Comparison of HDR Versions (PS3, 360, DC, CPS2)

[Thelo]

Yeah, I know that’s the theory (non-zero smoothing = less time to react). But the thing is, I’m pretty sure that in practice, having the game jump around all the time from zero smoothing is going to shoot your reactions way worse than that extra frame or two would.

As I said, I just see a definite correlation between being badly affected by lag and disabling smoothing, it’s not just you. I recommend you at least try it out for a while with default smoothing, since my intuition is that you might have disabled it at the very start and not actually tried playing with it set at default. Maybe you’ll still feel that zero smoothing is best for you and revert it, which is fine, but it’s possible you’ll find it just makes for a smoother experience overall.

[Ganelon]

I actually used default smoothing the first few days I played HDR and do remember it being smoother as you said. However, it also felt like I had a harder time escaping from tick throw mixups. But sure, you could have worthy advice right there; I’ll definitely give smoothing a go again next time I play HDR and see how it feels.

[orochizoolander]

I also prefer 16:9. 4:3 just looks really ghetto, like straight from the hood. Anyways, thanks for the testing Ganelon.

The current champ doesn’t have a problem with it why should anyone else?

Seriously though lets be honest it looks ugly as hell compared to 16:9, the poll showed the majority prefers it, and not being able to see your character fly off screen for a fraction of a second isn’t going to be your downfall if you lose and with that said I’m surprised evo decided to go 4:3.

I know I’ve seen Blanka’s electricity glitch (where the electricity currents remain a la DBZ super saiyajin for the rest of the round) on the PS3 version pre- and post-patch.

You know they left it in purposely and we all love it and thanks for taking the time:tup:

[Coth_X]

Yeah, I know that’s the theory (non-zero smoothing = less time to react). But the thing is, I’m pretty sure that in practice, having the game jump around all the time from zero smoothing is going to shoot your reactions way worse than that extra frame or two would.

As I said, I just see a definite correlation between being badly affected by lag and disabling smoothing, it’s not just you. I recommend you at least try it out for a while with default smoothing, since my intuition is that you might have disabled it at the very start and not actually tried playing with it set at default. Maybe you’ll still feel that zero smoothing is best for you and revert it, which is fine, but it’s possible you’ll find it just makes for a smoother experience overall.

I actually used default smoothing the first few days I played HDR and do remember it being smoother as you said. However, it also felt like I had a harder time escaping from tick throw mixups. But sure, you could have worthy advice right there; I’ll definitely give smoothing a go again next time I play HDR and see how it feels.

I played with smoothing too for a few days but after 10 or so rollbacks, I’d rather see the teleporting then not be able to “trust” what my eye’s are seeing.

I only get rollbacks now if the ping is like 200+ and I don’t even try when the connection’s that bad.

[Dark_Gaiden]

I performed my testing on whatever the default output is for HDMI and composite on either machine. I didn’t change the visual settings at all, nor did I change any other setting on either console or in the game itself (meaning speed was default when performing the PS3-360 tests) except to make the game display in 4:3 resolution (it’s a possibility 16:9 has differences but I wasn’t able to test that with a CRT).

I’m pretty sure the default HDMI output to a 1080p display (e.g. the Evo monitor) is 1080p on both machines. The game is drawn in 720p so the systems themselves must be upscaling that image to 1080p when I set it that way. If you don’t use the same resolution as your HDTV, the TV handles the upscaling into its native resolution for you, which is almost always slower than letting your system do the task.

As stated, I also reversed monitors with identical results. I’m not sure how you’re testing using 1 monitor (timing the difference isn’t exact enough IMO) but unless you’re recording everything, you’ll need to see the images side-by-side to be certain. I held the up direction with multiple characters in mirror matches across multiple stages and the results were always the same, with the exact same jumping rhythm up through the very end.

I definitely think you should change TVs, make sure settings are reset or defaulted, and use a control in your tests if not already.

I’ll definitely try a different TV and use some more control methods as numerous variables can lead to different results. One thing to note when I did my “pseudo” tests (4:3 mode on both systems) earlier in the morning with that Sony LCD; I wasn’t using HDMI cables for either system, instead I was using the HD Component Video Cables that support 1080P resolution. The cable came standard with the 360 (well older model 360s anyway) and as for the PS3 I purchased a HDTV Component Cables that also supported 1080P (from Walmart…I forgot the brand name). With those type of cables, you have to manually select the output resolution for the systems or they will just stay in the factory default of 480P. My Sony LCD only has one HDMI slot and that is already being occupied by my PC.

I’ll try to round up a CRT with video out and capture the footage. It may take me awhile since I’m not too familiar with capturing software on the PC (I prefer the MAC for situations such as this but I would have to completely rearrange my office just to get the Mac in the other room).

[x64]

The game is drawn in 720p so the systems themselves must be upscaling that image to 1080p when I set it that way.

Every piece of literature I’ve read that mentions resolution says HDR is a 1080p native game.

[Ganelon]

Yeah, you’re right, my mind was on SFIV when I wrote that. The graphical assets in HDR were designed to be displayed in 1080p 4:3 format. So a 1080p monitor should be the best looking and fastest processing (no scaling) viewing device for this game.

DG: Sounds great. I know composite is the same (using the official cables that came with the system) so hopefully, you’ll come to the same results.

[nomrah]

To add some fuel to the fire - HDr at Season’s Beatings was run on PS3, speed 4. A group of us played about 8 hours of casuals on speed 4 and we all agreed that it felt “right”. Daigo, while not being the end all of HDr knowledge by any means preferred speed 4 as well.

I’d really like to see where all this testing ends up so we can agree for future tournaments. I’m personally of the camp to just switch over to 360.

[Ganelon]

Well, Sirlin stated that the HDR speeds were supposed to be based on ST arcade speeds:

14. Game speed. The game speeds match the arcade version of the game, but this is confusing so bear with me. In SF HD Remix, speed 3 is the default and is intended for tournament play and online play. It?s the same speed as Japanese arcade speed 3, which is also known as US arcade speed 2. You don?t really have to understand what?s going on with all that, just play at the default speed 3 and be happy that it matches the arcade.

I suppose if there’s a serious interest about matching the seemingly ideal arcade ST T2 (X T3) speed (even though US ST tourneys were always on US T3, one step faster), I can get my supergun and ST board out to see how HDR compares.

[Mavrick]

VF4: Dropping inputs is always a possibility. However, what possible operation could there be in the programming that would cause such a thing? The analog-digital glitch was a result of a new feature combined with a programming omission.

If you use a autofire switch on LP to perform LP chains, there’s not a single instance where the LPs are cut off by a dropped input. Try it; even the slow setting on the TE (probably around 10 Hz, if that) never yields a single miss. Now, there’s still a possibility of a particular situation causing an issue but if you have an HRAP, leave LP on full auto (which is set to 23 Hz), mess around with the other side, and see if the LP chain ever pauses when you’re not hitting. I never saw it make a pause while I was messing around. It’s a logical impossibility for me to prove every single case so if anybody can find a reproducible input glitch, that would be helpful.

As for input delay, I’d be very interested in seeing what it is but I’m afraid right now, I don’t have the equipment to accurately count it. Hopefully, somebody else has the will and means.

Regarding tournament standards, HDR will always be at T3 speed with default dip switches, Udon sprites, and OC ReMix music. However, I still think that 4:3 is the optimal aspect ratio but I’ll leave that beaten horse alone and trust ponder to maintain his wise decision next year.

As for using AE, its speed is also off, old characters can’t be throw-softened, and o.Sagat and claw have issues, amongst other differences. And I doubt Evo would be willing to use the JP PS2 Best version of HSF2 for some fixes. But again, this topic wasn’t about ST vs. HDR but rather a look into whether HDR’s classic mode has noticeable differences from ST and primarily a comparison of the 2 HDR versions.

