It’s launch day again, and this time it’s for the second card in NVIDIA’s 600 series line-up, the GTX 670. EVGA is one of NVIDIA’s most well-known and well-regarded add-in board (AIB) partners, and we have one of their GTX 670 Superclocked cards to run through our testing.
Specifications & Features
A few things to note that aren’t displayed in GPUz are PCIe 3.0, TDP, thermal threshold, and reference clocks. The GTX 670 supports PCIe 3.0, but since my board is PCIe 2.0 that’s what is shown in GPUz. The TDP of the GTX 670 is 170 W, 25 W lower than NVIDIA’s single GPU flagship, the GTX 680. The thermal threshold is set to 98 °C, and is when the GPU begins to throttle clocks due to core temperatures being too high. GTX 670 reference clocks are 915(980)/1502 MHz, base(boost)/RAM, but this is EVGA’s SC version. The GTX 670 SC has clocks of 967(1046)/1552 MHz, which is a 52 MHz higher base clock, 66 MHz higher boost clock, and 50 MHz higher RAM clock.
- NVIDIA SMX Engine – Next generation streaming multiprocessor built from the ground up for incredible performance and power efficiency.
- NVIDIA GPU Boost – Dynamically maximizes clock speeds based on workload of the game to push performance to new levels and bring out the best in every game.
- NVIDIA Adaptive Vertical Sync – Dynamically enables vertical sync based on your current frame rates for the smoothest gaming experience.
- Supports Four Concurrent Displays – Two dual-link DVI connectors, HDMI, and DisplayPort 1.2
- PCI Express 3.0 Support – Designed for the new PCI Express 3.0 bus architecture offering the highest data transfer speeds for the most bandwidth-hungry games and 3D applications, while maintaining backwards compatibility with existing PCI Express motherboards for the broadest support.
- NVIDIA FXAA Technology – Shader-based anti-aliasing technology available from the NVIDIA Control Panel that enables ultra-fast anti-aliasing in hundreds of PC games.
- NVIDIA TXAA Technology – Support for new temporal anti-aliasing technique that delivers the ultimate combination of image quality and performance.
I went more in-depth on GPU Boost, Adaptive VSync, and FXAA/TXAA in my GTX 680 review, so please see that article if you are interested in more information on them. Quickly though, here are some key points about those features.
GPU Boost is analogous to Turbo Boost from the CPU side. What the GPU wants to do is get the most performance it can while remaining around its Power Target. GPU Boost does this by increasing core frequency when the GPU isn’t being fully utilized, this brings it closer to the Power Target. GPU Boost decides on where to set frequencies based on the power consumption and temperature of the GPU, so the overclocking will be limited to the power the card can pull and how cool the GPU can be kept. The same limitations as usual, but the clocks are dynamic and determined by offsets.
Adaptive Vertical Sync does exactly as it sounds, it turns VSync on or off depending on current frames per second. If the FPS is 60+ then VSync is turned on, otherwise VSync is turned off. This helps minimize the downsides of having VSync either on or off only. When VSync is on, microstutter can be introduced when the FPS need to drop below 60 since the FPS will jump straight from 60 to 30 and back without a smooth transition. When VSync is off, screen tearing could happen when the FPS exceed the monitor’s refresh rate. Adaptive VSync turns on VSync when the FPS is high to eliminate tearing, and it turns VSync off when frames are below 60 to keep the FPS high as possible without jumping to 60-30-60. So, in essence, Adaptive VSync narrows the gap between max and min FPS resulting in a smoother gaming experience.
FXAA & TXAA
FXAA or Fast Approximate Anti-Aliasing recognizes edges in-game by contrast comparison, and then “smooths” those surrounding pixels by forming a gradient between contrasting colors. FXAA does its thing quickly and without consuming many resources because it’s a post process pixel shader. Something that can be considered both a pro and con of FXAA is that it effects all pixels on the screen and smooths things are are not effected by MSAA; the con part being that it can effect small text on the screen causing it to be blurry.
