Today's the day that DX9 cards are put out to pasture. NVIDIA has unleashed the first DirectX 10 cards in the GeForce 8800 GTX and GeForce 8800 GTS. These two behemoths of a card represent the first in a line of a new architecture for NVIDIA in the G80. While we have a prototype of Leadtek's implementation of the GeForce 8800 GTS on hand, we'll start off with a brief overview of what's new with the G80. Let's see what 350+ people spending $475 million dollars for four years can come up with.
With the announcement of the two cards, the NVIDIA line from top to bottom now stands at:
- GeForce 8800 GTX - $599
- GeForce 8800 GTS - $449
- GeForce 7950 GT - $299
- GeForce 7900 GS - $199
- GeForce 7600 GT - $159
- GeForce 7600 GS - $129
- GeForce 7300 - <$99
You can see NVIDIA has cards across every price range to suite anyone's needs. The new enthusiast cards today sit at the top. So what's new with the GeForce 8800 architecture? Let's take a look at some of the new features that the two cards have.
Unification seems to be the way to go with video cards as NVIDIA has a unified shader design for the G80 rather than the old separation of pixel and vertex shaders. The new unified processors are now called stream processors. The GTX card features 128 individual stream processors running at 1.35GHz while the GTS knocks off one grouping with 96 individual processors running at a slower 1.2GHz. From the picture below you can see how the new architecture is setup.
So NVIDIA has done away with individual pixel and vertex shaders with this design. Previously, some parts of the card would be sitting idle if a scene is more intensive in terms of one or the other. Now since the stream processors can do anything really, the whole card can be used and nothing would sit idle as each processor would be doing something to help render the scene. It all comes down to the card being more efficient and being used to its full potential.
What helps make the GTX card so expensive is that the card holds 768MB of GDDR3 memory. The GTS has a little less at 640MB. There's no word on lower memory cards yet but you can bet they will be out in the near future to bring some of these features down to the mainstream price range. I asked NVIDIA why they didn't go with GDDR4 and they replied that a combination of availability and cost drove them to use GDDR3 for now.
You'll be hearing about the Lumenex Engine when the new cards are out. The Lumenex Engine consists of:
- 16x Coverage Sampling Anti-Aliasing (CSAA)
- 16x near perfect angle independent anisotropic filtering
- 16-bit and 32-bit floating point texture filtering
- Fully orthogonal 128-bit High Dynamic Range (HDR) rendering with all the above features
- A full 10-bit display pipeline
Let's talk a little about Coverage Sampling Anti-Aliasing. Introduced in the GeForce 8800 series, the new anti-aliasing technique was designed to provide the highest quality image little hit on performance. NVIDIA states that you can use CSAA with close to the same speed as when using traditional 4x MSAA (multi-sample anti-aliasing). Supposadly, you'll be experiencing a 10-20% decrease in performance with CSAA. The new algorithm is suppose to perform the 16 samples per pixel anti-aliasing without bogging down the memory. You can also now use HDR with AA, something that wasn't possible in previous NVIDIA cards.
To render textures at an angle accurately, a video card can use a process called anisotropic filtering to achieve this effect. In the past, the GPU would only filter textures at certain angles so as the performance hit wouldn't be as great. The shortcoming is that not all the textures go through the anisotropic filtering for accurate rendering. The Lumenex Engine doesn't have this limitation as it accounts for all angles of surfaces providing a more cleaner texture at different angles.
For High Dynamic Range, the Lumenex Engine offers 32-bit floating point precision for each color component for a total of 128-bits for HDR rendering. As mentioned earlier, you can now use HDR with AA without any problems in the GeForce 8800 line.
Along with support for over a billion colors, the features mentioned combined together form the Lumenex Engine. NVIDIA developed the set of features to offer great image quality. NVIDIA is really pushing image quality in this generation and this is part of their initiative to develop the highest quality graphics card out there.
Physics is something that NVIDIA is tackling more and more of and they're introducing Quantum Effects GPU-based physics in the GeForce 8800. This will help physics computations by using stream processors on the card to do the calculations. Things such as hair, water, and smoke can be all rendered by the GPU and free up some CPU power for other processes.
DirectX10 is getting a lot of hype recently with showings of the technology in games such as Crysis. The API is a big part of Windows Vista. The GeForce 8800 series are the first cards to provide support for such features as geomatry shaders and overall, Shader Model 4.0. With Vista coming out in January (November for corporations), the GeForce 8800 looks to be the only card to take advantage of DirectX 10 in Vista unless ATI does something in the next two months.
I've been a big proponent of PureVideo as all my HTPCs use NVIDIA's video processing. PureVideo HD takes it a step further as it's made for 720p, 1080i, and 1080p output. If you're going to be using an HD DVD or Blu-ray drive to watch HD video, PureVideo HD will help deliver smooth and crisp picture quality. The cards will be HDCP compliant so you'll be able to view HD content on HDCP compliant displays.
If you look at the pictures of the GTX card, you'll see not one but two SLI interfaces on top. While nothings announced officially from NVIDIA on support for more than two in SLI for the GeForce 8800, the secondary SLI link is there for future expansion. You can bet that NVIDIA will be hounded by questions of Quad-SLI support now but alas, there's nothing yet to talk about in that area for this card.
So all these new features, new architecture, new design adds up to a pretty fast card. How fast you ask? Before we get to the tests, the company states that a single GTX card will be faster than two GeForce 7900 GTX in SLI mode and the GeForce 7950 GX2 card as well. Can you say video card speed demons?
No doubt about it, these cards are made for extreme resolutions in gaming. With the GeForce 8800, you should be gaming at minimum of 1600x1200 NVIDIA states. If you have a high end LCD capable of 2560x1600, this card is made for you.Unfortunately , monitors with those specs are pretty pricey but then again if you can afford one of those you can probably afford a GeForce 8800 card as well.
So now that we went through what's new with the GeForce 8800 series, let's take a look at the card that Leadtek has kindly provided us. First off, the card is built on a 90nm process and contains 681M transistors. Core speed sits at 500MHz with the GTX version 75Mhz faster. As mentioned earlier, the shader clock sits at 1.2GHz compared to 1.35GHz for the GTX version. Memory is clocked at 1.6GHz DDR while the GTX has memory clocked at 1.8GHz DDR. The prototype has 640MB of memory.
You can see it uses a dual slot cooling solution. The all cooper cooling unit features a large fan that exhausts the warm air out the back. Like their recent NVIDIA cards, the GeForce 8800 GTS is pretty quiet running. After all the tests and running through some games, I didn't hear the fan at all. It does speed up a little when 3D applications start up but for the most part you won't hear much coming from the cooling system. The card measures a full 9.5" making it one of the longer cards out there. It is atleast 2" less than the GTX version. Unlike the GTX version, there's one power connector on the GTS you have to worry about.
Seeing as this card is made to run at high resolutions, I decided to forgo anything below 1280x1024. We don't have a monitor capable of higher than 1600x1200 yet but when it comes in I'll be sure to rerun the tests for this and other future cards. Beginning this card, I've replaced Half-Life 2 Lost Coast with Company of Heroes. I figured that I needed a little more variety in testing than FPS games and Company of Heroes is an RTS that really pushes video cards as well.To compare, I will test against Leadtek's GeForce 7950 GTS card with 256MB of memory. We don't have any ATI X1950 XTX cards in as I had to send them back so if we get one back in I'll do a follow up article with a comparison. The drivers were supplied by NVIDIA and they are version 96.94.
