Today is a going to be a great day to be an OCF reader. We have a rare chance to review one of the top of the line products from our friends at Galaxy, the GTX 780 Ti HOF+ (Hall of Fame). Galaxy recently started coming out with these high end cards to compete with the likes of ‘the big three’ and their high end enthusiast offerings. They have been pretty successful in doing so with their previous GTX 780 HOF among others, which made some waves on the benchmarking scene. The reception for these robust, unique white PCB based cards was pretty warm looking at other review sites and the web, and now we get our anxious little paws on it to put it through our testing suite and see what stands after the dust settles.
Specifications and Features
Below is a list of specifications from the Galaxytech website for the GTX 780 Ti HOF+. As we can see from below, there really isn’t much different outside of the clock speed. The card still has 2880 CUDA cores, 48 ROP’s, and 240 Texture units on the full GK110 die. The clock speed on this version comes in at 1020 MHz with a minimum boost to 1085 MHz. I found the actual boost at stock speeds will ramp up to 1124 MHz and stick there.
As far as the memory side, that of course does not change either coming in at 3072MB of GDDR5 resting on a 384bit bus, and comes in at 1750 MHz (7000 MHz DDR5 speeds). That totals to 336 GB/s bandwidth.
Some other things to note is NVIDIA’s GPU Boost 2.0, 3-WAY SLI ready, and the GTX 780 Ti HOF supports up to 4 displays on a single card. The TDP is dialed in at 250W with a 600W PSU recommended to push the entire system.
|Galaxy GTX 780 Ti HOF+ Specifications|
|GPU Engine Specs:|
|Base Clock (MHz)||1020|
|Boost Clock (MHz)||1085|
|Standard Memory Config||3072MB|
|Memory Interface Width||384-bit GDDR5|
|Memory Bandwidth (GB/sec)||336|
|Bus Support||PCI-E 3.0|
|Certified for Windows 8||Yes|
|Supported Technologies||GPU Boost 2.0, 3D Vision, CUDA, DirectX 11, PhysX, TXAA, Adaptive VSync, FXAA, 3D Vision Surround, SLI-ready|
|Multi Monitor||4 displays|
|Maximum Digital Resolution||4096×2160|
|Maximum VGA Resolution||2048×1536|
|Standard Display Connectors||Dual Link DVI x2, HDMI x1, DisplayPort 1.2 x1|
|Audio Input for HDMI||Internal|
|Standard Graphics Card Dimensions:|
|Maximum Graphics Card Power (W)||250W|
|Minimum System Power Requirement (W)||600W|
|Supplementary Power Connectors||Two 8-pin|
Below is a list of some high level features for NVIDIA cards in general, that of course the Galaxy 780 Ti HOF+ also supports.
- NVIDIA® TXAA Technology
- NVIDIA GPU Boost 2.0
- NVIDIA PhysX® Technology
- NVIDIA FXAA Technology
- NVIDIA Adaptive Vertical Sync
- NVIDIA Surround
- Support for four concurrent displays including:Two dual-link DVI, HDMI and DisplayPort 1.2
- Microsoft DirectX 11.2 API （feature level 11_0）
- NVIDIA® 3D Vision® Ready
- NVIDIA SLI® Ready Technology
- NVIDIA CUDA® technology
- PCI Express 3.0 support
- OpenGL 4.4 support
- OpenCL support
- NVIDIA SHIELD Ready
Getting into the Galaxy’s card specific features below, we see a lot of hardware improvements over the reference model such as the 8+2 power delivery configuration, push button “Hyper Boost” technology for easy overclocking (turns fan speed up among other things). Galaxy uses high end ICs from CHil and IR (International Rectifier) to control power, which are some of the things that set this card apart from its competition. Check out more details below from the reviewers guide we received from Galaxy.
One of the exclusive features of the card is the advanced dual fan cooling solution, which includes a 70 mm x 70 mm hybrid vapor chamber with heatpipes designed to keep temperatures and noise to a minimum while still providing (more than) adequate cooling. Part of this solution includes a fire resistant anodized die-cast aluminum housing to hold the fans. Speaking of the fans, Galaxy chose to use two 90mm liquid bearing fans to quietly move air through the heatsink.
