the Radeon HD 4830 is equipped with Qimonda HYB18H512321BF-10 chips that have a capacity of 512Mb (16Mb x 32), a rated frequency of 1000 (2000) MHz and a voltage of 2.0V. The card’s memory frequency is 900 (1800) MHz, though. The memory bandwidth is 57.6GBps just like that of the GeForce 9800 GT. The overall amount of graphics memory is 512 megabytes. We guess such fast memory is going to be installed on cards that follow the reference PCB design. Nonstandard versions of Radeon HD 4830 developed by graphics card makers to reduce the manufacturing cost further will probably come with slower memory and lower overclocking potential. They will be less interesting for overclockers, of course.
The GPU is an ordinary RV770 chip that is no different from chips installed on the Radeon HD 4850. Without knowing how to decipher ATI’s marking, you can’t tell them apart. Two out of the chip’s ten SIMD arrays are turned off along with the corresponding texture processors. Thus, the total amount of execution units is reduced from 800 to 640 in comparison with the Radeon HD 4850 and the amount of TMUs is reduced from 40 to 32. Anyway, the Radeon HD 4830 seems to have higher potential than the GeForce 9800 GT although the latter partially makes up for its fewer ALUs by clocking them at a higher frequency. The significantly reduced core frequency agrees with our supposition that the card may use those RV770 cores that were not certified for installation on the more advanced Radeons. The core frequency is only 575MHz as opposed to 625MHz of the Radeon HD 4850.
Like the Radeon HD 4850, the Radeon HD 4830 is equipped with two dual-link DVI-I ports and a universal 7-pin mini-DIN connector for analog video output. HDMI support is implemented via an adapter. The card also has two CrossFireX connectors for building a graphics subsystem out of four cards but we doubt anyone would want to join more than two Radeon HD 4830 in CrossFire mode.
The Radeon HD 4830 uses the same reference cooler as the Radeon HD 4850.
We described it in our theoretical review of the Radeon HD 4000 architecture. The key point of this cooler is that it has single-slot form-factor, yet this is also its main drawback. Small size means lower cooling performance. To keep the latter at an acceptable level, there is a flat heat pipe pressed into the base. It ensures uniform distribution of heat in the heatsink consisting of thin aluminum plates. Considering the cut-down configuration of the core that is also clocked at a reduced frequency, this cooler should do its job just fine, but the hot air will remain inside the system case. Therefore you shouldn’t install this card into a poorly ventilated system case. The cooler’s fastening is secure and unlikely to damage the GPU die.
Now we are going to proceed to the practical part of this review. We’ll first check out how good the above-described cooler is and if the cut-down RV770 consumes less power than the full-featured core.
We measured the power consumption of our Radeon HD 4830 on the following testbed:
The 3D load was created by means of the first SM3.0/HDR test from 3DMark06 running in a loop at 1600x1200 with 4x FSAA and 16x AF. The Peak 2D mode was emulated by means of the 2D Transparent Windows test from PCMark05. This test is important as it simulates the user’s working with application windows whereas Windows Vista’s Aero interface uses 3D features. Here are the results:
Despite the 55nm tech process, the peak power draw of the Radeon HD 4830 proved to be somewhat higher due to the complexity of the core incorporating 32 texture processors and 640 ALUs. That’s not much considering the potential of the card, though. The card is easily cooled by the reference single-slot cooler borrowed from the Radeon HD 4850. The GPU temperature is about 50-52°C when idle and 80°C under 3D load – that’s quite normal for a modern graphics card. Still, we wouldn’t recommend you to install your Radeon HD 4830, like any other card that doesn’t exhaust the hot air out of the system, into a cramped system case with poor ventilation.
The reference Radeon HD 4830 uses the same cooler as the Radeon HD 4850, so we can’t spot any difference in terms of noisiness.
The card is not noisy at all. The reduced power consumption means that there should be no overheat because the same cooler copes well with the more advanced HD 4850.
For our performance tests of ATI Radeon HD 4830 we used the following testbed:
According to our testing methodology, the drivers were set up to provide the highest possible quality of texture filtering and to minimize the effect of software optimizations used by default by both: AMD/ATI and Nvidia. Also, to ensure maximum image quality, we enabled transparent texture filtering. As a result, our ATI and Nvidia driver settings looked as follows:
For our tests we used the following games and synthetic benchmarks:
First-Person 3D Shooters
Third-Person 3D Shooters
We selected the highest possible level of detail in each game using standard tools provided by the game itself from the gaming menu. The games configuration files weren’t modified in any way, because the user doesn’t have to know how to do it. The only exception was Enemy Territory: Quake Wars game where we disabled the built-in fps rate limitation locked at 30fps. Games supporting DirectX 10 were tested in this particular mode.
