AMD Radeon HD 3870 X2 Review

AMD’s R680 is a huge graphics card with huge potential. Does the HD 3870 X2 mark a return to competition for the Green Team?

Two is always better than one, right? Nevermind the performance decrease that occurred when SLI technology was enabled in its infancy. Forget about how much data is lost in high-performance, high-heat RAID-0 arrays. Don’t even bat an eye when huge computer cases are released touting two built-in power supplies. While doubling up has been the source of a fair amount of trouble in the hardware industry, it seems to most often provide a performance increase when all is said and done. But what happens when it doesn’t? What happens when a company spends thousands if not millions in R&D reworking and refining a product that ends up flopping when it hits the market? You can rest assured that it has happened in the past, and even though the companies responsible managed to persevere and today are doing quite well for themselves, one needs only to look to AMD and their infamous Quad FX platform to realize just how big of a risk doubling up can be.

It’s not that AMD’s Quad FX platform was inherently bad, or that their competition’s multitasking was a lot better, but that the way the technology was implemented led to other problems that a vast majority of buyers must have considered a pretty big deterrent. In a world where die shrinks and increased efficiency initiatives would have you thinking that perhaps the emphasis has finally switched from ‘bigger and better’ to ‘less is more’- a delusion if you really take a look at what’s going on – it is always curious when a company decides it necessary to release something totally against the trend. And yet, we have seen these products – many of them – released in the past year alone. That’s why more than a few heads were turned when “R680” showed up on AMD roadmaps and details began to emerge in the second half of 2007.

R680, otherwise known as Radeon HD 3870 X2, is AMD’s third attempt at a dual-GPU graphics card. ATI’s Rage Fury MAXX launched way back in December of 1999 to a generally mixed field of reviews. Many reviewers applauded the exceptional DVD playback and high-resolution 32-bit gaming abilities of the card while criticizing its astronomical price tag (at launch). Perhaps unlike today, however, ATI’s MAXX was not a direct port of a pretty well-designed and respectable existing GPU. Basing a new product on halfway decent technology rather than a poorly executed amalgam of antiquated parts is probably a pretty good idea. The Radeon X1950 Pro dual was another dual-GPU graphics card, but it saw very limited exposure in the press and presumably limited success in the presence of genuinely high-end cards at the time like the GeForce 7900GTX and Radeon X1950XTX (not to mention the fact that it only fit in the very biggest cases). The Radeon HD 3870 X2 on the bench today is a very early sample of the card and the drivers even more so. That’s not to say the numbers you’re about to see aren’t accurate, it’s just we wish we could have tested it closer to launch with proper software support.

Before getting down to the nitty gritty, let’s take a look at the current graphics card market so we can figure out where the Radeon HD 3870 X2 fits in. The past few months have been pretty interesting in terms of new graphics products. Launches from both AMD and NVIDIA have effectively doubled the number of video card genres for better or worse. Whereas we used to have the enthusiast-level 8800 series from NVIDIA that may or may not have had company in the form of the Radeon HD 2900 XT, depending who you asked, we now have 6 or 7 different cards filling the same performance segment. In fact, there are 4 different NVIDIA cards that could make their case for being enthusiast-level parts. The GeForce 8800Ultra, 8800GTX, 8800GT and 8800GTS, while radically different in price, all offer performance that could legitimately be called enthusiast level. However, instead of just accepting all four cards as enthusiast grade products, hardware analysts and consumers tend to break them into their own separate market segments. At the very high-end we have the 8800Ultra, with its unbeatable performance record and astronomical price point.

Below that in a sort of upper high-end area we have a card that is more than a year old in the 8800GTX. Where things get really interesting, however, is just under 8800GTX in the high-end market where the 8800GTS 512 and 8800GT both reside. What makes both of these products so interesting is the incredible price differential between them (specifically the 8800GT) and the two cards situated just above them. At anywhere from $210 to $250 USD, the 8800GT offers really similar performance to the 8800GTX at half the price. Needless to say, the card has been flying off shelves everywhere and it is pretty hard to find it on sale for the suggested retail price. Accompanying the 8800GT and 8800GTS 512 in the high-end area are two or three different cards from AMD. The old guard Radeon HD 2900 XT can still hold its own against the 8800GT, as can the new Radeon HD 3870 and the hybrid Radeon HD 2900 Pro. Below these cards in the mid-range everything goes back to normal, with a limited number of offerings from both companies offering impressive price/performance ratios.

So where does the Radeon HD 3870 X2 fit in to all of this? Well, at a price point of $449, we would expect it to compete with the 8800GTX. Then again, the performance of two HD 3870’s in CrossFire trumps the 8800GTX quite handily most of the time, and since the HD 3870 X2 is pretty much the same thing, we might actually have a card from AMD that is reasonably priced and performs great. What’s more, if the HD 3870 X2 performs better than the $600 8800Ultra, the new card might just represent a price/performance revolution on par with the one the 8800GT pulled just a few short months ago.

Features and Specifications

See what we did there? Without the inclusion of those pictures of the card, the specifications alone probably would not have made a whole lot of sense. Looking at the pictures however, it is quite clear that the Radeon HD 3870 X2 really does have 2 physical GPUs and a full 1GB worth of memory. Unlike the 7950GX2 and the upcoming 9800GX2 from NVIDIA, AMD has managed to put all of that graphics horsepower on one PCB. What’s more, they’ve done it without being too excessive. Before the card found its way into our hands, we fully expected the product to be just as humungous (if not more humungous than) as the HD 2900 XTX cards we saw back in April 2007 that never made it to market. The card is certainly not small, but it’s far more compact than we had imagined possible.

It all starts to make sense when you compare the fansink of the HD 3870 with the one on the X2. The heatsink underneath the plastic shroud on most HD 3870 cards is basically just a chunk of copper fins for the GPU and a few smaller tendril areas that are supposed to take care of the memory chips and other ICs on the PCB. The X2 has essentially the same thing… times two. The biggest difference between the cooling systems on the two cards – besides one being basically twice the other one – is the actual fan itself. On the HD 3870 the fan is centrifugal of course, but more of a noise-centric design. The fan is designed to move decent amounts of air across the convection fins on the heatsink effectively but with speed and noise generated being primary concerns. The fan on the X2, on the other hand, is a balls to the wall volume fan designed to push as much air as possible with little regard for noise or speed. Since the X2 fan is a much more open design, more air can enter through the intake. The orientation of the blades and their size, however, mean it has to spin ridiculously fast to be effective. And it’s pretty effective. Not once during our testing did the HD 3870 X2 clearly overheat. Of course, the tradeoff is noise. Our testing occurred in an office adjacent to one of the main railroad arteries in southern California, and it was not uncommon that the sound of the passing trains were drowned out by the X2’s fan when under a full graphics load. Not exaggerating.

Going down the list of the actual specifications for the card there are a couple of things that stand out. First of all, AMD claims that the X2 possesses over 1 TeraFLOP of GPU compute power. Unfortunately, GPGPU applications that actually take advantage of all of this processing power is still nowhere to be seen. While AMD’s R520 and R580-based graphics cards are being used as Folding@Home supercomputers with tremendous results, widespread implementation of similar applications seems to be an afterthought. Nevertheless, there is no reason to doubt that the X2 actually has over 1 TeraFLOP of GPU compute power, considering G80-based cards from NVIDIA in multi-GPU configurations can do pretty much the same thing (and actually do something in Tesla form). On the official specifications sheet, AMD also lists “Game physics processing capability” as a feature for the X2.

This so-called GPU-based physics topic is getting pretty old at this point. Both companies have touted the physics processing capabilities of their GPUs for more than two years by now, and so far nobody has seen any real-world implementation of the technology. Talk is cheap, and while AMD and NVIDIA are listing these features without showing us the goods, AGEIA is left to completely dominate the physics hardware market.

Noteworthy on the specs list is the 640 stream processing units. This is a pretty obvious specification, considering the X2 is just two HD 3870s, which themselves have 320 stream processing units. The X2 also brings to the table AMD HD 3000 series features such as DirectX 10.1 and Shader Model 4.1 support, as well as CrossFireX scalability and 128-bit HDR rendering. Probably the most important thing on the list is the memory interface. The X2 is listed as having “2 x 256-bit memory interfaces”, which effectively means the two cores and each memory bank operate independently of one another (insofar as two cards in CrossFire operate independently of one another). It would be brash to speculate about the effect this design will have on overall performance, so we should probably just let the numbers do the talking.

Something you won’t find in the specs is that this video card will work on any PCI-Express X16 capable motherboard. It was widely assumed in the early-going that the card would only work on motherboards that support CrossFire technology. In reality, however, the CrossFire option does not even become available in the Catalyst driver when using this card. All CrossFire connectivity appears to be handled internally, and the computer recognizes the X2 as a single video card.

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 8800 GTS 512MB, AMD Radeon HD3870X2
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 169.01, AMD Vista Sample 8.45 RC4

By now our video card repertoire has grown quite large. However, you may noticed the absence of the Radeon HD 3870 from the list. This is a problem that we are working on resolving by building a near-identical testing platform using a CrossFire-capable motherboard. This will allow us to comprehensively test AMD’s video cards, which up until this point (release of RV670 and R680) have not been competitive with NVIDIA at the high-end. We received this card very shortly after AMD started sampling to board partners, and as such the driver we managed to download from the AMD ftp was an early build of what we believe wound up being Catalyst 7.12.

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 and HD 3870 X2 graphics cards 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, 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. More astute readers will note the absence of World in Conflict from our test suite this time around. We encountered a graphical error when trying to run WiC on the X2, which we are fairly certain can be attributed to driver problem. These driver incompatibility issues are relatively common with new graphics card releases, and we fully expect this one in particular to be resolved come launch. The benchmarks presented in this review are Vista exclusive, as we were unable to secure XP drivers in the short time we had with the card.

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).

Futuremark - 3DMark06
Monolith - F.E.A.R.
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

3DMark06

Every time we fire up 3DMark06 to test a video card we do so with perhaps slightly less than unbridled enthusiasm. At this point we use the test as a “sanity check” only –that is to ensure the card is operating correctly and that general performance is somewhere in the ballpark of what we expect it to be. 3DMark06 actually becomes useful when quality settings and resolution are cranked up, but when you do that the results are automatically meaningless unless a good amount of other websites use the same configuration. On the same token, we are eagerly waiting for more sites to adopt 3DMark Next to their testing suites, as having a widely-used, Futuremark-written test in your arsenal is generally considered a necessity by readers (at least we think?). That said, our 3DMark06 test is run “straight out of the box” at 1280×1024 resolution.

The results here are not particularly surprising, with the new Radeon HD 3870 X2 trumping all competitors by a pretty good margin. Given the emphasis that AMD seems to somehow place on synthetic benchmarks (read: HD 3870 3DMark06 scores disproportional to real-world game results), we fully expected the X2 to do well. What is interesting, however, is that the X2 also managed to beat out the 8800GT in SLI configuration. If this was a strong indication of how the X2 would perform in game tests as a dual-GPU solution then we would come right out and say these results are very promising, but it’s not, so we won’t.

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

Again the results here are not too surprising. Multi-GPU configurations from both companies tend to scale very well in F.E.A.R., so the dual-GPU X2 does expectedly well in this test. As a single-card solution (and we have serious reservations calling it that, but meh) the X2 beats out the 8800GTX by a pretty significant amount. As a multi-GPU solution, the X2 produces some solid results but is no match for the 8800GT SLI setup. It should be noted that F.E.A.R. remains playable at all resolutions with all cards tested, which is saying a lot considering the game still looks pretty good when everything is set to maximum.

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.

Here we see the tremendous graphical horsepower of the HD 3870 X2 absolutely demolish the competition. The X2 manages to pump higher frames at 1920×1200 resolution than any other card we’ve tested does at 1280×1024. The new AMD cards based on RV670 tend to do fairly well in Call of Juarez by themselves, so it is not too surprising to see the X2 pull these kinds of numbers. Though impressive, we’ve seen numbers well in excess of these in our preliminary testing of 3-way SLI on 8800Ultras. Then again, this article is not about 3-way SLI -the X2 takes the top spot by a landslide in CoJ. 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.

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 FPSLabs 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 for those of you who have a hard time letting go). 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.

Before everyone starts yelling about these scores, let us just say that we probably ran each test for this card about 20 times to make sure we weren’t doing something totally stupid. The fact that we got higher scores at 1920×1200 resolution than we did at 1280×1024 really only points to one thing: at the settings with which we test the game, the Radeon HD 3870 X2 turns Team Fortress 2 into a CPU limited application. Regardless, no other card we have tested manages to cause this kind of behavior from TF2, which can only mean the HD 3870 X2 takes the top spot in this game as well. Perhaps all that canoodling on Alcatraz back in 2004 is finally paying dividends for AMD/ATI.

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.

Unfortunately it looks like the HD 3870 X2 suffers from AMD’s tradition of sucking at Company of Heroes. The X2 narrowly beats the 8800GTS 512 and loses substantially to the 8800GT SLI setup in DirectX 9, and adds the 8800GTX to the list of superior-performing solutions when we switch to DirectX 10. Granted, this probably has something to do with the whole “NVIDIA The Way It’s Meant to be Played” thing plastered across the Company of Heroes startup, but still. There are a lot of Company of Heroes faithfulls that I am sure would love to see a performance increase with AMD cards, even though the game is an RTS and high FPS is not exactly a prerequisite of success.

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.

The HD 3870 X2 doesn’t do particularly well in Oblivion, losing to both the 8800GTS 512 and the 8800GT. The card does manage to beat out the 8800GTX, which leads us to believe the higher clock speeds (and fill-rates) of the 8800GT and 8800GTS 512 make a big difference in this game. Interesting to note here is that Oblivion seems to be CPU limited at lower resolutions, and only really starts to stress the graphics card to a higher level at resolutions above 1600×1200. Hmm… perhaps we will have to fill out an expense report for a 30” monitor…

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.

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.

So there was this mandatory update to Bioshock that increased performance in general across the board for all graphics cards, but for some reason the 8800GTS 512 seemed to benefit particularly well from this patch. The HD 3870 X2 can’t even hold a candle to the (single) GTS at lower resolutions, and still comes in second at 1920×1200. At no point during testing did frame rates drop to a point we would ever dare consider unplayable, but things were noticeably smoother on the 8800GTS 512. It is very possible that the X2’s performance in Bioshock will increase with more mature driver releases, as AMD has a history of manufacturing significant performance gains via software.

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.

Crysis might be a pretty fun game that everyone should experience at some point if only for its pure, unadulterated graphical prowess, but it would be so much better if it didn’t completely destroy every system we try to run it on. Our scores for the HD 3870 X2 seem extremely low to us, so we advise you take them with a grain of salt. The low scores are likely the result of immature drivers, as Crysis was only released a few short weeks before the production of this driver version. Even still, it is strange to see the seemingly far less powerful 8800GTS 512 perform so much better than the X2 across the board. Neither card allowed for playable frame rates at high resolution and/or detail, which totally sucks.

Final Thoughts and Conclusions

In the introduction we said that if the HD 3870 X2 performed better than the 8800Ultra then, considering price points, its release would effectively be as impressive as the launch of the 8800GT. Well, what we’ve seen today is a clear indication that the R680-based video card is definitely superior to the 8800Ultra, which maintains a street price of well over $600. MSRP of the HD 3870 X2 is set at $449 –a bargain considering the HD 3870, essentially half of the X2, goes for about $230. Now, given AMD’s recent history of availability woes, we don’t expect very many of these cards to be available for $449. Then again, a trip to the major e-tailers doesn’t exactly produce too many results for 8800GT’s at MSRP. All of these things considered, it is hard to gage whether or not the HD 3870 X2 will be successful in the marketplace.

But that’s only due to the inevitable limited availability and subsequent price gouging –the HD 3870 X2 would sell like hotcakes if success was based on performance alone. To say the X2 is an enthusiast level card would be a drastic understatement. Our tests have shown that the Radeon HD 3870 X2 belongs at the very top of the graphics performance chart. The R680-based part is clearly the very best single-card solution available, and will remain there at least until later in the quarter when NVIDIA releases their product refreshes. Ahh, but there’s that nagging “single-card solution” term again. Is the HD 3870 X2 really a single card? Surely it is more of a single card than the GeForce 7950GX2 was a single card. It only takes up two slots, just as a typical single card would. Then again, it certainly consumes the power that a dual-card setup would, and it definitely makes enough noise and produces enough heat to fool people into thinking it’s really two cards. When referring to the X2 as a “single card” to AMD reps, they replied, “well, it’s two GPUs…”, alluding to the fact that calling R680 the best single-card solution available is somewhat of a questionable statement in and of itself.

Whatever you call it, it most definitely has earned AMD the graphics performance crown that it has been craving for well over a year now. This is a pretty significant feat considering the cadence with which the crown has changed hands recently. The Radeon HD 3870 X2 is an exciting product, not only because it provides incredible performance, but also because it marks the first time that a dual-GPU solution has been implemented successfully to the point where people will actually consider buying this card for reasons other than novelty. It marks the first time that two graphics cores have actually been married onto one PCB without bringing forth all kinds of scary (and material) heat and power concerns. NVIDIA’s dual-GPU solution should be coming out pretty soon, and it will be interesting to see how the GeForce 9800GX2 stacks up to the HD 3870 X2 on the performance and price fronts. Things are definitely heating up in the graphics world, both figuratively and literally, and the next four or five months should mark the return of competition in the high-end graphics market.

Popularity: 100% [?]

Support FPSLabs! Share and Enjoy: