The cards

The cards

For this test, Asus and MSI each provided us with their GeForce 8600 GTS. The two cards are identical, because they are both Nvidia reference models and all they had to do was place their names on the package.


The card is short, but needs a supplementary six pin PCI Express-type power source, because the 289 million transistors, of course, are more demanding than the140 million of the GeForce 7600 GT. In practice, power consumption is, in fact, comparable to a 7900 GT, and it is therefore more than a 7600 GT’s. The official numbers are a max of 71 watts for a 8600 GTS and 43 watts for the 8600 GT. In terms of noise, it’s satisfactory at best. It’s a little better than the GeForce 7600 GT in 3D, however, at times slightly more noisy in 2D, when the PC or room temperature is higher.

Asus

Asus’ card is the EN8600 GTS TOP, in other words an overclocked version of the 8600 GTS. Thus, frequencies go from 675/1450/1008 to 745/1620/1145 with respective overclockings of 10, 12 and 14%.

MSI

For MSI, the NX8600GTS-OC is also overclocked but less so, because frequencies only increase 2, 0 and 4%, respectively, or 690/1450/1050. In other words, an overclocking that is rather pointless.

Differences in performance

Video, drivers, the test

A new video engine

With the GeForce 8600 and 8500, Nvidia introduces an evolution in its video engine, the VP2. Thanks to it, a bit stream processor, and a hardware decoding of content protection systems, Nvidia finally claims to offer a solution that will take into account the complete specifications of HD-DVD and Bluray.



After having vaguely used HD video and the GeForce 6 and 7 as main marketing points and without really providing viable solutions, will we finally have a card that lives up to its promises?

This is something still to be verified. We were unable to carry out this test due to a lack of time, optimized applications (a new video player version is necessary) and available drivers. If a beta driver is indeed available for Windows Vista, Nvidia says that we will have to wait until June for Windows XP, which is quite a long time.

The drivers

For this test, Nvidia provided a beta version of its 150 drivers, the 158.16 for Windows XP and 158.14 for Windows Vista. This supposedly represents a return to normality after the setbacks in the last few months of rarely updated drivers and/or those solely intended for specific graphic cards. Let’s hope this is the case, and either way these drivers appear to be on the right track.

The test

In this test, we decided to leave aside basic rendering modes in games and to systematically concentrate on HDR or assimilated modes when available. HDR or rather overall image quality processing is more and more common and adds a non negligible benefit to rendering. It would be unfortunate to play without it, as well as anisotropic filtering, which is always activated.

The test configuration

Intel Core 2 Extreme X6800
EVGA nForce 680i
2 GB of memory
Windows XP SP2
Forceware 158.16
Catalyst 7.3

Please note that we tested all Nvidia cards, including the GeForce 7, with the 158.16 in order to have comparable results.

Quake 4

Quake 4

We saved a demo of an action scene. Unlike Doom 3, there are fewer shadows, but more monsters and textures, which changes the rendering load.

8x Anisotropic filtering is automatically activated via the game. Tests were carried out in high quality mode.

Half Life 2 Lost Coast

Half-Life 2 Lost Coast

For this test, we use an internal demo recorded in 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 cards with MSAA.

4x Anisotropic filtering is activated in the game.

F.E.A.R.

F.E.A.R.

We use 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.

All graphic options are pushed to the maximum except for soft shadows, which were deactivated. 4x anisotropic filtering is activated in the game.

S.T.A.L.K.E.R.

S.T.A.L.K.E.R.

We carry out a movement that is always the same and measure the framerate with fraps. First, we test the game with average settings, in other words with partial dynamic lighting and intermediary details. Then we increase the quality, adjust to complete dynamic lighting, maximum detail, and grass shadows. Anti-aliasing isn’t used, because S.T.A.L.K.E.R. has an engine based on deferred rendering, which is essentially incompatible with MSAA. You may activate a type of antialiasing based on edge detection, which is carried out via a shader, however, the result isn’t too convincing. The 1.00001 patch is used.

Rainbow Six Vegas

Rainbow Six: Vegas

We measure performances in the introductory scene with 4x anisotropic filtering activated via drivers. The HDR mode is activated as it is more or less obligatory, because without it, banding is very noticeable. Shadows are set to “low” and setting them to a higher quality lowers performance too much in certain areas. We add that ForceWare 158.16 allows the activation of AA even in HDR, which is something not mentioned in the game. Let’s hope that ATI will follow NVIDIA in this way, even if we have to admit that this type of extra isn’t that useful four months after the game’s release.

Tomb Raider Legend

Tomb Raider Legend

We saved a game where we carried out a regular and reproducible movement. 4x Anisotropic filtering was activated via drivers. We exclusively used the Next Gen mode, as the normal one seemed to display quality from a former era.

Oblivion

Oblivion

We saved a specific movement in order for it to be always identical and the test reproducible. HDR is activated.

4x Anisotropic filtering is activated via drivers.

Age of Empire III

Age of Empire III

For this test, we saved a game, in which we carry out a specific movement in pause mode. HDR was activated. Note that HDR rendering produces a similar result for both ATI and Nvidia but is carried out differently. Based on FP16 for Nvidia, it’s in FX10 for ATI., While it only has 2 bits for transparency, the FX10 mode allows staying in the classic 32 bits, which is ideal for performances. The last patch, the 1.11, is used in a way to support HDR FP16 and 4x MSAA for the GeForce 8. A SSAA 2x is emulated for this game for the GeForce 7.


Anisotropic filtering is activated via the game.

Need for Speed Carbon

Need for Speed Carbon

We saved a specific movement in order for it to be always identical and the test reproducible. All game options are pushed to a maximum, except the rain, which is something that can vary.

Anisotropic filtering is activated via the game.

Conclusion

Conclusion

This first batch of mid-level DirectX 10 graphic cards still leaves something to be desired. The GeForce 8600 GTS is comparable to the Radeon X1950 XT 256 MB in price, but is far from its performance. It’s similar to the much less expensive X1950 Pro, and is actually a little ahead without antialiasing, and behind with its memory bus limited to 128 bits.

Compared to Nvidia’s previous offer, the GeForce 8600 GTS is of course better because performance is satisfactory in addition to DirectX 10 support. Now, the previous generation has lost a bit of interest these past few months. An improvement wasn't that difficult and for this reason we are bit disappointed by this GeForce 8600 GTS. It’s not that it’s bad, it gives us the strict minimum.


While Nvidia had us used to a mid-level card that was closer to high end, this one does the contrary. Gaps between lines are even bigger now and the explosion of performances that came with GeForce 8800s didn’t happen here. If some believed that the 8600 GTS would be an interesting alternative to the GeForce 8800 GTS 320 Mo, they will be disappointed with the results of this test as the latter’s performance is far off. This is true even with two 8600 GTS in SLI ! The GeForce 8600 GT should have provided results similar to a 7900 GS at an equivalent price and with additional features. However, once again the comparison with ATI’s solutions is problematic for Nvidia.

The new video engine could add to the value of this card, if it keeps its promises by finally giving a reliable solution to users in the domain of HD video reading. This is something to be verified.

The release of these entry and mid-level GeForce 8s leads us to believe that Nvidia is paying the price for its complex architecture, which imposes a high fixed cost in terms of transistors. This can make it difficult to make a chip of reasonable size with good performances, and is a problem similar to what ATI encountered with the Radeon X1000. This manufacturer needed several months and a more advanced fabrication process to finally be competitive in terms of raw performances with the GeForce 7.

AMD, which now includes ATI, should also propose its solutions for DirectX 10, including a mid level one next month. Unless you urgently need a mid level DirectX 10 card, we recommend you opt for a competitive DirectX 9 solution or wait a few weeks to find out how the competition is going go react. Either way, Nvidia should review its pricing so that its GeForce 8600 GTS has more of a chance.