Digital Foundry Article Technical Discussion Archive [2014]

[Deleted member 86764]

And yet the PS4 is still a solid 60fps. I only saw a drop to 59 in the footage, but I didn't watch every second.

I wonder whether HBAO could've been used.

 

[TomRL]

Bare in mind they have had over half a year to further optimize the game. Not the fairest comparison.

 

[DrJay24]

Or they had half a year from scratch, we have no idea.

 

[Arwin]

Or they had half a year from scratch, we have no idea.

We have some idea ... Porting the PC version to PS4 and sustaining a half decent performance is easier than on Xbox One. But the version on One was an early port not benefitting from various SDK improvements. I have a feeling the average difference will go down to 20-30% max until we get to the next level of Compute application / optimisation. With the odd exception of course of a quick port here and there.

 

[Shifty Geezer]

I raised the PvZ PC results in the comments, and others are reporting far better performance from their PCs. I think DF's PC results are anomalous.

 

[pjbliverpool]

I raised the PvZ PC results in the comments, and others are reporting far better performance from their PCs. I think DF's PC results are anomalous.

Yep these benchmarks certainly paint a very different picture more in line with what we'd expect based on the hardwares capability and reports from PC gamers:

http://www.overclock.net/t/1499072/ggpu-plants-vs-zombies-garden-warfare-cpu-gpu-benchmark

 

[Globalisateur]

Yes but on PC gamers know when their game runs at +100fps, who knows at what fps runs the consoles versions as they are capped at 60fps?

What count for DF in their test is probably the absolute minimum fps, not the average ~100fps or max ~150fps people are probably getting from their PCs.

One short drop to min 55fps is enough for any average ~100fps PC build to be "worse" than the PS4 version (which drops at min ~58fps occasionnaly).

 

[Deleted member 86764]

It's a fair point. Also, digitalfoundry are all about consistency, so they're likely to cap their framerate at the hz of their monitor/TV to avoid tearing at higher rates.

 

[pjbliverpool]

Did either of you look at the link? It included minimum frame rates. Even the 750Ti could achieve 60fps minimum.

Adding in the fact that we know the PC is rendering at a higher level of quality and it seems to me that drawing any sweeping conclusions from that 1 DF sentence is hardly sensible.

 

[pjbliverpool]

Did either of you look at the link? It included minimum frame rates. Even the 750Ti could achieve 60fps minimum.

Adding in the fact that we know the PC is rendering at a higher level of quality and it seems to me that drawing any sweeping conclusions from that 1 DF sentence is hardly sensible.

 

[Deleted member 86764]

All things being equal perhaps. Were both tests equal? Do we know whether PvZ includes a standard demo and analysis tool on PC? They always used to back in the day...

I agree that a 680 should perform better. I'm wondering if we're seeing the advantages of a closed box.

 

[Globalisateur]

Did either of you look at the link? It included minimum frame rates. Even the 750Ti could achieve 60fps minimum.

Adding in the fact that we know the PC is rendering at a higher level of quality and it seems to me that drawing any sweeping conclusions from that 1 DF sentence is hardly sensible.

But were their tests consistent with what was shown on the consoles versions, like triple buffering (no screen tearing)? I have just seen on the source link that they tested the game with Vsync OFF apparently, so already the tests aren't consistent and those 62fps on the 750ti must be with constant screen tearing.

Also what areas/maps of the games exactly did they test? Did they really do some reasonnable framerate stress test like DF most probably did on different levels?

Thoses tests, without a video and consistent consoles settings (at least with triple buffering), still prove nothing.

 

[-tkf-]

But were their tests consistent with what was shown on the consoles versions, like triple buffering (no screen tearing)? I have just seen on the source link that they tested the game with Vsync OFF apparently, so already the tests aren't consistent and those 62fps on the 750ti must be with constant screen tearing.

Also what areas/maps of the games exactly did they test? Did they really do some reasonnable framerate stress test like DF most probably did on different levels?

Thoses tests, without a video and consistent consoles settings (at least with triple buffering), still prove nothing.

No vsync? How pointless.

 

[betternatethanlever]

I'm wondering if we're seeing the advantages of a closed box.

Highly unlikely. The GTX 680 has been consistent in performing better than both of the next-gen consoles.

 

[Shortbread]

Highly unlikely. The GTX 680 has been consistent in performing better than both of the next-gen consoles.

Game for game is there any proof of this? Meaning; graphs/articles showing GTX 680 compared to PS4/XB1? I'm not talking about approximation graphs, nor PS4/XB1 PC GPU variants graphs... but actual titles running on both GTX 680 and PS4/XB1 showing these "consistent" performance/frame-rates that the GTX 680 holds over them.

Not that I'm doubting brute force rendering which most PC GPUs tend to do... but I don't subscribe to the PC mantra that PCs will always sustain "consistent" frame-rates. There are way to many variables in the PC world that can affect frame-rates/performance. For every 120fps or greater game that the PC can achieve, there is always some section/area within a game that will knock it down to 30/45fps... which in my book isn't consistent performance.

 

[Shifty Geezer]

I think betternatethanlever's use of 'consistent' is 'across pretty much all games' and not 'across pretty much all frames'. The GTX 680 consistently performs better, not performs consistently better.

In fact, I think there's a fair degree of wire-crossing with people's use of the word 'consistency'. eg. "DF are all about consistency." Is that consistency in their testing methodology, or reporting framerate consistency with their averages?

 

[-tkf-]

Game for game is there any proof of this? Meaning; graphs/articles showing GTX 680 compared to PS4/XB1? I'm not talking about approximation graphs, nor PS4/XB1 PC GPU variants graphs... but actual titles running on both GTX 680 and PS4/XB1 showing these "consistent" performance/frame-rates.

Not that I'm doubting brute force rendering which most PC GPUs tend to do... but I don't subscribe to the PC mantra that PCs will always sustain "consistent" frame-rates. There are way to many variables in the PC world that can affect frame-rates/performance. For every 120fps or greater game that the PC can achieve, there is always some section/area within a game that will knock it down to 30/45fps... which in my book isn't consistent performance.

This, there is so much going on on a normal PC compared to a Console, the best way to secure constant framerates is with plenty of brute force. I guess that having a operating system along with countless of little bit and pieces running in the background simply causes the odd hiccup, and from that interview on DF

"And let's not forget that programming close to the metal will usually mean that we can get 2x performance gain over the equivalent PC spec."

Which would be what kind of GPU?

With consoles it's easier, there is a well known limit on what the engine can push through the CPU and GPU, the assets are tailored to that specific limit. On the PC the games have to support weak setups and high end setups, making it a much more tough job to balance assets/framerates/resolution.

And lets face it, the amount of games that is played on weak ass laptops doesn't help either, the number of times i have told fathers that "no, that isn't gonna be worth playing on" when they showed me laptops offers without discreet graphics cards tells a story of very very low end market that has to be covered.

 

[Shortbread]

I think betternatethanlever's use of 'consistent' is 'across pretty much all games' and not 'across pretty much all frames'. The GTX 680 consistently performs better, not performs consistently better.

In fact, I think there's a fair degree of wire-crossing with people's use of the word 'consistency'. eg. "DF are all about consistency." Is that consistency in their testing methodology, or reporting framerate consistency with their averages?

Ok. Understandable now.

So, if a title on PS4/XB1 is unlocked at 60fps, averages about 55fps in most scenes, with low's around 40's in heavy action scenes.... and the PC hits 120fps, averages about 90s in most scenes, lows hitting 50s in heavy action scenes. So the question becomes; which is more consistent (enjoyable) for the gamer?

Mind you; I'm not trying to pi** over the PC superiority crowd, I own a GTX 690, GTX 590 and R9 295X2... but I have a problem when I hear "consistent performance", especially out of demanding titles.

 

[betternatethanlever]

I think betternatethanlever's use of 'consistent' is 'across pretty much all games' and not 'across pretty much all frames'. The GTX 680 consistently performs better, not performs consistently better.

In fact, I think there's a fair degree of wire-crossing with people's use of the word 'consistency'. eg. "DF are all about consistency." Is that consistency in their testing methodology, or reporting framerate consistency with their averages?

Yes.


I should have elaborated further. On the topic of whether the frame-rates are more consistent on consoles compared to PC, I can't say much as I don't own any of the consoles (yet), but we learn from DF's performance analysis videos that they're just as consistent as any game on PC can be with a frame-rate lock and/or v-sync.

 

[pjbliverpool]

But were their tests consistent with what was shown on the consoles versions, like triple buffering (no screen tearing)? I have just seen on the source link that they tested the game with Vsync OFF apparently, so already the tests aren't consistent and those 62fps on the 750ti must be with constant screen tearing.

Also what areas/maps of the games exactly did they test? Did they really do some reasonnable framerate stress test like DF most probably did on different levels?

Thoses tests, without a video and consistent consoles settings (at least with triple buffering), still prove nothing.

What does it matter if vsync wasn't on if the minimum framerate was above 60fps? Obviously if vsync were on we'd see a locked 60fps.

And yes there were differences between what the PC and the consoles are rendering, since it's already been established that the PC is rendering more.

So on the one hand we have a PC version proven to be doing more graphically, plus a whole suite of benchmarks covering several dozen data points and hardware configurations all showing performance exactly in line with what we'd expect from such hardware.

And on the other hand we have one data point from one hardware configuration from DF which shows performance far away from what we would expect and what PC gamers are reporting first hand.

It seems a stretch at best to conclude from this that consoles are now performing at or above 680 levels when they clearly haven't in any other game yet released.