I remember a GPU called the 6600GT that offered so much bang for the buck that no one was able to follow at such an aggressive price point. I remember it well because it was the first GPU that I ever bought that made me feel like somebody out there actually wanted me to play my games without having to spend a small fortune. It allowed me to play my favourite games at reasonable frame rates and at decent resolutions, but most importantly left me enough cash to buy more games. For most of today’s mainstream gamers, that sort gift hasn’t come from nVidia in quite some time, but this week everything changed as the GeForce GTX 460 was announced to the world.
A New Fermi for the Masses
nVidia really did their homework with this new GPU striking a balance of performance and price that makes the rest of their GTX 4xx line feel a bit silly. While all the rest of the line up is based on the notoriously hot and inevitably noisy and power hungry GF100, the new GF104 GPU has some unique tweaks that make it the polar opposite of the old regime.
For starters, the new core is re-engineered to be more agile and efficient. nVidia themselves call their GTX 480 a “tank” while the GTX 460 is more like a “hunter”: Agile and fast, but just as dangerous in some respects. Personally, I would have called it a “hot hatch” like a Ford Focus or an Abarth 500, but they must not watch Top Gear at NV HQ.
Moving along, instead of simply cutting a GTX 480 core in half, they made some very important modifications that make it very much its own chip. They gave the GTX 460 336 CUDA cores (48 per SM) instead of crippling it with 240 which is half what the GTX 480 has. They also kept the number of texture units to 56 which is only four less than the GTX 480 and only cut where they needed to which was in the polymorph engines. The GTX 460 has 7 while the GTX 480 has 15. In other words, a lot of the same grunt without the massive coronary causing calories.
Most importantly though, the GTX 460 GPU is cheaper to make, very high performance and is far more efficient than the GTX 4xx’s that came before it. All the jokes about frying an egg on a Fermi end with this new GPU.
Since it’s not fair to compare a “tank” to a “hot hatch”, we parked it against it’s slightly bigger brother, the GTX 465. As you can see, the GTX 460 will come in two flavors with the differences being in memory amount (768MB vs 1GB), ROP units/partitions (24/3 vs 32/4) and memory bus (192 vs 256 bit). Both however make the more expensive GTX 465 look like a diesel powered VW Scirocco. This becomes more embarrassing for the GTX 465 when you factor in the magical price of $199US for the 768MB version of the GTX 460 ($229 US for the 1GB version), which we’ll be checking out today.
Who Dares Challenge at the $200 Price Point? Hello? ATI?
Obviously the GTX 460 already looks like a winner on paper. In fact, during our press briefing last week, I was really sad to find out that everyone else had their GTX 460s a week ago hence the delay of our own look at this new GPU. However, while waiting around patiently for our card to show up, I got familiar with ATI’s $200 price point GPUs: The HD 5830 and the slightly less expensive HD 5770 which we unboxed very recently on our YouTube channel. The arrival of this GPU is a total game changer so we decided to wait until we had the band together before playing this gig. Here’s a quick rundown on their bios…
The GTX 460 card that finally showed up for us to lay our paw prints all over it was from MSI and is called the N460GTX Cyclone 768D5. This card is a 768MB version of the GTX 460 that is slightly overclocked from stock clocks of 675Mhz to 725MHz. It also comes with a customized cooling solution that looks somewhat like a Zalman design and you can see us peel it apart in our other unboxing video. For the purpose of our testing though, we will be returning the card back to stock clocks. We will be overclocking the snot out of the MSI when we review it separately later this week.
I don’t know about you, but I like to play games, not stare at specifications all day. So if there are no complaints or comments, let’s move on to introducing our test setup and then seeing what sort of playable frame rates we get at the magical $200 price point!
Test System Setup
For the purposes of our testing, we put together a system that hopefully reflects that needs and budget of a mainstream gamer…
- CPU(s): AMD Phenom II X4 965 Black Edition CPU
- Motherboard: GIGABYTE GA-890FXA-UD7 (F2 BIOS)
- Memory: Kingston HyperX DDR3 1333 MHz 8GB Quad Channel Memory Kit
- Video Card 1: GIGABYTE GV-R5770SO-1GD Super Overclock (HD 5770)
- Video Card 2: GIGABYTE GV-R583UD-1GD (HD 5830)
- Video Card 3: MSI N460GTX Cyclone 768D5 OC Edition
- Hard Drive: INTEL X25 80GB SATA2 SSD
- CPU Cooler: Thermaltake V1 CPU Cooler
- Power Supply: Ultra Products X4 750 Watt Modular ATX Power Supply
- Optical Drive: LG SATA DVD Writer
- Operating System: Windows 7 Ultimate 64 Bit
- Other: 2 x Silverstone AP121 Air Penetrator 120mm Case Fans
With the exception of a grossly excessive amount of DDR3 memory and the INTEL SSD, we think that the system does a pretty good job of representing that. We already know that excessive amounts of memory do nothing for gaming and we only used the SSD to speed up the time to load up our benchmarks. Neither should have an effect on the big picture we’re trying get from the results. All of our components were installed in an open test bench (High Speed PC Top Deck Tech Station).
Testing Notes
All of our tests were conducted using the latest available drivers and BIOS revisions. As of the date of this article, we are running ATI Catalyst 10.6 drivers for the HD 5770 and HD 5830. The the GTX 460 gets run with the latest public release of the GeForce driver release 258.96 which was released on launch day to support the card. Since we are using a non reference GTX 460 board from MSI, this board has been underclocked to mimic the performance of cards that you would find at the $199 US price point.
For GPU temperature testing we will be using FurMark 18.2 to push each GPUs to their limit and reveal max temperatures and noise levels. Idle temperature will be taken 30 minutes after booting into the system from the off position with each video card while load temperature will be taken after each card endures 30 minutes of torture. Ambient temperature within the room is maintained at 25C for all temperature readings. GTX 460 readings will be for the MSI card only as it does not use a reference cooling solution.
Noise levels will be measured at both load and idle using a noise level meter pointed directly at the middle of the video card and placed 6 inches away. We are looking for the relative noise level between each video card in terms of higher or lower noise between the three video cards. Again, we are not using a stock reference cooler so the results are only representative of the MSI card.
Power draw will also be measured as a whole system consumption at both idle and load conditions using a power meter plugged into the PSU then into the wall outlet. Idle load will be measured at 30 minutes after booting in from the off position. Load measurements will be taken before the conclusion of our FurMark stress test when the GPU is producing the highest level of heat.
Unless otherwise stated, benchmarks will be performed using default settings and we use a standard resolution of 1920×1080. Please see each test for specific modifications. Let’s get cracking!
Synthetic Benchmarks
There’s a place for synthetic benchmarks in all testing. They give us a rough idea of how things will play out as we progress through out testing with other real world benchmarks. If something is out of whack, we can quickly check back to our synthetic benchmarks to see if we need to test again. We start off our testing with a few of the greatest hits…
3DMark06
We start off with Futuremark’s 3DMark06 as our warm up benchmark. While an older benchmark, it sets the tone for how DirectX9 tests should pan out. Yes, there are still people that play DirectX9 games out there and we’ll make sure they are taken care of. As it would seem, the new hotness of the GTX 460 does seem very eager early in this round of testing. I’m hoping to see this translate into playable frame rates when we get to the game benchmarks.
3DMarkVantage
Futuremark’s 3DMarkVantage is the standard when it comes to synthetic benchmarks. Representing DirectX 10, the benchmark gives us a rough idea of how things will play out in more modern games. We find it interesting that while the GPU Score favours the HD 5830, the GTX 460 mops up in the overall performance score. It is inconclusive whether the GTX 460 will continue due to the lower GPU score which is the test of video card capability.
Cinebench 11.5
Cinebench is a favourite for CPU and memory tests, but it does house an OpenGL test component. While few games these days use OpenGL, some folks out there may be using OpenGL apps and knowing how the cards perform may be important information. It would seem that ATI’s strong workstation background with OpenGL has carried into the Radeon HD lineage as both the HD 5830 and HD 5770 clearly command this benchmark.
Heaven Demo v1.0
Unigines Heaven Demo v1.0 has been replaced by their version 2.1 edition. We ran all the benches before the realization had set in. However, this version is still relevant in the context that it wraps up all our synthetic benchmarks into one neat bow, testing from DirectX 9 through to DirectX 11 and even OpenGL.
As the previous benches have shown, OpenGL isn’t the GTX 460’s strong suite. We are however surprised by the GTX 460 in the DirectX9 column though but are pleased to see the new $200 kid in town edge out the other $200 kid in town in more modern DirectX 10 and DirectX 11 tests. nVidia has made no apologies about optimizing it for today’s games and future games so we may be seeing this as a reflection of what comes in the next few benchmarks.
Stone Giant
Stone Giant is a great benchmark for showing off the prowess of a video card when it comes to tessellation and DX11. These are both areas where the Fermi architecture thrives. As we can see, it definitely does do that by a very convincing margin. Although ATI was the first out the gate with a DirectX 11 card, it would seem that nVidia is doing it better today.
Game Engine Benchmarks
Game engine benchmarks are extremely important to the evaluation of the performance of a video card. It provides a consistent measuring stick and lets us know which cards work well for which games. This allows us to select the right tool for the right job. Obviously we’ll pick the card that performs best for the games or genres of games that we play the most. We’ve also selected benchmarks that are mostly free and readily available to the public so that you’re able to compare at home, just like our selection of synthetic benchmarks. We’ll be concentrating on playable frame rates on each of the tests.
Crysis
If you have a video card review, you have to have Crysis. Despite its age, it is still one of the most punishing game engines out there that has stood the test of time. We used HardwareOC’s Crysis Benchmark tool to set the quality to Ultra High and we used the built in Crysis Map for our test…
As you can see even today’s newest cards can’t handle Crysis at 30 frames per second. We didn’t even bother to try anti aliasing because we’d just be looking at unplayable frame rates. Having said that, DX9 was very playable and it’s nice to see the GTX 460 does indeed do well in some older titles. DX10 however is another matter as none of the cards reach the magical 30 fps window of playability, but the HD 5830 does show that it’s the boss here.
Far Cry2
Where we have Crysis, we have Far Cry2 not that far behind. This time we dive a little deeper into which GPU reigns supreme and where…
It’s nice to see that both $200 cards and the slightly cheaper HD 5770 all produce playable frame rates at both AA and non AA settings in both DX9 and DX10. It’s also great to see that even minimum frame rates are at or exceed the 30 fps magical barrier which ensure smooth gameplay throughout. The GTX 460 shows that it is the King here in a commanding lead overall.
Resident Evil 5
Resident Evil 5 provides some extra test material for our look at how anti aliasing can affect performance. We are happy to report that 4x AA is something you should turn on in this title as the impact does not kill frame rates enough for you to not enjoy the extra picture quality. This game is completely playable and smooth with any of the three candidates tested here. Here, we see the GTX 460 take the middle spot in DX9 as eluded to earlier by some of our benchmarks, but takes off again in more modern DX10 scenarios.
STALKER – Call of Pripiyat
Like Crysis, this game is not only a decent, but it also represents one of the most punishing tests a GPU can go through. We refrained from hitting the EXTREME setting and just settled for Ultra High while running the gamut of tests. The emphasis on newer DX10 and DX11 APIs should bring out the best in the GTX 460…
The HD 5830 and GTX 460 seem to trade blows throughout the DX10 test scenarios. It is great to see that both cards do offer enough power to play this game smoothly. Even the HD 5770 can make a go of it with the exception of a couple scenarios. While the GTX 460 does have some impressive average frame rates, it’s worth noting that the HD 5830 has higher minimum frame rates which translates into fewer situations where things just seem to stutter and get stuck, translating into an overall smoother game play experience.
The DX11 side of this benchmark however, shows that GTX 460 is the GPU of the future and mostly comes up on top in both important minimum and average frame rates. It’s also nice to see here that all three cards provide very playable frame rates above 30 fps, even min frame rates.
Alien Vs Predator
Alien vs Predator or AvP is another benchmark that is based on DX11 where we expected the GTX 460 to do well. While it does provide us with playable frame rates above 30 fps, it doesn’t beat out the HD 5830 in this game. nVidia did provide us with tweaked config files to use, but we decided to just use the default settings and go where the benchmark took us. Although the GTX 460 doesn’t win, all three cards win for being $200 (or under) cards that can play this game.
World in Conflict
I know a lot of people are waiting for StarCraft II to come out which is the sequel to one of the best real time strategy (RTS) games of all time. Since we don’t have StarCraft II here, we’ll use our old standby World in Conflict to assess potential performance in RTS titles. With the settings cranked to max, from what we can see from the results, everyone is in trouble when it comes to minimum frame rates with AA or AA off. However, average frame rates are good and keep the game fun for the most part in my hands on testing. GTX 460 seems to take less of a hit when enabling anti aliasing.
Battleforge
Battleforge is another great RTS benchmark that’s free to play and is put out by EA. It uses DX11 and looks really great and is fun to play when you get sick of Zergs, Protoss and Humans. The benchmark gives us an absolute worst case scenario with some crazy amounts of characters on the map. As you can see, it definitely stutters a bit with some very low minimum frame rates, but all three cards produce playable averages. Again, we see GTX 460 react well to the increase in quality a little anti-aliasing adds.
Street Fighter IV
This timeless classic gets played at LAN parties all the time. Based on DX9 we can see that this game plays well for all of the GPUs at the highest possible settings, both with AA and AA off. We do take a pretty major hit in performance for turning it on though. GTX 460 is clearly ahead by a long shot so this is one DX9 game where it is optimized well for this engine.
H.A.W.X.
