ATI Radeon 9800SE Video Card Review

By: Aleksey Burdyko

The emergence of 3D flagships like ATI Radeon 9800 Pro and NVIDIA GeForce FX 5900 Ultra (and even their renewed "overclocked" versions of the mentioned chips ATI Radeon 9800 XT and NVIDIA GeForce FX 5950 Ultra) couldn't help provoking the emergence of cut-down solutions built on the base of generic Hi-end, which is simply beyond means for many. But nowadays we can witness an increasing popularity of DirectX 9.0 games, so you can't agree with the latter statement, unless you are an inveterate fancier of "mine-sweeper" who does not care about those Doom's of the third series or Half-Lives 2. This is quite natural both in terms of economy and marketing as well as technology since modern technologies of graphic chips manufacture imply a certain amount of defective articles, and which can be put to use through applying some cunning tricks.

It's quite a different thing when the release of cut-down solutions in fact is not sanctioned by the chips manufacturer. If you look into ATI's website, you won't find even a mention of the chip Radeon 9800SE among the currently available Radeon solutions. "Why so?" - the user will ask. Isn't the chip of ATI's or shouldn't it be responsible for that? All that would have been that way but only if it concerned the chip produced from process-rejected R350 chips sent for the manufacture of Radeon 9800 Pro video cards. Things about that are more tickling than it may seem at first glance. That's why ATI looks like it is not to blame for that. All the operations of manufacturing video boards on the base of the Radeon 9800SE chip are handed over to the partners themselves. Of course, that couldn't help provoking the partners themselves to make more cunning tricks with boards bearing the Radeon 9800SE name. This is why there are two (even three) retail versions of boards on the base of the Radeon 9800SE chip, which (as our tests will tell) offer radically differing performance at the same retail price. There is a definite mix-up and mishmash for the end users who wish to buy a board of the "a-little-bit-lower-than-the-hi-end" level =)" but can't do that simply because ATI partners start fitting culled ATI R350 chips on various PCB. So, what do we get in the upshot? Below, we'll try to sort out which boards are there to date on the market and how to identify it by eye what sort of a product we got.

Radeon 9800SE: Background...

First of all, let's remember how such chips like Radeon 9800SE come into the world and what potentials (and the related problems =) ) such boards offer.

And the possibility of emergence of such chips like ATI Radeon 9800SE is embedded in the technology for the manufacture of graphics chips. In the production process, there come up chips that are unable to run at the intended frequencies. For instance, this is how the ATI Radeon 9800 appeared that runs at speeds reduced relatively to ATI Radeon 9800 Pro. However, there exists another kind of production flaws when the chip is able running at the intended frequencies but in so doing one of the rendering pipelines fails instead. In this case, to make these cores be of use in the production, half the core is disabled (i.e. half of the rendering pipelines become inactive) and then they start making graphic boards with such a chip. Since the chip was originally manufactured for operation with the full number of pipelines (for the R350, their number is 8) and physically there are just 8 pipelines, with 4 are disabled programmatically, then "great folk craftsmen" get a real opportunity to switch on the disabled pipelines back programmatically. Then we'll review all the aspects of the software re-making of Radeon 9800SE into full-featured ATI Radeon 9800 Pro, with a small reservation though. Clearly, the re-make does not give you a 100% functional board, since if it is culled for Radeon 9800SE boards, then the chip ALREADY has a flaw in one of the rendering pipelines. Sure you won't see anything much after the re-make, but nevertheless the flaw will anyway come out. The thing is even chips of minimum "deviations" are subject to culling, and the lucky owners can anyway get a board which after re-making will behave in the "right" way, and of course things may turn unlucky and the chip will fail to operate properly.

Nevertheless, the prospect of getting an ATI Radeon 9800 Pro out of an ATI Radeon 9800SE "for free" will be saddened by the mess among ATI partners, which we mentioned in the beginning of the review. ATI let its partners "loose" and handed the design of boards with defective R350 chips over to the manufacturers. Originally, the culled R350 chips were to be installed on PCBs of Radeon 9700 Pro and offer a 256-bit memory exchange interface. That's the way some video cards manufacturers followed. But then there appeared the idea of making the boards on the base of ATI Radeon 9800SE cheaper due to installation of the already culled R350 chip onto the simpler PCB of ATI Radeon 9500 Pro with the 128-bit memory exchange interface. Reduction of production costs was attained, but that did not imply lower retail prices for the boards, which are missing in the retail. The memory bus width was not indicated on boxes, so the users in fact were buying a "pig in a poke". It is absolutely clear that in the latter case you won't get a full-featured ATI Radeon 9800 Pro out of ATI Radeon 9800SE through enabling extra 4 rendering pipelines because the memory bandwidth restricted to 128 bit will hinder, but which should be equal to 256 bit the way it is in the full-featured ATI Radeon 9800 Pro. After re-make you won't get an ATI Radeon 9800 Pro but something that ATI marketing people are still unaware of - a board whose performance is well behind that of ATI Radeon 9800 Pro.

So, how to tell ATI Radeon 9800SE boards apart? In practice, the PCB designs of ATI Radeon 9700 Pro and ATI Radeon 9500 Pro are very easy to distinguish, so below we are presenting sort of a brief guide on how to tell the "right" and "wrong" ATI Radeon 9800SE apart:

To the left, there is a photo of a PowerColor Radeon 9800SE board. To the right, there is a photo of the ATI Radeon 9700 Pro reference board. As we see, both the boards are absolutely identical by the PCB design, that is the PowerColor Radeon 9800SE is a "right" board based on the Radeon 9800SE after the re-make of which we can get an ATI Radeon 9800 Pro.

And this is how the reference design of the ATI Radeon 9500 Pro looks:

All this reasoning implies that Radeon 9800SE boards built on this PCB can be regarded "wrong" =) merely because of the 128-bit memory bus.

In practice, it's not difficult to distinguish the PCB designs used for the manufacture of both versions of Radeon 9800 SE. As a reference, take the positioning of the memory chips. The memory chips on the PCB of Radeon 9700 Pro are positioned in pairs around the graphics chip, while the PCB of ATI Radeon 9500 Pro chips provides that the memory chips be positioned in line under the graphic chip.

Radeon 9800SE: software re-make in action

Below is sort of a brief FAQ for beginners on how to re-make an ATI Radeon 9800SE board into a full-featured ATI Radeon 9800 Pro (of course, provided the ATI Radeon 9800SE offers a 256-bit memory bus).

We will open the locked pipelines with the RivaTuner program of the most recent version. The program already offers various scripts including those for activating locked pipelines.

To enable the pipelines, apply a patch to the original drivers with the script available from and the setup folder RivaTuner/ PatchScripts/ ATI / SoftR9x00.

The dialog box looks like this:

Then you can install drivers with the already patched file. If the driver is WHQL-certified, then once the above actions are complete the setup program will inform the WHQL-certification is no longer there:

Whether the pipelines are enabled, you will know from here:

As we see, all the pipelines have been enabled, but the board hasn't passed all the tests in such a mode.

Two graphic boards built on the ATI Radeon 9800SE chip took part in the tests: one on the base of the PCB from ATI Radeon 9700 Pro (PowerColor Radeon 9800SE) and one on the base of the PCB from ATI Radeon 9500 Pro (Sapphire Radeon 9800SE).

PowerColor Radeon 9800SE 256-bit

Package bundle:

Since the board came to our test lab not in the retail version, but as an OEM version, then there's nothing much to talk about the package bundle:

• The graphics board itself;
• A user's manual in 5 languages;
• 1 CD with the WinDVD 4 software suite.

This is an absolutely standard delivery package for a board supplied in the OEM version.

Design and layout

As we have already noted, the ATI Radeon 9800SE video card is manufactured on the base of the PCB from Radeon 9700 Pro. We couldn't find any differences from the reference ATI Radeon 9700 Pro board - the reference PCB design was used.

The board offers ATI's traditional bright red color of the PCB, has 128 MB DDR memory onboard, the AGP 2x/4x/8x interface and a standard set of outputs: one analogous, one digital, and one TV-OUT. The board does not provide any third-party codecs to implement the TV-OUT - the features to support it are already integrated in the Radeon 9800SE chip. The good old two-phase SC1175CSW of Semtech is used as a voltage regulator.

The video card is equipped with 128 MB DDR memory packaged in 8 chips (4 chips on each of the sides - front and rear) within the advanced BGA packaging, with the 256-bit memory bus. Note that the number of rendering pipelines is 4, not 8 (the other 4 are disabled programmatically). The memory produced by Hynix (HY5DU283222) offers a 2.8 ns access time, which is equivalent to approximately 355 MHz of memory operation (710 MHz), but the memory runs at its intended frequency 340 MHz (680 MHz). That is, there is a small overclocking margin for the memory. The graphics chip also runs at 380 MHz as per the specifications.

There is absolutely no cooling for the memory chips. To cool the graphics processor, a low-profile cooling system is used which hardly can be regarded as effective enough. A standard reference small fan is fitted on the radiators. Nevertheless, in the nominal mode during the long 3D testing session there were no stability problems found. At the same time, the radiators were heated up quite immensely.

On the whole, the layout and design are good in that they replicate the design of ATI Radeon 9700 Pro, not ATI Radeon 9500 Pro (these boards, as we already mentioned, will soon prevail on the market), so the owners of ATI Radeon 9800SE on the base of the PCB of Radeon 9700 Pro can safely get round to enabling the 4 programmatically locked rendering pipelines. Nothing prevents the owners of Radeon 9800SE on the base of the PCB of ATI Radeon 9500Pro from doing the same, but because of the rigid restriction of the memory bus width (128 bit) and lower frequencies the performance will be SUBSTANTIALLY lower than that in the unlocked Radeon 9800SE on the base of the PCB of ATI Radeon 9700 Pro that came to our test lab.

Sapphire Atlantis Radeon 9800SE (128-bit)

Retail bundle

We found the following items in the package box:

• User's manual;
• A drivers CD;
• A CD with the proprietary RedLine utility, PowerDVD;
• Game Tomb Raider: Angel of Darkness;
• TV extension cords;
• DVI-to-D-Sub adapters;
• SVideo-to-RCA;
• A splitter adapter for external power supply.

Quite a rich package bundle. Once again, Sapphire makes the users happy.

Design and layout:

The board offers ATI's traditional bright red color of the PCB, has 128 MB DDR memory onboard, the AGP 2x/4x/8x interface and a standard set of outputs: one analogous, one digital, and one TV-OUT. The board does not provide any third-party codecs to implement the TV-OUT - the features to support it are already integrated in the Radeon 9800SE chip.

The video card is equipped with 128 MB DDR memory packaged in 8 chips (4 chips on each of the sides - front and rear) within the advanced BGA packaging, with the 128-bit memory bus. But the chips are positioned in a way different from that in Radeon 9800Se. Note that the number of rendering pipelines is 4, not 8 (the other 4 are disabled programmatically). The memory produced by Infinion offers a 3.3 ns access time, which is equivalent to approximately 355 MHz of memory operation (710 MHz), but the memory runs at its intended frequency 340 MHz (680 MHz). That is, there is a small overclocking margin for the memory. The graphics chip also runs at 380 MHz as per the specifications.

There is absolutely no cooling for the memory chips. To cool the graphics processor, a low-profile cooling system is used which hardly can be regarded as effective enough. A standard reference small fan is fitted on the radiators. Nevertheless, in the nominal mode during the long 3D testing session there were no stability problems found. At the same time, the radiators were heated up quite immensely.

Benchmarking

Test configuration:

We remove all the decorative "niceties" out of the Windows GUI and set the operating system to maximum performance.

Disable the Vsync forcedly via the drivers both in OpenGL and in Direct3D applications. The S3TC texture compression was also disabled.

Test software:

• 3DMark2003;
• 3DMark2001SE;
• Codecreatures v1.0.0 (a DirectX 8.1 application, shaders on, Hardware T&L);
• Return to Castle Wolfenstein v1.0 (OpenGL, multitexturing. The image quality was set to the maximum. Demo Checkpoint was used);
• Unreal Tournament 2003 (Direct3D, Hardware T&L, vertex shaders, Dot3, cube texturing. "Antalus Flyby" demo);
• Gun Metal Benchmark 2 v1.20s (a DirectX 9.0 benchmark, Vertex Shaders 2.0, Pixel Shaders 1.1, Hardware T&L);
• X2: The Threat Demo (Direct3D, multitexturing, Dot3, running in the benchmark mode embedded in the demo version);
• Final Fantasy XI Official Benchmark 2 (a benchmark for assessing the performance in the future game Final Fantasy XI. Unfortunately, the developers haven't presented any data on the gaming engine);
• HALO: Combat Evolved 1.2 (DirectX 9.0, Vertex Shaders 1.1/1.4/2.0, Pixel Shaders 1.1/1.4/2.0, Hardware T&L, quality set to the maximum possible);
• AquaMark 3 (DirectX 9.0, Vertex Shaders 1.1/1.4/2.0, Pixel Shaders 1.1/1.4/2.0, Hardware T&L, AquaMark3 Triscore mode);
• Half-life 2 leaked beta (DirectX 9.0, Vertex Shaders 2.0, Pixel Shaders 2.0, default quality);
• Unreal II: The Awakening (Direct3D, vertex shaders, Hardware T&L, Dot3, cube texturing, quality offered by the BenchemAll suite).

Benchmarking Results

The tests with the PowerColor Radeon 9800SE cards were run in two modes. In the nominal mode and with the 4 additional pipelines unlocked (more precisely, the legitimate ATI Radeon 9800 Pro pipelines). A few boards by Sapphire made on the PCB of ATI Radeon 9500 Pro proved amenable for the re-make, but failed to pass the tests. That is, we get a typical example of that "not all ATI Radeon 9800SE boards are equally useful =)". For comparison, we present the results for an NVIDIA GeForce FX 5900 overclocked to speeds of the Ultra version of the chip so as to get a visual idea of the relation of results for the patched PowerColor Radeon 9800SE versus the NVIDIA counterpart. In none of the cases we will get boards running as fast as the original ATI Radeon 9800 Pro and NVIDIA GeForce FX 5900. The comparison is more interesting in that Hi-end boards albeit bought are purchased as a basis for further modernization and re-make, so the users would be much interested in comparison of the results for boards supplied by both 3D flagships to date which somehow are amenable to re-making/overclocking.

Synthetic benchmarks:

In this review, we are not bringing in the results of tests for the already traditional synthetic benchmarks like ShaderMark v2.0 and D3D RightMark, since the review of these boards is more consumer-oriented and is unlikely to add more clarity if loaded with abundance of useless synthetic benchmarking diagrams versus forecasted results.

3DMark 2001SE

Let's look at the results produced by the "old-timers" of our benchmarking sessions. They are quite predictable. As was expected, the Sapphire Radeon 9800SE with 128-bit memory bus shows the worst results of all the tested boards. The 128-bit memory bus makes itself felt. The NVIDIA counterpart in the person of ASUS V9950 overclocked to 450/850 MHz for the core and memory respectively takes a middle position between the PowerColor's ATI Radeon 9800SE (256-bit) in the nominal mode and with 4 additional rendering pipelines enabled.

3DMark 2003

As you can see from the diagrams, the performance difference between the original ATI's Radeon 9800SE 256-bit and its "patched version" is quite substantial (in the previous test you couldn't see that drastical difference because of the greater processor-dependence of the test as compared to 3DMark 2003). But the 128- and 256-bit ATI Radeon 9800SE versions are priced very high.

Codecreatures

The alignment of forces and percentages of absolute values in this test remain the same with the only difference that the NVIDIA board shows a more distinct leadership.

Real gaming applications

From synthetic applications, we are now moving on to analyzing the performance of the graphic boards in real gaming applications.

Unreal Tournament 2003

Unreal Tournament 2003 traditionally "fond of" NVIDIA cards shows a clear superiority of the representative of the Californian company at this test. In all the other respects: again the same poor results of ATI Radeon 9800SE 128-bit and the same outstanding results for ATI Radeon 9800SE with all the rendering pipelines enabled.

X2: The Threat Demo

The leadership of similarly positioned NVIDIA products in this test is traditional. The processing of stencil shades in NVIDIA cards in the X2: The Threat engine requires less passes. The alignment of forces in the camp of the two versions of ATI Radeon 9800SE is the same.

Unreal II: The Awakening

At low resolution, the boards by NVIDIA and ATI (PowerColor ATI Radeon 9800SE 256-bit with 8 rendering pipelines enabled) demonstrate a complete parity. At high resolution, ATI board takes a lead. Again the situation with results for Unreal 2 tests relative to the results for Unreal Tournament 2003 is still unclear. The results are different, although the engines of the games are the same (the Unreal 2 engine is merely a "cosmetically remade" Unreal Tournament 2003 engine). In our previous materials, we noted such strange results for these benchmarks. And this time as well we've got to point to the strange results and remind the readers that there is an embedded benchmark in Unreal Tournament 2003 and is not in Unreal 2, which suggests certain ideas =).

Final Fantasy XI Official Benchmark 2

The alignment of forces is generally the same. There is a slight leadership of ATI board with the pipelines enabled.

AquaMark 3

This is a really indicative test to date. NVIDIA's leadership at that is evident.

Traditionally, we take detailed readings from the AquaMark 3 test.

ASUS FX 5900 o/c 450/850	Sapphire Radeon 9800SE 128-bit 4õ1
PowerColor Radeon 9800SE 256-bit 4õ1	PowerColor Radeon 9800SE 256-bit 8õ1

Gun Metal Benchmark 2

At that quite demanding graphics test where pixel shaders 1.1 and vertex shaders 2.0 are used, the cut-down functionality of Sapphire Radeon 9800SE 128-bit makes itself felt through the drastical percentage difference relative to PowerColor ATI Radeon 9800SE 256-bit. The re-make of PowerColor ATI Radeon 9800SE 256-bit gives the PowerColor board substantial advantages.

HALO: Combat Evolved

With the release of new ForceWare 52.16 drivers, the alignment of forces at this real gaming DirectX 9.0 benchmark has turned to the better for NVIDIA. As we see, at both the benchmarks the ASUS V9950 @ 450/850 takes a sure lead. Also note the unprecedented effect produced from enabling 4 additional rendering pipelines.

Half-life 2 Leaked Beta

At Half-Life 2, the situation for NVIDIA is not straightforward at all. On the one hand, there is a certain loss at one of the tests (the results are on the level of Radeon 9800SE 128-bit!). On the other hand, the picture is much better in the other of benchmarks. However, ATI Radeon 9800 Pro takes a lead in the long run at both first and second benchmarks.

Image quality - AntiAliasing 4x

Image quality – Anisotropic Filtering 8x

Image quality – Anisotropic Filtering 4x + Anisotropic Filtering 8x

We have already reviewed the performance for AA and AF. Again we note the faster AA in similarly positioned NVIDIA and ATI video cards as well as approximate parity both in the AF mode and in the most demanding mode with Aniso and AA simultaneously enabled.

Image Quality – AntiAliasing 6x/8x + Anisotropic Filtering 8x/16x

In our new mode of testing the image quality with the maximum possible details, we note a leadership of PowerColor Radeon 9800SE 256-bit with 8 rendering pipelines enabled.

Final Words

The considered two varieties of boards on the base of ATI Radeon 9800SE vividly demonstrated that not all SE cards are equally useful. The Sapphire card based on the ATI Radeon 9800SE chip using the PCB design of ATI Radeon 9500 Pro at all of the tests shows an essentially lower performance than in the 256-bit version of ATI Radeon 9800SE by PowerColor. The purchase of a card based on the PCB of ATI Radeon 9500 Pro can be recommended only in case the price is substantially lower than that for the "right" ATI Radeon 9800SE based on the PCB of ATI Radeon 9700 Pro. Fortunately, vendors of Radeon 9800SE cards begin to specify what exactly a product is there at the warehouse and are not trying to sell "a pig in a poke". .

3DNews Forum:

• Discussing ATI Radeon 9800SE

Read more on this topic: