Contents

1. Introduction
2. Video cards' features
3. Testbed configurations, benchmarks
4. Test results: Serious Sam: The Second Encounter
5. Test results: Code Creatures DEMO
6. Test results: Unreal Tournament 2003
7. Test results: Unreal II: The Awakening
8. Test results: RightMark 3D
9. Test results: TRAOD
10. Test results: FarCry
11. Test results: Call Of Duty
12. Test results: HALO: Combat Evolved
13. Test results: Half-Life2(beta)
14. Test results: Splinter Cell
15. Test results: Quake3 Arena
16. Test results: Return to Castle Wolfenstein
17. Test results: AquaMark3
18. Test results: DOOM3 (Alpha)
19. Test results: 3DMark03 (marks)
20. Conclusions

We proceed with our tests of video cards based on NVIDIA GeForce 6800, which are gradually appearing on the market. It's a Low End model, which is intended to replace such previous High End cards as RADEON 9800 XT and GeForce FX 5950 Ultra. ATI does not yet have anything to offer in this segment, which could extrude RADEON 9800 XT from the market (or could at least dramatically drop its prices, which is however hardly possible because these cards are long out of production and stores offer the cards that were purchased by the dealers long ago at high prices).

But NVIDIA already has such products on sale at prices even lower than that of GeForce FX 5950 Ultra or RADEON 9800 XT, which is very pleasing.

So, what is 6800? It's a 12-pipeline chip dressed into narrow CPU bus bandwidth. Indeed, you can see that the same memory operating not at 350 (700) MHz in GeForce FX5900 XT is installed on GeForce 6800. But for all that FX5900 XT has only 4 pipelines and 8 texture units (compare with 12 pipelines and 12 texture units). The NV40 core itself is not different from GeForce 6800 Ultra or GT in its features, it only has fewer pipelines.

Theoretical materials and reviews of video cards, which concern functional properties of the GPU ATI RADEON X800 (R420) and NVIDIA GeForce 6800 (NV40)

• NVIDIA GeForce 6800 Ultra (NV40). Part 1: Architecture Features and Synthetic Tests in D3D RightMark
• ATI RADEON X800 XT and X800
• A Battle of ATI RADEON X800 XT and NVIDIA GeForce 6800 Ultra - Part Two: 450 MHz NV40 and Tests of Both Cards
• Battle of ATI RADEON X800 XT and NVIDIA GeForce 6800 Ultra - Part Three: Trilinear filtering (synthetic examples)
• Battle of ATI RADEON X800 XT and NVIDIA GeForce 6800 Ultra - Part Four: RightMark3D-based Filtering Tests Sapphire RADEON X800 Pro Review
• Battle of ATI RADEON X800 XT and NVIDIA GeForce 6800 Ultra - Part Five: Game-based Filtering Tests
• PowerColor RADEON X800 PRO Limited Edition, X800 PRO Made Into X800 XT Platinum Edition
• NVIDIA GeForce 6800 GT Review
• PowerColor RADEON X800 PRO Limited Edition, hardware reconstruction of X800 PRO into X800 XT Platinum Edition
• NVIDIA GeForce 6800 GT
• Leadtek WinFast A400 TDH, Leadtek WinFast A400 Ultra TDH based on NVIDIA GeForce 6800/6800 Ultra
• GeCube RADEON 9550 XT Extreme, GeCube RADEON X800 PRO
• Brief account of FarCry v.1.2 tests and of the first Shader 3.0 implementation
• FarCry v.1.2 Tests Continued and Shader 2.b

So in this material it'll be interesting to learn how strong the influence of CPU bus bandwidth is on 6800 (in my tests I overclocked the memory on the card from 700 to 840 MHz, and I didn't touch the core). And due to the fact that the core of the Galaxy card operates at 350 MHz instead of 325 MHz, we shall be able to learn how strong the influence of GPU frequency raise is with the same CPU bus bandwidth (it's really interesting how effective the memory resource saving technologies are with this relatively low bandwidth).

Yes, I have touched above the topic about the card from Galaxy. This company is almost unknown in our country, though its products can be found on the shelves, but very often as noname. Though you can see that it has even High-End products, with original designs at that. The powerful cooling system allowed the specialists from this company to increase the core frequency by 25 MHz to make their product deliberately a tad faster than the other cards. Our tests will demonstrate exactly how faster it is.

I think that there is no need in telling you about the MSI company, it's the oldest brand in our market, its video cards have been known since 1998.

MSI products stand out by the fact that MSI plant produces not only copies of reference cards but cards of own design, especially it concerns cooling systems. You shouldn't also forget that MSI offers almost the best bundle for its video cards we have ever seen.

So, in our laboratory we have tested two video cards based on GeForce 6800, which, according to our research, is the best buy for those who have 350-370 USD for a powerful state-of-the-art accelerator.

Video Cards

Galaxy Glacier GeForce 6800
MSI NX6800-TD128

Galaxy Glacier GeForce 6800
The cards have an AGP x8/x4 interface, 128 MB DDR SDRAM allocated in 8 chips on the front PCB side. Hynix memory chips. 2.2ns memory access time, which corresponds to 450 (900) MHz. The memory operates at 350 (700) MHz. GPU frequency in MSI NX6800-TD128 is 325 MHz, and in Galaxy Glacier GeForce 6800 the GPU frequency is 350 MHz. 256bit memory bus. 12 (3x4) pixel pipelines.
MSI NX6800-TD128
The cards have an AGP x8/x4 interface, 128 MB DDR SDRAM allocated in 8 chips on the front PCB side. Hynix memory chips. 2.8ns memory access time, which corresponds to 350 (700) MHz, at which the memory operates. GPU frequency in MSI NX6800-TD128 is 325 MHz, and in Galaxy Glacier GeForce 6800 the GPU frequency is 350 MHz. 256bit memory bus. 12 (3x4) pixel pipelines. Later we found out that this video card actually has 16 pipelines (4x4) instead of 12. We don't know what was the reason for such chip in GeForce 6800 and whether all such cards will be equipped with a 16-pipeline GPU.

Comparison with the reference design, front view
Galaxy Glacier GeForce 6800	Reference card NVIDIA GeForce 6800
MSI NX6800-TD128

Comparison with the reference design, back view
Galaxy Glacier GeForce 6800	Reference card NVIDIA GeForce 6800
MSI NX6800-TD128

Both cards are obviously of the same design based on the reference one. In fact the PCBs differ only in lacquer colors. The layout of memory chips is similar to GeForce FX 5900, actually they have a similar design. The only external power connector is located in the upper part of the card. A 300Wt power supply unit will be sufficient for the card's operation in an average system block. I want to note that the memory in the video card from Galaxy offers great overclocking potential and thus overclockers are sure to like this card.

But the cooling systems in these cards are different.

MSI NX6800-TD128
Look at the very effective system from ARX (well-known cooler manufacturer) based on a narrow copper heatsink sprawled on the PCB. The heatsink is narrow enough for the cooler not to cover the slot next to AGP. A powerful fan is on the left, it can operate at 5000 rpm, but this revolution frequency is either enabled by the driver in case of graphic processor overheating, or it is specified by a slider on top of the heatsink. The cooler is similar to those we have previously seen in ATI RADEON 9800 XT (X800) - see the bottom photo on the right. At least it has the same operating principle.
Galaxy Glacier GeForce 6800
The video card from Galaxy possesses a cooling system from Arctic Cooling. We have already come across the same system in HIS video cards based on RADEON X800. Galaxy Glacier GeForce 6800 just has a different plastic color (blue instead of colorless/transparent). It's a very effective cooling system, which, despite its high rpms, has a silent fan-turbine blowing out the air from the system block, unlike the previous one. It goes without saying that the card blocks the slot next to AGP because of its huge size.

Let's return to the cooler installed on the card from MSI. Looking at this strikingly effective cooling system, which occupies only one slot, we decided to check (1) what will happen to the temperature condition if we set the fan rpm slider to minimum; (2) how this cooler will operate in GeForce 6800 Ultra, considering that the PCB mounting holes are identical.

The device is firmly attached to the card with 8 screws (4 screws at chip corners, and 4 screws between memory chips).

Thus, with maximum fan rpms in GeForce 6800 we got the following temperature curve (we used RivaTuner to take temperature readings (written by Alexei Nikolaichuk AKA Unwinder)):

We can see that the maximum temperature (in 3D, FarCry) was 62 degrees centigrade at GPU. This is very good and there is much in reserve. And now let's set the slider to minimum:

The temperature rose to 68 degrees centigrade. This also has much in reserve so we can fearlessly recommend the minimum position of the slider, in this case you will not hear the fan at all.

Now let's install this cooling system to 6800 Ultra (slider to High, that is to maximum rpm):

You can see the temperature quickly rise to critical 85 degrees, and... the card froze (you can see the break in readings).

Alas, the design of this cooler does not really fit 6800 Ultra, it's obviously not sufficient. Perhaps we should have increased rpm, but this cooler can hardly handle it. Besides, for a sufficient cooling effect 6800 Ultra needs separate heatsinks for the memory and the chip. Why? - It's very simple: a slightest misfit of the heatsink for the chip heights will result in a microgap between the chip and the heatsink. In 6800 this gap can be compensated by the thermopaste viscosity, but in 6800 Ultra it cannot be compensated.

Below is a temperature curve for 6800 Ultra with an original cooler:

Yes, the temperature here does not rise above 72 degrees. It should be specially noted that all the experiments were conducted inside the case with disabled case fans.

You can see the chip under the coolers. It has an FCPGA package, the crystal size is impressive (we have already talked about it before). It's clear that it physically contains 16 (not 12) pipelines, 4 of them are disabled (to think logically the crystals are sorted by their nonoperability; but they may be sorted that way by marketing reasons, we are still looking for an option to enable these pipelines in a smooth software way)

Let's examine the package contents.

MSI NX6800-TD128
The box contains a user's guide, a lot of CDs (14 items!) with various software (see the photo on the right), TV extension cords, DVI-to-d-Sub, SVideo-to-RCA adapters, external power supply splitter. The bundle is fit for a king!
Galaxy Glacier GeForce 6800
The box contains a user's guide, CD with drivers, 3 CDs with games (Chaser, MotoGP, game demos), PowerDVD, TV extension cords, DVI-to-d-Sub, SVideo-to-RCA adapters, power supply cable. Note that the bundle contains two modern games: Prince of Persia and SCPT (the latter is on DVD!), which is a good thing.

And now look at the boxes:

MSI NX6800-TD128
A huge box with a carrier handle. It has a flap lid, which covers a window demonstrating the card. The box is colorful and attractive.
Galaxy Glacier GeForce 6800
The box is made of millboard covered with holographic gloss.

Now about overclocking the video card from MSI. The tests conducted are quite enough, the memory was overclocked to 840 MHz. I didn't specially overclock the core (in fact, I'm not up to overclocking, this data is provided only as a reference, because you cannot draw conclusions from the overclocking potential of a SPECIFIC card sample).

Concerning the video card from Galaxy, its memory managed to operate even at 500 (1000) MHz! This is possible obviously due to 2.2ns memory chips. So I repeat, this video card is a real boon for overclockers.

Installation and Drivers

Testbed configurations:

• Athlon 64 3200+ based computer
  • AMD Athlon 64 3200+ (L2=1024K) CPU
  • ASUS K8V SE Deluxe mainboard on VIA K8T800
  • 1 GB DDR SDRAM PC3200
  • Seagate Barracuda 7200.7 80GB SATA HDD

• Operating system: Windows XP SP1; DirectX 9.0b
• Monitors: ViewSonic P810 (21") and Mitsubishi Diamond Pro 2070sb (21")
• ATI drivers v6.458 (CATALYST 4.7); NVIDIA v61.76

VSync is disabled.

Test results: performance comparison

We used the following test applications:

• Serious Sam: The Second Encounter v.1.05 (Croteam/GodGames) - OpenGL, multitexturing, ixbt0703-demo, test settings: quality, S3TC OFF
  
  
• Unreal Tournament 2003 v.2225 (Digital Extreme/Epic Games) - Direct3D, Vertex Shaders, Hardware T&L, Dot3, cube texturing, default quality
  
  
• Code Creatures Benchmark Pro (CodeCult) - game test demonstrating how the card operates with DirectX 8.1, Shaders, HW T&L.
  
  
• Unreal II: The Awakening (Legend Ent./Epic Games) - Direct3D, Vertex Shaders, Hardware T&L, Dot3, cube texturing, default quality
  
  
• RightMark 3D v.0.4 (one of the game scenes) - DirectX 8.1, Dot3, cube texturing, shadow buffers, vertex and pixel shaders (1.1, 1.4).
  
  
• Tomb Raider: Angel of Darkness v.49 (Core Design/Eldos Software) - DirectX 9.0, Paris5_4 demo. Maximum quality, we only disabled Depth of Fields PS20.
  
  
• HALO: Combat Evolved (Microsoft) - Direct3D, Vertex/Pixel Shaders 1.1/2.0, Hardware T&L, maximum quality
  
  
• Half-Life2 (Valve/Sierra) - DirectX 9.0, demo (ixbt07. Tested with anisotropic filtering disabled, and also in a heavy mode with AA and anisotropy.
  
