Dual-GPU Battle: Does Frame Pacing In Catalyst 13.8 Turn The Tide?

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.
Why is it in the past you say that frame variance of less than 15ms is practically imperceptible but in this article you suddenly can notice that when the AMD card approaches 10ms is appears "more choppy" than the Nvidia card?
 
It's good to see AMD/ATI holding true to their word about working on these issues, as opposed to sweeping them under the rug and quietly moving on to the 8000 series without resolving the issue.
 
great. now ill be lookin froward to the 9000 series (or the R9-xxxx the rumours once said)
with even more performance, and more importantly, more power efficiency while not increasing the pricing much
 
"There seems to be some disconnect between the severity of what FCAT reports and the video output used to create those benchmark results. I do see a handful of sequences where Catalyst 13.8 looks smoother than 13.6. However, it certainly doesn't appear that 13.6 is hovering around 40 FPS like the frame rate over time suggests."

So you say FCAT should be a better framerate benchmark for end-user gaming performance than FRAPS-like framerate benchmarks, and then you say this?

Personally, the numbers that FRAPS show are directly correlated to my qualitative analysis of the game. If FCAT isn't, then why use FCAT at all?

I think we need to see a graph of the numbers FCAT reports vs the numbers FRAPS reports.
 
As the owner of a 7990 + 7970 tri-fire setup I can say that 13.8 fixed almost 100% of my issues and the difference in percieved stutter is night and day.
 

This may just be the most sensible thing i'll read on this thread.

You will, unfortunately, have to face the wrath of the fan-people.
 


I've been in the same boat. I owned ATI cards through the 5850. I ran with two and then three 5850s. Here we are 3+ years later and finally AMD is working on issues I was having with my crossfire setup.

At any rate, I stopped buying AMD with the 5850 because my experience with driver-related issues with crossfire was terrible. I went on the AMD forums and was basically ridiculed for trying to fix my issues. I was accused of trolling and even had my ID banned (of course I established a new ID, but was still unable to solve the problem). Pretty awesome support community. I just wanted my cards and crossfire to work as advertised. Microstutter was only part of the equation and even then, it took a 3rd party with enough of a voice (in 2013!) to call AMD out on the runt/dropped/microstutter driver problems before they'd even do something about it. This is why I won't consider AMD at this time.

After becoming fed up with seeing the high framerates and flickery video, I dropped in two GTX 580s a year later. Problem solved. I've been running with two Nvidia cards since (subsequently 2x 680s and 2x 780s).

I feel like AMD devotes its time to its single-card customers because there are far fewer multi-card customers (as with Nvidia). If you have more than one card, you're kind of a back-burner customer to AMD. This has not been my experience with Nvidia.

If there is well-documented info regarding AMD getting their stuff together, I will definitely put them back in the mix. Even now, though, it still seems like they're a work-in-progress. When you spend that much money on a multi-GPU solution, though, you shouldn't have to deal with the quality issues.

The last thing I want is to not have choices in the marketplace, but until it's documented that AMD has done away with their well-documented driver issues, I can't even consider spending my money on their product. It's not that I don't like AMD. I want them to succeed. As a matter of fact, I would like to see other players enter the market.
 
So, did we just discover that FCAT is not reliable? Shocker! Makes me question the legitimacy of this entire "Frame Rating" fiasco that has been happening for the past 6 months.
 



I know right? People don't realize that we're the ones in control, we decide the success of the company by deciding which product to buy. Being loyal to a company is ridiculous since they're at our mercy and should always deliver the best working products and not some half-assed option. I've owned almost everything from every manufacturer, Intel, AMD, ATI and Nvidia and at some times one is better than the other and other times the other one triumphs, so I just go with the best. ATI has been suffering from not so great drivers ever since their X800 series cards! Only fanboys will deny this and while Nvidia also have had their issues they're usually much better at driver support than AMD.

I stopped taking people's advice a long time ago because of their pre-determined bias towards a product, I just use Tom's and other reputable hardware review sites to draw my own conclusions.

People don't be fools, you owe these companies nothing! They're the ones who owe us the best experience our money can buy.
 
So basically every test they perform shows the less expensive 7990 to be a considerably faster card and that the 13.8 drivers did indeed iron out alot of the dropped/runt frame issues. Despite all this every time the 690 gets stomped they still gotta drop some little comment for the nvidia to try and save face... It lost kick it straight. Also metro last light is a HORRENDOUSLY CODED game for AMD. This has nothing to do with AMD, their cards or their drivers. This is a 110% 4A dev issue that they addressed MONTHS ago and the game now plays totally fine on AMD, funny toms didn't know about this https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&cad=rja&ved=0CCwQFjAA&url=http%3A%2F%2Fwww.pcgamer.com%2F2013%2F05%2F16%2Fmetro-last-light-fov-patch-released-also-fixes-amd-issues%2F&ei=0GQnUozTBYOGyQHD0YDYDA&usg=AFQjCNHj2NphgzErHxGirvfx4KcFGUSXrw&sig2=zmTKLzKdGezOk4ZmbYwC-Q&bvm=bv.51495398,d.aWc
 
A test of this type can take weeks to complete. Can you say for certain that the solution you posted has anything to do with the poor showing here, and that Toms have slipped up?
 
Nice to see AMD start making Dual GPU work smoother than it used to be. I just hope that they continue ironing out the kinks make it even better for people who have CF setup.
 
Great article!, Don.

Please can you consider to talk in a future article about AMD and downsampling?, the situation is terrible as NVIDIA downsampling has horrible IQ and AMD user developed tool (AMDDownsamplingGui) is no longer working with latest Catalyst (the IQ is astounding with that tool).

We need to make aware AMD of the huge IQ gain that we will have (for all games, no restrictions!) if they implement downsampling in their drivers (and also the huge graphic power [$$$] that we will demand if they do so). To me the situation is similar to this frame pacing issue that has taken so long to identify and solve.
 


Funny you didn't read the link you posted. This has nothing to do with the CrossFire problem:

"“A patch has been released on Steam to fix issues with AMD hardware, and improve performance for these cards,” writes Deep Silver’s community manager on the Steam Forums. “It also fixes a shadow visual corruption bug on AMD 7xxx cards"

We talked to AMD about the metro CrossFire problem, they are aware of it and are working on a fix. We're not out to get them, we're trying to improve the user experience by letting them know about the issue. It doesn't help anyone if we bury problems to make a company look better.

By the way, we run our benches off of Steam, so the game is automatically patched. And we ran that one at the end of last month, it was the last bench we took - retook actually, because of the issue people saw when playing it.

 
Maybe Lara's hair is a SLI bug? I enabled TressFX on my GTX 570 and it looked perfectly fine, but it did drop the fps too much.

That sort of thing can happen.

I remember opting to use a single 580 while playing Skyrim, b/c SLI totally glitched the water in a similar manner. Eventually, a mod came out with more realistic water and it looked good again in SLI mode.
 


That is very odd as I was trying to diagnose the Metro LL issues before there was any fix or patches. I tested with 2 separate GPU configs, my current 7870 LE, and my previous 6950 CFX. The game suffered from low GPU utilization (10-50% on both setups) and low FPS (15-40 AVG about 28 with 7870 LE) (20-50 AVG about 34 with 6950 CFX) It wasn't as though the CFX was suffering from any type of different issue than the 7870 LE as game play was extremely choppy with massive fps spikes and drops constantly regardless of config. Now a couple weeks later here comes 4A with an AMD patch and what do ya know both configs are massively improved and perfectly playable. Although the 6950's push a higher FPS with a little more candy the experience was a obviously slightly smoother on the single card 7870 LE due to the then present runt dropped frame issues. Yes I do read my links before I post them... I never noticed a CFX problem with LL, I noticed an LL problem with AMD that was thankfully resolved on both single and multi GPU configs. So in your benchmark with my eyes i saw LL+AMD= Horrid performance= No patch. Maybe this CFX issues only effects 7XXX cards ?? I am not partial to any brand I have a 560ti in my backup rig and a GTX-260 that used to be one of my favorite cards still hanging around as 3rd string. This being said why does it seem that toms is really hard pressed to give true credit to amd. Take this article as a perfect example, it is chock full of positive benchmarks almost across the board for the 7990 showing higher avg/max FPS and a DRASTICALLY reduced frame latency. Despite this it's like (insert face saving pro Nvidia comment) after almost every positive remark you make about the AMD card, even though nvidia lost... I just want straight up no BS benching, to be honest regardless of company affiliations and brand preferences.

All the same I am very happy to see that the new drivers were put to the test and CFX as a whole was improved. Hopefully positive development continues and CF gets gets even better ! Might persuade me to use my 6950's again.
 

13.8 has actually been available for a while now. As always, articles take time to research, time on the bench, and time to write up. You're lucky if an article hits within four weeks of when it was assigned. That means this was started right around the time 13.8 even became available.

I'm looking forward to part two. I want to know if this addresses Pitcairn and Cape Verde as well.
 
If the results are skewed because of an in game glitch, then you can't post those results. That's simply unfair. It's also not just a radeon problem with this game. My bro had this same problem on his AMD desktop AND his nvidia laptop. All it takes is redownloading the game.
 
Nobody plays Metro anyway, outside of bench-testing their equipment. Sorry, but that doesn't count as "real world" application.

I always get a kick out of the games most commonly used for these tests. Sure, they are super power-hungry games capable of crippling most hardware, but they aren't much fun, and don't sell that well to gamers. They just give PC gamers some leverage of "booyah!" over console gamers (I use both extensively). They can brag about how pretty their games are...even though it's the games no one plays and it took 4 years for PC's to pass the consoles in graphics quality.
 
Bravo AMD, I'm definitely noticing a visually smoother experience in some of those videos. If AMD adds some Crossfire tiers to Never Settle Forever soon, I might just go with a Crossfire setup.
 
Status
Not open for further replies.