 | |||
|
|
|
|
|
 |
Videocard Display Quality: Rapid Communication November 07, 2003 Alan Dang |
|
 |
|
| |
Summary: "2D or not 2D that is the question." -Alan
In this short preview, Alan documents the performance differences between plugging in a monitor into a regular surge protector and plugging it into a high-end line conditioner. He also starts to look at the differences between ATI and NVIDIA cards. You might be surprised with what he found out.
 Introduction | Page:: ( 1 / 4 ) |
Over the weekend, I spent an hour doing some initial image quality tests on my own just for fun. After an hour of testing, I discovered that the differences were noticeable enough that I had plenty of data for an initial article. Rather than have you wait until next year for a definitive article, I believe there's enough data to get started with an initial look at the current set of data. I'm going to keep my commentary to a minimum and be cautious with my conclusions so that they will not change with additional data. This is a peek at a work-in-progress, a "rapid communication" article of sorts.
Test Setup
iiyama i90A Monitor at 1600x1200, 32-bit color, 85 Hz, 6500K
In order to effectively evaluate the source picture quality of graphics card, we need to make sure we have a good way to display the picture and objectively analyze the output. I don't have results from all of the monitors, but today I'll show you the results from my iiyama i90A. The iiyama monitor is a 19" Diamondtron NF screen with an excellent 300MHz bandwidth, and 115kHz horizontal sync. There is a rock-solid power supply behind the electron guns making it a very good CRT.
Canon EOS Digital Rebel
The Digital Rebel is a 6 megapixel 1.6x crop digital SLR. For still images, this digital SLR offers superb image quality and contains Canon's latest firmware. I have spent much more time with the D30 but this bargain priced D-SLR is what I grab when I want to go stealthy or when I need something lighter.
Canon BG-E1 Battery Grip
The BG-E1 doesn't add anything to the test, but since it was used, I've included it in the system spec.
Canon EF 100mm f/2.8 Macro USM lens
The Canon EF 100mm f/2.8 Macro USM is among Canon's sharper prime lenses. We've stopped down to f/8.0 for all our shots and chosen a conservative 1:5 reproduction ratio.
Manfrotto 3021BPRO Tripod Legs and Manfrotto 488RC0 Ball Head
The best camera is useless if the final image is blurred from camera shake. Our Manfrotto setup is more than adequate for the equipment we're using.
NVIDIA GeForce FX 5900 Ultra Reference Board
FIC ATI Radeon 9800 Pro
Our colleagues at Extremetech have noted that the reference boards from NVIDIA are better than most retail boards when it comes to digital DVI output. No one's sure how this affects the analog output at this time.
Monster Power HTS 3600 Stage 3 Version 2.0 Line Conditioner
We're going to test no-line conditioner versus a Monster Stage 3 Version 2.0 line conditioner.
All shots taken at f/8.0, 1" exposure, ISO 100 under identical lighting conditions. 1:5 Macro Reproduction Ratio
Some of us may question the ability of our system to resolve 2D image quality differences. To demonstrate the optical quality of the setup, I present you these two samples:
[image]
| <% print_image("01"); %> | <% print_image("02"); %> |
SIDEBAR: I'm so poor, I don't even have the new $20 bills
| Line Conditioning | Page:: ( 2 / 4 ) |
For today, we're just going to look at the results from a few test images. As you know from our previous article, we had a lot to say about AC line noise and the importance of conditioning for system stability. To begin this article, I'll now show you the difference between image quality with a high-end AC line conditioner. Like Monster Cable for consoles, the improvement is small and it's up to you to decide whether or not if the improvement is worth the price to you – we're just showing that there is a measured difference.
In these series of tests, we use an ATI Radeon 9800 Pro from the Dual Opteron workstation built in my previous article. The system is always powered through the Monster Power line conditioner. The only difference is where we plug the monitor in. In one case, it goes to a Curtis 9 Outlet Surge Protector with 2163 Joule Energy Rating, and an EMI/RFI filter. This is your run-of-the-mill middle-grade surge protector. In the second test, we plug the monitor into the HDTV AC outlet from the Monster Power HTS 3600 power conditioner.
We used a series of test images and then looked at the histogram output from the full-resolution samples. We're not posting those samples here in the interest of not-wasting-bandwidth. Each image is 2-3MB each and it's really the histograms that tell the story.
Green
[image]
| <% print_image("03"); %> | <% print_image("04"); %> |
[image]
| <% print_image("05"); %> | <% print_image("06"); %> |
We start with green because that is where our camera is most sensitive. With a good monitor, you can actually see the difference with your eyes, but it is slight. Looking at the histograms, it's much easier to see what the Monster Power conditioner brings to the table. The areas under the curves on the left are smaller, and the blue and red overlap better, meaning better color neutrality.
Red
[image]
| <% print_image("07"); %> | <% print_image("08"); %> |
[image]
| <% print_image("09"); %> | <% print_image("10"); %> |
You really cannot see a difference between the two images off-hand. Looking at the histograms, it's clear that the line conditioner produces a "redder" red in the picture. This is seen not in the big peaks on the left, but the tiny one on the very right on the screen. The low-luminance peaks on the left look paradoxically wider with the line conditioner, but we really have to determine the area under the curve.
Blue
[image]
| <% print_image("11"); %> | <% print_image("12"); %> |
[image]
| <% print_image("13"); %> | <% print_image("14"); %> |
The area under the low-luminosity blue and green curves appears to be smaller. The amount of aberrant red is also less with the line conditioner.
Black
[image]
| <% print_image("15"); %> | <% print_image("16"); %> |
There's no need to include screenshots of black level tests ;). Once again, the line conditioned output is superior, with a "blacker black."
Power Conditioning Conclusion
The differences are measurable, though slight. On the other hand, it is sort of impressive to think that there's a difference (measurable with a digital camera), depending on where you choose to plug in your monitor.
This brings me back to the same discussion I made in the console cable article. Everyone has a home theater and a computer. Not everyone will want to buy an AC line conditioner. We've shown you that, yes, if you want the absolute best possible picture quality, an AC line conditioner can provide a measurable benefit. Whether or not this benefit is subjectively significant enough is a personal decision for you. From our previous article, we've said that we found the differences to be meaningful with a HDTV, and less so with our standard non-HDTV NTSC XBR TV.
Clearly, before you buy a fancy line conditioner, you will want to make sure you also have a good monitor in the first place. The analogy would be spending thousands of dollars on an HDTV, but electing not to wear glasses/contacts if you need them.
For now, we'd recommend getting AC line conditioning for computer stability, and to consider a videophile/audiophile grade line conditioner if you're passionate about your audio and video performance.
Since the Monster Power proved to offer better picture quality, the rest of article is a system powered entirely by the Monster Power line conditioner.
SIDEBAR: Don't forget to go Siteseeing.
| Sharpness | Page:: ( 3 / 4 ) |
So far, we found no measurable differences in the horizontal lines being output by the graphics cards. There were differences in the reproduction of the vertical lines. As mentioned above, this is highly sensitive to the monitor settings. To let you better understand what we're looking at, we're including one high-resolution sample. Look at the vertical lines and notice that some are double:
[image]
| <% print_image("17"); %> |
We took a close look at the 11 vertical lines to see and counted which ones appeared to be a single solid line at the top of the screen. Here are our results.
Green
ATI: 1
NVIDIA: 1
Red
ATI: 9
NVIDIA: 5
Blue
ATI: 10
NVIDIA: 7
We then looked at the bottom of the screen and counted again
Green
ATI: 1
NVIDIA: 0
Red
ATI: 2
NVIDIA: 2
Blue
ATI: 9
NVIDIA: 7
Does this mean that NVIDIA isn't outputting as sharp of a signal? You cannot tell from this test alone. On this particular setup however, the ATI card is producing a sharper picture. That's what the data unquestionably shows.
Then again, my iiyama monitor has a variable aperture grill of 0.25 - 0.27 mm. A perfect output of evenly spaced lines may look worse because the fences are spaced out unevenly. What's sharper on this setup could mean that it gets distance increases from the center. In other words, in the case of my older monitor, one card may very well produce better sharpness – but on something like a 0.24mm aperture grill monitor or something like an Invar Shadow Mask, the results may be reversed…or maybe it's a non-issue.
We don’t know if the variable AG versus fixed AG (i.e. 0.24mm) reflects upon the graphics card, but that's why this is just a "Rapid Communications" article, and not the definitive article.
SIDEBAR: Would you rather wait for the final article? Or is it more exciting to read about the discoveries as they happen?
| Color | Page:: ( 4 / 4 ) |
I'm a big believer in having good color. The look of a 3-megapixel photograph from a D-SLR with accurate color is still far more pleasing a 6-megapixel consumer image. Photographers often talk about this quality as being images that are sharp, but not brittle.
We have a much more comprehensive color suite planned, but using the same tests, we decided to look at the purity of each color.
[image]
| <% print_image("18"); %> | <% print_image("19"); %> |
The greens seem similar enough, but since our test images are only supposed to contain green and "black", we would expect the overlap in the blue and red curves. The ATI seems to be producing a slightly warmer image, which is consistent with what is experienced subjectively.
[image]
| <% print_image("20"); %> | <% print_image("21"); %> |
[image]
| <% print_image("22"); %> | <% print_image("23"); %> |
Figuring out how the red and green differ is going to require more work and we'll need to turn to more advanced tools than just histograms.
Conclusion
As a rapid communication, there's not much we can say. The one conclusion I feel strongly about is that line conditioning can provide a small, but measurable benefit in picture quality. Again, the standard disclaimer exists -- I live in San Francisco, known for its old power grid. It's hard to assess how this affects the change, but these tests show that line conditioners aren't smoke and mirrors when it comes to performance.
As for the NVIDIA versus ATI image quality question. This article establishes the fact that there are noticeable differences in signal timing (sharpness) and color balance, and so more research will be needed on our part to characterize the differences. Right now, based upon our one monitor, the ATI is producing a sharper picture, but NVIDIA is producing more accurate color.
Whether it is contrast, color, or sharpness that makes for the best picture overall is something photographers have long debated and is something we'll try to address in the next version.
SIDEBAR: Questions or comments? Have you noticed a difference between ATI and NVIDIA’s 2D display quality and have some experiences you want to share? Feel free to do so in the news comments!