It is easy to get caught up in the frenetic battle for the top video card slot. AMD and NVIDIA are constantly slugging it out to see who reigns supreme in the graphics arms race. The media side of this battle is fought with the Enthusiast-class or "Halo" cards. These cards often come to the ring wearing a price tag of $500+ and represent each companies’ no-holds-barred attempt at graphic domination. While this battle makes for great headlines and tag lines for the PR guys, the reality is that neither company really sells that many of these cards and truth be told, they never intend to. For your information, $500 boards account for three million cards sold each year, while the overall market is just shy of 80 million board mark.
The battle for supremacy in the $200 range of video cards is somewhat akin to a Cold War, its constantly brewing but not always in the forefront of everyone’s mind, that space is instead reserved for the Enthusiast-class cards with their $500+ price tags. The truth of the matter is that the $200 price range holds a little something for everyone. AMD and NVIDIA focus on it because the price is low enough to ensure significant volumes of cards are sold, yet high enough to guarantee a solid profit. The benefit for consumers is that we get a card with solid performance and many of the features of their larger siblings, but at a fraction of the price.
The two rivals in the $200 space have been out for some time now and as such the numerous add-in board partners have had a chance to apply their own secret sauce to the reference design in an effort to develop the perfect card. Today we will be looking at two very competitive cards from MSI, the NVIDIA-based N460GTX HAWK and the AMD-driven R6850 Cyclone. These two cards represent AMD and NVIDIA’s primary weapons in the mainstream graphics battle and as both cards are an obvious departure from the standard reference-design offerings they should provide for an interesting comparison.
MSI is quick to tout the overclocking ability of both cards and has built up certain features of these cards with that purpose in mind. Aside from custom cooling solutions, both cards feature what MSI calls Military Class Components. MSI’s own description of the Military Class standard is as follows:
"MSI graphics cards adopt the high quality components that meet the Military Class standard, ensuring the best stability, longest life span and no buzz noise under full load."
There are three main components that are featured as Military Class, these include Highly-conductive Capacitors, Solid Capacitors and Super Ferrite Choke. It’s quick to see that the core of this military class standard basically relates to power and more specifically power stability and efficiency. The Highly-conductive capacitors tout a Tantalum core (what, no Adamantium? Seriously disappointed, Ed.) and 15% less leakage, while the Solid Capacitors boast a 10 year lifetime and no explosions of which I would have to say the 'no explosions' appears to the better feature.
What is interesting to note is that the packaging for the N460 GTX Hawk lists the card as having Military Class Components while the packaging for the R6850 Cyclone specifies the card as having Military Class II Components, however their does not seem to be a difference in the actual components listed. I have never heard of a Military Class standard when it comes to video cards but if you are heading to a war zone then evidently these are the cards you should take with you.
The first thing you notice about any graphics card is the packaging and these MSI cards are no different. Flashy graphics and bold performance claims abound, not to mention the fact that there are enough logo stickers on the front to make any 4 year giddy with excitement. Both offerings feature the now-common front flat box with plastic window that allows you see that there is indeed an actual video card inside the box, should there have been any worry that you were buying an empty box.
The interior of the box is similar for both cards. The cards themselves are protected inside an anti-static bag and nestled securely inside a custom foam cutout. Additional adapters are placed in separate compartment next to the card while all documentation, software CD and the like are located beneath the foam tray surrounding the graphics card.
The software bundle included with our retail review samples was similar for both cards and can be described as modest at best. A single disc is included with each card, the contents of which include the requisite drivers as well as MSI’s Afterbuner utility, no bundled games to be found here. A recent check of Newegg.com does show that there is a current promo offering a coupon for a free copy of Just Cause 2 with the N460GTX or a $10 gift card with the R6850.