The test setup consisted of:
- AMD64 X2 3800+
- ABIT NF-M2 nView
- 2 GIG Crucial Ballistix PC2-6400 (2 sticks of 1GB each)
- Seagate 160GIG 7200RPM HDD
- Windows XP w/ Service Pack 2
is the worldwide standard in advanced 3D game performance benchmarking. A fundamental tool for every company in the PC industry as well as PC users and gamers, 3DMark06 uses advanced real-time 3D game workloads to measure PC performance using a suite of DirectX 9 3D graphics tests, CPU tests, and 3D feature tests. 3DMark06 tests include all new HDR/SM3.0 graphics tests, SM2.0 graphics tests, AI and physics driven single and multiple cores or processor CPU tests and a collection of comprehensive feature tests to reliably measure next generation gaming performance today. We tested at the standard 1280x1024 resolution.
is Raven Software's true sequel to the id classic. The game uses an improved Doom 3
engine for some great graphics. For the test we ran a demo featuring a few enemies and some squad mates. We set the graphics qualities at maximum and ran it on three different resolutions.
One of the surprise hits out of Monolith was F.E.A.R.
This supernatural FPS looks incredible and really pushes a video card to its limits. For the benchmark, we ran three resolutions using the in game benchmark with all the settings set at max.
-Updated 12/18/06- I didn't turn on soft shadows the first time I ran this so this is what the real score with the options I normally use.
has been in development for many years but the folks at Human Head finally released the game this year. The game utilizes the Doom 3
engine like Quake 4
and features the really cool Portal technology to garner some interesting game play aspects. All settings were set to maximum and three resolutions were chosen for the test.
Company of Heroes
is an RTS that really pushes video cards. The game by the fine folks at Relic Entertainment is set in WWII and features deformable terrain as well as great physics. The level of detail in the game for an RTS is amazing. For the tests, we set everything at maximum or ultra to ensure that the card was taxed as much as possible.
While games like Quake 4 seem to have my CPU as a bottleneck, the newer ones show a greater range of scores. You can see how much faster the GF8800 GTS card is in these tests over an overclocked GeForce 7950 GTS card. We'll be getting some Core 2 Duo CPUs in soon so I'll revisit the tests again with the faster CPU so that we're not as CPU bound and you'll be able to see how much faster the card really is.
While there's still the traditional anti-aliasing and anisotropic filtering modes, the GeForce 8800 introduces CSAA. We'll test the regular modes first and then I'll show you how NVIDIA's CSAA doesn't bog the card down as much while offering some very nice anti-aliasing quality. First up is 4xAA with 8xAF.
Next up we turn it up a little with the traditional 8xAA with 16xAF.
Now we move onto NVIDIA's latest achievement in image quality and that's Coverage Sampling Anti-Aliasing. I ratcheted the anisotropic filtering back down to 8X for these tests and used 16x for CSAA. You might be suprised by the results below.
You can see that the new Coverage Sampling Anti-Aliasing technique doesn't impact the card as much. The tests do indeed show that it falls within the 10-30% performance range of 4xAA using the same ansiotropic filtering. That's quite an impressive feat and I dont't see a reason to hold back on turning on 16x CSAA when playing some current games at 1280x1024 or 1600x1200. CSAA also works at 8x so if 16x is bogging your machine down a bit you can scale back a little and still get better performance over the traditional 8x MSAA. I ran some tests to make sure this was the case and the benchmarks were faster with 8x CSAA than 8x MSAA.
What can I say about the GeForce 8800? NVIDIA really did a great job on this card. The GTS version that Leadtek supplied me really showed off the speed and great image quality that this card can accomplish. Yes, it's pretty pricey and not many can afford to have one of these in their machines but if you can you'll be gaming at the fastest speed possible right now. ATI's got a lot on their hands to try and top this one. First to market with a DirectX 10 card and featuring a hard launch whereby you can purchase the cards right now, the GeForce 8800 series is a big success for the green team. I'm anxious to see how a GTX card performs over the GTS one and I hope to have one soon to test out. In the meantime, enthusiasts should definitely pick this puppy up. Thanks to Leadtek for helping us out on this one and providing us with a prototype.