Review Index:
Feedback

Frame Rating: Eyefinity vs Surround in Single and Multi-GPU Configurations

Author: Ryan Shrout
Manufacturer: Various

Discovered Problems with Eyefinity

Hopefully you followed (or just read) our previous Frame Rating stories and learned about the problems that AMD CrossFire had with runt frames and artificial performance results.  As it turns out though in a CrossFire + Eyefinity system, the issues are very different and likely not as simple to fix.  I found three individual concerns when running our AMD Radeon HD 7970s in CrossFire at 5760x1080: dropped frames, frame interleaving and stepped tearing. 

 

Dropped Frames

This is the easiest problem to understand and demonstrate.  With our overlay that is applied to game frames as they leave the game engine, but before the graphics system takes hold of them, we are not only to accurately measure performance on the screen but discover missing frames.  This is possible simply because we know the expected pattern of colors and can easily detect when one of them is missing. 

Our test pattern follows a sequence of white, lime, blue, red, teal, navy, green, aqua, maroon, silver, purple, olive, grey, fuchsia, yellow, and orange.  These colors then loop and we can use the scan line length of them to measure frame rates very accurately.  But if a color is missing in the order, say white then blue, we know that lime was missing and completely dropped from user’s point of view.

View Full Size

Thanks to The Tech Report for the tiled screenshot idea!

These successive captured frames from Bioshock Infinite skip the lime, red, navy, aqua and silver colors indicated that every other frame in this sequence that is being rendered by the game is NOT being displayed.  With single display testing we would see RUNT frames in those places but it appears that with Eyefinity the problem is worse and those frames that the game engine submits are never displayed.

Why is this important?  Tools like FRAPS that measure performance at the same point as our overlay is applied are essentially DOUBLING the frame rate in this instance, giving unfair and unrepresentative results.  In our Observed FPS metrics those missing frames are accounted for correctly with lower frame rates.

 

Interleaved Frames

Dropped frames were easy to understand but we are about to get more complicated with our discussion.  Look at this:

View Full Size

Click to Enlarge

What you are seeing is one frame being displayed at the same time as another, interwoven together.  This new problem is being dubbed “interleaved frames” and is unique to Eyefinity.  I have witnessed this happening in every game we have tested though the games do shift between dropping frames completely and interleaving them together in this way. 

This problem not only causes lower perceived frame rates but also can cause some visual anomalies.

View Full Size

Click to Enlarge

This image from Skyrim shows the problem at work once again and if you look at the vertical lines of the structure you can clearly see the result of an interleaved frame: portions of the image alternate between positions back and forth rather than once with typical Vsync tearing.

After seeing this specific issue I spent a lot of time making sure it was not caused by or influenced by the overlay itself.  To double check we captured the output of the screens without the overlay enabled and through simple frame grabs we found interleaving continued. 

View Full Size

Click to Enlarge

This animated image shows (in a very slow motion fashion) how the frame interleaving is seen in gameplay. 

http://i.imgur.com/9IJ1MVS.gif

For a full resolution of the above image, click here. (5MB)

 

Stepped Tearing

As it turns out, the interleaved frames also result in another visual anomaly we are calling stepped tearing.  Take a look at the image below:

View Full Size

Click to Enlarge

Not only are there intermixed frames in this screenshot but the horizontal tears are NOT level, instead they appear to grow or shrink.  As a result, and combined with the interleaving, these tears are more noticeable to the end user than the standard vertical sync tears and also are affecting observed frame rates.

View Full Size

Click to Enlarge

Look at this same Skyrim screenshot from above.  Notice on the left hand side you’ll see there are clearly five different frames slices (looking at the left side wood strut).  But the color bars on the overlay on the far left hand side only see three frames.  The first “slice” of that wooden support doesn’t start until somewhere in the middle of the mountain range area in the background. 

 

I was also fearful that perhaps our capture equipment was responsible for the errors and problems seen in these screen grabs and recordings.  To enforce the problem I picked up a camera that could record 120 FPS at 1280x720 resolution and recorded the Eyefinity configuration.  In this image below you can obviously still see the frame interleaving as well as the stepped tears.

View Full Size

Click to Enlarge

Clearly this problem is real and isn’t a result of our testing methods.  It was just nearly impossible to find before the creation of this overlay and the testing methodologies at work here. 

 

We have developed a working theory of what is happening and why we are seeing these interleaving and stepped pattern issues.  When running in a multi-GPU configuration of any kind the contents of the buffer from each of the secondary GPUs must be copied back to the primary card that has the display physically connected to it.  That card’s frame output buffer is then scanned out to the displays.

I believe that interleaving is caused by incomplete frame buffer copies from the secondary GPU to the primary GPU.  If the copy is slow or stalls at any point, the current scan out process (being drawn at the screen at that specific moment) starts to re-read the primary GPU’s frame buffer rather than the secondary GPU’s as it simply didn’t make it in time.  When you see a back and forth pattern of frame interleaving it indicates that the copy is actually “catching up” a bit but then falling behind yet again.  This may happen a few times before a third frame (from the primary GPU) is ready and is pushed to the buffer easily.

This is likely also the reason that we see dropped frames – the copy of data from the second GPU to the primary is slow enough that the next frame from the primary GPU is ready before a copy takes place.

It is also possible that the interleaving is an issue of synchronization and not bandwidth, but without more data and information from AMD it is hard to tell as an outside observer.

I do want to note before anyone comments ask this question: no enabling Vsync in the game or the control panel does NOT fix or change these problems.

The stepped tearing issue is caused by a different but related property – AMD does not appear to have a synchronization step on each scan line but instead on each pixel.  Rather than wait for each scan line to output before updating the frame buffer AMD allows the buffer to update at any point which would cause the variable height frame results we are showing you above.

I asked AMD for more information on why or how this is happening but they decided not to comment at this time.  When I approached NVIDIA they said only that they have enabled “locks and syncs” for their frame copies and scan line outputs on SLI.

UPDATE: We did finally get some feedback from AMD on the subject after quite a bit of back and forth.  Essentially, AMD claims the problems we are seeing are due only to synchronization issues and NOT from bandwidth limitations.  Since the file copies are done over the PCIe bus, only an instance of near 100% utilization on it would cause degradation – and the only instances of that would be from heavy system memory access.  However, if you are accessing main system memory with the GPUs in your PC then other performance bottlenecks are going to creep up before CrossFire scaling.

If that is the case then AMD should be able to fix the CrossFire + Eyefinity issues in the coming weeks or months.  A bandwidth issue would be much harder to deal with and could mean a fix would have never arrived for HD 7000-series users.

September 17, 2013 | 07:22 PM - Posted by Alien0227

PcPer.com, I am one of those "that are invested in the HD 7000-series of graphics cards and have already purchased their Eyefinity configurations on the promise of a great experience".

I thank you for your hard work PcPer! I am truly grateful!

Because of it so far, I have seen a dramatic improvement in single screen Xfire game play with my 2 HD 7870s.

I did however invest in triple 1920 x 1080p screens, 2 HD 7870s and I wish I could be rewarded with the performance expectations I paid for. The experience would be awesome.

I only hope AMD will strive to keep me as a loyal customer by listening to your findings and offer a solution shortly to this Eyefinity-CrossFire problem.

You and AMD will have my undying loyalty and gratitude for it.

If AMD does, it will surely make them the king of the hill in value, for dollar per dollar performance...

Are you listening AMD? Is this make or break time for your company? My next GPU purchase depends on how you react. I wish AMD great success!

p.s. Ryan and PcPer.com Team. Please be confident that you are doing a great service to AMD in the long run and their customers.

Marc Senecal
Fort Myers, FL

September 17, 2013 | 09:42 PM - Posted by Ryan Shrout

Marc, thank you very much for your comments and feedback.  We greatly appreciate the support!

September 18, 2013 | 04:10 AM - Posted by j (not verified)

I'm sorry but I have to correct you: gaming equipment is not an INVESTMENT. Unless you're a pro gamer is a consumer spending.

Anyway, I wish you all the best fun with your new hardware :)

September 18, 2013 | 10:21 AM - Posted by Aranyszin (not verified)

This is no place for a grammar Nazi, and you aren't even correct in the first place.

Yes, the word "investment" is often associated with money spent expecting a monetary return. However, that is not the only usage of the word.

The poster clearly is "invested" considering the amount of money he spent, which he spent anticipating a return - in this case, a return in performance.

Honestly, contribute something useful to the topic or stay quiet. This thread is about obtaining a correction to a product flaw, not "How to parse nomenclature".

September 18, 2013 | 10:03 PM - Posted by Roy Taylor (not verified)

Dear Marc, yes we are definitely listening. Drop me a line direct.

best regards
Roy

January 28, 2014 | 08:20 AM - Posted by Roy Taylor (not verified)

hi Marc, I can assure you that myself and several other senior executives and engineers at AMD are deeply committed to making sure you get the experience you deserve.

We believe that Ryan treats this subject fairly, reports accurately and he works closely with us. We are grateful of his recognition of our leadership where we have it, even if I dont personally agree about our competitors 'sexyness'!

We will back with more on this shortly,
thanks

September 17, 2013 | 09:33 PM - Posted by Jonny (not verified)

hello
there is 13.10 beta did u test it?

September 17, 2013 | 09:42 PM - Posted by Ryan Shrout

Nope, we were using 13.8 beta.  But 13.10 did not add any changes for Eyefinity / Frame Pacing.

September 18, 2013 | 01:43 PM - Posted by Jonny (not verified)

i got 7950 X2
I had a chance to buy another one at $ 100

Now I'm all out of it and save the next generation of NVIDIA ...
I'm very disappointed that AMD are selling a defective product ..

Thanks for the information.

September 18, 2013 | 04:44 PM - Posted by NvidiaPWNS (not verified)

AMD is always all about selling defective and inferior products. They did it before with their "Tri core" CPUs. Those were just Quad cores that had a 4th core that they couldn't get to work. AMD GPUs and drivers are shit. You get what you pay for.

September 19, 2013 | 06:05 AM - Posted by Anonymous (not verified)

You do realize that binning is a key part in the silicon-industry? By your logic, the Titan is crap because it only has 14 of the GK110's 15 SMX units activated.
When a produc-line is announced, there are actually few different dies being produced. 3820, 3930, 60 and 70 are all 8-core "xeons" with cores disabled due yield issues.
EVERYONE do this.

And please, stop spreading the false claim that the drivers are bad. Maybe they were back in good ol' '04, but that is long gone. AMD has actually had better drivers than Nvidia this generation...

September 21, 2013 | 11:13 AM - Posted by NvidiaPWNS (not verified)

In your dreams retard! NVIDIA PWNS YOUR AMD, NUB!

September 18, 2013 | 05:22 PM - Posted by mAxius

Are you sure http://support.amd.com/us/kbarticles/Pages/AMDCatalyst13-10WINBetaDriver...

I'm still wondering about this little gem... "PCI-E bus speed is no longer set to x1 on the secondary GPU when running in CrossFire configurations"

September 19, 2013 | 12:24 AM - Posted by JJ (not verified)

"But 13.10 did not add any changes for Eyefinity / Frame Pacing." - I am curious about this too, as the press release clearly states that it updated something regarding Eyefinity.

September 19, 2013 | 12:50 AM - Posted by JJ (not verified)

I'm fairly technical, but I am getting a little outside my level of knowledge here

1. http://www.bit-tech.net/hardware/2010/11/27/pci-express-3-0-explained/
Each lane of PCIe 3.0 only has 1 GB/sec

2. http://www.tomshardware.com/reviews/pci-express-scaling-analysis,1572-8....
This is a very old article, but THG did testing graphics cards by limiting the PCIe lanes available to graphics cards, and you can see a very large performance degradation.

3. Back to the "PCI-E bus speed is no longer set to x1 on the secondary GPU when running in CrossFire configurations" Note in the press release.

Would one of the cards in the setup being starved for bandwidth be able to account for these anomalies?

September 19, 2013 | 01:35 AM - Posted by Ryan Shrout

By AMD's admission, no. This problem listed in the Catalyst 13.10 notes only affects Crossfire configurations that do not use a bridge adapter across the graphics cards. This coming from AMD's own Andrew Dodd.

If this were a fix for our problems AMD would surely be crowing about it.

September 17, 2013 | 10:53 PM - Posted by Anonymous (not verified)

You guys rock pcper. Been here for years. Love amd, but they need to get their act together. Emphasis on the good faith part ya know. Don't become the crappy option. Amd has so much great iP if they could only get their software side together they would be SIGNIFCANTLY more competitive.

September 17, 2013 | 11:05 PM - Posted by Nacelle

I hope this fix AMD says they are working on will help with my 6970's too. I know they aren't worth that now, but I paid $700 for the pair, a couple years ago, to power my 3 screens. I'm about to sell them on craigslist and get an nvidia card, if it doesn't.

September 18, 2013 | 12:10 PM - Posted by Ryan Shrout

I honestly don't know if they will fix it, but the 13.8 beta did fix single screen for 6000-series parts so you might get lucky here too.

September 17, 2013 | 11:10 PM - Posted by Paul Keeble (not verified)

Keep hitting them until they fix this. I bought 2x 7970's in January 2012 and I noticed immediately the problem. Its not like AMD has only known about this since the beginning of this year, thousands of users were reporting this problem a year before that. It really put me off dual cards until I got a pair of 680's and found that the grass was much greener on the Nvidia side with SLI.

We need this fixed so we have some competition in the high end of GPUs.

September 18, 2013 | 08:40 AM - Posted by Anonymous (not verified)

Ryan...let us see your Tiger-Bus !

September 18, 2013 | 12:13 PM - Posted by Ryan Shrout

How about this?  http://screencast.com/t/uTeQAykAYhJF

:D

September 18, 2013 | 09:51 AM - Posted by Anonymous (not verified)

with any luck a fix will be out from AMD a few after the release of the 9000 series, they are so slow with driver updates that I'd see it taking that long ... if they dont just give up ...

September 18, 2013 | 09:52 AM - Posted by Anonymous (not verified)

a few *months* after

September 18, 2013 | 10:23 AM - Posted by NLPsajeeth

So you can't run 3 PQ321's on AMD in a 6x1 eyefinity configuration?

I'm glad to see AMD is putting focus on 4K and hope they have a 4K eyefinity solution soon.

The only way NVIDIA will ever support anything other than 3x1 surround is if AMD turns up the heat. NVIDIA if you are listening, you need 2x1 and 2x2 surround support at any resolution to stay competitive on the consumer side. No one is dropping thousands of dollars on Quadro's just to get that one feature.

September 18, 2013 | 12:15 PM - Posted by Ryan Shrout

I think technically YES you can support 3 4K monitors like the ASUS PQ321Q with AMD today...but the main issue is going to be PERFORMANCE.  Even without frame pacing and interleaving problems how much horsepower would you need for that?

More than they have available today.

September 18, 2013 | 02:50 PM - Posted by NLPsajeeth

Just wanted to make sure it was a performance issue not a driver issue. I agree you wouldn't be able to run anything more than simple 3D demos in such a setup.

Agreed, I really hope we will see large GPU performance increases with the next round of silicon so that multi 4K gaming becomes a reality.

Keep up the good work, you guys are really pushing the boundaries of what is possible with 4K!

September 18, 2013 | 10:30 AM - Posted by gamerk2 (not verified)

I can't help but wonder, if the issues seen really are due to lack of proper synchronization, then would there be a FPS impact when AMD makes their changes?

Or in other words: Is AMD cheating on performance (knowingly or not)?

September 18, 2013 | 12:16 PM - Posted by Ryan Shrout

I don't think they were doing it knowingly, no.  That doesn't excuse them though; they should have been able to see these problems before and been trying to fix them before this all started to happen this year.

September 18, 2013 | 12:05 PM - Posted by fhmuffman (not verified)

Hi Ryan, I hope 4K connectivity is scheduled to be included in all future reviews of hardware, like laptop reviews. Back in June I needed to buy something quick when my system crashed and it would have been great to know if any low to mid range laptops could at least drive a 4K display. I am not expecting benchmarks of a Chromebook running Metro Last Light at 4K but it would be nice to know if I could display Sheets on a Seiki display with an A10 based laptop.

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.