This is not to detract from the cards themselves as I know when it comes to purchasing a new graphics card, the included software bundle does not even rank in the top five of my deciding factors. The included MSI Afterburner utility is actually a great bonus. This handy utility allows you to track all of your cards vital signs, from fan speed to Memory load to GPU temperature and more. When it comes to overclocking, Afterburner is easily the go-to app. Everything is laid out in a visually appealing GUI interface and adjusting speeds and voltages is simply a matter of dragging a slider bar, with all values taking affect in real time once the Apply button is clicked. The utility also offers the ability to save overclock profiles which is an incredibly handy feature as it allows you to quickly switch between different performance scenarios at the click of a button.
One of the immediate indications that these cares are no mere reference offerings is their outward appearance. Each card features a custom cooling solution and it is that cooling solution that gives each card its unique look.
The full name of the AMD powered card is the MSI R6850 Cyclone IGD5 Power Edition, quite a mouth full indeed. One of the most predominant features of the card and therefore part of its namesake is the Cyclone cooling system. The heart of the Cyclone system is essentially a heatsink/fan combination which in itself is nothing incredibly new to the world of graphics card cooling. What is interesting however is how MSI has chosen to go about implementing their heatsink/fan setup. The Cyclone features one of the largest single fans I have ever seen on a video card. The benefit of such a large fan is that it can move larger volumes of air at a slower rotational speed, in turn creating much less noise. This large PWM fan is mated to what MSI calls a Hybrid Heatsink. The heatsink combines aluminum extrusion, heatpipes and fins to complete it cooling tasks. The two thick heatpipes draw the heat out of the GPU plate into a fin array that encircles the cooling fan, effectively moving the heat into a prime location where it can be quickly dealt with by the fan. As if that weren’t enough the top of the GPU plate is adorned with a circular heatsink fin array that sits directly below the cooling fan, allowing it to actively combat any heat the heatpipes may have missed. Its clear that MSI has done it’s homework on the Cyclone cooling system.
When it comes to cooling, the N460GTX Hawk boasts MSI’s familiar Twin Frozr II solution. Unlike the single cooling fan of the Cyclone cooling system on the R6850, the Twin Frozr II sports twin 80mm cooling fans mounted side by side, a feature that MSI claims will provides 50% better efficiency and direct airflow to not just the GPU but also the memory and power module. The N460GTX’s Twin Frozr II cooling also boasts four heatpipes, with two heatpipes diverting to fin arrays under each fan. The cooling system features somewhat of a shroud surrounding the fans and heatsink combination giving the card more of a buttoned-up appearance than the open air look of its competitor.
Both cards sport the standard Dual-DVI setup, albeit in different configurations as the DVI ports on the the R6850 are grouped vertically while the ports on the N460 GTX are arranged horizontally. HDMI is also standard on both cards, however the N460GTX chooses to use a mini-HDMI connector while the R6850 features a standard size HDMI output. The R6850 does take things a step further by including DisplayPort and given AMD’s proven track record with DisplayPort and it’s necessity for Eyefinity this does not come as much of a surprise.
Modern video cards are power-hungry and these two are no diferent with each recommending a 450W power supply or greater. You have to get the power to the card somehow and 6 pin PCIe connectors are the norm. The N460GTX requires two 6-pin connectors while the R6850 only needs one. The R6850 connector is conveniently placed on outside edge of the card which lends itself well to easy plugging and unplugging. On the other hand the N460GTX’s connectors are nestled underneath cooling shroud on the rear of the card in an orientation which generally has them facing the front of the case. This may not seem like much but the placement of these connectors does make them more difficult to plug in and the close proximity of the cooling shroud makes them downright annoying to unplug. While this may be a one-time issue for most users, enthusiasts that find themselves spending more time under the hood of their PC can become quickly aggravated by the setup… take it from a reviewer who is constantly swapping cards.