Overall the solution does work quite well. The only curiosity is the lowest speed one can put these fans is 40%. Most other fans are much lower than that. This isn’t a problem at all because frankly it is not noisy at 40% or even at 100%. I am left wondering why such a high base speed though.
Below is more information regarding their cooling implementation for this card.
Photo Op – Meet the Galaxy GTX 780 Ti HOF+
Next up we get to see Galaxy’s retail packaging for their GTX 780 Ti HOF+. We can see the mostly white box with the Galaxy HOF nomenclature, of course what card is inside, its 3-year warranty, and a couple of high level features. Flipping the box around shows even more high level features.
The top shows what is included in the box while the bottom shows system requirements. One thing you will notice is even though its a high end card, it comes in typical packaging. To me, this is not a bad thing in the least. Please save me $1, $5, or $10 dollars on these things and don’t ship them in fancy cases, or even a small metal briefcase like we have seen some do. The packaging does the job well by being informative and sticking out on the shelves, as designed.
As is quite typical, we see a box in a box environment for the card. Opening the box up reveals the card inside an anti-static bag sitting securely within the internal foam packaging. Towards the top is where most of the accessories are hiding.
Speaking of accessories, take a look below to see what was included. Everything one needs to get started. Driver disk, a manual, some Molex to PCIe 8-pin adapters, as well as a DVI to VGA adapter. And finally, the most important part, a do not disturb door hanger… because overclocking is in progress!
OK, here we go… you finally get to see the beast! We can see the cooler design is what I would consider a typical Galaxy design in that there are a lot of sharp angles and corners that make up the matte grey cover. In the first image below, it does look a lot brighter with a bit more sheen if you will, due to the bright sunshine I shot these photos in, note. In the middle is the now familiar ‘HOF’ name plate flanked by their two 90 mm low noise EBR liquid bearing fans.
Flipping the card over we see the shiny brushed aluminum back plate that adds a bit of rigidity, some additional heat dissipation, and some flare. I really like the color scheme of the card here and the brushed finish on the back plate… excellent choices to me.
Next are some shots from the top of the card giving you a glimpse of the four heat pipes, the “Hall of Fame” wording which glows white in a slow pulsing fashion. Unlike other cards, this pulsing action is not by load. We can also see a shot of the two required PCIe 8-pin power connections. There is not too much to see on the bottom really, other than where it connects into the motherboard’s PCIe slot.
As far as the display outputs go, the Galaxy GTX 780 Ti HOF+ has what is now a pretty typical (from left to right) DisplayPort, HDMI, and two DVI ports (DVI-I, and DVI-D). As mentioned above this card can drive four monitors. As you may recall from earlier, the card requires two PCIe 8-pin power leads, which combined with the PCIe slot gives the card a total of 375W input power.
A Closer Look
In this section we get to take apart the card and see how she looks in its birthday suit. The first picture is taking off the heatsink and exposing the four 7 mm heatpipes and fin array, as well as the liberal application of TIM out of the factory. You also see the copper base of the 70 x 70 mm vapor chamber in this shot.
Next up I took off the back plate and the memory and VRM coolers. Though you cannot see it in the picture, they made good solid contact with any ICs they were supposed to, so we are good on that front.
Here we see our first good shot of the white PCB. I’ll be up front. I didn’t like the white PCB when I first saw it last year sometime. I have to say though, as time goes on it grew on me.. I think this would look especially good inside of a white case, which seems to be a popular trend among builders these days. But back to the PCB. As you can likely tell, Galaxy did their share of fortifying the power delivery and chose to go with a 8+2 configuration for the GPU and memory. We see glimpses of the CHil controller on the right, and after we remove the memory and VRM heatsinks, its all there to show the world! The back of the board shows the Posicaps (in orange). If you look close enough you can see voltage read points scattered throughout the backside. While I am thankful a card of this caliber has them, the positioning is awfully curious. You can’t reach them without removing the back plate. While the read points are really for extreme overclocking, it would be quite nice to have this in a location that is easily accessible regardless of cooling solution.
One thing that is a curious omission from this card is a dual bios. The card only has one. A card of this caliber should really have this feature as the demographic that is buying this card would have a much greater propensity to flash. So, be careful with any flashing. Make sure you have another GPU handy or are good at blind flashing in case something goes wrong.
We’ll zoom in a bit on the power delivery area and show off the 8+2 setup.
The last three shots of the card will be the memory IC, the PWM Controller, MOSFETs, and some of the caps. Galaxy has chosen to use Hynix modules, specifically H5GQ2H24AFR R2C which are rated at 1.6 V and 1750 MHz (7 GHz GDDR5). I hope to see some nice headroom out of these. Next up, the PWM controller is the popular CHiL 8318. The MOSFETSs used on these cards are the International Rectifier, or IR 3550 whose output current capabilities are 60A.
Software – GPU-Z and Galaxy’s Xtreme Tuner Plus
As always, we have the gratuitous shot of GPU-Z. This time we have moved on to version 7.8 as the latest release. Looking below confirms the above specifications with the card coming in at 1020 MHz with 1085 MHz boost and 1750 MHz memory. All the shaders(2880), ROPs(48), and TMUs(240), and memory(3072MB) are accounted for!
As a side note, at the time of publishing this was the only application to accurately monitor voltages. MSI Afterburner sat at 1.1 V through the entire time I had with the card while GPU-Z changed with what I set it to.
Galaxy has their own software based overclocking and monitoring program in Xtreme Tuner Plus (latest version 3.0.03). Like other software of its type, it does the usual monitoring/displaying of critical information such as clock speeds, temperatures, and fan speeds among others. You have the necessary tools to overclock and adjust clock speeds, power limits, and even voltages where applicable. The only two small peeves was the size, and the lack of granularity in the sliders, in that the smallest movement tends to jump the values up or down larger than I would like to see. Other than those two minor issues, the application works as advertised.
Performance and Benchmarks
- Intel i7 4770K @ 4 GHz, 1.1 V
- Gigabyte Z87-OC
- Kingston Hyper X Predator 2 x 4 GB 2666 MHz CL11 @ 1866 MHz 9-9-9-24
- 240 GB OCZ Vertex 3 SSD
- Seasonic 1000 W PSU
- Galaxy GTX 780 Ti HOF+ @ 1020/1085MHz boost (1124 actual)/1750 MHz Memory, and Overclocked @ 1106 MHz (1206 actual)/1900 MHz memory
- Windows 7 64 bit Operating System
- NVIDIA 335.23 WHQL
Other cards used for comparison are as follows (links are to the OCF reviews).
Note all testing below uses 1920×1080 screen resolution.
- All Synthetic benchmarks were at their default settings
- Unigine Heaven (HWbot) – Extreme setting
- Crysis 3 – Very High settings with 8xMSAA/16xAF (2nd level when you procure and use the Crossbow to get across the level and kill the Helicopter)
- Metro:LL – DX11, Very High, 16xAF, Motion Blur – Normal, SSAA Enabled, DX11 Tessellation – Very High, Advanced PhysX – Disabled, Scene D6
- Battlefield 4 – Default Ultra setting (Tashgar level – ‘on rails’ car scene)
- Bioshock: Infinite – Ultra DX11, DDOF (through Steam – option # 2, then option #1 assuming your are at 1080p)
- Batman: Arkham Origin – 8xMSAA, Geometry Details/Dynamic Shadows/DOF/Ambient Occlusion: DX11 Advanced, Hardware PhysX: OFF, the rest On or High
- Grid 2 – 8xMSAA, Ultra defaults + Soft Ambient Occlusion: ON
- Final Fantasy XIV:ARR – Default Maximum setting
- More detail is in our article: Overclockers.com GPU Testing Procedures
Finally getting on to the benchmarks, we start off with good old 3DMark Vantage. In this benchmark the Galaxy GTX 780 Ti HOF+ started off with an impressive showing scoring 44,105 falling less than a percentage point behind the EVGA 780 Ti Classified. The differences between these cards on paper are null, however what you will see happen is the Classified be ever so slightly faster to due more aggressive boosting on the Classified by around 38 MHz over the HOF+ (1124 MHz versus 1162 MHz). I tried raising up the power limit and pressing the hyper boost to no avail (temps and power limit were not breached to cause throttling). But as you will see the difference is quite small in all tests, not to mention you are going to be overclocking this anyway, and at that point it’s a toss up as to who can reach the highest clock speed.
Moving on to 3DMark 11, the HOF+ pulls out a 14,820 out of its hat, falling 2% behind the Classy, but besting the 290x Lightning by almost 3% and easily trumping an aggressively overclocked GTX 780 Lightning and a reference Titan.
Our next benchmark is the latest from Futuremark in 3DMark Fire Strike. Here the HOF+ scores 10,510 points, again a couple percentage points behind the higher boosting Classified, but easily besting the 290x and 780 Lightning by almost 7% and 15% respectively.
In Unigine Heaven Extreme, the Galaxy card posts a 3,551.7 on the board and just a bit over 1% behind the Classified. And again besting a 290x, 780, and Titan by considerable margins, maxing out at almost 15% again.
Moving on to our gaming suite, I start with the ever GPU heavy Crysis 3. Here the Galaxy GTX 780 Ti HOF+ manages to eek out 41.1 FPS falling less than two FPS short of the Classy, but again besting the rest of the field.
In Metro: Last Light, the HOF+ pulled down 56.5 FPS essentially matching the Classy in this benchmark while still easily besting the rest of the field.
Last up in this graph is my ongoing ‘glitchy’ favorite, Battlefield 4. This beautiful game and card combo with our settings achieved 86.9 FPS. One FPS difference from the slightly higher boosting Classified and again making quick work of the lesser cards.
Moving on to Bioshock:Infinite first, we see the 780 Ti HOF+ getting 129.4 FPS. In this game, it actually beats the Classified with lesser clocks. I would imagine the driver used is the difference here considering the previous (and future) results. It really, REALLY beats out the rest of the field in this benchmark.
Next up, Batman: Arkham Origins. In this thriller, the HOF+ gets 125 FPS here losing slightly to the higher clocked Classy as we have come to learn, and still consistently beats out the lesser cards in the field as it should.
The last game is Grid 2… arguably my favorite driving game (love the Dirt series)! Here the Galaxy offering hits 129.4 FPS again a couple of FPS behind the other 780 Ti in the test and still beating the rest of the pack.
And finally, the last gaming benchmark in our current test suite, Final Fantasy XIV: A Real Reborn. In this benchmark, it is TOTAL FPS instead of an average. Here the total for the Galaxy card comes in at 17,020 with the trends above continuing (ever so slightly behind the Classified and besting the rest easily).
Pushing the Limits
Finally to my favorite section… Here I have the chance to really see what any GPU will do. I have to admit I had some high expectations of this card, especially considering I got my hands on a modified bios (thanks Skyn3t!) to help bypass the power limit NVIDIA imposes on their partners. However, at least for the core, I was a little disappointed in the results. The highest I could get to was 1233 MHz actual on the core. I was really hoping to reach 1300 MHz with the tools I have (allows voltage to go over 1.25v). So the short of it is that I settled on 1233 MHz GPU/2004 MHz memory. Now, the memory clocked to the moon with only 100 mV added to it (so 1.7v total). So that was a great result. But for some reason this card would simply flame out at that clock speed. No matter what voltage I added (set 1.3 V and it shut off the card) or took away, it just would not budge past 1233 MHz on the core for this sample.
I had other peculiar issues with how the card responded to too high of an overclock. It simply dropped a monitor signal, and presumably hosing the PC forcing a hard reset every time. I recall seeing the ‘display driver recovered’ message all of once during my time with the card. I am not sure what the deal is there… but putting back in a different NVIDA card, it tended to play ‘nice’ and more gracefully ‘back out’ of a bad overclock. I am unable to pin down the source of that behavior at this time though.
