Frame Rating: High End GPUs Benchmarked at 4K Resolutions

Final8y · 1 May 2013 at 05:37

Our experiment with early testing of high end graphics cards on a 4K display has definitely been interesting. After using $3000 graphics card configurations on a 50-in 3840x2160 monitor it is going to be a struggle to go back to smaller display and single lowly graphics card. But alas, that is part of the job!

Problems creep up for the Radeon HD 7970 if you consider going to the route of SLI or CrossFire. While the GTX Titan and GTX 680 scale very well in most of our titles at 3840x2160, the Radeon HD 7970s in CrossFire suffer from runt frames and high frame time variance that result in either much lower than expected animation smoothness and/or stutter. The GTX cards in SLI do not have this behavior resulting in a much better multi-card scaling proposition. Yes, the prototype driver would help the HD 7970s out quite a bit, but that option is still weeks or months away.

If we take into account the prototype driver for the Radeon HD 7990, it makes a good case to take that spot away from the GTX 690. With much more even frame distribution, the dual-Tahiti card looks like a better option in several games even though there is still much work to be had for the driver team to produce frame times as evenly as NVIDIA SLI does. Long story short, even if the prototype driver were available today to consumers in its current form, I still think the GTX 690 would have the advantage. But AMD is on the right track and with some more development they could make their card the better choice.

http://www.pcper.com/reviews/Graphics-Cards/Frame-Rating-High-End-GPUs-Benchmarked-4K-Resolutions

LtMatt · 1 May 2013 at 07:16

Muahaha i beat you to it, this time!!

http://forums.overclockers.co.uk/showpost.php?p=24206716&postcount=138

I'll copy my last posts over though since i posted that in the other pcper time frame thread.

LtMatt said:
I think if i had an xfire setup and had the choice between what it is now raw fps or to have added latency to improve frame times but lower overall fps id stick with raw fps in most cases, certainly in twitch shooters. If it looks fine and is perfectly smooth i can't see a reason why you'd want extra latency. We've seen already from multiple xfire users here that battlefield 3 is perfectly smooth on xfire.

Phixsator said:
My Crossfire setup of 2x Sapphire 7950 never seemed to have any issues with Crysis 3 using the 13.2 beta driver(havent played it since 13.3 came out) like the one showed off in the video. Has been a very smooth in the 50-60fps+ area, with no runt frames. Im usually very picky about such things. Something im missing here?

LtMatt said:
You need the tool specially created by Nvidia to be able to detect such things im afraid.

pgi947 · 1 May 2013 at 07:21

I'm taking this with a pinch of salt, this is the same reviewers that only a week ago were saying dual 7970's on 13.5b was essentially a stuttering mess. The prototype driver does sound promising however.

By all means they can send me a 4k display to do my own testing

LtMatt · 1 May 2013 at 07:25

pgi947 said:
I'm taking this with a pinch of salt, this is the same reviewers that only a week ago were saying dual 7970's on 13.5b was essentially a stuttering mess. The prototype driver does sound promising however.

By all means they can send me a 4k display to do my own testing

You'll need FCAT to tell you theres a problem, a 4k display won't do anything to help in that regard.

A 4k display would be wasted on PG tips. Send it to me and ill get stuff running at fairly high details on 1 gpu.

Like an albatross :cool:

bru · 1 May 2013 at 07:50

I'm quite surprised that the AMD cards are suffering that much with the frame times in crossfire, especially the official 7990 seeing as it has been so long in coming. I would have thought they could have got the drivers sorted out by now.
Well at least this new prototype driver looks like a definite improvement.

ICDP · 1 May 2013 at 08:12

bru said:
I'm quite surprised that the AMD cards are suffering that much with the frame times, especially the official 7990 seeing as it has been so long in coming. I would have thought they could have got the drivers sorted out by now.
Well at least this new prototype driver looks like a definite improvement.

AMD cards are not suffering from frame times apart from in CF. Single 7970 is getting much better results than a single GTX680 in those tests. In fact the single 7970 is right between GTX680 and Titan for speed. Correct me if I'm wrong but that puts HD7970 right in the new rumoured GTX780 territory.

bru · 1 May 2013 at 08:25

ICDP said:
AMD cards are not suffering from frame times apart from in CF.

Quite correct, I did mean in crossfire, edited my post accordingly.

Gregster · 1 May 2013 at 09:23

Interesting read thanks.

Diagro · 1 May 2013 at 09:58

This is odd, All this flapping about frame rateings... they make it sound like this has been around for years, yet right now it seems people are complaining, Also you have to use a program made by nvidia to test for it??? well... yea

bru · 1 May 2013 at 10:12

Are you suggesting that it isn't an issue?

If so why are AMD working on prototype drivers to fix this non issue.

The bigger problem in my opinion is that cards from either company can be allowed to be released when the drivers are not up to scratch in the first place, I mean it's not as if the companies don't have plenty of time to write the drivers, with the development times these modern cards have.

pgi947 · 1 May 2013 at 10:18

It is an issue but it's not as bad as it's being made out to be, the 13.5 vs prototype video had to use a high speed camera slow down to highlight the issue. How many games do you play at that speed? None.

I guess if it is something myself and others aren't noticing just now maybe we will when the new driver go's public & we can then say 'yes there was a noticeable problem'.

ubersonic · 1 May 2013 at 10:20

Not really tht interested in 4K, so its 2x the res of 1920*1080 and 1.5x the res of 2560*1440, its hardly the massive jump a lot of sites are panning it as, 1280*720 to 1920*1080 was just as big.

IMO more effort should be spent on bringing out higher res 21:9 screens instead of more 16:9 aspects. My 29" 2560*1080 Dell screen is awesome, much prefer it to the old 27" Apple 1440p panel as its basically the same width but without the wasted vertical space (when you have to start tilting your head a lot to see the top/bottom of the screen that's an issue), 21:9 is much better than 16:9 for games, movies and also for desktop work as a pair of windows tiles is like a pair of old school 1280*1024 screens

layte · 1 May 2013 at 10:30

Diagro said:
This is odd, All this flapping about frame rateings... they make it sound like this has been around for years, yet right now it seems people are complaining, Also you have to use a program made by nvidia to test for it??? well... yea

It has. It's just that people got mocked and accused of being fanboys if they tried to talk about SLI being smoother than xfire. The difference being we can now test for it.

As foot the tool coming from Nvidia, so what. If it was broken or biased in anyway I'm sure either other sites or AMD themselves would have advertised that.

layte · 1 May 2013 at 10:33

ubersonic said:
Not really tht interested in 4K, so its 2x the res of 1920*1080 and 1.5x the res of 2560*1440, its hardly the massive jump a lot of sites are panning it as, 1280*720 to 1920*1080 was just as big.

IMO more effort should be spent on bringing out higher res 21:9 screens instead of more 16:9 aspects. My 29" 2560*1080 Dell screen is awesome, much prefer it to the old 27" Apple 1440p panel as its basically the same width but without the wasted vertical space (when you have to start tilting your head a lot to see the top/bottom of the screen that's an issue), 21:9 is much better than 16:9 for games, movies and also for desktop work as a pair of windows tiles is like a pair of old school 1280*1024 screens

So much wrong in a single post

. Vertical res is king. Which is why I'm happy to drop a grand on decent 1600p displays.

ubersonic · 1 May 2013 at 10:42

layte said:
So much wrong in a single post . Vertical res is king. Which is why I'm happy to drop a grand on decent 1600p displays.

For CAD/graphics work I don't deny its useful, but for gaming/movies/general use its just wasted space.

Rroff · 1 May 2013 at 12:06

Diagro said:
This is odd, All this flapping about frame rateings... they make it sound like this has been around for years, yet right now it seems people are complaining, Also you have to use a program made by nvidia to test for it??? well... yea

Its been under investigation for over 2.5 years its far from a simple issue i.e. you have to rule out it being a specific application or software support problem with a specific title.

Despite the damage control and some prople genuinely not noticing it it is there and is noticeable without "high speed" cameras (the videos are mostly normal speed with some sections only slowed down by 0.5x so that it can be seen on 30fps youtube).

tribz · 1 May 2013 at 12:12

ubersonic said:
Not really tht interested in 4K, so its 2x the res of 1920*1080

Its 4x 1080p isnt it? 1920x1080 = 2073600 pixels, 3840 × 2160 = 8294400 pixels.

Kaapstad · 1 May 2013 at 12:35

ubersonic said:
For CAD/graphics work I don't deny its useful, but for gaming/movies/general use its just wasted space.

I really enjoy using a 30" dell u3011 and would not want to game on anything smaller, I am looking forward to 4k and an even bigger screen, I would rather use this than game @5760x1080.

tribz said:
Its 4x 1080p isnt it? 1920x1080 = 2073600 pixels, 3840 × 2160 = 8294400 pixels.

Yes

1600p is double the resolution of 1080p and 4k is double the resolution of 1600p.

ubersonic · 1 May 2013 at 12:51

tribz said:
Its 4x 1080p isnt it? 1920x1080 = 2073600 pixels, 3840 × 2160 = 8294400 pixels.

Its 2x the resolution and 4x the pixel area.

Kaapstad · 1 May 2013 at 13:05

ubersonic said:
Its 2x the resolution and 4x the pixel area.

4x the pixel area = 4x the resolution.