FPSLabs : NVIDIA GeForce 8800GTS 512 Review

NVIDIA is not ready to ease up on their relentless pummelling of the high-end graphics market just yet.

Introduction

{multithumb}
A little over a month ago we posted our review of the GeForce 8800GT. That video card, based around the all new “G92” or D8P graphics core, was heralded very much unanimously as nothing less than a revolution in the GPU market. The card not only came out in time to steal the limited thunder that AMD’s RV670 (the graphics leg of their spider platform) was bound to accrue; it managed to offer near-8800GTX performance for less than one half the price as well. It also cannibalized a good amount of NVIDIA’s own graphics line, all but eliminating the 8800GTX as a viable option for the reasonable consumer and relegating the 8800Ultra to a land of absurd price/performance. But that wasn’t enough for NVIDIA. For some reason, and counter-intuitively to what you would expect from a business (after all, NVIDIA is a business, right?), the company wants to make damn sure that sales of their flagship 8800Ultra plummet to the point of irrelevance. But for that they needed a new product; perhaps one that would do what the 8800GT did the GTX. After all, NVIDIA can’t very well have AMD’s Radeon HD3870 masquerading as a reasonably-priced high-end graphics card for more than a couple of weeks, can they? By enabling the 12 extra shaders that were disabled on the 8800GT and bumping up clock speeds across the board, NVIDIA has managed to find a spot in their ridiculously robust range of high-performance video cards with the GeForce 8800GTS 512: a glorified 8800GT.

[ArticleAdvert]

Although our modus operandi would usually call for some sort of “but” statement right about now, it would be daft to even think about leading you away from the ultimate point this article is going to make. By this point in time you more than likely have read more reviews of NVIDIA graphics cards than you care to remember. You know damn well that right now, everything NVIDIA touches turns to gold. Well, you can also know that the GeForce 8800GTS 512 does not disappoint. You can know that with driver updates, the 8800Ultra and 8800GTX will become (almost) obsolete. But (there it is! Did you think we would leave it out?) when we take into consideration everything we know about NVIDIA’s plans for the next couple of months, about their platform updates, about their product launch schedule, and about the products themselves, there is one question we can’t seem to get around: Why? Without an NVIDIA contact to ask for a straight answer, we naturally tried to come about the reason for the 8800GTS 512 on our own. The first paragraph of this article is one possible explanation for its existence. The results of the wonderful FPSLabs testing you are about to read might just be another.

If you’ve not read our review of the 8800GT, it might be a good idea to at least give the first page a once over. Doing this will give you a better idea of our opinion of these new NVIDIA GPUs and the current graphics card market in general. Now then, on to the GeForce 8800GTS 512.

{mospagebreak heading=Introduction&title=Features and Specifications}

Features and Specifications

{multithumb popup_type=none enable_thumbs=0}
{multithumb popup_type=default enable_thumbs=default}

After careful deliberation, we have decided that quoting, paraphrasing, and drawing comparisons to parts of our 8800GT review is the most effective way of presenting the following information.

So if you’re like us, after looking at the specs you might be asking yourself, “How is this any different than the G80 of the 8800GTX just with 16 fewer shaders, different clock frequencies, and a smaller manufacturing process?”. Answer: No idea. Clearly the GeForce 8800GT’s D8P adds in compliance with PCI-Express 2.0, which we hear from industry insiders does very, very little to increase performance on the high-end. Other than that? Presumably nothing. However, the only thing this would have any bearing over is the name of the core. In the past NVIDIA and AMD have called updated versions of a family of GPU’s what would effectively be a G85 or a G87. OK, so “D8P” doesn’t make a whole lot of sense, but let’s take a closer look at what has changed from the original G80.

We still don’t understand the nomenclature here. With NVIDIA’s roadmaps indicating graphics cores following the “D8*” scheme we can’t help but feel a bit confused. Not that any of this really matters at all…

We held the unified shader architecture first implemented in G80 in high regard due to its all-around superiority to the separate vertex/pixel shader designs used in every (modern) GPU prior. The GeForce 8800GTX has 128 of these unified shaders. The GeForce 8800GTS has 96. The GeForce 8800GT is to have 112. This is a pretty straightforward and basically linear comparison to make: the 8800GT has 16 more shaders than the 8800GTS and 16 less shaders than the 8800GTX. This suggests that, all else equal, the card D8P-based card will perform somewhere in between its siblings. Interestingly, there are actually 128 shaders on the D8P core, but 16 of them have been disabled. When producing the actual GPU die, many of the units on the silicon wafer turn out to be partially defective. This is true in every brand of chip making, and is handled differently by different companies. IBM ran into this problem when producing Cell processors for Sony’s Playstation 3. The end result was the announcement that there would actually be fewer SPE’s on production PS3’s (a.k.a change the spec). AMD is reportedly running into problems with their upcoming Phenom processors. As a result they will bring 3-core processors to the market, which are effectively quad-core processors with one core disabled. This practice ends up reducing the manufacturing costs of the chip, and decreases the amount of chips that are thrown in the garbage. For the consumer, lower shelf prices are the result of creative rebranding and remarketing of less than perfect parts. NVIDIA’s approach to increasing usable yield is very similar: D8P cores with 112 shaders used on the 8800GT are just 128 shader cores with 16 disabled. More D8P cores per wafer can be used if 16 defective shaders can be disabled. There is nothing wrong with yield-improvement techniques, and certainly if it means lower prices for the consumer we are all for them.

OK so basically the 8800GTS 512 is the same thing as the 8800GT, just with an extra 16 shaders (ala 8800GTX).

Core, memory, and shader clockspeeds are all refreshed on the 8800GT. One benefit of a die shrink (65nm from 90nm) in any core is the ability to increase clock speeds with little to no additional heat overhead. We saw this when Intel moved its Pentium D line to 65nm, and we are about to see it again from them as they roll out Penryn, the long awaited Conroe refresh. The reference D8P board we sneaky snaked from NVIDIA is has a core clock of 600MHz, a memory clock of 900MHz (1800MHz effective), and a shader clock of 1.5GHz. All of these numbers are equal or beyond what stock 8800GTX numbers, 575, 900, and 1.35, respectively. Not only this, but several of NVIDIA’s add-in board partners will be shipping the 8800GT with factory overclocks, up to 660MHz on the core. This is a pretty large factory overclock, and is no doubt the result of a tremendous amount of headroom on the D8P core. We did not run any tests when our cards were overclocked, but we had two 8800GT’s running in SLI pretty comfortably at 690/1900.

Clockspeeds on the 8800GTS 512 have been raised even more than they were on the GT. The GTS enjoys a core frequency of 650MHz, memory frequency of 970MHz (1940MHz effective), and a shader clock of 1625MHz. Since the same D8P core is being used on the 8800GTS 512, it is safe to conclude that original assumptions about headroom were valid. We can tell you pretty definitively that the a bump in shader clockspeed of this amount on the G80 core would result in heavy artifacting and general chaos on the wrong chip and without excessive cooling. From what these stock clockspeeds indicate, the D8P is a more robust core than the G80. It should also be noted that numerous NVIDIA board partners have expressed interest in releasing 8800GTS 512 parts with factory overclocks. Since this happened with the 8800GT, we have no reason to believe it won’t happen again.

The last really big changes from G80 to D8P are the amount of memory and the memory bus width. On the 8800GTX, there is 768MB (512+256) of GDDR3 memory supplying the core through a 384-bit (256+128) bus. On the 8800GT, there is 512MB of memory (with a 256MB supposedly soon to follow) supplying the core through a 256-bit bus. The amounts here are basically the same if you disregard that weird extra stuff on the G80. I don’t know if anyone ever really figured out what the hell that was for anyway. However, numbers being numbers, there is a clear advantage here to the G80 based 8800GTX. The increased memory amount should lead to increased performance with games that use very high texture detail with very large environments (expect Call of Juarez and Oblivion numbers to benefit). On a side note, the 640MB of the 8800GTS also has more memory than the 8800GT, so we might expect that to make up for some of the ground lost as a result of fewer shaders and slower clock speeds.

Nothing to change here. The 8800GTS 512 has a virtually identical memory configuration to the 8800GT, the only difference being faster frequency.

[ArticleAdvert]

Now in terms of physical features, the 8800GTS 512 is a pretty unique card. Whereas the 8800GT is characterized by a sleek, single-slot fansink design, the 8800GTS 512 has a dual-slot fansink. However, the new design is not reminiscent to what you may have seen on the 8800GTX, 8800GTS, or 8800Ultra. The portion of the 8800GTS 512’s fansink that houses the actual fan is slanted inward, seemingly directing the flow of error toward the the GPU directly, rather than the traditional horizontally mounted design. Whether or not this design is beneficial to cooling performance is something that we unfortunately do not have the means to test, but although we are certain there will be reviewers out there that call this design attractive, we like to think that the design is a result of someone dropping the fansink mold during the final stages of production when it was too late to fix. Like the 8800GT, the 8800GTS 512 is powered by a single 6-pin PCI-Express connector (in addition to the PCI-Express bus). We were not provided with a press kit full of official information by NVIDIA, in fact we totally snaked our sample in the first place, so we can’t offer any exact numbers on power requirements. However, given the increased clock frequencies across the board on the 8800GTS 512, we would expect the “minimum power supply requirement” for this card to be somewhere in the 450W area, as opposed to the 8800GT’s 400W.
{mospagebreak title=Testbed and Methods}

Testbed and Methods
Test Setup

Case: Vigor Force
Power Supply: Cooler Master Real Power Pro 850
Motherboard: Foxconn N68S7AA
Processor: Intel Core 2 Extreme QX6700
Hard Drive: Western Digital WD2500KS 7,200RPM
Video: NVIDIA GeForce 8800GTS 640, GeForce 8800GTX (stock, reference design clocks), Diamond Viper Radeon HD2900XT 1GB, 2x NVIDIA GeForce 8800GT, NVIDIA GeForce 8800GTS 512
Sound: Creative X-Fi XtremeMusic
Memory: 2048MB (2×1024MB) G.Skill PC2 6400
Optical Drive: Lite-ON SHW160P6S05
Cooling: ASUS Silent Square Pro

Software Configuration

Operating System: Microsoft Windows Vista Home Premium 32-bit
Video Driver: NVIDIA ForceWare Version 169.09 (Vista 32-bit), AMD Catalyst Version 7.10

Believe it or not we were able to get our hands on two 8800GTS 512 cards. Unfortunately, we were unable to get said cards with enough time to spare before the NDA was lifted to provide you with a full review of the cards in SLI. You can fully expect a supplement to this article, or at least a mention of the performance of these cards in SLI in a future article. Also, we were not provided with drivers for these cards so we had to resort to rather unsavory practices to acquire the modified ForceWare 169.09 from a private FTP. It goes without saying that this is not going to be the driver provided for the 8800GTS 512 at launch, and it also goes without saying that performance differences with the new driver could very well be significant. And, once again, the only drivers we could scavenge were for Window Vista. We are not completely ready to migrate our benchmarks to Vista-Only, but that time is most certainly approaching.

All tests were conducted at least three times, with the average score being that reported in the graphs. For Futuremark’s 3DMark06, our only synthetic test, settings were left on default out-of-the-box values. When testing the Radeon HD 2900XT graphics card with 3DMark06, we had to add the –nosysteminfo tag to the target line in the shortcut because for some reason R600-based video cards STILL have a problem with that. In our game tests, built-in benchmarks were used for F.E.A.R., Call of Juarez, World in Conflict, and Company of Heroes. We recorded a custom timedemo for Team Fortress 2, which for various reasons will have to be revamped in the near future. Finally, Fraps was used for Bioshock and Oblivion tests. Detailed methodology for these tests will be provided in their individual sections.

Vsync was turned off for all tests. If there were any special kind of graphical quality settings used in any game, they will be noted in the section for that game or a screenshot of the settings will be provided. Below is a complete list of the test suite, some games were used twice to test different graphical APIs (DX10 vs. DX9).

[ArticleAdvert]

Futuremark - 3DMark06
Monolith - F.E.A.R.
Sierra/Massive Entertainment - World in Conflict DX9
Sierra/Massive Entertainment - World in Conflict DX10
Valve Software – Team Fortress 2
Ubisoft/Techland – Call of Juarez
THQ/relic – Company of Heroes DX9
THQ/relic – Company of Heroes DX10
2K/Bethesda Softworks – TES IV: Oblivion
2K Boston – Bioshock DX9
2K Boston – Bioshock DX10

{mospagebreak title=3DMark06, F.E.A.R.}Tests
3DMark06

It seems like in every video card review we write we criticize the fact that the industry uses 3DMark06 as the go-to standard for graphics benchmarking. The test is old by now, and it has various deficiencies that just make it not a very good test for modern cards. And yet, here we are, using it again. We really wish Futuremark would hurry and release 3DMark Next (replacement for 06) so we could get a more meaningful synthetic test running on new cards. Please keep in mind with these scores that we are running windows Vista 32-bit, which results in a score decrease from Windows XP. Also, most reviews you will see are using slightly faster processors than our non-overclocked QX6700.

{multithumb popup_type=none enable_thumbs=0}
{multithumb popup_type=default enable_thumbs=default}

[ArticleAdvert]

It’s times like these we really wish we could afford an 8800Ultra. Oh wait, not it isn’t. Here we can see that NVIDIA’s new GeForce 8800GTS 512 trumps the 8800GTX by a solid 200 points, making it the single fastest card in 3DMark06 that we have tested in this system. Not that that is saying a lot. At FPSLabs we tend not to care too much about benchmark results in a form other than FPS, as such results would indicate that the test is not necessarily representative of real-world gaming. As such, we only recommend you use 3DMark06 as a sanity check to make sure your video card is performing somewhere in the relative area of where you expect it to.

F.E.A.R. (v1.08)

You might ask why we still use F.E.A.R. for benchmarking. The game came out forever ago and most decently powerful cards handle the game just fine. Well, F.E.A.R. is extremely sensitive to minor changes in hardware. Even if the end FPS scores were not valuable (they are), the tweaking you can do to the image quality settings within the game options is invaluable for singling out various under-performing features on any given card. Also, F.E.A.R. seems to run exceptionally bad in Vista when compared to XP, so it adds an extra element of use to the game. Finally, the built-in benchmark makes it quick and easy ;).

{multithumb popup_type=none enable_thumbs=0}
{multithumb popup_type=default enable_thumbs=default}

Here you can see that the GeForce 8800GTX remains the top dog. We cant really offer a concrete explanation of why the 8800GTX outperforms the 8800GTS 512 in this test, but it is worth pointing out that the new card only barely beats out the 8800GT. This would suggest that the extra memory bandwidth (86.4GB/s vs 62.7) on the 8800GTX is more beneficial in a game like F.E.A.R. than are bumps in core and shader frequency.

{mospagebreak title=Call of Juarez}
Call of Juarez

The DirectX 10 Benchmark incorporated into the 1.1.1.0 release of Call of Juarez is a great way to test a graphics card’s ability to operate under DX10. There are a series of “tests” where the fly-through demo stresses various features that are new to DirectX 10. This is by far the most stressful test we can remember putting a graphics card through in the gaming space, and it is for a pretty good reason too. Call of Juarez in DirectX 10, and indeed the Call of Juarez DirectX 10 Benchmark, is one of the nicest looking games we have ever been allowed to play. What games do we think look better than Call of Juarez? Well, any game based on the UE3 engine besides Rainbow Six: Vegas like Bioshock, and of course Crysis, which very recently crushed the hell out of every video card in our labs at very high AND high settings.

{multithumb popup_type=none enable_thumbs=0}
{multithumb popup_type=default enable_thumbs=default}

Here you can see that the 8800GTX once again takes the top spot. Also, we can’t ignore the fact that both of NVIDIA’s newer cards perform a bit worse than the some of the older cards in the test. We can’t really attribute this to anything besides memory bandwidth, so there you go. Call of Juarez is a great benchmark because it is clear throughout the entire process that every single frame is hard fought. The test takes you on a flyby that goes through various areas to demonstrate the graphical prowess of the game. Below you will find screenshots of each of these areas, along with a shot of the settings.

{multithumb popup_type=none enable_thumbs=0}

[ArticleAdvert]

{multithumb popup_type=default enable_thumbs=default}

{mospagebreak title=World in Conflict, Team Fortress 2}
World in Conflict

Storytime. When the WiC beta was released, several GotFrag people were onboard to test out the game and give feedback. Before the beta was released, news came of a $10,000 WiC tournament at the 2007 Summer CPL. Being the ambitions sons of guns that we are, we created a WiC team called the Silver Snakes (tag: [S~], name from Legends of the Hidden Temple (Nickelodeon)). For a while we really dominated the unorganized, public server circuit. Several us got up really high in the ranking system even though we did not play nearly as much as some of the top ranked people. We then sort of became less dedicated and totally missed going to CPL, then Fry’s cancelled our preorders of the full version of the game and our WiC adventures were pretty much over. That said, we still love the game. It offers incredible gameplay, lifelike animations and effects (with the exception of the nuke), and the zoom in/out function works amazingly well. It is a tremendously fun game that you should all try out. Also, if you buy the Collector’s Edition box, it comes with an authentic (yeah right) piece of the Berlin Wall! All settings in this game were the default ones for “Very High”, with the exception of “shadows from clouds”, as that option does not appear to be available under DirectX 9.

{multithumb popup_type=none enable_thumbs=0}

{multithumb popup_type=default enable_thumbs=default}

[ArticleAdvert]

Once again we have the 8800GTX beating out everything else both DirectX9 and DirectX10 modes in World in Conflict. Scaling in WiC for SLI configuration appears to be very marginal, if not completely absent. The performance hit across the board when in DX10 mode is really not too bad when compared with other DX10 capable games (like Company of Heroes and Call of Juarez). It should be noted that WiC was very much playable on all of these cards when the noted FPS was above 30. WiC is an RTS game, and as such does not require super-high frames per second to run smoothly.

Team Fortress 2

Ever since the open beta became available to those who pre-ordered the game, TF2 has been a favorite of most of the GotFrag staff. We wrote an article about our impressions of the beta and both editors behind that piece pretty much loved the game. As a result of how awesome the game is we even changed our CS:S server to TF2 (we also have a CS1.6 server now, to be announced soon I expect). The game also runs on a much newer build of the Source engine than our previous entry from Valve, Half-Life 2: Episode 1. We recorded a custom timedemo on the cp_gravelpit map, but due to a really weaksauce internet connection at the time, we will have to re-record the demo under more favorable network conditions to ensure a stable framerate. All of the settings were set to the max, with AA and AF both at 8x. We did not click the Bloom (If Available) box for HDR bloom effect, and of course, vsync was disabled.

{multithumb popup_type=none enable_thumbs=0}
{multithumb popup_type=default enable_thumbs=default}

Since the graphics settings screen prominently displays the ATI logo and Valve and ATI have a long history together dating back to at least the 2004 Alcatraz launch of Half-Life 2 and the Radeon 9800XT, we would expect ATI to perform exceptionally well in TF2 or any Valve game. However, that does not appear to be the case, as the GeForce 8800GTX takes the top spot in our tests. The 8800GTS 512 turns in a good performance here, especially considering its D8P brethren basically bombed on this test. We feel somewhat inclined to throw out our results for the 8800GT in Team Fortress 2, as they seem inexplicably low. Either way, the new 8800GTS 512 beats out the only ATI card we had on hand to test, but the 8800GTX still reigns supreme.

{mospagebreak title=Company of Heroes, Oblivion}
Company of Heroes (v2.102)

We’re not the biggest fans of Company of Heroes or the built-in benchmark test that it offers. The game has always seemed very particular about which graphics settings you select and which resolution you run. The benchmark also freezes very often under Windows Vista, which is not a pleasant experience. Nevertheless, it is widely used by other publications and will continue to be used by us until a better candidate for our testing suite reveals itself. To set our graphics settings, we selected the “High” shader quality (DX9) and moved all fields to the highest available values. This would be our DirectX9 test. For DirectX10, we only changed the shader quality to Direct3D 10 – we did not make use of the extra settings that become available in other fields.

{multithumb popup_type=none enable_thumbs=0}

{multithumb popup_type=default enable_thumbs=default}

[ArticleAdvert]

Oh my gosh! What’s this? Results that actually make sense? Yay! Here we can see that they 8800GTS 512 is right on pace with the 8800GTX at lower resolutions, but drops off above 1600×1200. This makes perfect sense considering the smaller memory bus and the resulting decreased memory bandwidth. We can also see from these results that scaling in SLI on the D8P-based GPUs is good in Company of Heroes, as it is with other NVIDIA cards.

Oblivion

Our Oblivion test consists of repeatedly running along the coast to the right of the very first sewer exit. We cut across the little water inlet, jump up and over a bunch of rocks, and eventually end up jumping off a huge rock and into the water, where our 60 second recording length ends. Oblivion tests are ran at least 5 times per reported value to ensure the most representative frame rates are taken. Graphics settings are left on whatever clicking “Ultra High” in the opening options selects by default, but vsync is disabled and resolution is changed to whatever we want it to be. We do not enable AA or AF in Oblivion.

{multithumb popup_type=none enable_thumbs=0}
{multithumb popup_type=default enable_thumbs=default}

The 8800GTS 512 takes the top spot here just as we would expect it to. The 8800GT also triumphs over the 8800GTX, leading us to believe that higher clock speeds (and fill-rates) make a big difference in this game. The recurring trend at 1600×1200 and 1280×1024 resolutions of them producing very similar results seems to suggest that the game is CPU limited at these resolutions, which is kind of a weird thing to say when you consider how damn amazing Oblivion still looks on high settings (we’ve been saying this for more than a year now… we know). SLI seems to lead to increases only at higher resolutions, where the extra graphics memory probably does wonders for the vast landscapes portrayed on screen.

{mospagebreak title=Bioshock}
Bioshock

When Bioshock came out, a couple of us stayed up for like 17 hours or however long it took to play the game the whole way through. We started working on our own review of the game shortly after completing it, but progress on that has been very slow due to us attending so many events in the past month. However, if you ever get a chance to read IGN’s review of the game on PC we strongly suggest you do so, as it parallels our thoughts quite well. The game represents a level of artistic and graphical splendor never before seen on any game in any genre while maintaining pretty decent gameplay and a mindblowingly fantastic story. If you have not played Bioshock, make sure you at least download the free demo over the Steam network. It is certainly an experience we would recommend to every gamer, whether they are a first person shooter fan or not. The fact that it is cross-platform only enhances its awesomeness. All graphical settings are set to high and every effect is enabled except for “Force Global Lighting” because for some reason that is turned off by default, we suspect because it doesn’t make the game look any better and adversely effects system performance. Vsync has to be manually turned off with each game restart, which pretty much sucks but is bearable. We ran a set of tests in DX9 and in DX10. Bioshock in DX10 with full settings doesn’t hold a candle to Crysis at very high settings, but we will only consider that a deficiency when you show us a computer that can run Crysis on those settings without it being nearly impossible to aim in real-time.

[ArticleAdvert]

Our Bioshock test is a new one and we do it a bit different from other sites. Because such a crucial (and frequently occurring) part of the game is killing splicers, we leave enemies alive when we run the test. Our test occurs in the Medical Pavilion area of Rapture and starts right after you hack the flying security bot to dislodge the door. We then run and cap the chick in the back of the head and pick up her gun, then turn to cap the doctor dude on the stairs in his head as well. After flicking the security switch thing we wait at the door with the wrench in hand ready to beat down the chick in the hallway. After doing that with ease we run through the next door and pick up the machine gun and flip the other switch. We then book it back down the stairs to where the security bot room is, where there should be two splicers about ready to jump at you. If you time it right, you should be able to own both of them with one shot each and head up the stairs in the atrium. This is where you will find several splicers, some of which have pistols, the rest of which have a pipe or a pipe wrench. With a fair amount of skill you should be able to headshot these guys as well. After disposing of the enemies you should have about 10 seconds left of the programmed 60. We usually find our friendly security bot and destroy it, then proceed to shoot out the windows of the control room across the atrium. If time is not out by then we hold down fire on the machine gun until fraps indicates that the benchmark has completed. This process is repeated at least 5 times, as Bioshock demonstrates the same kind of variance as Oblivion when testing using fraps. We did, however, notice that our tests for Bioshock were fairly precise, as all 5 or more tests were usually within a few tenths of an FPS.