Dude your so wrong, at EVO, Final Round and, SB4 everyoneincluding myself can see that the PS3 version plays way diffrent then the XBOX one. For one the combos do not link I play BISON as a main and, on the 360 i can TOD like cake but, on the PS3 its not so. For a long time i thought it was me so i did the same thing side by side with tv. I could not land TODs with bison on the PS3, if i did 20 on each i could do 5 out of twenty on the PS3 and, 18 out of Twenty from the Xbox version were not talking about speed becuse, the XBOX runs slower offline the sameway the PS3 does and, thats not me just saying that it’s a fact. I asked Sirlin myself at EVO about that and, he said it does run faster online does it does offline.

As for input dropping the PS3 doesnot allow mashing to link to combo like the 360 does so i was inputing the combos my self one of them that does not link that should on the PS3 is if you take DeeJay and RYU and use DJ jump MK cross Ryu then try and, like the cr. HP which should link does not DJ will pass through ryu every time like clock work. Certin people the game works fine with like SHOTOs but others like Chun if you do a super low cross MK to a standing MP or MK the attack will go in the wrong direction but, will not link even if its Gief or Honda. MEssing up your combo so, Ya the PS3 is Diff inferior to the Xbox Version.

[Mavrick]

Its on people $1000 HDR lest do this

http://forums.shoryuken.com/showthread.php?t=212102

[Ganelon]

I had a chance to perform some more testing tonight:

I just tested the times between every speed on PS3 and 360 and they were identical. The difference between T1-T2, T2-T3, and T3-T4 is always ~3 timer ticks apiece. The difference between T0 and T1 is a whopping ~16 timer ticks, which means there are still a lot of frames being skipped in the meantime.

I also got out my supergun and tested out 2 stages on the CPS2 board (speed: turbo 2) but they were so variant that there was no point continuing. The difference in timer ticks in China was ~3 timer ticks on HDR T4 and ~5 timer ticks on HDR T3. Surprisingly, in U.S.S.R., speed varies ST because T3 was different by ~20 ticks and T4 by ~22 ticks… except that HDR was faster by that much. Note that HDR stages are constant and that T.Akiba found China 4th fastest and USSR 5th fastest in his tests on the JP version. I have no idea what to say about that.

And some more possible PS3 issues brought up by Grog after SB4 that I tried out:

1. dictator’s crossup j.MK,c.MK where the c.MK didn’t come out: This worked fine for me under a few different wakeup situations. I even did it 10 times in a row in practice mode without a hitch. Since moves aren’t buffered and any button press while still in recovery is ignored, this seems like a normal execution error.

2. Honda’s LK splash where no splash or LK came out: Again, no problem at all. I did it in multiple positions and in different timing and with other strengths but there were no issues.

3. DeeJay’s crossup j.MK,c.HP where the c.HP went through the opponent: I tried this and sure enough, if the j.MK was performed too high to combo, the c.HP always whiffed. After investigating further, the reason was clear. Dee Jay’s standing hitbox is slightly bigger than his crouching hitbox (you can check this out in training mode).

If the opponent walks forward up to Dee Jay while Dee Jay is crouching, he/she can get in slightly further than if Dee Jay was standing. And if the opponent then crouches, neither character gets pushed back. If the opponent’s crouching hitbox is thin enough (Chun is very thin, shotos are thin enough, Fei Long is too fat), Dee Jay’s c.HP starts too far (just like claw and Dhalsim’s c.HP) and misses the opponent altogether.

So when Dee Jay executes a crossup and immediately crouches, the opponent is never pushed back. Dee Jay is too close to the opponent and thus whiffs the c.HP. This happens the same way in normal ST, as well as the 360 version.