Cooling and Power Consumption
As far as cooling goes, I will be brief in this section and say the cooler does a fine job. It is quiet, and keeps load temperatures down even when I was overclocked to the card’s limit. Idle though, now that is an interesting story. The fans run by default at 40% out of the box which is curiously high to me. However, the idle temps were a bit higher than I am used to, coming in at 41° C stock. I am used to seeing my GPUs in the low 30’s at idle. I tried remounting and reapplying TIM just to rule that out as well, but both times I managed complete coverage and good pressure.
So, we all should know idle temps do not really matter but the load temps do. The load temps while using the auto fan profile were kept under the throttling point of 80° C and the fan was essentially inaudible among everything else going on in my room. At 100% you do hear the fans, but it really isn’t that loud at all. I have heard way worse, no doubt!
As far as power consumption, we managed to peak at 397W at the wall when fully overclocked. A slightly higher number than we are used to being the top of the heap card, but still almost a dream compared to the power versus performance even a generation ago. Again, in the Intel world, a quality 550W PSU will be plenty, though if you want to play it really safe, grab the recommended 600W.
So where do we stand with this card is the big question, isn’t it? Well, let’s get the minor bad things out of the way before we start waxing (not so) poetic. To be honest, there is not really anything bad to say, at least nothing major, which is a great thing. I have to admit the most glaring omission from a card of this caliber is the lack of a dual bios switch. Again, a competition caliber card, one would want that type of flexibility certainly. In the same thought as benchmarking, the voltage read points are scattered around the PCB. While this is fine for extreme (read: Dry Ice/LN2) overclockers, those that are on air or water can’t reach them without removing your incredible looking back plate. Clearly not a deal breaker, but something to be aware of if you expect to be able to know those values and push the card hard enough to want to know the actual voltage. The last minor thing, and this is a curiosity more than anything, is why the fan runs at 40% stock. Its not loud, but I have never seen such a high fan speed stock. Like I said, no deal breakers, but like every card there are always areas that can be improved upon.
Now the good… Galaxy has taken a solid card in reference form and worked their magic on it by fortifying the PCB layers for optimum signaling, as well as going from the reference power bits to better quality parts and a 8+ 2 power configuration, which should allow more and cleaner power getting to the GPU and memory. The cooling solution does a good job and quietly. While it isn’t the best cooling solution I have seen out of the box, it is not a slouch under load by any stretch. There is some good headroom available to push the limits or game quietly for that matter. As far as overclocking, the use of Hynix memory ICs was a good start, and my sample showed pretty good scaling reaching over 2000 MHz (8000 MHz DDR5) with a bit of a voltage bump. The core clocks however didn’t get as much as I may have wanted at 1233 MHz, but that said, I need to put things in perspective as well. The reference clock speeds on the GTX 780 Ti are 875 Mhz with 928 MHz minimum boost. So, out of the box this card comes in at a whopping 145 MHz base clock over stock already. Anything on top of that is already what I would consider ‘gravy’. Not too shabby in the end, but my personal goal (1300 MHz) wasn’t able to be met with this particular sample. Lofty goal, I know.
About the only thing I haven’t mentioned yet is pricing, but don’t be scared! If you are looking at a GTX 780 Ti you already have an idea of where we are going to land I am sure. Drum roll please…! The Galaxy GTX 780 Ti HOF+ comes in with an MSRP of $749.99. In looking around at its primary competition, a 290x Lightning, which it beats across the board, comes in at $50 less while the EVGA Classified version comes in at $739.99 at Newegg.com. I will admit that I want to see this price a bit lower, but it is in line with other cards like it.
Regardless of the meager price differences, Galaxy has really brought a solid offering to the table in their GTX 780 Ti HOF+. Its intent was to compete with some of the more established ‘big boys’ in this exclusive playground of the high end, and I feel that it has stood up for itself quite well. This card is Overclockers.com approved!
Click the Stamp for an explanation of what this means.
– Joe Shields (Earthdog)