If you have been following the GPU landscape this last generation, you may have observed a bit of the “Goldilocks Paradox” with some recent releases. AMD fired first with their HD 7790 to fill a price and performance gap, and most recently, NVIDIA fired back with the release of the GTX 650 Ti Boost to do the same. NVIDIA’s aim was to put out a card for the ‘sweet spot’ resolution of 1920×1080 while using ‘high’ settings in games. It’s time to kick the tires and take the Galaxy GTX 650 Ti Boost for spin. Let’s see if their goals were met and what other goodies Galaxy has under the hood.
Specifications and Features
In the table below we can see the Specifications from NVIDIA on this dual slot configured card. We see it’s sporting 768 CUDA cores with a back end of 64 texture units and 24 ROPs. The base clock comes in at 980 MHz on the core boosting to an average of 1033 MHz (it’s a lot more than that as you will read later). The 2GB 192-bit memory comes in at 1502 MHz or 6008 MHz DDR5 data rate. For comparison sake, the reference GTX 660 comes in at the same clockspeeds, but has more CUDA cores.
NVIDIA recommends at least a 450 Watt Power Supply to power this 140 W max TDP card. Note however, that most loads will be around 115 W stock and 127 W with the power limit increased to its maximum of 110%.
Listed below are some of the key features of this card (exclusive features are the three up top):
- Force Air Bracket
- Whisper quiet custom cooler
- DHT (Direct Heatpipe Touch) cooler
- 768 CUDA Processors
- Kepler Powered
- NVIDIA GPU Boost
- NVIDIA Adaptive Vsync
- NVIDIA Surround
- Support for three concurrent displays (DL-DVI, HDMI, and VGA)
- MS DX11 w/DirectCompute 5.0 Support
- NVIDIA PhsyX
- NVIDIA 3D Vision Ready
- NVIDIA SLI Ready
- NVIDIA CUDA
- PCIe 3.0 support
- OpenGL 4.3 support
- OpenCL support
The Force Air bracket is really nothing too special outside of having a lot of space to move warm air out of the case. I know some of those brackets are pretty small and potentially restrictive, so it’s nice to see larger spaces in there to get some of the heat out. Galaxy also put their own Direct Heatpipe Touch (DHT) cooler on top of this GK106 core to help get the heat away from it. Their fan is also designed to be ‘whisper quiet’ and it does so at normal speeds (less than 50%). Some pictures and discussion will follow a bit later. Otherwise, listed above are typical NVIDIA card features in this generation such as Adaptive Vsync, GPU Boost, NVIDIA Surround, and being SLI-ready among other things.
Photo Op – Meet the Galaxy GTX 650 Ti Boost
Our first look, as usual, will be the retail packaging that Galaxy puts its cards in. There are no attractive female ninjas, or swords, just a large “G” on the front with what looks like a carbon fiber pattern behind it. Of course it shows the card type, how much Vram (2GB), the warranty (3 Years), the Force Air Bracket feature, custom cooling, and its ability to support a tri-monitor setup out of the box. Flipping the box around to show the backside, you see more marketing goodies. On the bottom we see some features (listed below), System Requirements, and what’s included in the package. The top of the packaging simply shows the name of the card and what’s included inside.
When we open up the package to get at our new toy, we are greeted by box in box style packaging. An internal box holds the GPU and included accessories. Once we crack that open to expose the card, we see the Galaxy 650 Ti Boost sitting in an anti-static bag with what should be snug fitting foam. In this case, one can plainly see the card does not fit in this foam packaging very well, which can allow for a significant amount of movement inside the packaging.
I have reached out to Galaxy to ask if the internal packaging I received was actual retail, or just something to get out to a reviewer since it was just released a bit less than a week ago. Galaxy responded by saying this is the final packing product. That said, the card arrived just fine, but I would like see the packaging a lot more form fitting to protect it better during shipping.
Last up, we see the included accessories. A pretty plain stack, but enough to get you up and running with the included driver disk, manual, molex to 6-pin PCIe converter, and a quick installation guide.
Some extra pictures of the box…
And now, the Galaxy GTX 650 Ti Boost for all to see. One of the first things you may notice is the custom, single fan cooling solution Galaxy has chosen to use. This Direct Touch Heatpipe configuration (pictured later) with a single (what looks to be an 80mm fan) cools down the 140 W TDP (115 W actual in non TDP applications) through the heatsink array. At idle, the card is quiet to around 50-60%, after which you begin to hear some fan and wind noise. Nothing obnoxious, but it really lets you know it’s working over the 60%.
On the rear of the card we do not see much special outside of the SLI connector up top, and the holes in the PCB to enhance MOSFET cooling. I’m not sure if that works well or not, but it’s there to aid airflow. They stay warm to the touch though, even after overclocking and looping benchmarks.
Next up is the output area that features a full sized HDMI port, VGA port, and a DL-DVI port. As mentioned earlier, this card can support up to three monitors on just this one card. I wouldn’t try to game on one of these cards across three monitors, at least with modern games, but the option is there for less resource intensive uses.
Last up, this card requires one lone 6-pin PCIe power plug in order to work.
Below are a couple more pictures showing the card off…
Next up we take the card apart and see whats doing under the hood. As was mentioned above, this card does not use the reference PCB, but gives users a more robust power delivery area from the 3+1 configuration, to a 4+1 configuration. This can help deliver more and cleaner power to the core, potentially allowing for higher overclocks. Of course, it’s core dependent really as to what clocks you will reach, but its nice to have a more robust back end to support your overclocking adventures.
Flipping over the heatsink we now see the Direct Heatpipe Touch (DHT) cooler. As the namesake describes, we can see the two copper heatpipes making direct contact with the core helping to whisk away the heat throughout the aluminum fin array. One thing to note is that even with the Force Air Bracket, some of the heat generated by this card will still go in to the case. This isn’t a worry at all though, especially since the power draw is so low.
The last set of pictures here show the power delivery with the heatsink on it, and then with the heatsink removed. The heatsink base was found to make good contact with the GPU, and the TIM application was just fine. This is all cooled by the single fan hovering over the GPU die.
Last up is a shot of the GK106 core this card is working with, as well as the Hynix H5GQ2H24AFR-ROC memory IC’s used (rated at 1.5 GHz).
Performance and Overclocking
- Intel i7 3770K CPU @ 4 GHz, 1.2v
- Asrock Z77 OC Formula
- Kingston Hyper X Predator 2 x 4 GB 2666 MHz CL11 @ 1.65 V
- 60 GB OCZ Vertex 2 SSD
- Seasonic 1000 W PSU
- Galaxy GTX 650 Ti Boost
- Stock @ 980 MHz Core (Boost 1097) – 1502 MHz Memory
- Overclocked @ 1053 MHz Core/(boost 1163) – 1600 MHz Memory
- Windows 7 64 bit Operating System
- NVIDIA 314.22 Drivers
Other cards used for comparison are as follows (links are to the reviews):
- All Synthetic benchmarks were at their default settings
- Unigine Heaven (HWbot) was run with the “extreme” setting
- Alien vs. Predator – 1920×1080 with highest settings offered (4x AA, textures set to highest)
- Batman: Arkham City – 1920×1080, 8xMSAA, MVSS and HBAO, Tessellation HIGH, Detail Level: Extreme
- Battlefield 3 – 1920×1080 at Ultra settings (4xAA/HBAO by default)
- Dirt 3 - 1920×1080 with 8x MSAA and all settings enabled and at Ultra where possible
- Metro 2033 – 1920×1080, DX11, Very High, 4x MSAA/ 16x AF, PhysX OFF, DOF enabled, Scene: Frontline
- Civilization V – 1920×1080, 8x MSAA, Vysnc OFF, High Detail Strategic View: Enabled, Other Settings: High, using full render frames value ( / 60)
- More detail is in our article: Overclockers’ Updated Video Card Testing Procedure
There really isn’t too much to see here that wasn’t covered already. We are using the 314.22 drivers and we can see the ‘average’ boost comes in at 1033 MHz. Of course, we see the 980 MHz stock core clock and 1502 MHz on the memory, all on its 192-bit bus.
If you remember the previous Galaxy card we reviewed, a GTX 660 GC, we explored their overclocking utility Xtreme Tuner Plus (XTP moving forward). Since that time, XTP has undergone a face lift and below is the fruits of that labor. Galaxy has gone with a blue theme and ‘dial’ in the middle, which shows the important things like your Power Target percent, and core and memory clock speeds. Flanking this on each side are displays showing the GPU temperature (left), and the fan speed (right). Your sliders for adjusting these items are at the bottom. The software also has a graphing tool for monitoring, as well as profiles to save your overclock and fan settings.
That said, this display is a bit confusing to me in that the GPU clock offset value by the slider displays your 2D core speed in this case. Normally, with other applications, the offset starts in the zero position and you add to the base clock. I reached out to Galaxy to let them know what was up, and they promptly sent a new version that resolves this problem. Until that is released (very soon), one can use other software to adjust the clocks.
My previous concern with the old version was that it was a bit too large for my personal taste. This… is even bigger and takes up more real estate on the screen. I can’t say I’m a huge fan of it getting bigger.
You can download this software at the Galaxy website.
As is the usual in my reviews, we start out with the synthetic benchmarks. In this case, we are using 3DMark 11 and Unigine Heaven (Hwbot version). The Galaxy GTX 650 Ti Boost managed to score 5,898 in 3dMark 11, which runs a hair behind its direct competitor, the AMD HD 7790. We can see it beats out the ‘base’ 650 Ti by almost 14% at stock speeds. A sizable increase.
Stepping over to Unigine Heaven, this card pulls down a 1,261.4 score and just smoked the HD 7790, which is around 13% slower at stock speeds. Again we see the GTX 650 Ti Boost easily besting its little brother by almost 27% here.
Moving on to our ‘oldie but goody’ synthetics, we give you 3dMark03 and 3DMark Vantage. First up is 3DMark03 where the Galaxy card manages a 71,120 score and is getting crushed by the AMD offering. NVIDIA cards of this generation have never really done well in this benchmark. That said, it still beats out the GTX 650 Ti by almost 8% here.
Moving on to 3DMark Vantage, the GTX 650 Ti boost manages a score of 23,372 at stock speeds. This beats out the HD 7790 by nearly 12% and the regular GTX 650 Ti by almost 28%.
The first game we are going to check out here is Civilization V. Here we can see the GTX 650 Ti Boost cleaning shop, hitting 57 FPS. It only lost out to its bigger brother, the GTX 660. Sliding over to Dirt 3, we see the boost version again besting everything outside of the GTX 660, coming in at 63.6 FPS. The last game in this graph is Metro 2033. The pecking order remains the same, but we don’t see playable frame rates with our maxed out settings. 20.9 FPS… That game, Metro… ooof!
Next up in the game benchmarks, we see Aliens vs. Predator, Batman: Arkham City, and Battlefield 3. In these benchmarks… wait… ya know what? Instead of trying to figure out a creative way to say the same thing, in all the games here we see the same exact story. The Galaxy 650 Ti Boost easily bests the HD 7790 and is only a bit behind the GTX 660. What can you say?
As far as the games go, it’s plain to see NVIDIA’s goal to put out a card that is perfectly playable at 1080p resolutions has been pretty successful, at least through the titles we tested. Let’s hope the pricing is in line as well.
Pushing the Limits
And last we arrive at my favorite section, Pushing the Limits. As you know, this is where we crank up the CPU and crank up the GPU, and see what happens. This being an NVIDIA card, we were limited by the lack of voltage adjustments, so we didn’t get too much more out of the core compared to the overclocking section. Here we managed a 1202 MHz boost clock, +99 offset, stable through these benchmarks. As far as the memory goes, I managed a very solid overclock there hitting 1722 MHz (6890Mhz DDR5) in 3DMark 11. Not too bad at all there. As we all should know, your mileage may vary.
Cooling and Power Consumption
We have mentioned a couple of times now the DHT heatsink used on this cooler assembly, so I suppose it’s about time to see how it did. As you can see below, idle temps sat around 32 °C while the maximum temperature in 3DMark 11 hit 55 °C using stock clocks. When overclocked the temp rose to 57 °C and the fan ramped up to a quiet 41% speed (auto profile). Hwbot Heaven temperatures hit 57 °C and 59° C respectively with the fan hitting 42%. Overall, even when overclocked, I didn’t need to stray from the auto profile for noise or cooling reasons. This card kept well under the 70 °C threshold we like to see for NVIDIA cards. With normal ambient temperatures and solid case airflow, keeping things in this range shouldn’t be an issue.
The fan does get a bit noisy above the 60% mark with some mechanical whine and wind noise. Left on auto, you wont hear that noise as the fan doesn’t need tramp up to that speed in order to keep temps well below 70 °C.
Power consumption wise, the card is listed as a 140 W TDP. However, in non TDP applications (TDP applications being Furmark and things of the like) the power use as stated by Nvidia is around 115 W. Of course, you have the 10% power limit headroom to add to that, and the card is said to put out around 127 W in that case. We can see below the maximum power hit in both tests was a measly 218 W ~ 228 W at the wall stock/overclocked. This was tested with the trusty (heh) old Kill-A-Watt meter.
Now we get to put all these pieces of the puzzle together and see if we like the card or not. Overall, NVIDIA has brought to the table a nice answer to AMD’s HD 7790. While the GTX 650 Ti Boost Edition is more expensive than the HD 7790 by around $20, in my opinion the performance you get over the HD 7790 makes up for that difference.
So what about the Galaxy offering specifically? Well, Galaxy has made the choice to beef up the GTX 650 Ti Boost PCB and give it a bit more in the power delivery area. This can be seen with a 4+1 power phase configuration as opposed to the 3+1 on reference boards. In order to keep things cool with the added overclocking potential, they use their DHT based cooler to help keep the card under NVIDIA’s magical 70 °C ‘boost neutering’ temperature threshold. This card does a solid job at that, even with the stock fan profile. There are no worries there.
As far as pricing among its peers, Newegg.com has the Galaxy GTX 650 Ti Boost for $169.99. Galaxy has the MSRP between $169.99 and $179.99, so Newegg falls at the bottom that range. The other cards can start out at $169.99 too, which makes this card appropriately priced. What you do get with this card over some of the rest are a more effective and quiet cooler, as well as having a more robust power delivery area to support your overclocks. Let’s also not forget NVIDIA also has some free games that come with select cards ($75 in-game coupon towards Hawken, World of Tanks, and Planetside 2).
So, let’s to wrap things up here. Galaxy has brought to the table an option in the crowded midrange landscape that by its pricing and performance, deserves a long look when choosing a card of this type. Galaxy, along with NVIDIA, has achieved their goal of creating a card that will play in the sweet spot of 1080p with ‘high’ settings in games. This card is Overclockers.com Approved!
~Joe Shields (Earthdog)