At this point, I don’t think it’s worth my checking for more routine input issues and attributing it to the PS3 platform. If folks will recall, similar “imagining” of issues was present when the PS Street Fighter Collection version of ST was played and again when the DC version was used.

If there are very specific circumstances someone saw that produced obvious results, I’ll still test. For simple combo mess-ups, I can only attribute that to unfamiliarity with PS3 controllers, muscle memory from inaccurate online speeds, or just plain human error. I welcome other folks to test for themselves though.

[Murderbydeath]

Ganelon, I appreciate all the testing, but honestly, people are just full of shit. It’d be best to just ignore them. I saw Grog bitching about crossup mk -> low fierce whiffing, but he was watching some ultrascrub trying it in training mode and he was drunk. I do it all the time, PS3 and X360. Anyways, thanks again.

[Rufus]

9. The difference between T1-T2, T2-T3, and T3-T4 is always ~3 timer ticks apiece. The difference between
   …
10. Even HDR’s max speed of T4 (1 above default) is still a couple ticks slower than the DC version’s T3 (default). And the DC version still has turbo speeds up to T6.
11. HDR T3 (default) is a couple of ticks slower than DC T2 (1 below default) but a tick faster than DC T1.
12. HDR is a few ticks slower than DC ST at the same turbo level. HDR T3 is slower than DC ST T3, same for T2, same for T1. This is true whether the DC version is set to fixed or free speeds.
13. However, HDR T0 is actually the exact same speed as DC ST T0. Different characters, different stages, and induced slowdown made no difference. It seems if the developers did a speed comparison between the DC version and HDR, that was only checked on T0.

I posted this over in the frame data thread, but it’s interesting for you to come up with that. The XBox360’s HDR turbo settings work like this:
Turbo 0: 32 frames displayed, 0 frames skipped
Turbo 1: 32 frames displayed, 6 frames skipped
Turbo 2: 32 frames displayed, 7 frames skipped
Turbo 3: 32 frames displayed, 8 frames skipped
Turbo 4: 32 frames displayed, 9 frames skipped
(Tested for samples on the order of 5000 displayed frames. The skip pattern’s period is always a divisor of 32.)

[EA_Megaman]

The graphical assets in HDR were designed to be displayed in 1080p 4:3 format. So a 1080p monitor should be the best looking and fastest processing (no scaling) viewing device for this game.

What resolution the textures were designed for has nothing to do with performance.

The fact that the assets arent large enough to render pixel accurate into the frame buffer is really nothing new. It happens in 3D games all the time. Because you can move around the world pixels projected to the screen always occupy different areas. This is filtering hardware is built into graphics chips. The cost (performance) of this is hidden and the only thing it causes are filtering artifacts.

Either aliasing (shimmering, noise) when the texture gets smaller on the screen, or magnification (blurring) when it gets larger. Sure it would be possible for this game to ship textures that were 1:1 with the display, but then they would have to have textures for every possible resolution/aspect that people could set on their device. Instead we have bi-linear/tri-linear filtering and mip-mapping support built into graphics chips to handle this for free.

The only time there may be performance implications is when the framebuffer is rendered larger/smaller than the output resolution and the output scaler is activated. If the game is already running close to 16ms per frame, the hardware scaler might add half a millisecond to the rendering of the frame, reducing the output speed lower than 60fps.

When setting up the renderer on the PS3 or Xenon, games typically (but not always) query the hardware to see what resolution it is currently set at. They will create the backbuffer at the highest resolution that doesnt exceed the output setting, or (more of a concern on the 360/EDRAM) exceed the memory capacity of the graphics hardware. I highly doubt the hardware scaler is even active on this game.

Scaling is a big problem with TVs however. If you set your console on 480p, and upscale that to 1080p using your TV the scaler probably doesnt run at 60hz, so the image you see may be a few frames behind what the console is outputting.

On a side note, I was wondering if you might be able to test/capture the screen jump in ST and 16:9 HDR. I’m curious what the relative difference is between the two versions. You would expect the HDR 16:9 screen jump to be just ST but scaled up. However, it doesnt look like that and it’s either my eyes, or something else. Plus I dont have a way to record Super Turbo, only HDR.

[Ganelon]

A few days back when HDR was first announced to be in 16:9 (ugh, but it is what is is), I tested around in practice mode for a few hours to try and get used to it. During that time, I was trying a lot of claw’s combos including his B&B (j.HP,c.MK,c.MP). Being a link, you must time each button after recovery within a limited number of frames. Between c.MK and c.MP is about 4 frames to combo, not incredibly difficult but easy to mess up if you’re not on point.

I normally perform the combo partially by muscle memory and partially based off visual cues. I tried it first of all on 360 for awhile and all was well. When I moved onto PS3 right afterward however, my c.MP would often not combo, seemingly with the same timing. I was sort of confused why that’s the case and realized I was depending on the visual indicators for when claw would retract his leg from the c.MK. Once I started timing the c.MP slightly earlier than what I was just doing on 360, my muscle memory took over and I could perform the link consistently again.

Well, that got me thinking of other folks complaining about combos dropping between versions. As explained above, I extensively tested the speed differences inside the actual matches for 360 and PS3 and found both to be identical. The only matter left to test was input lag and nobody had appropriate means of testing it. But now, we know SFIV and SSFIV on PS3 is confirmed to have 1 frame more input lag than on 360. Folks who attended Stunfest have told me that NKI (one of the first to report on input lag in CCC2) could feel lag on PS3 HDR as well. Many US players (quoted above) have felt that things were different and this time, spending lengths of time in practice mode on both systems one immediately after the other, I have to personally lean with the notion that there appears to be input lag on the PS3.

Now, this is not empirical proof of any difference (I could be imagining it for all I know), just another opinion cautioning that it’s potentially possible the 360 version may in fact be superior for the input reason many have felt. We’ll need an actual test to prove the existence of input lag. Of course, whatever the outcome is, it’s too late to change Evo (plus, SSFIV players will be dealing with the same). So as many have said, it’s a good idea to play the PS3 version offline if you haven’t already. Players who don’t use strict link combos and don’t rely on visual cues probably won’t notice anything. DGV alluded to another thread where people do have the means to perform input lag tests and hopefully shed more light into this issue for future tourneys: http://shoryuken.com/showthread.php?t=231128

[CWheezy]

I think evo might be able to run hd remix on xbox if enough xboxes are brought, but if not enough are brought then hdremix stays on ps3.

Maybe if the word is sent out it can be done?

[Coth_X]

:wow: Even Ganelon is saying it now.

[Rufus]

Now, this is not empirical proof of any difference (I could be imagining it for all I know), just another opinion cautioning that it’s potentially possible the 360 version may in fact be superior for the input reason many have felt. We’ll need an actual test to prove the existence of input lag. Of course, whatever the outcome is, it’s too late to change Evo (plus, SSFIV players will be dealing with the same). So as many have said, it’s a good idea to play the PS3 version offline if you haven’t already. Players who don’t use strict link combos and don’t rely on visual cues probably won’t notice anything. DGV alluded to another thread where people do have the means to perform input lag tests and hopefully shed more light into this issue for future tourneys: http://shoryuken.com/showthread.php?t=231128

I could probably put together an input lag testing set-up, but I don’t have a PS3 and I’m …reluctant… to spend $370 for a PS3 testbed that I’d probably never use for anything else. I suppose I could try to track down someone in the area who has most of what would be necessary, but in practice, I’m not sure it’s going to be all that useful anyway - the people who play competitively will end up having to practice to adjust regardless, and the publishers are unlikely to patch for it - especially for SF2 games where we probably won’t see any publisher activity 'till the next generation or two of consoles have come out.

[Shari]

To clarify.

NKI did not say there was inputlag on PS3 HDR.

He said there was inputlag on HDR.