NVIDIA’s G80: GeForce 8800GTX

The next generation in graphics technology has arrived. Does it deliver the performance you would expect from a next gen part? Read on to find out.

Sigh. Don’t you just hate it when you come home after buying the biggest and best thing available, only to find your highly competitive neighbor proudly wielding the newly-released version up from the one you just bought? It’s a series of events we see unfold all the time in the entertainment world, whether on the small screen in cartoons, or on the big screen in movies like “Borat: …”. However common this scenario may be on TV, it probably doesn’t happen all too much in real life. People are generally complacent when it comes to expensive purchases. If you get “outdone” in the real world as described above, chances are you won’t be going back to the store to buy an even bigger or better model. Then again, check out the hardware scene. Try as we might, we can’t think of a day that has gone by in our forums during which someone has not asked what the biggest and best thing available was, where they could get it, and how much for.

It’s this vicious cycle in the hardware world. You buy something that is the top of the line for a hefty chunk of change, only to have something better, often times MUCH better come out just two or three months later. And, despite what we would like to think, there are plenty of those among us that take the bait time and time again. Why in the world of hardware do people more commonly go for the top of the line products than in most other market sectors? Perhaps it is the (relatively) inexpensive nature of computer parts. Perhaps it is because the high-end market, at its core, is driven by gamers who are always searching for the bleeding edge.

Whatever the reason, it happens, and it happens a lot. A few months ago gamers were eating up a top of the line video card called the GeForce 7950GX2. Produced by one of the staple names in the gaming scene, NVIDIA, the card, or cards as it were, boasted not one but TWO GPUs that offered the very best performance money could buy. Sure there were less expensive solutions out there that could muster performance in the same ballpark, but there can only be one “absolute best”, and the 7950GX2 was it. After reaching into their pockets and coming up considerably short of the $600 required to own one of these behemoths, gamers across the world ended up falling for one of the less expensive options, they had little choice. Priced at right around $300, NVIDIA’s GeForce 7900GT offered far more than half the performance of the 7950GX2 at half the price. What a great deal. Well, WE jumped on it at least.

So it is with heavy hearts that we announce to you today that there is a new monster out there: a video card that, without breaking a sweat, outperforms our trusty 7900GT by a factor of two. It is extremely rare in the hardware world that we see a product released that can ravage its predecessor two times over. As luck would have it however, it has happened twice this year. NVIDIA’s new GeForce 8800GTX, the incarnation of that “G80” core we have been hearing so much about, is that new card. And today, we take it for a spin.
{mospagebreak heading=Introduction&title=Specifications}

NVIDIA’s GeForce 8800GTX
Specs

With the insane amount of speculation and rumors circulating around the web about the specs of this card, let’s take a look at what they really mean.

DirectX 10 Compatible
Shader Model 4.0 Support
575MHz Core Clock
768MB of GDDR3 memory at 900MHz
384-bit memory interface (256-bit+128-bit)
86GB/s memory bandwidth
128 unified shaders at 1350MHz
Theoretical texture fill-rate at 38.4 billion pixels per second
HDCP compliant
Dual link DVI, VIVO, and HDTV outputs
Dual-slot cooling

Straight off the bat it is evident that this video card is unlike any other. Breaking down these specs step-by-step we see that the 8800GTX is DirectX 10 compatible. This is significant purely for the fact that it is the first such card to the market. Should DirectX 10 drop in a big way sometime in the very near future, gamers wanting to live the DX10 experience to its utmost will have no choice but to get a G80 based NVIDIA card, as ATI’s DX10 compatible cards based on the R600 core are still a couple of months out. Among other features, DirectX 10 will have this lovely thing called “Epic’s Unreal 3 Engine uses it”. To us that is reason enough to be giddy. Next on the spec list is Shader Model 4.0 support. SM3.0 is not very old itself. First implemented fully with NVIDIA’s 6 series, SM3.0 offered considerable advancements over SM2.0. SM4.0 is quite simply the next step in the progression of shader models. No other card is known to support SM4.0, so if a game drops using SM4.0, the G80 cards will again be the only way to go.