Besides ATI Radeon HD 4830 we have also included the following graphics accelerators to participate in our test session:
We used the most widely spread resolutions: 1280x1024, 1680x1050 and 1920x1200. We enabled “eye candy” mode everywhere, where it was possible without disabling the HDR/Shader Model 3.0/Shader Model 4.0. Namely, we ran the tests with enabled MSAA 4x antialiasing and anisotropic filtering 16x in all tests except 3DMark. We enabled them from the game’s menu. If this was not possible, we forced them using the appropriate driver settings of ATI Catalyst and Nvidia ForceWare drivers. As we have already said, we didn’t modify the games configurations files.
Performance was measured with the games’ own tools and the original demos were recorded if possible. Otherwise, the performance was measured manually with Fraps utility version 2.9.6. We measured not only the average speed, but also the minimum speed of the cards where possible.
The Radeon HD 4830 is good in this test, especially for a product priced far below $150. It is almost as fast as the Radeon HD 4850 at low resolutions and provides a comfortable speed at 1920x1200. It is about 13% slower than the GeForce 9800 GT, though.
The game engine prefers graphics cards with good computing capabilities and the surgery performed by ATI couldn’t but affect the performance of the Radeon HD 4830. It is only really superior to the GeForce 9800 GT at 1920x1200 where each of the tested cards is slow. Well, the speeds are not comfortable even at 1280x1024, actually.
The Radeon HD 4830 and GeForce 9800 GT match each other in this test, but the latter has a higher bottom speed at 1920x1200, which makes is preferable for owners of monitors that support this display mode.
ATI’s solutions are not generally good in Crysis, except for the topmost models. The Radeon HD 4830 is far from being one, so it is 23-38% behind the GeForce 9800 GT depending on the resolution. The difference in the bottom speeds is even bigger. The card may occasionally slow down to 10fps and lower, which makes it impossible to achieve smooth gameplay even at the reduced graphics quality settings.
The frame rate is fixed at 30fps in this game as this is the rate at which the physical model is being updated at the server. Thus, this 30fps speed is the required minimum for playing the game.
The Radeon HD 4830 looks good in Quake Wars, especially at 1280x1024 and 1920x1200 where it outperforms its main opponent by 10% and 14%, respectively. ATI’s solution isn’t any worse than the GeForce 9800 GT at 1680x1050, either.
Being third is an excellent result for an affordable gaming graphics card especially as it is only inferior to the Radeon HD 4850 and GeForce 9800 GTX+ that belong to a higher price category. The amputation of shader and texture processors doesn’t have a big effect on the performance of the Radeon HD 4830. It is no farther than 12% behind its elder brother.
The HD 4830 is not brilliant in this game although it manages to reduce the gap from the GeForce 9800 GT from 22% to 12% as the resolution grows up. This gap looks small considering the low overall frame rates of the two opponents. You just won’t be able to play the game normally on these inexpensive cards unless you turn off such resource-consuming options as volumetric lighting.
Being a multiplatform project, Dead Space is not a very demanding game. The Radeon HD 4830 will be quite able to submerge you into the atmosphere of horror on board the spaceship Ishimura at any popular display resolution including 1920x1200. The GeForce 9800 GT is somewhat faster but the difference cannot be felt in practice.
The Tomb Raider series prefers Nvidia’s hardware, probably due to the more advanced RBE subsystem. Even with the increased texturing capacity of the RV770 core the RV770-based solutions can’t compete with the previous, let alone the current, generation of GeForce cards.
The Radeon HD 4830 makes no exception, but you can use it at resolutions below 1920x1200 to achieve the same level of comfort as with the GeForce 9800 GT. At 1920x1200 neither card can deliver an acceptable bottom speed.
The Radeon HD 4830 has a higher bottom speed than the GeForce 9800 GT but both cards deliver playable frame rates at 1280x1024 only. Anyway, ATI’s solution is ahead at two out of the three tested resolutions and even beats the GeForce 9800 GTX+ at 1280x1024.
The ATI Radeon HD 4830 is quite good here. It is at least as good as GeForce 9800 GT, and in 1280x1024 and 1920x1200 is significantly ahead of it, especially if we look at the minimal performance numbers. It means that this solution is a pretty good choice for playing Mass Effect. Here I have to remind you that this game doesn’t’ formally support antialiasing. You may still force it from the driver control panel, however, it will cause a significant performance drop in this case.
There is no need for much commenting here. The Radeon HD 4850 and Radeon HD 4830 are the only cards to be able to deliver playable frame rates at high resolutions whereas the GeForce 9800 GT and GTX+ are limited to 1280x1024. Moreover, the Radeon HD 4830 is better than the GeForce 9800 GTX+ at 1920x1200.