The reference design versions of these cards were designed from the ground up to compete with each other so it comes as no surprise that the two competitors share some fairly close numbers when it comes to specifications. When compared to reference designs from AMD and nVidia, MSI's parts are factory overclocked; The GPU chip on N460GTX operates at 780MHz (nVidia reference: 675MHz), while R6850 operates at 860MHz (AMD reference: 775MHz). As you can see here, MSI clocked both parts around 10% higher than stock clock: N460 got 13.5%, R6850 got 10% boost. Both parts feature 1GB of GDDR5 high-speed memory but the difference is quite big: while the N460GTX operates at only 900MHz QDR, R6850 shows who created the GDDR5 standard in the first place (AMD team lead by Joe Macri) - 1.1GHz QDR. In terms on numbers, N460GTX gives you 112.5GB/s to play with, while R6850 pulls in 137.5GB/s.
N460GTX Hawk features 336 cores ticking at 1.56GHz, while R6850 Cyclone has 960 cores ticking at 860MHz. Do note that the way how both architectures work in regards to cores operation is entirely different, thus it is hard to make an apples-to-apples comparison.
Benchmarks by definition provide a measure of performance, the challenge however is selecting benchmarks that properly recreate the intended usage scenario each product. Additionally we strive to provide a balanced mix of both synthetic benchmarks and real world gaming scenario benchmarks.
The scope of this review is the battle for supremacy in the $200 graphics card position and as such we are focusing on the performance of these two cards in relation to one another. To focus on the fight for the $200 crown we have only included the performance numbers for these two cards in the following graphs.
• Intel Core i7 2600K Processor at 3.4GHz (Supplied by Intel)
• Intel DP67BG Motherboard (Supplied by Intel)
• 2x 2GB Kingston DDR3-1600 MHz Memory
• 160GB Intel X25-M SSD
• Enermax Revolution 1000W PSU
• Samsung SyncMaster 2443BW 24-inch 1900x1200 (1200p) display
Synthetic benchmarks often prove to be some of the most popular, They allow for direct comparisons between different cards by offering the same exact conditions during each test. These repeatable and consistent tests do not necessarily reflect exact real world usage scenarios but they do offer a glimpse into the hardware’s performance and are often able to illustrate subtle performance differences between varying hardware components and performance tweaks.
Futuremark 3DMark 11
The latest benchmark in Futuremark’s long line of synthetic benchmarking tools is 3DMark11. The introduction of 3DMark11 now gives us the ability to test Tessellation, Direct Compute Physics and more. As this is one of the newest full featured benchmarks on the scene it is proving to be somewhat of a system crusher and lower frames per second and the resulting lower 3DMark score is to be expected for the time being.
In our 3DMark testing we run the hardware through two testing cycles. The first test is Performance Mode which offers a moderate test load designed to simulate modern gaming commensurate with most gaming PCs. The second test is Extreme Mode which pushes the hardware even further with a heavy load designed to stress even the highest end gaming PCs.
The outcome of both the performance and extreme tests showed the cards hitting fairly close to each other score-wise with the R6850 just nudging out the N460GTX on both tests. While the R6850 did come out on top, the actual experience with both cards was fairly similar with the Extreme test running achingly and jittery. This is not a detraction from the performance of these cards but more of a statement as to the punishment 3DMark11 is able to hand out as the Extreme test can even bring some enthusiast level cards to their knees.
Unigine Heaven (DX11 and DX10)
Unigine was one of the first to offer a true DX11 benchmark and really showcase a graphics card’s DX11 ability. As DX11 cards are just now becoming more common we still test performance in DX10 as well to give a greater overview of performance.
The DirectX 11 test of Heaven unveils some startling results. The R6850 gets soundly trounced by the N460GTX, to the tune of a 58% performance discrepancy. The tests were repeated numerous times to ensure the R6850’s score wasn’t merely a fluke, however the scores repeatedly came in around the same mark. This is also one of reasons for benchmark controversy between AMD and nVidia. For years, we've been hearing that AMD (ex-ATI) has tessellation in 1st, 2nd, 3rd... all the way until today's 8th Generation Tessellator, while Fermi architecture is the first nVidia GPU architecture to really support Tessellation (we'll discard N-patches discussion, as that was a paper feature). What happened was that nVidia put Tessellation engine where the efficiency is highest, while AMD has to battle legacy positioning of Tessellation engine. We've also seen this results in MachStudio Pro, professional rendering tool that uses GPU to render: GeForce and Quadro cards eat Radeon and FirePro cards alive as soon as you crank up the Tessellation load.
However, take a look a the slide below:
We took the cards thru another spin of the Heaven benchmark, this time in DX10. Here we can see that the scores are much closer. Again we see the N460GTX coming out on top, albeit by a much smaller margin. The DX10 tests remove Tessellation from equation and the R6850 catches up. The question "Can Unigine Heaven DirectX 10 and 11 results impact your gaming experience" is a whole another story.
Synthetic game benchmarks hit a little closer to the core usage scenario of a graphics card. These benchmarks are built upon the game itself and generally run a scene or series of scenes from the game in order to garner a performance score. As these benchmarks are built upon current popular gaming titles they offer a better picture of the video cards real world performance.
First up we have Metro 2033, a game set in the post-apocalyptic former Soviet Union. Metro 2033 has become the new Crysis and while it is one of the most visually stunning games on the market it is also one of the most punishing, bringing even high end video cards to their knees. The benchmark is run at 1900x1200 resolution with all the eye-candy turned up. The test is repeated three times with the end result being an overall average frame rate measured in frames per second.
The R6850 edges out the N460GTX by a noticeable margin, even if both video cards fail to break the 30fps barrier. The sub-30fps score is not really all that surprising, as we mentioned, Metro 2033 is still pretty much a system crusher and can bog down even the fastest video cards. Additionally all of the eye-candy is maxed which obviously takes its toll.
Both cards would be more than able to provide smooth frame rates should the visual detail be reigned in a bit - as you can see in DirectX 11 mode, where the framerate is well below acceptable levels.
S.T.A.L.K.E.R.: Call of Pripyat
Call of Pripyat is set in the area around the Chernobyl incident. The test can implement both DX11 and DX10 elements during a run though of separate game scenes and as we have done in the past, we have sorted this test by the Sun Shafts test as this has proved to be the most taxing on the GPU’s.
Stalker CoP proved to be a knock-down drag out fight between the two cards with each competitor exchanging blows on a test by test basis. As we mentioned, the results are sorted by the Sun Shafts test, but it is not the only means of comparison. History has shown us that AMD based cards tend to score higher in the Day and Sun Shafts tests, while NVIDIA generally takes the lead in the Night and Rain tests. This particular matchup however showed a different story with the N460GTX managing to eek out a victory, albeit a small one, in three of the four tests. The R6850’s sole win in this round was in the Day test, where it bested the NVIDIA card by 7fps, conversely the N460GTX responded in kind by posting a near 7fps lead in the more rigorous Sun Shafts test.
We have mentioned before that the DX10 results are markedly low. Especially for NVIDIA cards where we can see the scores plummet. While scores for both AMD and NVIDIA-based cards suffer under DX10, NVIDIA cards demonstrate a definitively marked drop. This current performance delta renders the DX10 version of this benchmark as unreliable and therefore useless.
Colin McRae DiRT 2
Running the built-in benchmark in Colin McRae DIRT 2 shows both cards achieving solid frame rates near the 60fps mark. The N460GTX crosses the finish line first in a clear victory, however the R6850 isn’t eating too much NVIDIA dust turning in a still respectable 58fps average.
Just Cause 2
Just Cause 2 is a visually impressive game highlighted by bright and beautiful tropical location coupled with the sandbox style of play popularized by the Grand Theft Auto franchise. If you are looking for a great story line and visually stunning cut-scenes you will want to keep looking. The plot is thin and the banter between characters plays out with one-liners reminiscent of 80’s-era action flicks.
The game features three built-in benchmarks, with Concrete Jungle providing the highest GPU workload. Average frame per second scores are reported and the results have been sorted by the Concrete Jungle test. In an interesting twist of fate, just as DiRT 2 was an AMD-pushed title, Just Cause 2 is an NVIDIA "The Way It’s Meant To Be Played" title, so it is a little surprising to see that the AMD-based R6850 upsets the N460GTX in this scenario with the R6850 holding onto an average lead of roughly 10%. I guess "The Way Just Cause 2 Is Meant To Be Played" is on - an AMD card.
Click on next page to see the results in actual gameplay.
Synthetic benchmarks do an admirable job of allowing system to system or component to component comparisons, however they often lack a direct "real world" relevancy. Just because a video card does well in 3DMark11 for example, does not necessarily mean it will have a guaranteed level of performance in the next blockbuster gaming title.
In order to balance both sides of the coin we have developed a few of our own real world benchmarks utilizing some well known gaming titles. The process here is rather simple. We load up said game while also running FRAPS. The game is played for 30 minutes during which time we have FRAPS calculate the average frame rate. This is admittedly not as structured as a synthetic benchmark, but it is more akin to real world usage where gaming will see the GPU usage peak and valley rather randomly based on game play.
Call of Duty: Black Ops
One of the most highly anticipated, and consequently most highly purchased games of all times, CoD: Black Ops is a game most are familiar with. We have chosen to include this game in our suite largely due to its sheer penetration in the gaming market, most gamers either own the game or have played the game, and those that haven’t have most likely heard about it. While this test will fall under the real world gaming benchmark description, the "shooter on rails" moniker that the COD series have earned is well deserved in this case as there is generally only one way through a level. This real world benchmark will prove to be the closest to fixed-variable aspect of a synthetic benchmark.
Black Ops follows the exploits of its main character Mason throughout 1960’s including settings in Cuba, Vietnam, Soviet Russia and more. True to it’s namesake, the mix of missions are "black bag" or clandestine operations. For benchmarking purposes we play through the first full level of the game.
Just Cause 2
Here we are again with Just Cause 2. While we have already utilized the in-game benchmark we thought it would be interesting to bench some real world gaming with the title and see how closely the game’s in-game benchmark predicts real world performance.
The first thing you will notice here is a change in the leadership position. Running the canned benchmarks resulted in the R6850 grabbing the top slot, yet running the real world benchmark we see that the tide has turned and the N460GTX takes top billing. This works to illustrate the fact that synthetic benchmarks are only part of the story.
We chose Metro 2033 for its system crushing-abilities. It is one thing to run a canned benchmark and get a sub-30fps score, its quite different however to actually have to play a game at those frame rates. Much the same as Just Cause 2, it will be interesting to see how well built-in benchmark predicts actual game play performance.
The real world benchmark of Metro 2033 shows us a few things. First, the leader remained the same with the R6850 showing the lead in this benchmark as it had with the in-game benchmark we tested previously. The second thing we notice here is that the both cards performed significantly better on average FPS in the real world benchmark than they did in the previous in-game benchmark. Each card gained roughly 5-6fps over their built-in benchmark score, this may not sound like much but with scores that were sub-20fps before, this equates to roughly a 25% performance "gain". Obviously the cards themselves have not actually gained any performance but it does appear that the built in benchmark for Metro 2033 does tend to stress the graphics card more strenuously than does general game play. Again, this is not so much a revelation as a reminder it is important to look beyond mere synthetic/canned benchmarks in order to see the entire performance spectrum of a component.
Temperature and Noise
A common consideration when looking at video cards is noise and heat output, after all if you are like most of us, you are going to have to live with this video card operating a few feet from you on a daily basis. As we are more concerned with the real-world attributes of a cards temperature and acoustic characteristics, that is what we test for. Temperature is easy enough to discern and in order to create a level playing field we run each card through our own in house temperature test designed to simulate normal gameplay usage. If we wanted to max out the temperature on the cards we could easily fire up Furmark and set it to the Xtreme Burning Mode selection and watch the card fry, but seeing as even the most extreme normal usage scenario would not heat a card to that level we do not feel it is a fair representation of a video cards thermal attributes.
In the temperature test we measured the idle temperature of each graphics card by allowing the system to sit powered-on but inactive with the desktop visible and no screen saver running. The temperature measurement is taken after the system has been sitting idle for 15 minutes. After this the idle temperature reading is taken the system is launched into the under-load temperature test.
In this test both cards demonstrated that their non-reference cooling solutions seem to be benefiting them quite well. The N460GTX had somewhat of an edge posting a lower idle temperature at 25C (to the R6850’s 29C) and load temperature of 48C versus the R6850’s 54C reading. In all reality both cards performed very well and did a great job of cooling some hot GPU’s. The N460GTX’s better performance in this test is to be expected when considering the fact that it’s cooling solution employs two more heatpipes and an additional fan as compared to the Cyclone cooling solution of the R6850.
When it comes to the noise level or sound output of a graphics card we feel it is important to remain practical. Granted we could use a dB meter and a quiet room to measure the exact noise level of the card but that setting would be atypical of the average usage scenario for a desktop card. Secondly the dB scale is not an easy indicator to relate to as each increase is in order of magnitude and not easily comparable. How much more annoying is a graphics card with a sound level of 82dB versus one with 80dB? Its hard to tell. Therefore we feel it makes more sense to break down the sounds levels into four categories akin to real world experience, and these "measurements" are taken in a standard office/room environment with standard ambient noises such as HVAC present.:
Unnoticeable: At this level the sound of the card is not perceptible. Either completely silent or only perceptible when your ear is place directly next to the card itself
Noticeable: At this level the sound of the card is perceptible, generally as a low hum. The noise at this level is unobtrusive and generally blends in with other ambient noises, such as case fans, power supply exhaust fans etc. The noise from the card can be heard but you have to listen for it to really hear it.
Clearly Noticeable: At this level the noise output from the card is clearly evident. The graphics card is discernable as the source of the noise and tends to be of higher magnitude than the ambient noise around it.
Annoying: This moniker pretty much describes itself. At this level the sound of the card is distracting. The video card is clearly discernable as the source of noise and during gameplay/media enjoyment speaker and/or headphone volume must be increased to overcome the noise of the card.
The results of this test are a direct tie. Both cards performed well in terms of noise output, clearly falling into the Noticeable category.
The heavyweight bouts are not always the most interesting to watch, whether it be Professional Boxing, UFC or graphic cards. The N460GTX Hawk and the R6850 Cyclone both proved to be scrappy, knock-down drag-out fighters. Throughout our testing there was no constant victor, it seemed as soon as a pattern would emerge indicating one card was clearly dominant the lead would abrupt shift with the next benchmark ran. In true brawler style, neither competitor would give in, instead rallying back and forth with the lead changing hands multiple times.
It is true that even though each card earned a number of wins in different benchmarks, one card did come out on top with more checkmarks in their Win column. That card is the R6850 Cyclone. The R6850 can handle any of today’s latest games while still maintaining solid frame rates. The Cyclone cooling solution handles the heat of stressful gaming with ease and without the annoying fan noise found on some of its competitors.
Even though the R6850 Cyclone came out on top, the fact of the matter is that the gap between first and second place is minimal and the N460GTX provided a strong showing as well. When it comes down to it, I can easily recommend either card as neither card truly dominated the other in our tests, there were wins for sure but the performance from both of these cards is close enough that purchasing either is going to provide you with a great gaming experience. When you factor in that both cards support their own version of multi-card performance, be it SLI or CrossFire, the main deciding factor may come down to which technology your motherboard supports.
If you are truly looking for the best bang-for-your-buck performance, both cards deliver. It is clear that NVIDIA and AMD understand the market and it is evidenced by the fact that they have designed competing cards that so closely rival each others performance. MSI has taken the performance and value a step further on both offerings making one hard to beat package, whichever flavor you choose.