The next two bullets on the list are kind of normal…. Wait… 768MB of memory? This is far from a commonly used memory amount. Why does it use 768MB of memory instead of 512MB or even 1024MB? We have some ideas, but we can’t know for sure. We speculate that the extra 256MB of memory on top of the 512MB will probably act as some sort of buffer for one of the unmentioned features of the card, the 128-bit HDR + 16AA capability. This feature might be what puts NVIDIA ahead of ATI in terms of image quality, but only time will tell on that one since there aren’t really any games that have code for 128-bit HDR. The whole memory bus width deal is kind of weird too. 256-bit+128-bit? Huh? We can only assume this is for some other newfangled feature on the G80, namely NVIDIA’s Quantum Physics Engine. The name itself is kind of confusing, not only because it just sounds complicated but also because the whole “Quantum Physics” title is already kind of being used by a real life science. Technical documents that have been percolating across the internets have NVIDIA claiming that QPE will be a direct competitor to AGEIA’s PhysX. This isn’t something we are entirely convinced will work out too well, not after our recent interview with AGEIA’s CEO about their processor and the PhysX SDK. Since we doubt NVIDIA went out and coded a new proprietary physics middleware by themselves, we have to think Havok has played a role with QPE. In terms of hardware, NVIDIA has apparently allowed several of the 128 unified shaders on the G80 to carry out physics calculations. The debate over whether or not a specifically designed chip like the PhysX is more effective at physics calculations than a few shaders on a GPU has been mulled over for months already. We can tell you that if history is any indication, taking a page out of the GPU book will show that a dedicated chip is usually the way to go. We’re not taking any sides though, it is WAY too soon to speculate about which will come out on top.

The next bullet down concerns the memory bandwidth of the G80. While the listed 86GB/s (really 86.4GB/s but let’s not argue semantics) of bandwidth exceeds that of the predecessor by nearly 10GB/s, that is entirely the result of the significantly higher operating frequency than anything else. The 8800GTX is still using the tried and true GDDR3 memory modules, even though GDDR4 is already out and being used with impressive results on ATI’s flagship Radeon X1950XTX. Although the bandwidth on this X1950XTX is “only” 64GB/s, bandwidth isn’t everything. Easily the coolest spec on the G80 is that there are 128 unified shaders. A unified shader is one that is capable of executing both vertex and pixel shading instructions. Whereas in the past the number of pixel shaders and vertex shaders were carefully balanced and differed from card to card, unified shaders allow manufacturers to use more where more are needed. Imagine having a set of 10 dudes on either side of a wall. All of a sudden, a whole bunch of work is dumped on one side of the wall that will take the 10 dudes 2 hours to complete. Meanwhile, the other 10 dudes are just chilling on the other side of the wall, not only because there is a wall in the way, but because they speak a different language than the other 10 dudes so they wouldn’t be able to do the work that they are doing in the first place. Now imagine all 20 dudes are bunched together and all speak a singular hybrid language. That same work load is dumped on the 20 dudes and they are able to knock it out in 1 hour. If you just read all that and it made sense to you, congratulations. That, we believe, is the premise behind unified shaders. That’s enough about specs and features we think. Sure we didn’t go over HDCP, which is kind of a mixed blessing, but we’re anxious to get into the benchmarks. Aren’t you?
{mospagebreak title=The Card}

The Card
Taking a look at the actual card, you can see that it is quite big. The card is HUGE. It is not overly huge in terms of weight or even girth, but in length. This card is longer than your ATX motherboard. This card is longer than the long side of your favorite magazine. This card is 10.5+ inches. It is pretty long. It is so long that we dare say many users will run into problems installing it in their mid-tower cases, or on motherboards that make extended use of the area around the edge of the board directly behind the PCI-e x16 slots. The next thing to notice on the card is either the cooler or the power sockets, depending on… what you noticed first. The cooler on this card is nothing too far out of the ordinary. It consists of an oversized copper heatsink and a centrifugal fan. An interesting thing to note about this cooling solution is that there is only one heatpipe incorporated in the design. We typically see more elaborate heatpipe designs in cooling solutions these days, from NVIDIA in particular. This doesn’t really say anything about the effectiveness of the cooler, and seeing as how the card did not get overly hot during our testing, it certainly isn’t inadequate.

In terms of power, the G80 requires not one, but two 6-pin PCI-e power connections to operate nominally. An interesting tidbit of info that you more than likely will not find anywhere else but right here at GotFrag Hardware is that there was initially one 6-pin and one 8-pin power socket on the G80 in the early reference model. The idea for this 8-pin connector still eludes us however, as it was not a standard EPS12V connector, not that any looney would think to use that on a video card. Nevertheless, it was there and we should have some pictures up in the future to prove it if it hasn’t already been proven elsewhere. For the time being you can simply take a look at the back of the board where 8 leads can clearly be seen next to 6. We assume that the purpose of having multiple PCI-e connectors on the card is to distribute the load more evenly across the +12V rails of the power supply. The ATX 2.2 standard for power supplies allows for a maximum amperage on a single +12V rail of 20A. Since the maximum output of a single +12V rail is then theoretically 240A, far below the load consumption of the 8800GTX, splitting the load across multiple +12V rails (2 separate power connectors) is a pretty good idea. There is more to it than that, and in fact you can get more than 240W out of a +12V rail on the ATX 2.2 standard, but we believe that this is likely the basis for NVIDIA’s decision to use multiple power connectors.

While we would love nothing more than to dismantle this card and show you its guts, we will not be doing that today. You can see from the label on the fan that this card was manufactured by ASUS. Well, we did not receive this card from ASUS or NVIDIA, and it is not really ours to begin with. As confident as we are in our dismantling skills, we don’t feel comfortable with potentially destroying someone elses’ $550 video card. Also, since we did not receive this card from ASUS, and it is in fact using NVIDIA’s reference design, the manufacturer of the card is of little importance and will not be addressed further in this review.
{mospagebreak title=Test Setup and Methods}

Test Setup and Methods

Hardware Configuration

Case: Vigor Force
Power Supply: PCP&C Silencer 750 EPS12V
Motherboard: Intel Desktop Board D975XBX2 (Bad Axe rev2)
Processor: Intel Core 2 Extreme QX6700 (Varying MHz)
Hard Drive: Western Digital 2500KS 7200RPM, 250GB with 16MB Cache Buffer
Video: eVGA NVIDIA GeForce 7900GT KO Superclocked (580/1580) | NVIDIA GeForce 8800GTX (575/1800)
Memory: 1024MB (2×512MB) Corsair XMS2 PC2 6400 (TWIN2X1024A-6400)
Optical Drive: Lite-ON SHW160P6S05
Cooling: Vigor Monsoon II
Audio: Creative X-Fi XtremeMusic

Software Configuration

Motherboard BIOS: D975XBX2 Express
Operating System: Windows XP Professional with Service Pack 2
Video Driver: NVIDIA ForceWare Version 91.47 WHQL Certified (September 14 release) | NVIDIA ForceWare Version 96.89 (October 16 release)

We will be comparing the GeForce 8800GTX against an overclocked eVGA GeForce 7900GT KO in gaming and synthetic benchmarks. We chose the fresh off the press QX6700 Quad core processor from Intel for the sole reason that it is the same processor we most recently used with the 7900GT in the Kentsfield Performance Review. We kept the installed ForceWare 91.47 drivers for testing with the 7900GT and then used the 96.89 drivers with the 8800GTX, as these were the drivers that shipped with the card.

Methods

Testing methodology is pretty straight-forward on video cards. To test the 8800GTX we will use a sampling of synthetic and real-world gaming benchmarks. On the synthetic end, 3DMark06 will be used to get an idea of the potential performance of this card and it’s relation to the potential performance of the 7900GT. Since 3DMark06 is the only really good synthetic video benchmark available right now, with the exception of the outdated but still useful 3DMark05, it is the only synthetic benchmark we will be utilizing in this review. 3DMark06 makes extensive use of features and technologies found on modern video cards such as HDR, AA, AF, and SM3.0.

On the gaming side, we have selected six games that we feel represent the current market fairly well and are played by more than just one person in Siberia. These games will be tested at various resolutions up to 1280×1024. We know that we probably should have tested the card at a resolution of 1600×1200 and possibly above, but due to time constraints on this article and the seemingly depleted stock of 1600×1200 capable displays amongst local vendors, we were unable to do so. Rest assured: we will most definitely have a review of this card at higher resolutions for you sometime in the near future. That said, each resolution was tested in each game at least three times over, and averages were taken of the resulting scores.
{mospagebreak title=Test Suite, 3DMark06}

Test Suite

As mentioned above, we will be using a total of 7 applications in our test suite. They are:

Futuremark – 3DMark06
Infinity Ward – Call of Duty 2
Bethesda – Elder Scrolls IV: Oblivion
Relic – Company of Heroes
Id Software – Quake 4
Valve – Half-Life 2: Episode 1
Monolith – F.E.A.R

At this point it probably sounds like a broken record, but this list is shorter than what we wanted. It seems like every time we sit down to write a big-time review, our time is quite limited. Due to our testing procedures, benchmarking parts tends to take a pretty long time. It certainly isn’t ideal, but it is the price we pay for delivering truly accurate results. We also wanted to get this review out well ahead of the NDA, since we are not under the embargo on information regarding the G80, but sadly that was not really possible. We got this card just a few days ago, and we have literally been running benchmarks ever since. So then, enough with the bore and on to the score(s)!

Tests
3DMark06

As mentioned before, we will be using 3DMark06 to get a general idea of the capabilities of the 8800GTX. Although 3DMark06 scores are not entirely practical in terms of the real-world performance of a graphics card, it is, by far, the tool that more people use to compare their graphics subsystems with those of others. People often optimize their systems to maximize their 3DMark scores in an attempt to win bragging rights in a forum or other community. We thought it might be fun to engage in some friendly competition as well, so in addition to running tests at stock, we overclocked the system significantly (Kentsfield @ 3630MHz, RAM @ 419MHz) to get the highest score we could muster.

As you can see, the 8800GTX nearly doubles the score of 7900GT at stock settings. When the system was overclocked, the 8800GTX pulled even farther ahead. Our final score on this system of 12209 is pretty impressive, but we managed a score of 12931 on a different system that was also equipped with a G80 and a Kentsfield processor. That system made use of NVIDIA 680i chipset on the motherboard, and had various other optimizations, including an overclocked 8800GTX, that allowed it to score higher than our 12209. If the results of this test are any indication of the performance we will see out of the 8800GTX in the gaming tests, and we will tell you outright that they are, you may want to take a break and grab a glass of water.
{mospagebreak title=Quake 4}

Quake 4

Quake 4 is a game that is pretty heavily CPU limited on the low end. We have shown this many times in other reviews. Because of this, we don’t expect the 8800GTX to offer any tangible performance increases at high performance / low quality settings. However, the CPU limitation of any and all games tends to decrease with an increase in resolution, so let’s see if the 8800GTX makes Quake 4 at low settings even more CPU limited than usual.

Here you can see that while scores for the lower resolutions are essentially the same, the drop off at the higher resolution of 1280×1024 was diminished with the 8800GTX. Whereas the 7900GT experienced a drop of over 16 FPS, the 8800GTX suffered only a 6.7 FPS loss. These are not particularly impressive results since the numbers involved are relatively small, so let’s crank the settings up to high quality and see what the 8800GTX can really do.

From these results the performance advantages of the 8800GTX become far more distinguished. Not only is the 8800GTX beating the 7900GT by as much as 63%, the game becomes far less GPU limited. Without going into too much detail, the extent at which a game is GPU limited is evident from the slope of the line created when connecting the frame rates at various resolutions. Graphics cards will strain more to pump out high quality textures and lighting effects at higher resolutions than they will at lower resolutions, and as such they will experience a performance hit. Since the FPS of the game at this point is pretty much entirely dependent on the video card, it is then GPU dependent. These are some really impressive scores for the 8800GTX in Quake 4.
{mospagebreak title=F.E.A.R}

F.E.A.R

Like Quake 4, F.E.A.R is extremely CPU limited on the low end. As such, we don’t expect to see performance benefits with the 8800GTX at low details. Let’s see if the results in F.E.A.R are reminiscent of the ones we saw in Quake 4, with the 8800GTX making the game even MORE CPU limited than it already is.

Well, it looks like F.E.A.R is a little too CPU limited for the 8800GTX to have any affect in that arena. However, performance numbers DID increase by a good 20 frames across the board. We don’t know whether to attribute this gain to the 8800GTX, the graphics drivers or something else, but 20 FPS is significant enough to make a difference. Next, we tested F.E.A.R on medium settings to get an intermediate look at what would be happening on the high end.

At medium settings it is clear that the 8800GTX outperforms the 7900GT, that much is expected. However, the performance gap here seems pretty small. We can only assume that this is because medium settings only puts a small strain on the video card. So then, since we didn’t get any spectacular results in the middle, let’s crank everything up and see what happens on the high end.

Ahh there we go. A 79.4% performance increase with the 8800GTX in F.E.A.R on high settings is just the spectacular result we were looking for. Even at low resolutions the 8800GTX beats the 7900GT by at least 40 frames. It doesn’t matter if we’re talking about CS1.6 or CoD2; 40 FPS is a big difference.
{mospagebreak title=HL2:EP1, COD2}

Half Life 2: Episode 1

Our newly revised testing procedure for HL2:EP1 is a far better method than what we have used in the past. The three timedemos that we now use for our testing suite are situated throughout various levels in the game that offer dramatic lighting changes, lots of neato explosions, and an absurd level of texture details that are well suited for putting a sizable load on any video card. We started off testing the 8800GTX with HL2:EP1 on low settings, just as we did with F.E.A.R and Quake 4. HL2:EP1 is also a CPU limited game on the low end, although it is far more demanding on the CPU than either F.E.A.R or Quake 4, as is demonstrated by the lower frame rates.

As you can see, the performance advantage here is actually shifted somewhat toward the 7900GT, but the differences are really minimal. This isn’t anything out of the ordinary. Now let’s see what happens when we turn the settings up all the way.

These are some interesting results here. Unlike in other games, the performance with the 8800GTX did not really drop at all with increasing resolutions. This would suggest that the 8800GTX has taken a game that is clearly GPU limited with less powerful graphics cards, and has made it, in effect, CPU limited. So then, not only do we see a performance increase of as much as 49% with the 8800GTX, but we have also totally wiped out the GPU limitation of the game. Pretty impressive indeed.

Call of Duty 2

CoD2, though getting old, is still one of the most graphically intensive games in our arsenal. In our Kentsfield Performance Review, the 7900GT was taken almost to the point of “unplayability” (state of performing below 30 FPS) in CoD2. Not only is CoD2 graphically intensive with details turned up, it is also somewhat GPU limited on the low end. Running the 8800GTX should actually yield some significant improvements over the 7900GT.

As you can see, the 8800GTX yields a pretty solid gain of about 30FPS at all resolutions. You can also see that even with the 8800GTX, Call of Duty 2 is still GPU limited on the low end. Turning the details all the way up should give us more of the same results.

This time we have a pretty solid 20 FPS increase across the board with the 8800GTX. The G80 based card also gives a good 70% increase over the previous generation 7900GT on the high end in CoD2.
{mospagebreak title=Oblivion, CoH}

Oblivion

Oblivion is another game that will give a very good representation of the power of the 8800GTX at high settings. Like most of the other games we have tested so far, Oblivion is heavily CPU limited on the low end. The game, to put it simply, looks like crap with the settings turned all the way down. However, unlike F.E.A.R, the other game we have tested at medium settings, Oblivion starts looking pretty good right in the mid-range. As a result, we would expect to see a pretty decent performance advantage to the more powerful 8800GTX at medium settings. Futhermore, we would most certainly expect a huge performance increase from the 8800GTX at the highest settings.

Here we see an 85% performance increase with the 8800GTX at the high end. Breaking down the numbers further reveals that this difference is a result of an increase to 140 FPS from 75. In a game like Oblivion, you WILL notice stuttering at 75 FPS. The background textures and the lighting in the environments are far too detailed to not notice the choppiness of 75 FPS as opposed to, say, 100. The 8800GTX offers 140 FPS where the 7900GT offers 75. Being able to run Oblivion at the highest settings is one thing, being able to run it at 140 FPS is entirely another.