There is no clear winner at 1280x1024, but the average frame rate of the Radeon HD 4830 is higher than that of the GeForce 9800 GT at 1680x1050 and this gap grows from 22% to 28% at 1920x1200.
The GeForce 9800 GTX+ is ahead in open scenes of the game while the Radeon HD 4830 is as fast as the GeForce 9800 GT. It is only at 1920x1200 that the ATI solution has an advantage in terms of bottom speed.
Nvidia’s solutions have an advantage in terms of average frame rate but the bottom speed of the GeForce 9800 GT is below that of the Radeon HD HD 4830 at 1680x1050. It means that the latter ensures more comfort because the difference in average speed is only 13%.
Like some other projects from EA, Spore has an integrated frame rate limiter, so you should compare the bottom speeds of the cards in the first place.
Although the speed limit is set at 30fps, the Radeon HD 4800 series are the only cards to reach it at every resolution whereas the GeForce 9800 GT can be uncomfortably slow even at 1680x1050. Thus, this card is not a good choice for people who have a monitor larger than 19 inches. And we have to repeat it again that the potential of the G80/G92 architecture is virtually exhausted.
ATI’s cards are not brilliant here, but the Radeon HD 4830 is no worse than the GeForce 9800 GT, being slightly slower at 1280x1024 and slightly faster at the higher display modes. The overall level of performance is very low, so these results can be only of theoretical interest.
As we could expect, the Radeon HD 4830 is inferior to the GeForce 9800 GT in the SM2.0 tests but faster in the SM3.0/HDR tests thanks to its advanced computing capabilities. You shouldn’t make any far-reaching conclusions out of these results because 3DMark06 cannot really tell us much about the highs and lows of modern graphics architectures. It can only make a kind of a stop-watch for PC enthusiasts to fix their records.
The individual tests confirm our point: the Radeon HD 4830 is slower in the first test only. It wins in the second test even though the latter is limited to version 2.0 shaders.
The Radeon HD 4830 delivers high performance in the SM3.0/HDR tests, too. It is as good as the Radeon HD 4850 in the first of them, which indicates some speed limitation, perhaps due to the memory or RBE subsystem.
We minimize the CPU’s influence by using the Extreme profile (1920x1200, 4x FSAA and anisotropic filtering).
3DMark Vantage is not quite adequate to the realities of the current 3D world, yet it uses modern technologies and makes a good benchmarking tool. The Radeon HD 4830 deservedly has a higher overall score than the GeForce 9800 GT.
The Radeon HD 4830 enjoys an advantage in the first test which abounds in complex computations. The RV770, even in a cut-down configuration, is good at pure computing. The second test shows that the 32 texture processors of the Radeon HD 4830 are indeed equivalent to 56 such processors of the GeForce 9800 GT (well, it is an established fact that this card should be viewed as having 28 texture processors in real applications).
To all appearances the Radeon HD 4830 is a very good product. Of course, it is slower than the Radeon HD 4850 due to the cut-down configuration of the RV770 core – the difference amounts to 10-17% depending on the display resolution. But this is a very small loss considering the potential of the RV770 chip.
The Radeon HD 4830 is 100% worth its price which is below $150. You can get a detailed picture of its performance from our summary diagrams.
At a resolution of 1280x1024 pixels the Radeon HD 4830 loses eight out of the 17 tests to the GeForce 9800 GT but only in three tests it is indeed far slower than its opponent. Moreover, the GeForce 9800 GT could not provide a playable frame rate in S.T.A.L.K.E.R. Clear Sky and Crysis, either. And its advantage in Tomb Raider: Legend can only be seen in numbers but not in practice.
It’s about the same at 1680x1050: the Radeon HD 4830 has nine wins and eight losses, three of which are the above-mentioned S.T.A.L.K.E.R. Clear Sky, Crysis and Tomb Raider: Legend. Spore is the game where the Radeon HD 4830 is superior even to a naked eye.
The Radeon HD 4830 is not really meant for playing at 1920x1200 just because people who buy $100 graphics cards do not usually have an appropriate monitor. However, it can deliver a playable frame rate at that resolution, too, in a number of games including Dead Space, Enemy Territory: Quake Wars and BioShock. It now loses only six tests to the GeForce 9800 GT but wins 11.
Thus, the Radeon HD 4830 fulfils its purpose which was to fill in the gap between the Radeon HD 4850 and HD 4670. The card does its job well, being 100% worth its price. It may be slower than the GeForce 9800 GT in some games but the latter is based on an outdated architecture and doesn’t feel good in modern computations-heavy games. The Radeon HD 4830 is unlikely to become a sensation like the Radeon HD 4850, yet it is quite capable of becoming the best inexpensive gaming card featuring all modern multimedia capabilities.