Author : Daeguen Lee
English Translation : Wendy Kim
Contact : leedaeguen [at] kaist.ac.kr
(Any action violating either CCL policy or copyright laws is strictly prohibited)
It has been exactly four weeks since I have written an article on comparing game performances on plenty of different CPUs. If I remember correctly, I started the last one with similar words as this one("It has been..."), so I say I might, or should rename my blog as "Monthly" IYD or something. Well, let me get to the business and dare name this article as the trinity of unprecedentedness, as I am writing on an unprecedented subject with an unprecedented scenario by using an unprecedented amount of benchmark data, compared with my previous articles. Then, what makes this article so unprecedented?
The title of this article, that I am sure that has either consciously or subconsciously caught your attention at the moment that you clicked to read this, is "The Resolutionary War". For some of you who picks up things fast, the title would be vaguely familiar – as it would remind you of a frequent topic of modern history, a revolutionary war. The subtitle of this article, "with Gamer’s Liberation Army", is also a wordplay derived from Communists' "People’s Liberation Army". With these two combined, the article would shout out something radical, revolutionary and "unprecedented", isn’t it? Let’s delve into the details of this unprecedented matter, "The Resolutionary War".
Chatper 1 <What's 5K & Why?>
The first two of the trinity, "Unprecedented Subject" and "Unprecedented Scenario" are derived from this unprecedented resolution, 5K. Then, why 5K all the sudden? Or, rather, where in the world the 5K resolution came from? It all started at GTC (GPU Technology Conference) 2014 on March 26, two months ago.
At the conference, Jen-Hsun Huang, CEO of NVIDIA, has unveiled their new graphics card that will be released - GeForce GTX TITAN Z. This product is the very first one to employ two GK110s, which is the best GPU of NVIDIA that is out there in the market. According to Huang, the graphics card supports 8TFLOPS single-precision compute performance, as well as an enormous graphic memory of 12GB, 6GB for each GPU. In addition to that, what captured my attention the most was this one phrase, “5K-ready,” an unprecedented resolution throughout the history of graphics card, Though not even a half of a year has passed since GTX TITAN and AMD’s “Hawaii” series opened the new era of 4K-resolution, this random, yet monstrous “5K-ready” graphics card is about to be introduced to the arena. In reality, it will be a while before the 4K resolution will be widely used among the users, as 4K does not even have a standard (※ Though UHD = 3840 x 2160 using 16:9 display aspect ratio is widely accepted and commercially sold as 4K-supporting monitor, several monitor and optical equipment manufacturers regard 4096 x 2160 as the true 4K resolution, which is 256 pixels wider horizontally than the one aforementioned). Because of these issues, 5K resolution seemed and sounded like a very premature thing to expect at the time.
Well, I ranted enough about how it does not seem like the right time to have the 5K resolution available. But as long as I know that it actually exists, I could not let go of the thought, wondering “What exactly is a 5K resolution?” I have fervently googled for a few days and found the answer that the 5K resolution that NVIDIA assumed was 5120 x 2700, but having found the “answer”, it amplified my curiosity even more.
5120 x 2700.
5120 x 2700???????
I thought to myself, ‘The aspect ratio is not 16:9, nor 21:9. Then where did the resolution come from? How is this even possible?’ When I was bogged down by the curiosity, I received a report that the actual resolution is 4096 x 2160, in other words, it is an altered form of 4K resolution, which is 256 pixels wider than UHD. Then, the appearance of this rather strange ratio as if it would be the new standard for the 5K resolution, especially through a mainstream event as NVIDIA’s new product announcement, should be very significant. It is significant because a major GPU manufacturer is looking beyond to the era of 5K resolution when UHD monitors are not even popularly used in the general public, as well as their attempt of setting a totally new standard than UHD.
Anyhow, products that support the resolution of 5120 x 2700 (From now on, 5K refers to this resolution) will be released pretty soon, and I am obliged to test this new stuff out. Therefore, I planned an unprecedented thing throughout the world, a “gaming benchmark at 5K resolution.” But, no matter how good the idea would be, keeping the really good idea inside the head and actually doing it to get a visible result are two totally different things. But one thing promoted me for the actual realization of the “really good idea”. In order to explain the reason behind it, I would have to start with the fact that I had GTX 780 Ti for my computer’s main graphics card. Though for some reason, I felt sudden spasm on my fingers as I decided to talk about what has happened; but despite my unwillingness to unravel the story, I must go on:
1. (At first) I was examining the possibility of doing experiments on the resolution.