{multithumb popup_type=none enable_thumbs=0}

{multithumb popup_type=default enable_thumbs=default}

****It is very important to note that there was a mandatory update released for Bioshock very recently that resulted in huge performance gains. This update was not available when the majority of the tests you see here were conducted. As such, the results given here for the 8800GTS 512 are disproportionately high. It may be advisable to disregard this test altogether.****

Bioshock really eats up video cards like crazy. The game is based on the same Unreal Engine 3 technology that powers Rainbow Six: Vegas, Unreal Tournament 3, Gears of War, Huxley, Warmonger… and more. Nevertheless, the game is pretty playable at these frame rates, as lightning fast reaction times and lag-less gameplay are less important than technique and the use of all abilities when taking down big daddies.

{mospagebreak title=Crysis and Conclusions}
Crysis

When we heard about Crysis oh so long ago, we got pretty excited about the visual effects and realism it promised to provide. As the launch date grew closer, we heard from people within NVIDIA and ATI that early builds of the game ran smoothly on SLI and CrossFire hardware at high settings and ultra-high resolution. Then when the game came out and totally wrecked every computer that attempted to run it at high/very-high settings and a decent resolution, we were left questioning what the hell NVIDIA and ATI were talking about. Not only did the game look like a slideshow at ultra-high resolution and high or very high settings no matter what computer was being used, but SLI and CrossFire didn’t even work correctly until we got updated drivers. Whatever. For now our Crysis test consists of running the GPU_benchmark until it loops 4 times. We then take the first score out and average the remaining three.

{multithumb popup_type=none enable_thumbs=0}
{multithumb popup_type=default enable_thumbs=default}

We only tested the 8800GTS 512 in Crysis due to time constraints, but you can see here that anything over 1600×1200 is pretty much not playable. Crysis is tolerable at about 25 FPS, but if you seriously want to be able to aim and shoot at people, 35 FPS is about where you want to be. At 1280×1024 and high settings, the 8800GTS 512 will let you do just that. Aren’t you happy that your new $350 video card lets you run Crysis at 1280×1024 on high details!?!

Final Thoughts and Conclusions

[ArticleAdvert]

So we’ve got this 8800GTS 512 and it’s kind of in a weird position. It is definitely better than the 8800GT that we reviewed just over a month ago, and although it is close, you can definitely say it beats out the 8800GTX as well. But it still doesn’t dethrone the 8800Ultra as NVIDIA’s top performer. The price is also a bit weird. At somewhere in the $299 to $349 range, the 8800GTS 512 definitely constitutes what we would consider a worthy jump from the 8800GT. The new “GTS” (512) goes even further to eliminate the older GTS models from the list of video cards that a consumer should be looking at. However, looking at the price and seeing a 3 as the first number begs the question: why the 8800GTS 512? The high-performance sector is already being dominated by the 8800GT, which is supposed to be selling for $200 and offers near-GTX performance. If the 8800GTS 512 is pointed towards the ultra high-end market, then something doesn’t make a lot of sense.

One of the things NVIDIA is touting as of late is 3-way SLI. Having worked briefly on a system with 3-way SLI, we can tell you that you definitely need two SLI connectors on the top of each card. That means that only the 8800GTX and 8800Ultra are capable of operating in this configuration. We know that NVIDIA will be releasing D8E (E for Enthusiast) cards sometime in January, so perhaps this whole argument is invalid. When 780i eventually (re)launches, 3-way SLI will be tossed into the limelight and have only two really old cards at its disposal. Supposing NVIDIA was putting 780i back out at the same time as the D8E parts arrive, then everything makes sense, but that’s not the latest we’ve heard.

It’s all kind of confusing, as new tech launches always are. Consider this: your average Joe is walking down the aisle in a computer store and sees all three “8800GTS” cards lined up. Logically, he would think that the 640MB is the best card. But it’s not. The pricing won’t help him out either. What isn’t confusing, however, is that the 8800GTS 512 beats the 8800GTX in a lot of tests on very early drivers and can only get better (widen the gap) over time. At $350 max (for an overclocked version of the card we assume), it is clearly a much better option than the 8800GTX. Unlike the 8800GT before it, which totally cannibalized the 8800GTX, the 8800GTS 512 fails (if the word ‘fails’ can apply to a card this good) to gobble up the 8800Ultra at the top of NVIDIA’s lineup.

Popularity: 49% [?]

Support FPSLabs! Share and Enjoy: