Review Index:
Feedback

Frame Rating Dissected: Full Details on Capture-based Graphics Performance Testing

Capture Hardware and Extracting the Data

Capture Hardware

Once we have the overlay working, the next step in our Frame Rating process is to actually capture the video for later analysis.  While that sounds easy, it is far from the case.  Capturing the video on the testing system is out of the question. Instead we are using an external capture system built around an Ivy Bridge platform and a high-end capture card from Datapath, the VisionDVI-DL.

View Full Size

This card is capable of capturing video at 2560x1600 at 60 Hz with throughput as high as 650 MB/s.  We were actually the first customer for this card in the US and have spent quite a few days on the phone with driver engineers to make sure this kind of testing process was successful.  Meant mainly for single frame captures, getting the VisionDVI-DL to really meet the specifications turned out to be a more complicated process than we’d hoped.  The good news is we have it all figured out now and can reliably capture the necessary video.

The VisionDVI-DL essentially acts like and shows up as a monitor to the system.  It reports an EDID (extended display identification data) to the graphics card that is set through the driver on the capture system.  I am able to emulate a 1920x1080 monitor at 60 Hz refresh and up to a 2560x1440 @ 60 Hz as well.  Using a Gefen dual-link DVI splitter we can connect a single display output to the NVIDIA or AMD GPU (to avoid the issues of power consumption and GPU utilization with multiple monitors attached) and split the signal to our capture card and to a monitor for us to actually play the games.

With a data rate of nearly 500 MB/s when capturing video at 2560x1440 @ 60 Hz, some serious storage speed was needed for this capture system.  I decided to use our Pegasus R4 Thunderbolt array coupled with a set of four Corsair Force GS 240GB SSDs running in an RAID 0 array, giving us about 900 MB/s of write capability.

View Full Size

View Full Size

I cannot overestimate the importance of a completely correct capture in this step of the Frame Rating process.  If you deviate from the 60 Hz record rate even a single frame the data becomes basically useless.  Dropping a frame on the capture side would show up as dropped frames from the graphics card so avoiding that is paramount.  The software used for this capture will likely be VirtualDub on other sites that you see, and that was the program we used originally too.  However, the inability to automate the settings and having to re-adjust the options each time you load the program became a hassle so we developed our own that uses DirectShow filters and the Windows Platform SDK. 

 

Extracting the Data

Once you have the recorded raw AVI file, (that is about 25GB – yes GB – from a 60 second capture at 2560x1440) the next step is to actually read the colored bars and generate frame time data, runt data and drop data.  Using a basic DSP application, the extractor reads in each frame of the video sequentially and determines the number of scanlines in height that each color on the overlay bar has and records that information in an XLS file.

View Full Size

This is probably the most fool-proof part of the process.  I have uploaded a few example XLS files right here for you to see the output.

March 30, 2013 | 05:43 AM - Posted by Ryan Shrout

Interesting, hadn't considered the pros/cons of smoother or erratic input polling.

March 30, 2013 | 08:47 PM - Posted by bystander (not verified)

I have a hard time trying to grasp exactly how erratic input would affect the results. I have a feeling, based on my constitution (I get simulator sickness with poor latency), that the best case is which ever has the lowest worst case interval times.

March 31, 2013 | 01:55 AM - Posted by bystander (not verified)

..But then you have occasional latency increases. Of course those increases are to remove redundant frames, and once increased, they probably don't need much adjustments most the time.

This whole topic always gets me going back and forth, but my instincts is overall, even if latency is considered, even spacing matters more as it adds more useful points of input, assuming it adds only marginal/occasional increases of latency.

March 28, 2013 | 08:20 PM - Posted by Bob Jones (not verified)

Can you address the visual quality differences between the two cards? Specifically on Sleeping Dogs, the 660 Ti seems to be missing some lighting sources outside - most noticeable is the cafe/shop lights before you go down the stairs, and then the store across the street right at the end of the vieo.

March 29, 2013 | 10:46 AM - Posted by Ryan Shrout

Time of day in the game when those videos were recorded?  They should be basically identical, I'll check.

March 29, 2013 | 03:10 AM - Posted by I don't have a name. (not verified)

Fascinating article. I think it'll take a few reads to fully comprehend everything that is being said. Thank you indeed, I found it fascinating. Certainly, as a 7970 owner, I'll be holding off on a potential second 7970 purchase for the time being.

March 29, 2013 | 05:09 AM - Posted by rezes (not verified)

The last GeForce 314.22 Beta Drivers and last Radeon drivers 13.3 beta 3. Please using these drivers on yours test.

AMD driver may be more stable on this drivers!

March 29, 2013 | 10:46 AM - Posted by Ryan Shrout

Nothing has changed on the issues in question.

March 29, 2013 | 05:15 AM - Posted by technogiant (not verified)

Thanks for the great article and all the work you've done guys.

I run 2x gtx 460's is sli and while I dislike screen tearing I've noticed that options such as vsync, active vsyn and frame rate limiters actually make the experience less smooth as appears to have been highlighted in this article.

I've considered getting a 120Hz monitor just so I can run without any of those options at a decent frame rate but use sufficiently high settings so as not to go above 120Hz and so incur screen tearing.

Thinking further I'd like Nvidia to develop a variation of their gpu boost technology that would actually down clock the gpu to prevent frame rates from exceeding the monitors refresh rate.....think this would give the benefits of no screen tearing without the negatives of vsync and the like.

Thanks again for the article guys.

March 29, 2013 | 05:37 AM - Posted by technogiant (not verified)

Actually using gpu boost dynamically to both under and overclock the gpu to achieve a target frame rate could be a very nice way of producing a smoothed experience without any of the negatives of other methods as its occurring directly at the gpu instead of in the game engine to display time line.

March 29, 2013 | 08:01 AM - Posted by Pick23 (not verified)

So is this article saying that even with the new testing methodologies:
1.Single card 7970ghz is still slightly better than the 680
2.Crossfire absolutely sucks
?

March 29, 2013 | 08:18 AM - Posted by John Doe (not verified)

7970 Ghz is slightly better than a 680 ONLY at stock.

When you're comparing 7970 Ghz to 680, things ENTIRELY depend on clock speeds since the 7970 Ghz is nothing more than a pre-OC'ed 7970.

But yes, CF does indeed sorta suck. Still.

March 29, 2013 | 07:26 PM - Posted by Anonymous (not verified)

Sweet 7970 still the best card under.. well under $1000 lol

March 30, 2013 | 09:21 PM - Posted by steen (not verified)

What's stock for a 680? ;) 7970 GE is slightly slower than Titan...

CF sucks just like SLI. What's your poison, input lag or frame metering? Do poeple understand what "runts" are? CF is actually rendering the frames, you just don't benefit as they're too close together. One frame renders the top 1/4 of the screen when the next frame starts. Your top ~200 lines are the fastest on your screen. ;)

(Sorry for the previous multi-posts. Don't know what happened there.)

April 7, 2013 | 01:15 PM - Posted by John Doe (not verified)

I have ran far more CF and SLi setups than you did FOR YEARS and understand far more abouts these things than you and your little silly mind does.

March 29, 2013 | 11:39 AM - Posted by fausto412 (not verified)

interesting piece, good job pcper.com

now I wonder if when AMD does a global fix my 6990 performance will be dramatically improved on bf3?

and what is the effect of correcting this to latency and actual framerate? will we see FPS go down at the expense of frametimes?

It is Obvious Nvidia was on top of this for some time...I just don't see a proper fix in 120 days.

March 29, 2013 | 12:24 PM - Posted by Chad Wilson (not verified)

Just out of scientific curiosity, did you do a run without the splitter in the mix to confirm that the splitter is not adding any latency to your tests?

March 29, 2013 | 12:33 PM - Posted by Anonymous (not verified)

Do people actually use vsync without some form of triple buffering?

I don't see how. Jumping between 30 and 60fps or whatever is not an enjoyable, nor smooth experience.

So, if you can enable vsync AND allow the game to sweep through a normal range of available framerates, does this negate the increased frame times of constantly switching back and forth between high fps and low fps?

March 30, 2013 | 12:56 AM - Posted by bystander (not verified)

V-sync, even with triple buffering, still jumps back and forth between 16ms and 33ms, but it does it between frames. A 120hz monitor helps here, as you can have 25ms frames too, so it is less of a variance.

March 29, 2013 | 12:36 PM - Posted by Anonymous (not verified)

Furthermore, is playing a game without vsync enabled REALLY an option?

Are you sure gamers all over the world disable it to be rid of the latency issues? I'm not so sure.

I'll happily take a little latency in a heated round of counter-strike than end up dead, or missing my shot because 50% of the screen shifted 8 feet to the right. (screen tearing).

Pretty much all games are unplayable without the use of vsync and I'm not convinced it's a personal preference, either, if you enjoy your experience while you're tearing frames - I'd just call you a mean name that insinuates you're not telling the truth.

March 29, 2013 | 02:00 PM - Posted by Marty (not verified)

If you are a competitive player, VSync is not an option, you are lagging an extra frame behind.

March 29, 2013 | 12:45 PM - Posted by rezes

Where is the new tests? and when?

March 29, 2013 | 02:20 PM - Posted by Anonymous (not verified)

So how much did nVidia pay you?

While I can see the potential in this kind of testing, and some of the issues you have mentioned are valid, you have drawn quite a bold and one sided conclusion using the competitor's software. I'll save my judgements for when this becomes open source.

March 29, 2013 | 08:12 PM - Posted by Fergdog (not verified)

It's not purely Nvidia made software, if you read the article or paid attention to this site you'd know Ryan co-developed this benchmark with Nvidia and he's been working on it for a long time.

April 2, 2013 | 11:10 PM - Posted by Anonymous (not verified)

nVidia has to do almost everything, the amd fans need to get used to it.

AMD's years long broken bottom line and years of massive layoffs and closings mean they claim they "weren't even aware of this issue !? !! "

- like every other total screw up AMD gets caught on with their dragon drawers on the floor next to their spider platform, buck naked in epic fail and presumably clueless...

Maybe we need some depositions and subpoenas of internal emails to see just how much they covered up their epic fail here.

March 29, 2013 | 08:00 PM - Posted by Fergdog (not verified)

Quick question, for adaptive vsync, you put it on half refresh rate didnt you?

March 29, 2013 | 08:02 PM - Posted by Fergdog (not verified)

Half refresh adaptive only really makes sense on 120hz monitors, not sure why you used that setting on a 60hz monitor for benchmarking.

March 29, 2013 | 11:06 PM - Posted by Anonymous (not verified)

Hoping you post the titan and and info today as promised!!!!!!!!!

March 29, 2013 | 11:07 PM - Posted by Anonymous (not verified)

Titan and amd

March 30, 2013 | 12:56 AM - Posted by MaxBV (not verified)

Waiting on that GTX 690 and Titan article still, hope you guys haven't forgotten.

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

By submitting this form, you accept the Mollom privacy policy.