2. I accidentally opened the NVIDIA Control Panel, and then...
3. I made 4096 x 2160 using custom resolution option.
4. So the resolution option can be applied??????
5. I made 5120 x 2700 using custom resolution option.
6. So the resolution option can be applied, again??????
...for the reason, I ended up starting this project thinking that the 5K benchmark would be an easy task to complete. But, the project ended up being a huge catastrophe. “Though the beginning was small, the latter hours of the project were great difficulties.” The followings are the difficulties that I have encountered since I have started the test.
1. (After finishing NVIDIA's single GPU test,) the 4K /5K configuration won’t work after configuring SLI…?
-> The refresh rate was set to 29Hz, and it worked after changing it to 30Hz.
2. (After finishing the business with NVIDIA,) How come there is no such thing as “custom resolution” in AMD Catalyst Control Center?
-> I would have to configure Eyefinity in order to have a resolution that is higher than the monitor specification!
3. Even two monitors cannot handle 5120 x 2700?
-> I will make it three monitors..
4. Won’t work even with three monitors?!.
-> I will make it four. (Should contact CrossOver -a monitor manufacturer- and beg for their help...)
5. (After all the work to obtain the 4th monitor,) Why the 4th display isn’t recognized?
-> It has to be connected via Native DisplayPort.
6. Why it won’t work after all?!
-> The initial resolution of DP-display should be set at 1920 x 1080.
Just because I wrote down the solutions for each problem, they might seem no hard, however, I had no idea and rather felt devastated thinking how in the world I would resolve the problem when I first encountered each of them. I mean, seriously... think about how you would feel when you have completed the single GeForce Test and configured SLI, and all the sudden the resolution won’t work, and you have no idea what the reason is. Then, you have connected four monitors (yes, FOUR.) and the computer treats the fourth one as some sort of inexistent one, and you have no idea what the reason is... But in a retrospect, it was a fun progress to do this series of problem solving, and if I would make it sound more dramatic, I probably was “destined” to do this benchmark. Yes, the Lord probably allowed me to overcome these trials... You know, feelings like that.
Well, let me wrap up the first chapter with this. For the next two chapters, let me introduce my turmoil and trial to actually have the 5K resolution work. Note that I would have to "fold" all the detailed procedures over next two chapters due to the length of the whole article. Please "click" each folding (which is indicated as "Click to see more") and see the whole procedures, if you don't mind.
Chapter 2 <5K for AMD>
There is a legitimate reason for me to split the “5K resolution adventure” into two different chapters. As all of you would know, there is no sense of combining them into one since the ways to make it work for AMD graphics card and NVIDIA graphics card are, even if they both are aiming for the same resolution, are totally different from each other. Especially, for AMD graphics card, it is impossible to set a custom resolution and force “upscale” the resolution like NVIDIA. Therefore, in order to have a resolution that is above the standard resolution (I used 30" monitor, and the standard resolution for the monitor was 2560 x 1600), I had to use a multi-monitor configuration, which added more complexity to the experiment. Then, from now on, I will talk about how I have set the individual resolution in the ascending order of pixels.
Configuring 2560 x 1600 resolution (4 million pixels)
...Easy, isn’t it? ;)
A 30" monitor should be able to do this without any problem.
This is a screenshot of running a game with the monitor. The screenshot shows a typical 16:10 aspect ratio, so if you are accustomed with 16:9, you might feel that the screen is somewhat narrow.
Configuring 5760 x 1080 resolution (6 million pixels)
What you need is three monitors that supports a resolution of 1920 x 1080 (Yes, these monitors are common 1080p monitors with 16:9 aspect ratio). Just like the picture above, after making sure that all three displays are recognized, I set them as the same Eyefinity Group at the Catalyst Control center.
...Though it feels like I have omitted lots of details, the Eyefinity configuration is actually really simple.
This is a screenshot of running a game with the setting. Also, the aspect ratio here, 48:9, is the widest aspect ratio among all of the resolutions throughout the article. Since outputs for games are expressed in terms of aspect ratio, yet regardless of the total pixels of the monitor, this configuration is the best if what you want is the “wide” view.
Configuring 4096 x 2160 resolution (8 million pixels)
In fact, I have always had a scenario called "4K" up until this article, and the resolution used for the previous articles is 3840 x 2160 with an aspect ratio of 16:9, also known as UHD. The Reasons that I chose to use UHD configuration were first, most of the manufacturers for monitors produced products assuming UHD as the standard for 4K resolution, and second, the resolution of 3840 x 2160 is in the range of resolution for the monitor that I am using(= CrossOver 30"). To summarize, I used this specific resolution since there is no reason for me to not to use the "Legitimate Successor" of "HD" family, UHD, which has the 16:9 aspect ratio as a sort of "de facto standard" for monitors... Well, to jump to the conclusion, I will use 4096 x 2160 when referring to 4K resolution. The reason is simple: "AMD and NVIDIA are supporting THIS resolution."
The preface was a bit long, but the 4K (from now on, 4K refers to 4096 x 2160. I will refer to 3840 x 2160 as UHD) resolution actually has the same vertical resolution plus 256 more pixel columns for the horizontal resolution compared to UHD. But because of this tiny difference, up-scaling is impossible for AMD graphics card. Therefore, I had to combine two monitors using Eyefinity in order to make 4K resolution. After making sure the two monitors were recognized, let’s set Eyefinity at Catalyst Control center.
First of all, the above screenshot shows a combination of two monitors with 2560 x 1600 resolution using Eyefinity, and the resulting resolution is 5120 x 1600.
Since this model supports an up-scaling to UHD, I had no problem up-scaling the vertical resolution to 2160.
Since all of you can do the simple math, this resolution is much wider than the 5120 x 1600, because the vertical resolution is so much greater than before.
This is a screenshot of running a game with the monitor. Though you have more pixels, the view is narrower because the aspect ratio decreased (from 48:9 to 256:135)
Configuring 7680 x 1600 resolution (12 million pixels)
Just for the purpose of an argument, let's say the (previous) benchmark scenario of UHD up until now are pretty much the same as 4K that I have just defined. And since the resolution of 5760 x 1080 has appeared for other reviews before, there is nothing new under the sun for all of the scenarios that I have presented. But, the resolution that I am about to introduce is a totally new configuration, a terra incognita, and is the reason that I have said "unprecedented" at the beginning of the article. If I would allude from the title, I can probably name the resolution as "A Revolutionary Resolution".
The procedures to configure the surrounded display are the same as the previous one, the 5760 x 1080 Eyefinity.
This is a screenshot of running a game with the monitor. The view is pretty wide as you can see, but the output information is somewhat less than the gaming experience at the resolution of 5760 x 1080. (It is because the aspect ratio has decreased a bit from 48:9 to 48:10) Of course, the physical size of the display makes the screen look very overwhelming.
Configuring 5120 x 2700 (14 million pixels)
5K... A name that I both abuse and adore...
This article would have never appeared here only if I hadn't seen the presentation slides for GTX TITAN Z, only if I hadn't noticed the magic word "5K", only if I did not care what exactly 5K resolution is... Only if this DAMN monitor would have not supported the up-scaling option to that resolution at the NVIDIA control panel! But all of these factors, in other words, these "bad lucks" made me obsessed with this task for about a month and caused many sleepless nights. A worse thing is that this wouldn't be possible if I have started with AMD, but the reality was that I found out AMD graphics card does not support up-scaling to 4K / 5K resolution after finishing up the NVIDIA Comparison group (As I mentioned above, the upper limit of up-scaling for AMD graphics card was the UHD resolution). On top of everything else, though I was able to somewhat realize the 4K resolution by using two monitors, 5K resolution was impossible to achieve using the method.
Q : But why is it impossible?
A : The vertical resolution is too large (1600 to 2700 does not seem very practical after all…)
If I knew all those things that I ended up knowing after fact, then this article probably wouldn't have seen its end... Come to think of it, all the circumstances probably allowed me to finish this research.
I digress, and now back to the topic! The screenshot above is confirming all four monitors are successfully recognized. By the way, when you are hooking up more than three monitors using AMD graphics card, at least one monitor has to be connected via Native DP. Without knowing that, I used a DVI to DP cable, and I had a difficult time trying to get the fourth monitor hooked up. I probably became more of a saint trying to do that.
Up to this screenshot, the resolution is not yet 5120 x 2700. Since I have arranged four 2560 x 1600 monitors in 2 x 2 tile configuration, the resultant resolution would be 5120 x 3200. After configuring Eyefinity, the vertical resolution should be reduced to 2700.
For your information, a doubled resolution of 2560 x 1440, which is commonly referred as QHD, a resolution of 5120 x 2880 is specifically called as UHD+. For some crowd, this is the true 5K among other things. On the other hand, 4800 x 2700 is also considered as one of the 5K resolution standards among the ones with 16:9 aspect ratio (this resolution is also favored by some optical equipment manufacturers), and this resolution can be obtained by arranging nine HD+ monitors with 1600 x 900 resolution in 3 x 3 tile. That reminds me that the 5K resolution that this article describes (5120 x 2700) has the horizontal resolution of UHD+ and vertical resolution of 4800 x 2700.
So while I was adding a fifth wheel to my writing, I have finished the Eyefinity configuration. But up until now, the resolution is still 5120 x 3200.
...the resolution you see is finally 5120 x 2700.
This is the gaming screenshot. What you can see does not look that much different from the 4K screen shot that you have seen above, but if you click it to see the picture in its real size, you can see the scary, and realistic details of the Alien. By the way, while I was running the test, I now have a display of my dream: Samsung Curved 5K Television!
Chapter 3 <5K for NVIDIA>
For I have passionately expressed my ANGER in the previous chapter, this chapter is somewhat more succinct than the last chapter. Earlier I have described my journey of setting the resolution for AMD graphics card in the ascending order, but this time, I will start with 2560 x 1600, and continue on in the order of 4K -> 5K -> 5760 x 1080 -> and 7680 x 1600. The reason that I have decided on this order is that the first three resolution can be instantaneously made in a single monitor using up-scaling option in NVIDIA control panel.
Configuring 2560 x 1600 resolution
I will skip my explanation since this is the basic for the 30" monitor. You don't need any explanation, don't you? :)
The game that I used in the previous chapter is Alien vs. Predator, but I thought it would be better to have a display of an open space than a narrow hallway of USCSS Nostromo to show the difference in view angle according to the difference in aspect ratio, I chanced the game to Metro 2033(though technically this game also shows an indoor space).
Configuring 4096 x 2160 resolution
The screenshot above shows the reason that I was not able to describe the progress in the ascending order of the resolution. Also, this was the reason that enticed (!) me to begin my experiment as well. What it is, if you are using NVIDIA graphics card, it is so easy to have a custom-made resolution by adding a custom resolution in the control panel, as long as the monitor is able to support the resolution.
By going through these steps...
Just by entering the vertical and horizontal resolution and refresh rate, everything is set!
4K testing. No Problem.
5K testing. No Problem.
This shows the Windows recognizing new custom resolutions, which I made in the NVIDIA Control Panel, registered as its very own. That easy.
Once again, I applied the 4K for the game testing.
As you can see, this works great. Again, regardless of a huge increase of the pixel numbers, the information shown on the screen did not increase much, since the aspect ratio was increased by just a little bit (16:10 -> 256:135).
This time, I have applied 5K again for testing. It works great as well.
I'm going to emphasize this once more that this graphics card can realize 4K / 5K with a single monitor instead of 2 monitors / 4 monitors with the AMD counterparts, respectively. Well, the fact that one monitor can have such a high resolution enticed me to start the project... *sigh*...
There is no need to even explain this, but the readability is extremely low because of having such a high resolution in such a "small" space of 30" monitor. Not to mention a document, that you are supposed to read off it, I had difficulty setting options for the game just because even the in-game control panels are barely readible. The only significance of a 30" monitor being under 4K / 5K resolution is that this crazy resolutions are, actually "possible".
Since the aspect ratio is the same, you might think "same as before... What's the difference?" ...But I urge you to click and see picture in its real size to see what I am talking about.
Configuring 5760 x 1080 / 7680 x 1600 resolutions
Though NVIDIA is highly praised compared to AMD in terms of the driver's completeness, AMD is much more flexible when it comes to the setting of surrounded display. In fact, it is even simpler, too.
First of all, the "SurroundVision" by NVIDIA can only add three panels, and these panels have to be identical - No, not only for its size and supporting resolution but also for manufacturer. In fact, they are need to be exactly the same. On the other hand, AMD can have different panels together as long as the supporting resolution is same, and it can have two or four grouped together at the same time. The configurations such as 2 x 2 tile, which is the method that I made 5K for AMD, cannot be done here.
...These are the steps for configuring SurroundVision. I'll skip the details.
The screenshot shows a resolution of 5760 x 1080.
This is what it looks like when a gaming test was conducted.
7680 x 1600 works fine as well.
At a glance, it looks somewhat similar to 5760 x 1080, but when you look at the pillars on the right side, one less pillar is shown here compared to the former one.
Chapter 4 <Photo Gallery>
Finally, the prologue is over. In this chapter, I will show you the pictures of actual display that I had to configure in order to actualize the resolution by using the methods described in the previous chapters.
First of all, this is the configuration using a single monitor. The resolution of 2560 x 1600 (for both AMD and NVIDIA) and 4K / 5K (NVIDIA only) are displayed in this format.
The next one is the configuration using two monitors (2-monitor configuration). To display 4K resolution in AMD, each monitor's resolution was set to 2048 x 2160, and the two monitors were merged via Eyefinity. After all as I am uploading the picture, I started to wonder why I did not configure the monitors in portrait, because the vertical side is longer than the horizontal side (2160 > 2048). Yes, I already know I'm smart. Anyway, I have configured the 2-monitors setting by adding two landscape-positioned monitors (2 x landscape monitors)
Both AMD / NVIDIA require three-monitor configuration in order to realize 5760 x 1080 and 7680 x 1600 resolution. The picture shown is three landscape-positioned monitors (3 x landscape monitors).
For your information, I thought of pivoting three monitors and have those three grouped together in portrait position instead of what you see here if the 5K is not realizable by all means. In this case, since I have rotated the 2560 x 1600 by 90 degrees, the individual resolution will be 1600 x 2560, and 4800 x 2560 combining all three. This resolution, while getting a similar number of pixels of 4800 x 2700 (which is another 5K standard for 16:9 aspect ratio), will have a similar aspect ratio as the resolution of our favorite 5120 x 2700. In this case, I have even thought of overriding the problem of this (4800 x 2560) having the same number of pixels, 12 millions, as 7680 x 1600 by using three QHD (resolution of 2560 x 1440) monitors with surrounded configuration, thus making a 7680 x 1440 with 10.8 million pixels, and this was to differentiate the pixel number for each scenario (it is because the gaming performance is inversely proportionate to the pixel numbers). But, since the up-scaling was successful, this plan was not actually used.
This is what I was supposed to set as "Plan B", just as described above.
And this is... the long-awaited...
In order to level the height, I used eight graphics card box, 4 stacked boxes for one monitor.
At that point, the whole workplace scene was not captured by the camera lens. Maybe I should blame the fact that the space is quite small...
I say this again to make sure that this configuration was only for AMD. NVIDIA does not even have an option of having an even-numbered surround configuration. (I heard that NVIDIA Quadro supports the output of 4 and more monitors -don't tell me that's the reason why this graphics card was named Quadro-, but I am not sure if it is conceptually the same as Eyefinity or the SurroundVision.)
Chapter 5 <Test Setup & Methodology>
Let’s talk about the benchmark from now on. First of all, let me explain the test setup and methodology. The system setup that I used for the test are shown in the picture below.
In every game that I used, the output option was set to the maximum (highest possible) in each game, excluding anti-aliasing (AA). However, for the games that support PhysX, PhysX was deactivated for the fair comparison between AMD and NVIDIA products, and each resolution was compared with an option of applying AA and 4x AA to see the effect of having anti-aliasing on the graphics card performance.
If the game does not support the option of altering levels on anti-aliasing (i.e., Metro : Last Light), SSAA (which is the only option available in the game) was applied, and if there is no such quantitative leveling option but has qualitative options for the degree of anti-aliasing (i.e., Sniper Elite V2), the second-highest option was chosen and applied (For example: As for Sniper Elite V2, "Extreme" was the highest option, and the second highest option was "High" so that "High" was the option applied for the game). On the other hand, as for Sleeping Dogs, it is not a problem that the highest option available for anti-aliasing named "Extreme" is the 4x anti-aliasing, but it was impossible to "disable" anti-aliasing. Therefore, the lowest option (Normal; FXAA) was applied, regarded as no anti-aliasing, and recorded. Lastly, as for Bioshock : Infinite, the anti-aliasing setting could not be altered using the benchmark tool, and as a result, the anti-aliasing test was replaced with options of enabling and disabling DDOF. These conditions mentioned above are also recorded in the result section for each game.
The methodology is same as always. For each comparison group, a test was done for three times for each game / scenario / and anti-aliasing, then the median value of the three tests was taken. If there is any unusual circumstance that cannot be generalized, then the circumstance was dealt accordingly using previous experiments and experiences. If you have any specific question, please let me know, and I will be more than glad to answer those questions for you.
Chapter 6 <Result>
Here are the results. Have fun with them all :)
(1) 3DMark 11
(2) 3DMark 2013
(3) Aliens vs Predator
(4) Batman : Arkham City
(5) Battlefield 4
(6) Bioshock : Infinite
(7) Crysis : Warhead
(8) Crysis 3
(9) DiRT : Showdown
(10) Hitman : Absolution
(11) Just Cause 2
(12) Metro 2033
(13) Metro : Last Light
(14) Sleeping Dogs
(15) Sniper Elite V2
(16) Splinter Cell : Black List
(17) Tomb Raider
Chapter 7 <Performance Summary; Raw Data>
So, here we are at the Performance Summary after getting done with the *boring* Result chapter.
Up until now, I have tried to use two different methods for analysis of the result (in a single chapter). However, for this article, I have separated the summary chapter into two. The first one would be dealing with the "Raw Data", which includes the sum of the framerate for each game. The second section, named as "Normalized", is the average value of relative framerate (= normalized) for all games, obtained by the framerate of each graphics card for each game divided by the framerate of GTX TITAN Z for each game. The more reliable method would be the latter (as the raw data only provides what is obtained in the experiment, it cannot be called as an "analysis" per se), the Raw Data graphs were made to show that there are more ways to reach a conclusion by using a set of given data.
In summary, we could conclude the below after analyzing the result obtained at 5 different resolutions:
- As the resolution gets higher, the relative ranks of AMD graphics cards improve
- As anti-aliasing gets enabled, the relative ranks of AMD graphics cards improve
But when focusing on the details, the generalization above does not always work. For an example, when going from 5760 x 1080 to 4K or from 7680 x 1600 to 5K, though both of the cases went from lower to higher resolution (=more pixels), the relative rank of AMD family would remain the same, or lower. Of course, still, the result of AMD seems better when compared with 2560 x 1600.
Let's find out if the in-depth analysis would prove the same in the next chapter.
Chapter 8 <Performance Summary; Normalized>
Since I have already explained what "Normalized" means in the previous chapter, I will jump straight to the graphs.
In this chapter as well, the big picture seems identical as the previous chapter.
- As the resolution gets higher, the relative ranks of AMD graphics cards improve
- As anti-aliasing gets enabled, the relative ranks of AMD graphics cards improve
Also, you probably have noticed that AMD's relative rank has gone up in this chapter compared to the previous chapter throughout all scenarios - this is the third statement. This is caused by the games in favor of NVIDIA, such as Batman: Arkham City. As a result, the Raw Data, which are just a mere added value of frame rate of each game, could show that AMD's performance is not as good as NVIDIA. However, in "Normalized" analysis, AMD "performs" better as Normalized data signifies the average value of relative performance of the graphics card for each game. To summarize, the improvement of AMD graphics card is a result of losing a big time for a little number of games, and winning by a little in other majority of games.
Also, when results for each resolution were analyzed in-depth, it can be seen that the trend observed in the earlier chapters is repeated.
- The relative rank of AMD drops in 4K than 5760 x 1080 and in 5K than 7680 x 1600
There is no obvious explainable reason for this, but one thing that is suspicious for the result is Eyefinity itself. As I have explained through the lengthy prologue, while NVIDIA was able to realize the specific resolution with a single monitor, AMD had to configure Eyefinity using either 2 or 4 monitors to realize 4K and 5K, respectively. The fact that the Eyefinity had to be utilized for achieving such resolution probably had given AMD a penalty. When you think intuitively, it would be strange to have same amount of computational burden between having a single monitor as the output device and dividing the output among several monitors. Well, this is just an assumption of the cause, and the "facts" revealed from the results are indicated in the statements above. The result showed that higher the resolution, higher the relative rank of AMD, and in 7680 x 1600 resolution, two 290X graphics card showed 26% higher performance rate when compared with GTX TITAN Z. However, in 5K, the performance went down a little bit, and showed only 17% higher performance rate. The answer to the question shall be revealed not until when 4K/5K resolution can be realized in AMD products using a single monitor (Therefore, I should purchase Samsung Curved Television when I get my next paycheck...)
With a lengthy prologue and numerous graphs, I have dealt with the subject of "Gaming at 5K resolution". The meaning of this research can be summarized as the following:
- Gaming at 5K resolution was extremely slow, regardless of using AMD or NVIDIA. Simply, it sucks.
- (Though I have stated the above) when performance is the only factor, a pair of R9 290X was the best by far. In other words, R9 295X2 would be the best single graphics card out there.
- (Though I have stated above two) the only single graphics card that can run every single game at 5K is GTX TITAN Z.
As a result, AMD and NVIDIA got the same score. As NVIDIA stated when GTX TITAN Z was released, indeed it was the first graphics card in history that is actually a "5K-ready". Also, though AMD could not run some of the games under certain settings, it gets the title of "the fastest graphics card for any resolution" among the ones in the market for the games that it supports (Though I did not include R9 295X2 due to the circumstance, it is assumed to be similar to 290X CrossFire when performance is considered). Especially, the Hawaii family had difficulty in showing its strength in a low-resolution setting due to its characteristics (512bit memory bus and 64 ROP), but the true power of Hawaii products were shown in an extreme-high resolution setting. Brethren, this is the very first scenario and the result from it on the face of the planet, since no hardware website has dealt with this topic -5K- before.
When we step back for just a little bit from the cutting-edge graphics cards such as GTX TITAN Z and R9 295X2, the limitation of existing products, especially the GTX 700 series, are exposed through this article. The article gives out an obvious answer to the question of "Where should we use the 6GB version of GTX 780 / 780 Ti?". The answer is, that you should use it under the resolution of 7680 x 1600 or above.
Thinking realistically, while 4K and 5K resolutions need more time (maybe a couple of years) for any actual use, the resolution of 7680 x 1600 can be simply made with three 30" monitors, or even, three 27" QHD monitors. In other words, 7680 x 1600 is "today's" resolution that we might utilize right now. Also, it is evident that the 3GB of memory is starting to show its limitation, though the limitation has not been addressed since no hardware website dealt with such surrounded configuration. Looking at this, for those of you, the owners of GTX TITAN (Non-Black) who had been bitter up until now, should be proud of what you got.
...Finally, this long article is about to end.
Thank you for reading this lengthy article, and I hope you have a great day!
*Editor's note : I really appreciate your efforts on this work and yourself, dear Wendy :)*
'Graphics > Graphics Round-Up' 카테고리의 다른 글
|Catalyst War 2014 S/S (2)||2014/08/11|
|Watch Dogs : Benchmark & Analysis (2)||2014/07/01|
|The Resolutionary War : GTX TITAN Z @ 5K (17)||2014/05/20|
|MANTLE : A half-year achievement (10)||2014/03/30|
|Entry-level Gang Fight (6)||2014/02/11|
|Crossfire vs SLI 3 - (2) Mainstream Gang Fight (0)||2014/01/09|