Widescreen Gaming Forum

new drivers released --> AMD Catalyst 13.11 BETA6 (13.250.18.0 October 24)

they seem stable but still no phase 2 for frame pacing (for eyefinity)

*Edit: The following rant was posted because we're getting in to December without having heard from AMD on our Frame Pacing in Eyefinity resolutions despite the recent release of 13.11 Beta 9V5. The proceeding is a rant...

Getting real tired of your shit, AMD. OK, I'm really half-way joking there. I've been with AMD since the 4xxx series. I've had ATI products prior to that. So, no offense to Terry or Andrew or anybody else at AMD - you too LordJuanlo, you helped save the day on my X58 Crossfire Eyefinity bluescreens with your beta testing several months ago.

But, I am really, really anxious to get some Frame Pacing drivers for the 7xxx series cards. Originally, the HardOCP guys reported them as being expected later the same month when the first Frame Pacing revision was released. I imagine that Kyle or one of the new dudes over there jumped the gun on that one and I wouldn't hold you to it. Most other sources were reporting "Fall" for Crossfire+Eyefinity frame pacing. Then, the folks over at Anandtech reported that they'd heard from reps on the floor in Hawaii that it would likely be mid-November, but hopefully sooner.

Well, now Fall has come and gone again, at least in the northern hemisphere, and we've heard nary a peep from AMD on Frame Pacing in Crossfire+Eyefinity resolutions, where it is most badly needed by enthusiasts. The only thing we've heard is that it's better on the 290 X series due to hardware changes no longer requiring a Crossfire bridge. But those of us with the 2** series below 290 and the 7xxx series are still holding our breath.

I understand that a new product release like the 2xx series is gonna require immediate dev attention and that the 7xxx series updates will be delayed. I fully appreciate that Tom's got your attention with the fan RPM and resulting performance discrepancies on the new cards. But, c'mon, we've been patient. I may have brought it up previously, but I haven't ranted about it until now. So, here's my rant.

Another update from Anandtech. Looks like AMD is now shooting for some time in January for Phase 2 of the Frame Pacing drivers that are expected to add support for Frame Pacing on the 7xxx series in Crossfire+Eyefinity.

http://www.anandtech.com/show/7578/an-u ... ing-driver

We finally got our phase two frame pacing fixes in the 14.1 beta drivers on Saturday! They still only apply to DX 10 and 11 titles, but are reportedly now working in Eyefinity resolutions. I honestly haven't had the chance to benchmark frametimes with frame pacing enabled and disabled to compare, but a quick look at Far Cry 3 and Tomb Raider (2013) in Eyefinity+Crossfire both felt very smooth when running under 14.1 beta with frame pacing enabled.

*Edit: this was on 2 Radeon 7950s.

They can be downloaded here.

How did you even get xfire to enable? It just says my second card is a "disabled adapter". I don't get an option for it under performance. Bleh.

_________________
System Core: | Intel Core i5-2500K + ASUS P8Z68-V + 16GB Corsair XMS3 DDR3 1333 MHz | Win7 x64 | MSI R7970 Lightning 3GB [1105/1400] |
Display: | 3 x Dell Ultrasharp 3007WFP-HC @ 7680x1600 | Dell u3011 |

Weird. Enable AMD CrossfireX is checked under Catalyst Control Center->Performance->AMD CrossfireX and Frame Pacing set to On under Gaming->3D Application Settings.

I did a complete uninstall, reboot to safe mode, driver sweeper, reboot, install, reboot. Other than jumping through those hoops, I haven't done anything out of the ordinary. I have used MSI Afterburner's OSD to verify both cards are being used and to get a quick glimpse at frame times.

I did the same, there are other users having the same issue, but no official word from amd on it.

_________________
System Core: | Intel Core i5-2500K + ASUS P8Z68-V + 16GB Corsair XMS3 DDR3 1333 MHz | Win7 x64 | MSI R7970 Lightning 3GB [1105/1400] |
Display: | 3 x Dell Ultrasharp 3007WFP-HC @ 7680x1600 | Dell u3011 |

I'm attempting some more formal benchmarking today to see if there's any real difference in 14.1 Eyefinity+Crossfire with frame pacing enabled. Does anybody have a good method for doing so? Right now I'm using MSI Afterburner to export frame times to a file while running in game benchmarks and then using a spreadsheet to chart the results. I'm planning to use Hitman Absolution and Tomb Raider, since they are both DX 11 titles with benchmarks built-in.

If you've got any other games to suggest or any better way to measure frame times, let me know!

Also, has anybody seen any other analysis on 14.1 Eyefinity+Crossfire frame times elsewhere on the web? Looks like most of what's being presented right now is just concerned with how Mantle affects BF4 performance. I can't find anything else being discussed on 14.1 betas.

*Edit 1: So far, things aren't looking very good for 14.1 frame pacing with Eyefinity+Crossfire. I'm seeing little to no difference in Tomb Raider frame times.

My testing method was to enable frame pacing, reboot, and test. Then disable frame pacing, reboot, and test. I verified after each reboot that the setting for frame pacing stayed applied in Catalyst Control Center. So, either frame pacing is doing nothing for me or it is not really being enabled, despite saying it is in the CCC. I'm going to revert to a previous driver revision that doesn't support Eyefinity+Crossfire frame pacing and see if there is any difference.

Here's the rig I'm testing on.

Intel core i7 920 2.6 GHz @ 3.8 GHz (Bloomfield)
Asus P6T Motherboard (Intel X58)
6 GB RAM
2x Stock Sapphire Radeon 7950s
Windows 7 64-bit

I'm running at 6048x1080.

Below you will find my results so far. My apologies for the somewhat sloppy X axis on these graphs. It's time in 200 millisecond intervals.

Single Card with No Frame Pacing


Dual Cards with Frame Pacing Disabled


Dual Cards with Frame Pacing Enabled


Combined Graphs

I'll continue testing and report back (and try to get those graphs cleaned up a little more.) So far my initial results are not encouraging. I'm wondering if it could just be because I'm running an old X58 chipset and old core i7 that AMD probably wouldn't have tested on themselves or if it has something to do with my two 7950s being on different speed PCI-E slots.

Unfortunately, I don't have another rig to test on and compare. If anybody else has been testing 14.1 beta frame pacing under Eyefinity, chime in with your own results and lets see if anybody's gonna get some frame pacing benefits or not.

It's interesting to note that none of my frame times under any of the tested configurations are anything like what AMD has published:



Of course, they're not listing very many specifics there and may be measuring differently than I am...but under no configuration of 14.1 betas do I see anywhere near the terrible frame time variations they're showing as "Pre-14.1 Driver." Constant fluctuation between 15 and 80 ms every frame? Jeez...that's got to be an extreme worst case scenario...I don't think I've ever had it that badly.

Maybe my particular set up just doesn't suffer as much from frame time variations as some others and I'm not going to see a significant improvement anyway. Definitely interested to see more frame pacing benchmarks from different CPUs/GPUs.

Very, very interesting...when I put my second card in an adjacent PCI-E slot running the same speed as the first card, I lose my Crossfire and Frame Pacing settings altogether from CCC. It still shows the second card under the Hardware tab, but none of it's details such as the Core or Memory Clocks like I'm used to seeing.

CCC screenshot showing missing Crossfire:



Sounds like the same issue you're having, suiken_2mieu. Based on that, I'm going to make an educated guess that having the two cards in different speed PCI-E slots probably prevented frame pacing from truly working in my benchmarks above and having the two cards in the same speed PCI-E slots breaks crossfire altogether under 14.1.

Talk about a hot mess...I wonder how long it'll take before AMD sorts it out. We've been waiting long enough, eh?

*Edit: A reboot got my Crossfire and Frame Pacing settings back with both cards still in adjacent slots and no further changes having been made. WTF? Gonna test frame pacing now with the two cards running the same PCI-E speed and report back.

*Edit 2: OK, well, after getting Crossfire and Frame Pacing enabled while both cards are in the same speed PCI-E slot, I'm seeing marginally better frame pacing results than when they were in different slots. Fewer peaks and valleys, shorter peaks and valleys. But, it's not a huge difference. I'm also wondering what I can do to more closely emulate AMD's testing. It looks like they must be collecting data points much, much more frequently during the benchmark than I am. The benches I posted above are based on 200 ms polling. MSI Afterbuner will only poll as frequently as 100 ms. So, I may be able to get a little more accurate, but not much. Anybody know of a good way to poll frame times more frequently than every 100 ms?

*Edit 3: Looking in to FCAT now, since that seems to be what most sites were using around the time phase 1 of the frame pacing drivers were released.

*Edit 4: Wowza, FCAT is some serious business and requires a lot of investment to get set up and running. So, that's out!

*Edit 5: Alright, I'm calling it. As far as I can tell, there's no easy, accurate way for the average user to benchmark frame times. Take my benchmarks above with a large dose of salt. It's just too hard to measure exactly what impact frame pacing is having without better equipment and more knowledge than I've got on the subject. Hopefully some of the pros will chime back in soon with better measurements.