Temporal Approximate Anti-Aliasing, or TXAA, is like a combination of lower MSAA and filters to give looks comparable to higher MSAA without as much of a performance hit. TXAA is meant to be integrated into the game engines, so there’s not an On/Off switch in the NVIDIA control panel like FXAA (yet). Since TXAA is so new and needs to be in the game engines, there isn’t a game that I have to show comparisons between MSAA, FXAA, and TXAA.
Packaging & Accessories
As I’ve said in previous reviews, I like packaging that doesn’t distract you from the product, its specifications, and its features. EVGA’s packaging is just that. On the front of the box, we have the product name, model, and some of its specs and features along the top edge of the box. EVGA goes more in-depth on the GTX 670’s features, shows a picture of the GPU, and lists the contents on the back side of the box.
The GTX 670 SC is placed in the center of a plastic clam shell that isolates it from the edges of the box. I personally like the foam packaging used with the GTX 680, but as long as it works, it’s not a big deal to me. The included accessories are a massive EVGA poster, “Enthusiast Built” stickers, user guide, quick start guide, driver CD, case badge, two Molex-to-6pin power adapters, and a DVI-to-VGA adapter.
The EVGA GTX 670 SC
The GTX 670 SC has the same shroud as the Signature series of EVGA’s GTX 680. It’s very streamlined and almost perfectly rectangular. The front of the card has a mesh across the front in the black area to give it a unique texture. The light gray, almost white, banners displaying “EVGA” and “GTX 670” are also a unique look of the EVGA’s cards. So, the color scheme is pretty neutral and should look good in most builds.
On the top of the card, the GTX 670 has two SLI connectors for up to 4-way SLI and it has two 6-pin PCIe power connectors just like the GTX 680. The power connectors seems to be in an odd place that’s almost in the center of the card instead of on the corner like usual. We’ll see the reason behind this connector positioning a little further down in the review.
You may also notice that it looks like the card is thicker towards the video output end and thinner towards the fan end. If so, your eyes are not deceiving you. The card is actually around 3 mm thinner at the fan end to allow some room for intake air in multi-GPU setups where the GPUs are right next to each other on the motherboard.
The video outputs on the GTX 670 are the same as the GTX 680: DVI-D, DVI-I, HDMI, and DisplayPort. The GTX 670 supports for up to four monitors in the form of a three monitor Surround setup plus a fourth auxiliary display for web, music, chat, etc.
On to the back of the card, we see something very surprising. The PCB is quite a bit shorter than expected when looking at the overall size of the card. The PCB on the GTX 680 is ~10″ long, whereas the GTX 670 PCB is ~6.8″ long, that saves roughly 3.2″ * 3.875″ = 12.4 in2 on the reference PCB. Since PCIe power connectors are always located on the end of GPU PCBs, the shortened GTX 670 PCB is the reason that the two 6-pin PCIe power connectors aren’t on the end of the shroud like usual and are located more toward the center of the card. The PCB length of 6.8″ plus the fan/shroud extending 2.7″ further makes the overall length of the GTX 670 SC is 9.5″, which is 0.5″ shorter than the GTX 680.
Interestingly, four of the eight vRAM chips are located on the back side of the PCB. I’m not sure why all the vRAM wasn’t put on the same side of the PCB, but it is what it is. The vRAM consists of eight Hynix 2nd Gen GDDR5 3000 MHz @ 1.5 V chips for a total of 2 GB vRAM. The GTX 680 used the 1st Gen Hynix chips rated at the same speed and voltage, but whether that means better or worse overclocking capability remains to seen.
Okay, time to get naked! Removing the shroud from the card reveals the VRM heatsink, core heatsink, and the blower style fan. When the fan is removed it looks like the card is broken in half because it has a shorter PCB than we’re used to. The VRM heatsink is attached with two pins on either side of the heatsink with a thermal pad between the chips and heatsink. The heatsink is attached well, as can be seen from the deep indentations in the thermal pad. The core’s heatsink is attached by the typical spring-loaded screws, and it seems there was plenty of pressure that easily pushed excess TIM out the sides of the core.
As you can see below, NVIDIA made a number of adjustments to the reference board to save space, which can help the GTX 670 fit into SFF cases. They were able to move the GTX 670’s power section to the left side of the GPU. It is also now much closer to the GPU than on traditional boards, which improves power integrity and increases efficiency. With the GTX 670’s power circuitry moved to the other side of the board, the right side of the PCB was pretty much empty and was removed altogether to save space. Also, the fifth unused power phase on the GTX 680 reference board was removed altogether, which saved a little space. Even though the actual GTX 670 PCB is much shorter than the GTX 680 PCB, NVIDIA used the same type of cooling solution on the GTX 670 as they did on the GTX 680.
The power section of the card looks to be 4 phase for the core and 2 phase for the vRAM. This is the same as the reference GTX 680 with 4 of its 5 available phases active for the core and 2 phases for the vRAM. So, there should be plenty of clean power for the GTX 670.
A close up of the core shows that the GTX 670 uses the same GK104 core as the GTX 680. The difference in the GTX 670 core is that one of the eight SMX units is disabled resulting in 1344 CUDA cores as opposed to 1536 on the GTX 680.
|CPU||Intel i7 2700K @ 3.4 GHz (Mimic i7 2600K)|
|Motherboard||EVGA P67 FTW|
|RAM||2×2 GB Corsair Dominator GT DDR3-1600 6-6-6-20|
|Graphics Card||EVGA GTX 670 SC|
EVGA GTX 680
EVGA GTX 580 Classified
|Hard Drive||1 TB Samsung Spinpoint F3|
|Power Supply||SeaSonic SS-1000XP (80+ Platinum)|
|Operating System||Windows 7 x64 SP1 (Fresh Install)|
|Graphics Drivers||nVidia 301.34 Drivers (GTX 670 SC)|
nVidia 301.24 Drivers (retested GTX 680)
nVidia 285.62 Drivers (GTX 580 Classified)
|Tenma Sound Level Meter|
|Fluke 52 II Dual Input Thermometer|
Benchmarks & Settings
- Synthetics – Performance for 3DMark, Xtreme for Heaven, PhysX off when applicable
- Hawx2 DX10 – 1920×1080, settings maxed, tessellation off
- Alien vs Predator High – 1920×1080, settings maxed
- STALKER: Call of Pripyat – 1920×1080, 4x MSAA, maxed settings, tessellation on, Sunshafts test
- Dirt2 & Dirt3 – 1920×1080, 8x MSAA, settings maxed
- Metro 2033 – 1920×1080, settings maxed, PhysX off, DoF On, Frontline
- Battlefield 3 – 1920×1080, settings “Ultra”, manual runs of first mission in single player
Overall, we have a great showing here from the GTX 670 SC. The GTX 680 only averages 4.2% better than the GTX 670 SC in the synthetic tests. The widest difference between the GTX 670 and GTX 680 are in the newer, GPU intensive benchmarks, 3DMark11 and Unigine Heaven; resulting in the GTX 680 coming out 6.4-8.7% better than its younger brother. What’s most surprising to me is that the GTX 670 is on par or beats AMD’s flagship GPU, the HD 7970, in all the tests except for 3DMark03. Kepler just doesn’t seem to do well in 3DMark03 at all, getting beat by the HD 7970, HD 7950, and even the GTX 580.
The game tests are also very surprising. The GTX 670 results range from 5.5% worse than the GTX 680 up to 2.3% better than the GTX 680! As far as competing with AMD, the GTX 670 easily beats the HD 7950 is every test I have data on, and it even beats the HD 7970 in a couple of cases. Based on the percentages, it seems that the HD 7970 wins by a large margin on Metro 2033, STALKER, and Alien vs Predator, but the actual FPS between the cards are not very high, in the 2-4 FPS range. So, another good showing with the GTX 670 performing better than expected.
For your viewing pleasure, here are the recorded scores that were used to create the graphs above. Since I didn’t personally test the HD 7970 and HD 7950, I don’t have numbers for Dirt3 and Metro 2033.
NVIDIA GPU Boost
Here’s NVIDIA’s explanation of boost clock, and I replaced GTX 680 numbers with the GTX 670 numbers.
The “Boost Clock” is the average clock frequency the GPU will run under load in many typical non-TDP apps that require less GPU power consumption. On average, the typical Boost Clock provided by GPU Boost in GeForce GTX 670 is 980 MHz, an improvement of just over 7%. The Boost Clock is a typical clock level achieved running a typical game in a typical environment.
When testing and running 3D applications, the boost clock is usually even higher than the listed boost clock. This GTX 670 SC sample boosts to 1175 MHz during 3DMark11 and EVGA’s OC Scanner X stress test, that’s 129 MHz past its listed boost clock and 195 MHz higher than reference boost clock. This actual boost clock is 91.4 MHz higher than our GTX 680 sample, which is most likely a main reason the GTX 670 SC performed so well in our tests compared to the GTX 680 and HD 7970. The only other reason that comes to mind is possible driver differences in 301.24 and 301.34.
EVGA released new versions of their Precision X and OC Scanner X software since the GTX 680 review. Precision X is where all the GPU settings can be changed, such as power target, clock offsets, voltages, fan speed, etc., and it also monitors all the settings of the GPU in its Performance Log. Something new to these new software versions is the monitoring of the Power Target in both Precision X and OC Scanner X. The GTX 670 skin for Precision X matches the GPU’s looks with the circular mesh and white/gray/black color scheme with green accents for NVIDIA. OC Scanner X is a GPU stressing program that helps determine stable overclocks by scanning for artifacts during the test, and it gives you a quick idea of loaded temperatures as well.
Stock Air Results
The first thing I did when overclocking the GTX 670 SC is to increase the Power Target to the max of 122%, this is 10% lower than the max Power Target on the GTX 680. Once that is set the GPU is allowed to pull more power, which allows GPU Boost to reach higher frequencies and voltages. On the GTX 680, I was basically stuck to using 1.175 V, but on the GTX 670 SC I can make the GPU use 1.162 V or 1.175 V when clocking the core. Being able to use that little bit less voltage actually allowed me to clock higher since I didn’t reach the Power Target as fast.
On the core, I was able to reach 1230 MHz @ 1.162 V which is 263 MHz higher than the base clock and 250 MHz higher than the standard boost clock. It’s also 15 MHz higher than I got to on the GTX 680. However, I couldn’t increase my vRAM frequency at all when using 1230 MHz core because I would exceed the Power Target. So, I ended up dropping the core to 1215 MHz (same as the GTX 680) to be able to increase the vRAM clocks. With the core at 1215 MHz, I was able to get the vRAM up to 1752 MHz from 1552 MHz, which is 100 MHz higher than I reached on the GTX 680. The GTX 670 vRAM definitely clocked better than the vRAM on the GTX 680, not sure if this has to do with the 2nd Gen Hynix or not, but it’s possible.
Cooling Performance & Noise
Stock Cooling Performance
Cooling performance is measured by running 3DMark11 and recording temperatures shown in GPU-Z. Ambient temperature is measured with a Fluke 52 II thermometer by placing a K-type probe 1″ from the intake fan, and turned out to be 25 °C. The fan profiles used are Auto, 30% (min), 55% (middle), and 80% (max).
The cooling performance of the GTX 670 isn’t near as good as the GTX680. The percentages tested are as follows:
- GTX 670 – Min = 30%, Mid = 55%, Max = 80%
- GTX 680 – Min = 30%, Mid = 57%, Max = 85%
The 2-5% fan speed difference in the Mid and Max tests do not explain the 17-19 °C difference in temperature of the two cards. The GTX 670 reaches and exceeds the thermal threshold of 98 °C when set to Min 30%, the clocks did drop dramatically and the card turned to fan to Max 80% for a few seconds in this instance. The GTX 670’s cooling solution is just inferior to that of the GTX 680. However, there’s nothing wrong with running high temps as long as they are below the thermal threshold.
The GPU will be on an open bench table with all possible external sources of sound turned off (doors closed, A/C off, ceiling fan off, TV off, and CPU fan turned low as possible). Sound level is measured by placing a Tenma meter 10 cm away from the intake fan and recording the dBA reading with the fan set to 30%, 40%, 50%, 60%, 70%, and 80%. Then, dBA at other distances are estimated by using the formula L2 = L1 – 20 * log10(r2/r1), where L2 = dBA @ desired distance, L1 = dBA @ reference distance, r1 = reference distance, and r2 = desired distance.
NVIDIA says they used the same fan on the GTX 670 as they did on the GTX 680, so the noise of the the two cards should be very similar. There is some motor noise with the GTX 670 sample I have, that eventually goes away when the fan speed reaches ~50%. This can be seen from the “hitch” from 30 -40% or so in the graph below. So, at the lower fan speeds the GTX 670 is actually louder than the GTX 680, but once the GTX 670 reaches 50% it’s actually a little quieter than the GTX 680.
System Power Consumption
Peak system power consumption is measured and recorded during each of the tests using a Kill-a-Watt meter. The system power consumption is very similar between the the GTX 670 and GTX 680, as were the performance numbers. The GTX 670 averages at 267 W and the GTX 680 averages at 274 W, so only a 7 W difference or so between the two in testing. This doesn’t surprise me with the performance of the two cards being so similar and both of them using the GK104 core.
Performance per Dollar
The GTX 670 SC almost runs away with the performance per dollar numbers. It has the best performance for your money in all tests except for the older 3DMark software, 06 and 03. The HD 7950 comes close to the GTX 670 SC in a few of the other tests, but still loses out overall.
Here are the prices used when creating the following performance per dollar numbers:
- GTX 670 SC: $420
- HD 7950: $400
- HD 7970: $480
- GTX 680: $500
- GTX 580 Classified: $540
The bottom line as far as performance goes, is that the EVGA GTX 670 SC is a powerhouse that holds its ground even among the single GPU flagship cards from both NVIDIA and AMD. Not only is it great based strictly on performance, it also really shines in performance per dollar since it performs so close to the flagship cards from both camps while having an MSRP lower than both at $420. The reference GTX 670 has an MSRP of $400, and that could save you $20 if you don’t mind overclocking the card yourself; since the GTX 670 can easily overclock to the “SC” clocks offered.
GPU Boost on the GTX 670 SC goes much higher than the GTX 680 sample we reviewed. With final boost clocks on the GTX 670 SC being 1175 MHz, which gives a 91 MHz advantage to the GTX 670 over the GTX 680 out-of-the-box. This higher boost clock definitely had something to do with the card’s stellar performance in the stock testing. The GTX 670 SC overclocked really well too, up to 1230 MHz on the core and 1752 MHz on the vRAM, definitely beating out the GTX 680 on overclockability. I didn’t even have to increase the core offset much either because of its already high boost clock.
The stock cooler on the GTX 670 SC was pretty quiet, but it does have some motor noise at lower fan speeds. There is really only one con to the GTX 670 SC. Unfortunately, the cooling system isn’t very good compared to the GTX 680, with the GTX 670 SC core being ~18 °C hotter when using similar fan speeds. However, the card’s temperatures remain within spec unless manually set very low. Even then, if the card exceeds its thermal threshold, the clocks will decrease and fan will spin up to get the temps back down.
The power consumption of the GTX 670 really isn’t much different than the GTX 680, maybe a very small amount less. Even though the TDP is 25 W lower, the real-world power consumption measurements don’t show that, which is probably due to the higher boost clocks.
Don’t forget that the GTX 670 SC has a much smaller than usual PCB, especially for such a powerful GPU, coming in at 6.8″ long. This, combined with some aftermarket cooling, could results in some very powerful small form factor systems for those who like to pack a punch in a small footprint. If I used a metric for performance per PCB area, then this card would definitely be at the top of the heap, by far.
Overall, the EVGA GTX 670 SC vastly exceeds expectations in performance despite being built on a ridiculously small PCB. It performs so well and at such a good price, I would recommend this GPU over any other card out there, even the GTX 680. Another no-brainer Approved stamp for EVGA!
– Matt T. Green (MattNo5ss)