  
• Tom Clancy's Splinter Cell v.1.2b (UbiSoft) - Direct3D, Vertex/Pixel Shaders 1.1/2.0, Hardware T&L, maximum quality (Very High); demo 1_1_2_Tbilisi
  
  
• Call of Duty (MultiPlayer) (Infinity Ward/Activision) - OpenGL, multitexturing, ixbt0104demo, the test settings - Very High, S3TC ON
  
  
• FarCry 1.1 (Crytek/UbiSoft), DirectX 9.0, multitexturing, demo01 (research) (the game is started with the key -DEVMODE), the test settings - Very High.
  
  
• 3DMark03 v.3.40 (FutureMark) - DirectX 9.0, multitexturing, maximum possible test settings.
  
  
• AquaMark3 (Massive Development) - DirectX 9.0, multitexturing, maximum possible test settings.
  
  
• Quake3 Arena v.1.17 (id Software/Activision) - OpenGL, multitexturing, ixbt0703-demo, maximum possible test settings: detailing - High, texture details - T70;4, S3TC OFFcurved surfaces are smoothed by the variables r_subdivisions "1" and r_lodCurveError "30000" (note that the default value of r_lodCurveError is "250"!)
  
  
• Return to Castle Wolfenstein (MultiPlayer) (id Software/Activision) - OpenGL, multitexturing, ixbt0703-demo, maximum possible test settings, S3TC OFF
  
  
• DOOM III (alpha version) (id Software/Activision) - OpenGL, multitexturing, maximum possible test settings, S3TC OFF
  
  

If you want to get the demo-benchmarks, which we use, contact me at my e-mail.

ANIS mode on the diagrams means Anisotropic 16x Quality (FX5950 - 8x Quality), AA - Anti-Aliasing 4x Quality

I want to note that the diagrams below display the data obtained with "regular" GeForce 6800 (12 pipelines).

Serious Sam: The Second Encounter

Thus, on the whole:

• NVIDIA GeForce 6800 versus ATI RADEON 9800 XT - up to 66% advantage!
• NVIDIA GeForce 6800 versus NVIDIA GeForce FX 5950 Ultra - up to 51% advantage
• NVIDIA GeForce 6800 with memory overclocked to 840 MHz - the effect is insignificant, up to 3.7%, which is within the allowed precision of measurements. Interestingly, this game requires much RAM. Yet HSR operates well in 6800.
• NVIDIA GeForce 6800 with the core overclocked to 350 MHz - there is almost no gain even at heavy load. Comparing this with the previous effect, there is some contradiction. Perhaps even with a heavy load it all depends on system resources and the effect of the video card performance gain is minimal.

Code Creatures

Thus, on the whole:

• NVIDIA GeForce 6800 versus ATI RADEON 9800 XT - 34% advantage
• NVIDIA GeForce 6800 versus NVIDIA GeForce FX 5950 Ultra - much the same, the advantage is up to 44%
• NVIDIA GeForce 6800 with its memory overclocked to 840 MHz - similar to the previous test, CPU bus bandwidth expansion gave nothing, though in this case a larger part of load is on the core.
• NVIDIA GeForce 6800 with its core overclocked to 350 MHz - confirms the above, the effect is almost twice as great, though percentage wise the core was overclocked less than the memory.

Unreal Tournament 2003

The easiest modes without AA and anisotropy: on the whole 6800 is the leader, if not to take into account the parity in 1024x768. But here the game is up to system resources, everything depends on software, not hardware.

The final heaviest mode with AA and anisotropy: very much the same. The only discord is a performance drop of 6800 in 1600x1200.

Thus, on the whole:

• NVIDIA GeForce 6800 versus ATI RADEON 9800 XT - overall victory, but in heavy modes there are parity situations
• NVIDIA GeForce 6800 versus NVIDIA GeForce FX 5950 Ultra - earnest victory with the score up to 20%
• NVIDIA GeForce 6800 with its memory overclocked to 840 MHz - there is still no effect (HSR "rules")
• NVIDIA GeForce 6800 with its core overclocked to 350 MHz - the gain is more considerable, but still insignificant in light modes, and much more considerable with AA and AF (which again testifies that the game depends on CPU etc).

Unreal II: The Awakening

Thus, on the whole:

• NVIDIA GeForce 6800 versus ATI RADEON 9800 XT - victory, 6800 outscores with 12-40% depending on the resolution
• NVIDIA GeForce 6800 versus NVIDIA GeForce FX 5950 Ultra - the same results, but the advantage reaches 54%
• NVIDIA GeForce 6800 with its memory overclocked to 840 MHz - no effect
• NVIDIA GeForce 6800 with its core overclocked to 350 MHz - minimal effect.

RightMark 3D

Thus, on the whole:

• NVIDIA GeForce 6800 versus ATI RADEON 9800 XT - 6800 is the champion, its advantage increasing simultaneously with the load
• NVIDIA GeForce 6800 versus NVIDIA GeForce FX 5950 Ultra - also victory, the advantage is up to 27%
• NVIDIA GeForce 6800 with its memory overclocked to 840 MHz - null effect (it's quite clear, the core is loaded)
• NVIDIA GeForce 6800 with its core overclocked to 350 MHz - there is little gain, which is quite logical.

TR:AoD, Paris5_4 DEMO

Thus, on the whole:

• NVIDIA GeForce 6800 versus ATI RADEON 9800 XT - not a bad gain of 12-23%
• NVIDIA GeForce 6800 versus NVIDIA GeForce FX 5950 Ultra - phenomenal performance gain. It's all up to shaders. And we all know how slow they were in the GeForce FX series.
• NVIDIA GeForce 6800 with its memory overclocked to 840 MHz - minimal effect
• NVIDIA GeForce 6800 with its core overclocked to 350 MHz - the gain is a little more considerable, but still the low CPU bus bandwidth is the bottleneck.

FarCry, demo01

The easiest modes without AA and anisotropy: small gain.

The heaviest final mode with AA and anisotropy: these modes result in performance gain leap.

Thus, on the whole:

• NVIDIA GeForce 6800 versus ATI RADEON 9800 XT - from almost parity without AA+AF (it was all up to the system resources) to 41% performance gain in 6800!
• NVIDIA GeForce 6800 versus NVIDIA GeForce FX 5950 Ultra - a holiday for 6800: huge advantage!
• NVIDIA GeForce 6800 with its memory overclocked to 840 MHz - almost no effect
• NVIDIA GeForce 6800 with its core overclocked to 350 MHz - the game loads the core heavily, that's why its core overclocking results in more gain.

Call of Duty, ixbt04

Thus, on the whole:

• NVIDIA GeForce 6800 versus ATI RADEON 9800 XT - victory! With the score of 50%!
• NVIDIA GeForce 6800 versus NVIDIA GeForce FX 5950 Ultra - the same results, but the advantage already reaches 60% (just imagine the game running 60% faster!)
• NVIDIA GeForce 6800 with its memory overclocked to 840 MHz - almost no effect
• NVIDIA GeForce 6800 with its core overclocked to 350 MHz - the same results.

HALO: Combat Evolved

Thus, on the whole:

• NVIDIA GeForce 6800 versus ATI RADEON 9800 XT - 20-36% advantage
• NVIDIA GeForce 6800 versus NVIDIA GeForce FX 5950 Ultra - the same results, but the figures reach 30-40%
• NVIDIA GeForce 6800 with its memory overclocked to 840 MHz - no effect
• NVIDIA GeForce 6800 with its core overclocked to 350 MHz - there is little gain, the game mostly depends on the core potential.

Half-Life2 (beta): ixbt07 demo, ANISO 16xQ

Thus, on the whole:

• NVIDIA GeForce 6800 versus ATI RADEON 9800 XT - it's interesting to see the reverse picture. No, 6800 has not lost! But as a rule the gain is higher with AA+AF in comparison with easy modes, but here everything is vice versa. That's very strange, it's most likely the drivers' fault, they must have overloaded the card in heavy modes.
• NVIDIA GeForce 6800 versus NVIDIA GeForce FX 5950 Ultra - almost the same picture.
• NVIDIA GeForce 6800 with its memory overclocked to 840 MHz - nothing new
• NVIDIA GeForce 6800 with its core overclocked to 350 MHz - there is some gain, which is characteristic of this game.

Splinter Cell

Thus, on the whole:

• NVIDIA GeForce 6800 versus ATI RADEON 9800 XT - huge gain!
• NVIDIA GeForce 6800 versus NVIDIA GeForce FX 5950 Ultra - the huge becomes giant!
• NVIDIA GeForce 6800 with its memory overclocked to 840 MHz - nothing!
• NVIDIA GeForce 6800 with its core overclocked to 350 MHz - much the same, it's up to the system resources.

Quake3 Arena

Thus, on the whole:

• NVIDIA GeForce 6800 versus ATI RADEON 9800 XT - some quivering of performance gain depending on the resolution (actually this game has always depended on system resources in powerful video cards)
• NVIDIA GeForce 6800 versus NVIDIA GeForce FX 5950 Ultra - victory, but somehow very strange, the driver is fooling around again.
• NVIDIA GeForce 6800 with its memory overclocked to 840 MHz - zero
• NVIDIA GeForce 6800 with its core overclocked to 350 MHz - almost nothing.

Return to Castle Wolfenstein (Multiplayer)

Thus, on the whole:

• NVIDIA GeForce 6800 versus ATI RADEON 9800 XT - victory with 17-32%
• NVIDIA GeForce 6800 versus NVIDIA GeForce FX 5950 Ultra - the same results
• NVIDIA GeForce 6800 with its memory overclocked to 840 MHz - a little gain in 1600x1200
• NVIDIA GeForce 6800 with its core overclocked to 350 MHz - the same results

AquaMark3

Thus, on the whole:

• NVIDIA GeForce 6800 versus ATI RADEON 9800 XT - confident victory with 18-41%
• NVIDIA GeForce 6800 versus NVIDIA GeForce FX 5950 Ultra - the same results, the advantage is over 65%
• NVIDIA GeForce 6800 with its memory overclocked to 840 MHz - we can see that for the first time the CPU bus bandwidth expansion produced positive results, 14% gain (obviously in the highest resolution). I would have never thought that this test (strongly overloading the core) would depend on the memory clock.
• NVIDIA GeForce 6800 with its core overclocked to 350 MHz - the effect is even lower than the previous one.

DOOM III (alpha)

Unfortunately, this test couldn't be run on FX 5950 (driver), so we shall compare only with 9800 XT.

Thus, on the whole:

• NVIDIA GeForce 6800 versus ATI RADEON 9800 XT - excellent results, 88% gain!
• NVIDIA GeForce 6800 with its memory overclocked to 840 MHz - almost no effect
• NVIDIA GeForce 6800 with its core overclocked to 350 MHz - little gain

3DMark03 - MARKS

Thus, on the whole:

• NVIDIA GeForce 6800 versus ATI RADEON 9800 XT - the usual victory with 50%
• NVIDIA GeForce 6800 versus NVIDIA GeForce FX 5950 Ultra - the same results
• NVIDIA GeForce 6800 with its memory overclocked to 840 MHz - almost no effect
• NVIDIA GeForce 6800 with its core overclocked to 350 MHz - there is some gain, but...

Conclusions

First of all I want to note that strangely enough with its 12 pipelines and the core clock at 325 MHz (3900 billion texels/sec) GeForce 6800 is quite a balanced video card with CPU bus bandwidth implemented via memory at 350 (700) MHz. Considering that the fill rate of RADEON 9800 XT is 3300 billion texels/sec, plus its combination with 730 MHz memory clock and decent performance gain over 9800 XT (up to 50-60%), the effectiveness of GeForce 6800 speaks for itself. Every single MHz of this chip is much more effective despite the practically identical CPU bus bandwidth in 9800 XT.

Our tests proved that the card does not really need the raise in memory clock, which allows NVIDIA and its partners to considerably reduce the net cost of these cards in comparison with GeForce 6800 GT (high-performance GDDR3 is very expensive now and constitutes a great part of the net cost of video cards).

Now about the cards. Galaxy Glacier GeForce 6800 is the first.

As you can see from the test results, a frequency raise by 25 MHz produces little effect, but it's still pleasant and you can get moral satisfaction from it :). But I repeat that this card is equipped with 2.2ns memory instead of the standard (in such cases) 2.8ns memory, that's why it can be considerably overclocked. However, as we have previously seen, this will not have a brilliant effect.

The cards look wonderful as well as their operation. Besides, the Arctic Cooling cooler is almost noiseless, and overheating is not an option with 6800. What concerns the 2D quality, our sample was quite all right, it was comfortable to work in 1600x1200 at 85Hz.

MSI NX6800-TD128.

This video card has a nice design, the cooler perfectly matches the red PCB. It can also be made almost noiseless, our research proved that the heat dissipation will not deteriorate.

Astounding package contents makes this video card a cherished purchase for those who have 350-370 USD for a state-of-the-art 3D-accelerator (the retail price for early august in Moscow).

I note once again that this sample is equipped with 16 pipelines instead of 12.

Both cards are stable, there are no reprimands, the quality is good.

In our 3DiGest you can find more detailed comparisons of various video cards.

According to the test results MSI NX6800-TD128 merits the Excellent Package award (July).

Write a comment below. No registration needed!