How Shift 2 Really Drives (with videos)

  • Thread starter JJ72
  • 125 comments
  • 16,197 views
RomKnight
@imari
I don't think so. That USB is just for charging the DS3. I know this because i remember playing with it connected to my laptop (not to be on top of the TV)
Click to expand...

Still, it is the default intended control device. Any lag inherent to it should be taken into account when designing a game, despite it being wireless. I'll still try both if I can, we'll see how we go.
 
Imari
Still, it is the default intended control device. Any lag inherent to it should be taken into account when designing a game, despite it being wireless. I'll still try both if I can, we'll see how we go.
Click to expand...

The lag from the DS3 is reported to be in the 15ms range, or 0.015 of a second. Experts claim that for lag to be noticeable it needs to be greater than 160ms, a little more than 10x that of the DS3. I've a feeling that what we're talking about in the case of S2U is closer to 300-350ms.
 




Raw files are being uploaded now for those who want them to do their own analysis. Give it about 20 minutes. I am performing my own analysis now. Preliminary results: not quite as long as we thought, but long enough to be pretty serious.

Edit: For the curious and observant, note that although the car is at a dead stop when the brake is pressed the contact patches expand. What's that about? Not making a great case for being a great simulator here, SMS.
 
Spagetti69
Probably canned telemetry.
Click to expand...

As I've pointed out before, it's probably harder to make canned telemetry that matches up in a meaningful way than it is to just make a decent physics system.


Anyway, results. There was no real difference between the DS3 and the DFGT. No idea what lag times on my display (Benq E2200HD LCD Monitor) are, but the article linked above said that theirs was 3 frames of a 60fps camera, or 50ms. Let's be generous and say that mines a bit slower and it's 66ms. It was supposed to be a good gaming monitor when I bought it, so it shouldn't be too terrible. Internet wisdom suggests that my Benq E2200HD LCD monitor has input lag of 7~16ms, due to it being a TN panel.

I'm using a 30fps camera by the way, I've only got a crappy old digital and that's the best it can do.

From button press to the brake bar on the left side of the screen starting to move, I'm getting 5 frames. That's 166ms. Subtract the 66ms from the monitor, and that's 100ms. Subtract 16ms from the monitor, and that's 150ms. Not bad, not great.

From button press to the brake lights coming on, I'm getting 8 frames. That's 266ms. Subtract the 66ms from the monitor and that's 200ms. Subtract 16ms from the monitor and that's 250ms. That's not good at all.

We've seen a lot larger latencies than these in the steering. I suspect part of that is down to suspension response, which could conceivably add another 100ms. Input to display, you're looking at near 400ms then which is about right.

If anyone wants to confirm this for themselves, the movie files are here for download: http://uploadmirrors.com/download/HL4ACMZT/10610402.rar

We should probably now bring this to the attention of EA/SMS. Daredevyl, feel free to add this to your thread on the official forum.
 
Last edited:
While I appreciate the fact that you posted those videos I don't think they really do justice to the severity of the problem. The article I posted above nails it - lag increases when frame rates drop due to complex action in the game. Testing with the car standing still puts no stress on the game, basically giving it a free pass to show its best response. The thing that is making this game suck for me is that the controller isn't responding well while I am trying to execute turns or burning down a long straight - a moment of complex action for the game. I don't want to sound ungrateful for the effort - but if anyone else runs a test, try it while the game is really cranking.

PS - Just saw your post - apologies, you read the article. Much appreciated, will add a link to this forum in the EA thread. You the man!

PSS - May be able to test myself later. Will try.

-DD
 
DareDevyl
While I appreciate the fact that you posted those videos I don't think they really do justice to the severity of the problem. The article I posted above nails it - lag increases when frame rates drop due to complex action in the game. Testing with the car standing still puts no stress on the game, basically giving it a free pass to show its best response. The thing that is making this game suck for me is that the controller isn't responding well while I am trying to execute turns or burning down a long straight - a moment of complex action for the game. I don't want to sound ungrateful for the effort - but if anyone else runs a test, try it while the game is really cranking.

PS - Just saw your post - apologies, you read the article. Much appreciated, will add a link to this forum in the EA thread. You the man!

PSS - May be able to test myself later. Will try.

-DD
Click to expand...

I agree, it'd be good to see the worst that it gets to as well. I might have a go later, but it's tough because you'd want to be actively playing the game and trying to hold yourself in the middle of a pack of cars. That should tax the graphics, physics and AI the hardest. Tough to set up to get a clear shot of the screen and the pad while you're playing though.

Ideally, we'd also want to find which tracks are the heaviest on the console, some stutter more often than others.

But what we have is enough to start on. I'm doing a bit of research now on my monitor, random articles are suggesting that the input lag on the monitor is sub-20ms, which makes things a lot worse for Shift 2.

Edit: Yep, 7-16ms input lag on the monitor. http://www.wipeoutzone.com/forum/showpost.php?p=108647&postcount=5 Will edit the post above appropriately.
 
Imari
I agree, it'd be good to see the worst that it gets to as well. I might have a go later, but it's tough because you'd want to be actively playing the game and trying to hold yourself in the middle of a pack of cars. That should tax the graphics, physics and AI the hardest. Tough to set up to get a clear shot of the screen and the pad while you're playing though.

Ideally, we'd also want to find which tracks are the heaviest on the console, some stutter more often than others.

But what we have is enough to start on. I'm doing a bit of research now on my monitor, random articles are suggesting that the input lag on the monitor is sub-20ms, which makes things a lot worse for Shift 2.
Click to expand...

Good stuff. I agree - it's hard to get a video of it while in the middle of a game. The only reason I'm hesitant to create a video is that I have a bad habit - fingernail biting. I know this isn't a hand modeling contest and it isn't like my hands are THAT funky. Still, if someone out there with mad gaming skills and great looking hands can make this happen, you can be our poster child.

-DD
 
DareDevyl
Good stuff. I agree - it's hard to get a video of it while in the middle of a game. The only reason I'm hesitant to create a video is that I have a bad habit - fingernail biting. I know this isn't a hand modeling contest and it isn't like my hands are THAT funky. Still, if someone out there with mad gaming skills and great looking hands can make this happen, you can be our poster child.

-DD
Click to expand...

Gloves. I suggest some nice pink frilly ones with tassels hanging from the wrists. Bonus points if there's a floral pattern somewhere on them.
 
Imari
Gloves. I suggest some nice pink frilly ones with tassels hanging from the wrists. Bonus points if there's a floral pattern somewhere on them.
Click to expand...

This. Definitely. I'll send you a cupcake if you do it :)
 
More testing, and I think this nails it down once and for all.

Many folks have been testing various theories and eliminating possible causes and solutions along the way. One very interesting theory put forward by Imari (I believe) was that the suspension could be modeled in such a way that it takes a little time for the suspension to load up after the wheel is turned. That could account for the delay incurred before the car actually starts to rotate. Makes perfect sense as a soft suspension could indeed allow this to happen in real life. If that were true it could in theory point to a potentially very good physics engine, just with inaccurately modeled suspensions. eg. The car I was using for the videos is the Lamborghini LP640. I'm pretty sure they didn't reuse the suspension from a Crown Vic, as crazy as they are over there!!! :)

So, to test this all out I decided to run the same tests again but this time while viewing the actual tires on the car rather than the initial rotation of the vehicle as seen from inside the cockpit. The results are very interesting.

S2U : First hint of tire movement after input : 0.433 seconds
GT5 : First hint of tire movement after input : 0.233 seconds

That 0.2 second gap is pretty much exactly the same gap as seen from inside the cockpit with the car in motion. Therefore we can only conclude that the additional 0.2 second overhead occurring in S2U is taking place in the software and is not a side-effect of suspension loading or any other physics-related quirk. Simply put, it takes S2U almost twice as long as GT5 so translate user input into vehicle reaction. In my view that can only be sloppy coding and the only fix is a patch. I'd be happy to be proven wrong but I think we're running out of options here.

Here are the videos, first S2U :



And now GT5 :

 
So it is true. Its not "input lag" is how it is processed/transmitted by/to the game engine...

Well the good news its it can be fixed as told... the bad news are we have to wait... some... time...
 
RomKnight
So it is true. Its not "input lag" is how it is processed/transmitted by/to the game engine...

Well the good news its it can be fixed as told... the bad news are we have to wait... some... time...
Click to expand...

True, although I still find it easier to refer to it as input lag as it is a lag between my input and the vehicle reaction. Technically though ....

Yes, in theory it can be fixed - the big question is will it? I have not played other NFS games but others have commented that this problem has been around for a long time. Perhaps this is the first time it's really getting a lot of attention due to EA marketing it to a new audience. Who knows .....
 
oh great, so it is a input smoothener coded into the game, then it should be easy as heck to fix for SMS, I don't believe eliminating this feature will impair the ability to drive for pad users either. Someone start a petition over their official forum?
 
I have a thread over there with some video. If a bunch of us pop in and keep it visible that might help keep the pressure on. I also submitted a negative review on Amazon. I think they are likely to pay attention to anything visible that would impact sales.
 
JJ72
oh great, so it is a input smoothener coded into the game, then it should be easy as heck to fix for SMS, I don't believe eliminating this feature will impair the ability to drive for pad users either. Someone start a petition over their official forum?
Click to expand...

Shouldn't make a difference to pad users, there's already a kind of steering speed limitation that smooths out inputs somewhat. You can see it in telemetry, if you slam the stick over as fast as you can (lock to lock in fractions of a second) the steering only moves over at a slower rate. That should be enough to keep the pads drivable, along with suitable sensitivity adjustments.
 
JJ72
oh great, so it is a input smoothener coded into the game,...
Click to expand...
That's what it looks like. I tried to quantify the problem in this posting. At this point I believe it's a pass filter applied to the steering input, i.e. steering input => filter => wheel position. The filter appears to be a low order low pass; could be a simple leaky integrator or something like that. I could repeat the experiment in the linked to posting at different frequencies to generate enough data points to come up with a magnitude and phase plot - no real need to do that though.

I also believe that there's a good chance that the wheel graphic you see is the actual position of the wheel, i.e. it is directly linked to the steering angle of the front wheels - that matches the response I see from the car pretty well.

...then it should be easy as heck to fix for SMS,...
Click to expand...
Indeed. I assume it would be dead easy, as in a five minute fix. They could make it an option: Steering smoothing: on/off (or 0%-100%).

...I don't believe eliminating this feature will impair the ability to drive for pad users either. Someone start a petition over their official forum?
Click to expand...
As Imari mentions, the pad still needs a filter on the input. That's how you can pulse the pads steering buttons to keep the wheel in a certain position. Pad players are well conditioned to be human PWM generators :) Direct control by the pad would cause the wheels to slam left and right, making the car uncontrollable.

The really strange thing is: How on earth could a producer of a racing sim/game let this steering response through? How could the testers accept it? And what about their real-life-racing promoters? :confused:

DJ
--
 
DrJustice
The really strange thing is: How on earth could a producer of a racing sim/game let this steering response through? How could the testers accept it? And what about their real-life-racing promoters? :confused:

--
Click to expand...

About this, i think the explanation is simple. The same way people adapt playing a racing game with a pad, it easier for someone using the wheel to adapt.

I got the game yesterday and i've been reading and posting so i'm aware of the problem. I still find it difficult go feel the lag after start playing... you adapt to it and you compensate by turning early... i know i must do that but when i'm in a middle of the pack i just don't think about it :)

maybe i'm still sleepy...
 
RomKnight
About this, i think the explanation is simple. The same way people adapt playing a racing game with a pad, it easier for someone using the wheel to adapt.
Click to expand...
Well, even though it may be possible to adapt, it ruins the realism and immersion for a great number of players. Furthermore, I refuse to even attempt to adapt to a blatantly broken control system - if I did it might ruin my ability to handle correctly implemented systems. Besides I'm just too old for s**t like that :lol:

The pads should have filtering, yes - but the whole point of having a steering wheel in a program like this is for it to be directly connected to the steering rack in a 1:1 fashion. We now have what amounts to a steering column and rack made of soft rubber. A sad thing apart from the steering being buggered, is that if there's a good physics model and good FFB there, it's masked and ruined by the lag...

DJ
--
 
You can adapt, but it requires driving unnaturally and you forgo the possibility of making particularly aggressive moves because it's just not possible to react quickly enough if something goes wrong.

It's like writing on wet paper. You can do it if you adapt your style a bit, but it's just not as neat and clean as a real, dry piece of paper.

It's not bad, I'm still playing it and I'll continue to enjoy it even if it's not fixed. I'm just continually aware of how much better it could be if they sharpened up the response. I really think a lot of people's problems with the "physics" would be solved with a sharper input system.
 
Imari
You can adapt, but it requires driving unnaturally and you forgo the possibility of making particularly aggressive moves because it's just not possible to react quickly enough if something goes wrong.
Click to expand...
Right. If there's only me and the track and a nice sweeper I can manage the lag (I don't want to though!). But once you're in traffic and you need to react instantly and precisely, it falls apart. My biggest problem is when saving it, e.g. catching a slide, I then end up overcompensating too late (the basic problem with lag or deadzones). I've been having serious cases of PIO for half the main straight several times, after spiritedly catching slides in the last corner.

DJ
--
 
I didn't said adaptation is good. Only a possible explanation on to why this issue might have come overlooked. Not cool they missed this.

I still enjoying the game very much though :D
 
RomKnight
I still enjoying the game very much though :D
Click to expand...

Im also enjoying the game, despite these oddities. I have been GT5 enthusiast and still play it regularly. So I can't help but compare. I find S2U a decent drive. I get the floaty feeling too, but with some custom wheel settings and car setup, it does become a very enjoyable game to drive. It just takes some practice.

These videos are great and thanks for the info that you guys went to the effort for. It is really interesting viewing. I hope they tighten up the input lag (whatever is the cause), but its not a bad game and a bit of a welcome change from GT5. Although GT5 is still a stand out for me, purely for the physics.
 

Latest Posts

Back