H.A.W.X. doubles up as our flight simulator benchmark. It’s a really beautiful game that gives us access to eye candy from real aircraft models and real satellite maps. While playable on all three GPUs, the GTX 460 seems to reach in there and steal everyone’s lunch money. Percentage wise, turning on anti-aliasing has a negligible effect on frame rates and is completely playable on or off so you might as well leave it all on. This is also with settings cranked to high which is also impressive.
Final Fantasy XIV
Our last game benchmark is harder to decipher because we simply get a score at the end. However, the benchmark does have a chart at the bottom left hand corner that looks like a heartbeat, which I assume is the frame rate. Based on me sitting there for each card’s benchmark run, I could see visually just how well each card really did and how smooth the frame rate was. Despite this being a DirectX 11 benchmark, the GTX 460 was not smooth. The line created was rather herky jerky in some places where both the HD 5830 and HD 5770 were fine in. Since the game isn’t lated for release till September of this year, nVidia has lots of time to get some optimizations down to clean this performance up.
GPU Temperature
The original Fermi was hot. Too hot for most people. As we found out in our own review of the GTX 470 and GTX 480, they were both a real detriment to the thermal safety of components around them in an enclosed case. Well, the tests are in on the GTX 460…
Based on our observations, there is no heat issue. And no, I’m not making this up like Apple makes up stories to tell people that there’s nothing wrong with the iPhone 4. The GTX 460 is cool as a cucumber at idle and just barely hotter than the clearly less powerful HD 5770. It’s even a great deal cooler than the HD 5830 that it’s been trading blows with. Well done nVidia!
GPU Noise Levels
While not quite a leaf blower, the GTX 480 was one noisy bugger. Here’s how nVidia did today with the GTX 460…
It’s clear that nVidia’s work to reduce the heat has also reduced the noise. The fan doesn’t have to work as hard so the GPU stays quieter all the time. Though MSI is partly responsible for this audio performance, the GTX 460 is less noisy at idle than both the HD 5830 and HD 5770 which are also wearing custom cooling solutions. Under load, the GTX 460 is no louder than GIGABYTE’s excellent cooler on the HD 5830 and much quieter than the one on the HD 5770 which also has a fairly decent heat pipe cooler. Another checkmark for the GTX 460.
GPU Power Consumption
This basically ties it altogether. A more energy efficient GPU runs less hot and thus doesn’t need as much fan power to cool it down. Idle power consumption was the most impressive of the group at a measly 90 watts total system power consumption at idle. The power optimizations have clearly paid off in the new core. Even under load, the GTX 460 sucks far less power than the HD 5830. Looks like this “hot hatch” is also a hybrid as well when compared to the “tank” which is the GTX 480.
Overclocking the GTX 460
We weren’t going to mention any results in this article initially. We were going to save them for our full review of the MSI N460GTX 768D5 OC Edition coming up this week. But what the hell?
The bone stock GTX 460 sits at a core speed of 675MHz and a memory speed of 900Mhz. We took our MSI up to 825MHz and 1000MHz with no voltage tweaks whatsoever. This is straight out of the box, randomly picking a number that sounded cool, and let the burn in tests go. So does overclocking help the performance? Well, 3DMark06 scores went from 17,299 to 17,790 and 3DMark Vantages scores went from 11,228 to 13,444 on the GPU score and from 13,655 to 16,200 on the overall performance score. The card is completely solid at this speed and we’re going to have the whole gamut of overclocked benchmarks when that review drops later this week.
Final Thoughts on nVidia’s GeForce GTX 460
I’ve been waiting for nVidia to come up with something that really captured that 6600GT spirit. I’ve been looking for nVidia to come up with something that was such a great value that it would drive people back towards them instead of repelling them away. They failed to show people real value for the last couple years, making it easy for myself and many others to jump over to the greener pastures that ATI had to offer. Today, nVidia has mended wounds and rewarded their fans with a great peace offering. They listened to mainstream gamers, did their research, and came up with a $200 video card that plays my games at high quality and reasonable frame rates. Our tests have shown both the performance and the potential that this GPU has to offer.
Along the way, nVidia has also addressed the criticisms that plagued their former Fermi GPU. Instead of simply cutting down an existing chip, they reinvented it and reduced power consumption which in turn created a cooler and quieter video card. All the while, they made special adjustments to ensure that their tweaks would result in a GPU that offers the greatest value for the money.
Although ATI has a huge head start with the HD 5770 and HD 5830 GPUs, which have been out there for nearly a lifetime for tech, nVidia has shown that it’s back and that mainstream gamers like you matter to them enough to create a card just for you. That’s right! All you single card toting, $200 budgeted gamers out there have a new option to pick from and it’s a great one. Well done nVidia. Welcome back to the game!
Love This Review? Hate This Review? Leave a Comment or Hit The Forums!