Asus N7800GT Dual, power consumption, tests

Asus Extreme N7800GT Dual

Asus likes to see who is the strongest. After this manufacturers dual GeForce 6, they have now released a card equipped with two 7800 GTs. We feel that the production of these cards is better controlled, especially in its finishing touches and size. The size is much more reasonable even if it’s much bigger than a card equipped with a single GPU (3 cm higher and 1.5 cm longer than a GeForce 7800 GTX). It can fit into many towers if the hard drive isn´t in the way.


It comes with a power adapter, which looks like the Voodoo 5 6000´s, whose use is obligatory only if your power supply isn´t capable of powering the card by itself. This was the case of the Enermax 535W that we used for these tests.

As this article concerns very high-end graphic cards, Asus´ collosus definitively belongs here, because even if it uses SLI there is only a single card. The only thing is that you will have to spend approximately 1000€ / $1000 to obtain one.


The cooling system is relatively discreet despite the fact that it cools down two G70s, which even if they are GT models have 20 pixel shading pipelines and are clocked at the same frequency as the GTX (430 instead of 400 MHz). This is the same for the memory, which is clocked at 600 MHz instead of 500.

Radeon X1800 and GeForce 7800 GTX

We compared these 2 cards to the latest high end cards tested, the Radeon X1800 XT, Radeon X1800 XL, GeForce 7800 GTX and GeForce 7800 GT. As the 7800 GTX 512 MB chip is a G70 (identical to the 7800 GTX) we won´t linger on its architecture. If you want more information, take a look at GeForce 7800 GTX and Radeon X1800 tests.


We remind you that ATI benefits from more advanced architecture. Amongst other things there is its redesigned memory controller to improve the effective use of available memory bandwidth, and a more subtle branching management, which is more efficient than NVIDIA’s. The consolation of the GeForce 7800 is its higher calculation power. We won’t give the specifics on this subject, because of the major differences in architecture of these two companies and there being no identical way to calculate calculation power. We’ll just say that overall NVIDIA is in the lead.

We also included a Radeon X800 XT and GeForce 6800 GS to this test. The GS was added as it’s the new reference for an "efficient but not too expensive" card. It can give you a good idea about the differences between high mid-range and high end products. Why the X1800 XT? Simply because it has similar frequencies to the X1800 XL. Because of identical calculation powers and bandwidth, we can measure performance gains brought by ATI´s new architecture.

Specifications

Power consumption

We measured the power consumption of different graphic cards used in this test. Measurements were taken directly at the power outlet, which therefore represents the computer’s entire power consumption, here an Enermax 535W. We reported figures obtained under Window’s desktop and in use with a 3D scene and Prime95. Prime95 makes it possible to have constant CPU use regardless of a graphic card’s performance.


In 2D, the Asus Dual N7800GT has the highest power consumption and the X1800XL has the least. The 7800 GTX 512’s figure isn´t higher than the 7800 GTX’s.

In 3D, consumption clearly increases and in different ways depending on the load. In the fillrate test, ATI´s cards provide moderately good results and are probably capable of deactivating unused groups, whereas NVIDIA cards runs at full speed and have much higher power consumption. The 7800 GTX 512 figure is 38 Watts higher than the 7800 GTX. With 3DMark05, the situation is different and this time NVIDIA´s power consumption is lower, except of course for the Asus´ card which is twice as high. The GeForce 7800 GTX 512 is at the same level as the Radeon X1800 XT and is 25 watts higher than the GTX.

The test

For these cards, we updated our test protocol to include new games and modified some procedures. F.E.A.R., Serious Sam 2 and Age of Empire III are now included. Quake 4 replaces Doom 3, and Half-Life 2 Lost Coast is now used. Far Cry is still here, but with a different demo, which relies mostly on GPU power. Splinter Cell Chaos Theory, Act of War, Colin McRae 05 and Pacific Fighters remain the same.

We also decided to keep anisotropic filtering activated in all tests. It significantly increases texture sharpness and the higher the resolution the better it is. Its use is logical with high end graphic cards even if anti-aliasing is deactivated. When the game allows direct configuration, we selected the highest mode, or we forced 16x anisotropic filtering via the driver control panel. "Anisotropic sample optimizations" were deactivated, which increased NVIDIA´s shimmering. It doesn´t change performance much, but gives us similar graphic quality for ATI and NVIDIA in tests.

Test configuration:

Asus A8NSLI Premium
AMD Athlon FX 57
2 x 1 GB of Corsair XMS 3200
Western Digital Raptor 74 GB
Enermax 535W
Windows XP SP2
Catalyst 5.11
ForceWare beta 81.89

HDR, F.E.A.R.

HDR, reminder

HDR rendering is the new trend for 3D cards, but what is it exactly? Just nicer lighting effects? It is in fact more complicated. First, there is not just one "HDR rendering" but many, which all have their advantages and inconveniences in higher performances, better accuracy, blending support, etc. Developers have to chose the HDR rendering mode, which corresponds best to what they want to do and their overall engine.

The HDR rendering mode consists of an image or part of an image when it’s processed with a more extended dynamic format, which means a format with a wider range between the smallest and biggest values it can represent. In other words, HDR is a temporary rendering format with higher calculation accuracy. It preserves image quality in some operations but also applies more complex effects to the image, which require more accuracy. HDR, at least for now, is never directly displayed. You don´t see images in HDR format but rather images, which have benefited from HDR at some point in their calculation.


It’s possible to rewrite temporary HDR images to displayable ones simply by leaving out additional information included in the HDR format if they are no longer useful. It is however possible to go even further and use all data included in the HDR format to create a new better image to standard format. This is called tone mapping and is used in fields other than 3d, for example, photos. Tone mapping consists of interpreting HDR data to obtain a higher quality image. There are many tone mapping algorithms and developers include them according to their objectives such as increasing contrast, changing color palettes, displaying blinding effects, etc.

ATI and NVIDIA each support each a certain amount of different types of HDR. Here are the main differences:
- ATI supports FX10 blending, NVIDIA does not.
- ATI supports multi-sampling antialiasing in FX and FP, NVIDIA does not.
- FP16 texture filtering is free for NVIDIA not for ATI

Overall, ATI has more flexibility than NVIDIA for the number of HDR formats supported. However, as NVIDIA was the first to develop this technology, FP16 is the main format used. This can be a bit problematic for the Radeon X1K, which doesn´t normally filter FP16 textures and as long as shaders aren´t adapted to combat this deficiency.

In our tests, five games out of ten have HDR modes and two are exclusively based on FP16.

F.E.A.R.

We start the tests with F.E.A.R. using the integrated demo. Unfortunately, this only gives a whole number score, which can lead to a difference of one unit under the same conditions because of a normal variation of two tenths. For each card, we selected the best of three results. It’s important to point out that there was an error in ATI´s drivers. ATI´s optimisation is based on executable detection (like with all profiles), which in practice considerably reduces performances. We changed this in the ATI card tests and noted performance gains of up to 50%(!).

16x anisotropic filtering was activated via the game.


If Radeon X1800XL performances were similar to the 7800 GT’s and if the X1800 XT was ahead of the 7800 GTX, these two cards’ performances were much lower than that of the 7800 GTX 512 MB. The Asus N7800GT Dual finished first. You will also note a 10% higher performance gain between the X800 architecture and the X1800’s.


Once antialiasing is activated, the X1800 XT surprisingly takes the lead whereas its little brother is very close to the 7800 GTX 512. The gap between ATI´s 2 architectures is close to 20%.

Half-Life 2 Lost Coast

Half-Life 2 Lost Coast

For this test, we used an internal demo recorded with Lost Coast to test Valve´s HDR, which uses a quite complex rendering format. It doesn’t maximize the additional capabilities of the GeForce 6, 7 and Radeon X1K, but runs with all DirectX 9 with MSAA.

Anisotropic filtering x16 was activated via the game.


The GeForce 7800 had great results. The 7800 GTX is ahead of the X1800 XT, and the GTX 512 and dual N7800 are in the lead. Is this the end of ATI´s supremacy in Half-Life 2?


With anti-aliasing, the X1800 XT is back in the game and ahead of the 7800 GTX. The GTX 512 is still far beyond reach, however. The X1800 XL is ahead of the 7800 GT and clearly above the X800 XT. This shows the benefits of ATI´s new architecture.


In HDR, the gaps widen. The 7800 GTX is back ahead of the X1800 XT and the 7800 GTX 512 and N7800GT Dual increase their advance. In 1920 x 1200, the GTX 512 is 40% faster than X1800 XT!


With anti-aliasing in addition to HDR, ATI is back in front of the 7800 GTX. The 7800 GTX 512 and N7800GT Dual, however, are still in the lead. The gain for ATI´s new architecture is close to 50%.

Far Cry

Far Cry

We used an internal demo, which is a mix of outdoor and indoor locations recorded in the "catacombs" map. We activated the post process "cold" or "hot" filter, which noticeably increased image quality. With the post process and HDR, our honest impression is that rendering quality is from another era and we rediscover the game. This post process filter leads to +/-20% performance reduction.

Anisotropic filtering 8X was activated in the game.


If the X1800 XT provided great performances compared to the 7800 GTX, it can´t catch up to the GTX 512 MB.


Anti-aliasing is the Radeon’s favourite field and provides better performances. Even if close to the 7800 GTX 512, it cannot reach the same performance levels. The Asus´ card is almost 30% faster than the GTX 512.


The X1800s aren´t very comfortable with Far Cry in HDR. The Radeon and GeForce have a basic FP16 rendering quality, but it isn´t identical as the image is less "burned" for ATI. This is related to a more subtle processing of HDR data. For their activation with the Radeon X1800, we used the 1.4 beta patch. This was only on these cards and in HDR so as not to penalize the GeForce because of the beta of this patch. Compared to previous drivers, Radeon performances have improved (the X1800XT was behind the 6800 GT a month ago!), but the X1800 XT remains behind the 7800 GT. The other cards are clearly in the lead, for example, with 66% more for the GTX 512, which was for once ahead of Asus´ card. SLI isn´t entirely efficient with Far Cry HDR (in AFR2 mode, the efficient one).


Only the Radeon X1K supports MSAA when HDR FP16 is used. The X1800 XL can´t run the test in 1920 x 1200 as it "only" has 256 MB of memory.

Serious Sam 2

Serious Sam 2

Here we also recorded a demo and activated anisotropic filtering 16x in the game.


The X1800 XT and XL provide great results compared to the GeForce 7800, but the GTX 512 is once again in the lead. The Asus card dominates all. The X1K’s new architecture improves performances.


Once anti-aliasing is activated and unlike what we usually noticed, ATI´s gap with the 7800 GTX and GT was reduced. The GTX 512 and N7800 Dual are further ahead.


In HDR, here in FP16 format, ATI displays very good performances. The X1800 XT finishes in front of the 7800 GTX 512, whereas the XL is ahead of the 7800 GTX. It’s important to point out here that during the game’s developement, developers only based their code on the GeForce 6 and 7 and wrote a shader using NVIDIA´s native FP16 filtering. Once applied for ATI, it results in non-filtered elements, which produce “blooming” pixels. Nothing indicates, however, that these elements have to be in FP16 format and if they don´t developers could solve the problem by processing the filter directly in the shader. This would result in a performance drop for the Radeon X1K. By how much? It’s hard to tell, and we will have to wait for the release of a possible patch to know.

As NVIDIA doesn´t support MSAA with FP16, game developers haven’t implemented it and it isn´t possible to activate antialiasing with HDR on the Radeon. Once again, a patch could resovle this.

Splinter Cell Chaos Theory

Splinter Cell Chaos Theory

Once again we used an internal demo. SM 3.0 is used with all cards that support it. Soft shadows and parallax mapping were activated.

Anisotropic filtering 16x is activated via the game.


The X1800 provided great results compared to the 7800 GT and GTX, but the GTX 512 was ahead. Asus´ card fully benefited from its 2 GPUs and finishes 30% above the first card. X1K architecture brings a little less than a 10% gain, which we’ll gladly take.


Results here are similar to those previous ones except ATI increases its lead over the 7800 GT and GTX.


Here, it was the same situation as well. Note that HDR differs between ATI and NVIDIA. With the GeForce it is FP16, but with the Radeon it would rather be FX16. This is the reason it also works with the X800, noticeably reducing quality.

Age of Empire III

Age of Empire III

To test this game, we saved a scene with a movement of units.

Anisotropic filtering was activated via the game.


The Radeons were completely left behind with the "small" 6800 GS almost ahead of the X1800 XT. In addition, gains from the new architecture were reduced.


With antialiasing, the Radeons were back, but it wasn´t enough.


With HDR, the situation was identical. ATI was clearly behind. HDR produced a similar result for NVIDIA and ATI, but it wasn´t processed similarly, based on FP16 for NVIDIA and FX10 for ATI. If only 2 bits are left for transparency this FX10 mode is kept within the standard 32 bits, which is ideal for performances. ATI doesn´t seem to take advantage of this, however.


Like in Far Cry, the X1Ks are the only ones to combine HDR and MSAA.

Act of War, Colin McRae 05

Act of War

Like with Age of Empire III, unit movement in our test was always identical and reproducible.

Anisotropic filtering 16x was activated via drivers.


The X1800s were a bit behind despite a major performance gain of 36% from new architecture. The GTX 512 takes this opportunity run away with the lead once again. The N7800GT Dual was restricted by the CPU.


With anti-aliasing activated, ATI is back. The X1800 XT finishes ahead of the 7800 GTX but way behind the unreachable 512 MB.

Colin McRae 05

We drove a specific reproducible trajectory (straight ahead until the end of the track) in the Japan Rally.

Anisotropic filtering 16x was activated via the drivers.


With this favourite game of ATI, NVIDIA was closer. The X1800 XT remains ahead of the GTX. The new architecture doesn´t bring any improvement with this game. The GTX 512 and N7800GT Dual don´t have any difficulty in dominating this part of the test.


Anti-aliasing gives ATI better performances, but once again not enough to reach the GTX 512 MB.

Quake 4, Pacific Fighters

Quake 4

Here, we saved an action scene. Unlike Doom 3, there are fewer shadows but with more characters and textures. This changes the load in terms of rendering.

Anisotropic filtering 8x was automatically activated by the game.


Compared to Doom 3, the Radeons had better results. However, this game was still dominated by NVIDIA. The new architecture improved performances by a little less than 10%.


Once anti-aliasing was activated, the X1800 XT bounced back and used its 512 MB of memory to end up in front of the 7800 GTX. The GTX 512 also does the same and remains ahead by overtaking even the double 7800GT, whose GPUs only have 256 MB of memory. The improvement of ATI´s new architecture is 22%.

Pacific Fighters

In the second OpenGL game of this test, we measured performances while reading the recording of a combat scene. Anisotropic filtering 16x was activated via drivers.


ATI is left behind despite a 20% of improvement due to the new architecture. The 7800 GTX 512 fully benefits from its highest frequencies to be ahead of the GTX and even the Asus card in 1920 x 1200.


With anti-aliasing, ATI does a little better, but the new GeForce remains untouchable. Gains due to the new architecture reach 30%.

High quality AA and AF: cost

The cost of Transparency / Adaptative AA

NVIDIA and ATI have now included two "intelligent" anti-aliasing modes, which force a super sampling mode on the triangles needing it. This technology is used to soften objects simulated from alpha tests, which don’t benefit from standard multisampling which only filters real polygons. Reserved to GeForce 7800, this option is functional for all Radeon DirectX 9 starting with the Radeon 9700 and derived products.


Overall the cost of this option is greater for ATI, especially in F.E.AR and Act of War. In Far Cry, Serious Sam 2, Act of War and Colin McRae 05 it’s activation is unadvised as it strongly reduces performances in some areas. Keep in mind that this is an average performance loss that can be more significant in certain areas.

The cost of HQ anisotropic filtering

With the X1K, ATI is the first manufacturer to re-release a serious anisotropic filtering not based on an optimisation that consists of reducing the filtering quality of some angles. This can lead to serious reductions in quality for some games and for no reasons somes surfaces are blurred.


With the exception of Pacific Fighters, which almost exclusively relies on filtering, the aniso HQ cost is very low given that filtering latency, which increases with quality, can be masked by shader calculations. This can be done all the more efficiently by ATI because texturing units are decoupled from the rest. Don´t hesitate to activate this.

Benefits of the 512 MB

Benefits of the 512 MB

We wanted to know if the additional 256 MB of memory (generally quite expensive for high end cards) really improved performances while keeping in mind that more memory can stop monitors from freezing (not always obvious from the numbers) and that its use will continue to grow in the months to come.

For this purpose, we clocked the X1800 XT at the same frequency as the X1800 XL and modified the 7800 GTX 512 MB’s frequencies (including the "geometry delta clock") to make it correspond to the GTX. We had to change the bios to modify the "geometry delta clock" value as no software can do this and it would have distorted results considering that 7800 GTX vertex shaders are clocked 40 MHz higher than the GTX 512.


NVIDIA´s performance gains are greatly reduced and ATI benefits with an almost 20% performance gain in Quake 4.


With FSAA 4x, NVIDIA´s results increased but gains remained small. The maximum was slightly more than 13% in Quake 4. In comparison, ATI fully benefited from more memory with a 26% performance increase in Quake 4 and 15% in F.E.A.R., Act of War and Pacific Fighters.


In HDR, the 7800 GTX 512 MB doesn´t benefit at all from additional memory. There is a minimal gain for ATI, but it does exist.


In HDR with AA, the two cards benefited from 5-20% performance gains in Half-Life 2. In the two other games that ATI can run in HDR and AA, results were much different. The gain in Far Cry was 25% and null in Age of Empire III.

Overall, ATI seems to benefit more from local 512 MB memory. It’s hard to know the reason. Maybe NVIDIA hasn´t optimised their drivers yet for such a quantity of memory. Considering the price of a 7800 GTX 512 and the low performance gain, we would think that a version equipped with only 256 MB but clocked at the same frequencies could be of interest thanks to a much lower price. This wouldn´t have been good for NVIDIA, which isn´t planning on selling a massive amount of G70 chips clocked at 550 MHz.

In a nutshell

Performances in a nutshell

We calculated the average of all benchmark results, gave the same weight to all games, and attributed 100 to the GeForce 7800 GTX in 1600 x 1200.


If the GeForce 7800 GTX without antialiasing was already slightly ahead of the Radeon X1800 XT, the new one increases this situation and is 26.3% faster!

Asus´ dual 7800GT is in the lead with 5 to 10% higher performances than the GTX 512. A single 7800 GT is almost 15% faster than the X1800 XL. Average performance gains of 15% due to the new architecture aren´t enough!


With antialiasing activated, the X1800 XT is 18% faster than the 7800 GTX. The GTX 512 catches up and even takes the lead by being 20% faster than the X1800 XT. The X1800 XL finishes at the same level as the 7800 GT. X1K architecture improves performances by an average of 20% in this mode.


In HDR, the X1800 XT is comparable to the 7800 GTX. The 7800 GTX 512 is just in front. X800 XT performances are strongly reduced as it only supports HDR in Half-Life 2.


In HDR and AA, the Radeon X1800 obviously dominates because it’s capable of using it in three games compared to only one for other cards.

Conclusion

Conclusion

Even if the winner of this survey in terms of performances is the Extreme N7800GT Dual (the Asus collosus has become the current fastest graphic card by including two GeForce 7800 GTs), the new GeForce 7800 provides record impressive performances. By fixing its functioning frequencies, NVIDIA clearly hasn´t given much of a chance to the Radeon X1800 XT, which is, with a few exceptions (including the impossibility of the 7800 GTX 512 MB to display in HDR with anti-aliasing), completely left behind.


So with a much higher calculation power and bigger memory bandwidth for the 7800 GTX 512 MB what’s left for the Radeon X1800 XT? It’s capable of simultaneously processing HDR and antialiasing for more rendering formats and is far more efficient with more complex shaders using branching. It also has a new memory controller, which opens up the possibilities for new optimisations. But are these points sufficient when HDR combined with antialiasing is restricted to low resolutions to maintain a satisfactory framerate. And also when developers still haven’t lost the habit of avoiding branching to prevent low GeForce 6 and 7 performances and when ATI driver brings performance bugs (F.E.A.R, OpenGL et antialiasing) instead of optimisations? For now, NVIDIA is ahead and we will wait for ATI to demonstrate (or not) via new drivers that their cards still have headroom for more performances.

There is of course the problem of price. The GeForce 7800 GTX 512 MB costs 650€ /$649 compared to the 550€ / $549 of the X1800 XT. But even if this is high, it isn’t really an issue for high end products. On the other hand, with the low availability of these cards, in reality, prices tend to be much higher. It will be hard to know the exact figure for these cards in stores. This will be the same for the Asus´ N7800GT, which is announced at 950€-1000€ / $1000 in very restricted volumes.

A small drawback with the GeForce 7800 GTX 512 MB is its name. Only the difference in memory size is implied in the name GTX and GTX 512, whose card frequencies and performances are much superior. Moreover, if one manufacturer decides to release a 7800 GTX with 512 MB of memory or a 7800 GTX 512 MB with 256 MB the confusion will be even greater.