Digital Foundry Article Technical Discussion [2017]

John Linneman Digitalfoundry said the video comparison between Forza 7 and GT Sport will be available today 24 minutes long...

 

Its here
http://www.eurogamer.net/articles/digitalfoundry-2017-gran-turismo-sport-vs-forza-motorsport-7

 

In this thread, the discussion about the DF Car Comparison article will remain intelligent. Any senseless bickering will be purged or moved into the pointless Car Comparison thread.

 

The details on the headlights in GT is nice.

 

Got a break from my kids so I can post my thoughts on John's article finally.

You know, there is a reason why Linneman is a journalist/writer and the majority of us are not. I think this article was challenging to write, and he managed to do so perfectly. There was that EG article about games journalism and this quote comes up from John Updike '1. Try to understand what the author wished to do, and do not blame him for not achieving what he did not attempt.'

And that's pretty much how Linneman sets up the reader before he goes into his analysis

Linneman makes clear what both games do well, and ultimately stays away from the final judgement to the dismay of many; they want a winner but he doesn't provide one.

While Linneman stops short of making any really guesses of how it's technically accomplished, he does leave some things to the imagination. The burning question of how GTS does it?

Time of Day never changes during the game, and must be set up before loading the level vs changing TOD
Fewer Tracks (17 tracks vs 32 tracks)
Fewer Cars (70 vs 700)
Static and no rain vs Dynamic Weather

If you multiply out the number of cars by track by TOD settings possible, the permutations get rather large, several hundred lighting permutations.
If GTS wanted to be insanely accurate for lighting, they would adjust each car, for each TOD, for each track. Which I don't know if they did, but they most certainly are setup to do it (since everything is static). It would be a monumental amount of work to do. And perhaps this signals why there are less cars and less tracks than the traditional GT game.

Compared to Forza with just about as much as possible being dynamic, they also have significantly more cars and tracks and can support weather as s result of not being statically bound, but ultimately to be trumped graphically in about every aspect of the game.

Linneman leaves this thought out of this article, likely because he's not use if he can prove it or not, but as a reader I feel this was the message he was trying to get out there; it's a story of 2 games with 2 completely separate aims.

 

So what does John really think about both games graphically ? They use different solutions painfully listed in the article: for instance GTS has less cars but each model is using much more polygons, Forza 7 has higher resolution of textures.

But which game looks better according to him ? There is always a favorite even it's very hard to find his opinion in this very politically correct article:

He does state his opinion in this article actually, very rarely, but according to him:

GTS has trees that use a more advanced tech.

GTS looks more natural and more realistic (what most people who played both games actually think, not a revolutionary thought)

GTS has the best HDR implementation ever.

And Forza 7 running on a monster GPU using max antialiasing that won't be used on XBX (8xMSAA, 16xMSAA?) has better image quality.

 

He doesn't out right say it. It's pretty clear he writes that GTS is the graphical showcase. He doesn't construe this point. In just about most aspects, we find that GTS is more refined. We've known this, he does make clear that while the graphics are more refined GTS had significantly less content to refine.

In general this leads back to why DF does not often compare games that are not the same. If they don't have the same goals they will have separate focus on how they want their game to play and that would ultimate reflect on the presentation.

 

What is funny and interesting is that GT Sport technology is better suited for fully dynamic weather and TOD probably too much for the PS4 to chew but iterating on the technology for PS5 GT Sport 2 will probably implement fully dynamic weather and TOD...

Another interesting things is that the showroom car are more detailed than the gameplay/replay/photomode model in GT S. Kaz told that GT S cars are PS5 ready maybe it will be the model used on PS5.

 

Polyphony don't use outsourcing and don't forget that Polyphony were 100 when the development of GT Sports began and they are now 200 another fact the team work fully on GT S only after the last GT 6 update releasing end of September 2015. Turn 10 are 300 people and Playground games help too and they were 115 at the beginning of 2017. Don't know is if the outsourcing teams are inside this number probably not...

Forza 7 is the result of the development done since 2012 by Turn 10 on Forza 5 and after Forza 6, Forza 7 and playground games on Forza Horizon 2 and 3.

Edit: I think the skybox technology is the same on Forza Horizon 3 and Forza 7 for example...

 

The main reason is actually far more trivial. Both games have to run on base consoles and the PS4 is substantially more powerful than the Xbox One.

 

The development on Forza are more industrialised than on GT... GT is a team with much less people but only working in house...

 

Very true. Also, people tend to forget that Polyphony had to work on virtual reality.

 

I wouldn't be surprised at all if there is much more day per man into Forza series since 2012 than day per man into the GT Sport project since end of 2013...

 

For 12GB the PlayStation 4 Pro would need a 384 bit memory interface which would have an impact on everything. Including the interface between CUs and Blend units which changes the times. The PlayStation 4 Pro is 100% like a PlayStation 4 in basic mode.

Microsoft was assuming that old titles also run on non-100% compatible hardware. The "problem" for SONY is that the API is closer to the hardware. Microsoft has DX.

 

I dont see this as a proper excuse. Some of the approaches in rendering dont seem to be related to One's shortcomings but more of a conscious choice.

 

As per Linneman the lighting model for GTS seems to Follow a similar path to AC:Unity

 

But developers have to consider the hardware they are working on to make those choices.

If Forza 7 was developed on PS4 it would look different and vice versa.

 

Yes they do and thats when they make cut backs, speaking generally. Both consoles have similar feature set in terms of capabilities though. And in this specific case of F7 and GTS we don't see major cut backs in F7 where GTS didnt have to make any. We mostly see different rendering approaches that are not necessarily related to One's weaknesses compared to PS4's. Just a different approach not much different from the days of GT5 and Forza 4.
In addition the version they are comparing it with (X/PC) can still benefit from the additional performance regardless of the base model.

 

company sized noted!
However, also keep in mind that F7 is 73 GB. GTS is 67 GB.
F7 has 10x more cars (and they are very detailed, just not as much as GTS) and nearly doubled the tracks.

The 4K edition of F7 clocks in at 98GB. That's with a texture and model pack replacement.

Considering the content difference you would suspect a size difference to be significantly more.

So without a doubt GTS is storing a lot of other data there per car and per track than F7 is.

AC unity is 45GB, 8GB more than AC Syndicate, 37.5 GB, but syndicate has a world size 30% larger than Unity.

These static lighting models allow for amazing lighting, but the data must be stored somewhere.
I think this falls in line with Linneman's statement. 700 vs 70 cars is pretty drastic. How large would GTS be if they had 700 cars would be the real question.

 

But at the end, the result can't be the same with a 1,31 TFLOPS GPU... and third party developers have to cut the resolution to keep graphical parity.

Polyphony took a rendering approach allowed by a 1,84 TFLOPS GPU at 1080/60fps and Turn 10 allowed by a 1,31 TFLOPS GPU at 1080/60fps.

On console, everything is a matter of compromise and you can't say that a 500 GFLOPS difference is trivial.