Company of Heroes

The final game of our 8800GTX review is a relatively new addition to our testing Suite. Company of Heroes represents the first of hopefully a few new RTS-style games that place heavy loads on the graphics subsystem of the host computer. The detailed explosions and battle scenes offer plenty of things for your graphics card to struggle with that might not otherwise be seen in typical FPS games. We ran the built-in performance measurement test because we feel that it gives a good sampling of the different effects that run rampant in Company of Heroes. We also like watching it, because that plane crash is pretty sweet. Also, Hans’ coffee "sucks".

At low settings the 8800GTX offers marginal performance increases at lower resolutions. However, as the game starts to become GPU limited at higher resolutions with low settings, the 8800GTX performs admirably well, suffering only a 28 FPS loss as opposed to the 55 FPS lost on the 7900GT.

At medium settings we can see that the performance gap opens right up, with the 8800GTX beating the 7900GT by a good 40 FPS across the board and by 56% at 1280×1024. If this trend continues there should be huge performance increases at high settings.

Yep. That is a 141% performance increase ladies and gentlemen. At high settings and high resolutions, the 8800GTX performs A LOT more than TWICE as good as the 7900GT in Company of Heroes. When the 7900GT was stuttering along at 35 FPS, the 8800GTX was flying at 85 FPS. The difference between these cards really is astounding, and Company of Heroes displays it quite well.
{mospagebreak title=Final Thoughts and Conclusions}

Final Thoughts and Conclusions
The GeForce 8800GTX is nothing special. That is, if you consider video cards that enter the market and totally wipe the floor with anything else available to be normal. The 8800GTX does that. It is abundantly clear that NVIDIA has something amazing on their hands with the G80 core. As mentioned in the introduction to this article, it is quite rare that a product entering the scene can perform twice as good as its predecessor. Not even dual-core CPUs, you know those CPUs with 2 cores, as opposed to those older CPUs with 1 core, offered twice the performance of their predecessors when they hit the scene. With multithreaded applications becoming more readily available, that point can be argued, but we digress. When Intel dropped the Core 2 family on the hardware world back in July, it DID double the performance of its Pentium-D predecessors in a few areas. With the performance capability of NVIDIA’s 7 series on the G70 core and its derivatives, it would seem unthinkable that they could double it with their next core, the G80. Well umm, they kind of did. It’s not exactly twice as good, but it certainly is pretty close.

Moving away from the subject at hand for a moment, we would like to discuss something that happened this year that we took to heart. Consoles, not computers, consoles, were acclaimed around the world as being capable of producing the most impressive game graphics ever seen. PC Enthusiasts cried foul at the notion that a console, based almost entirely around PC technology using parts from PC manufacturers, could possibly be superior to its more diverse cousin. Well we didn’t like the sound of it either. But we played some XBOX360 games at Best Buy and we played some PS3 games at E3, and it slowly started to sink in: these systems are quite capable of producing some killer graphics. But it wasn’t TOO big of a problem for us. We knew that if ATI and NVIDIA were able to produce graphics chips for the XBOX360 and PS3, respectively, that they surely would have something even more advanced up their sleeves for use in the PC. Well it turns out that they did, and NVIDIA’s came first. Today we are proud to announce that with the G80, the graphics technology crown has been stripped from those wretched consoles and placed squarely back on the head of the personal computer, where it belongs, of course.

But what about price. You get all of this graphics muscle, and indeed, all of this dead weight of a very heavy video card, but for what price? The GeForce 8800GTX, NVIDIA’s new flagship part, is expected to debut with an MSRP of $549.99. This sure does seem expensive for just a video card. But we have shown here today that it is well near twice that of a GeForce 7900GT. The price of a 7900GT is right around $250 at the time of publishing, so twice that would be what….. we’re not very good at math, but we think it’s about $500. Add in a substantial premium for a card that utilizes brand new technology, is DirectX 10 Compatible, Supports SM4.0, is really huge… and all of a sudden $549 seems like a bargain. Should this price be too much for you to bare, and don’t get us wrong, it more than likely is, NVIDIA’s GeForce 8800GTS, the next step down on the G80 ladder, will be available for $499. We are unsure of the performance of the 8800GTS, but we obviously expect it to be quite good too. Is it a lot of money to ask for just a single video card? Absolutely. Is it a reasonable price to ask for what is quite simply the single most powerful video card ever (for now)? Probably.

Popularity: 5% [?]

Support FPSLabs! Share and Enjoy: