Gaming Homebrew Review 4IFIR / OC Suite - Showcases (performance / power consumption)

[Foxi4]

I don't get the math.

if it's 8x the performance. Wouldn't that mean the game if it runs in 60 fps, run in like 480 fps?

Nobody needs to be a brainiac to calculate this, we’re comparing like to like. In fact, we’re comparing performance on the same machine and the same die, just with different settings.

If you’re getting 480p@30 pre-overclock then you need double the compute to render at the same resolution with double the frames, so 480p@60. In fact, you’ll need a smidge more due to efficiency losses (double clock =/= double performance, it has to be measured as the increase is not linear).

“x12 performance” is voodoo. It’s stupid to even publish a number like that. What Cooler does is conflating his performance increase from a higher clock with his efficiency increase from lower power consumption. That’s not how that works, and it’s never how that worked since it’s two different measurements.

I’ll say this one more time in the vain hope that it will sink in - if you tune a car engine and manage to increase the RPM by 50% while decreasing your power consumption by 50%, that’s great, but don’t go around telling people that you’ve increased your performance by 4x, 8x, 12x or whatever x you arrive at with voodoo. The performance of the engine is measured in horsepower and your performance increase is your final increase in horsepower.

 

[Cooler3D]

Well spotted, I have VGA on my mind recently, Let me run the math again on my handy napkin, since the equation doesn’t change here.

540x360 = 230400
960x540 = 518400

Pixel difference = 518400 - 230400 = 288000
% difference = 288000/230400x100 = 125%

Wait a second. I'm confused, or you're wrong. Let's go first:

We have a game that renders at 640*360 (230400 pixels, 360p) on the stock console. The frame rate drops below 20 fps.
In order to render this game at 1280*720 (921600 pixels, 720p) with a frame rate not dropping below 60 fps, we need:
(921600 / 230400) * (60 / 20)
We get 12. We need at least x12 to the original performance for this game to work at 720p at a stable 60fps.
Right?
In my marketing crap, it is said that 4IFIR, with a number of reservations, in ideal conditions is able to give up to x8 in performance, compared to stock in portable mode.
This means that in an ideal scenario, testing should show something like:
720p at 40fps = 480p at 60fps.
So? You were right, I made a mistake when counting not in my favor, lol. Thank you for noticing.

Correct me if I'm wrong, just please be careful with calculation.
And I'm going to test it so we can find out for sure if 4IFIR is capable of "up to x8 to total performance (on x2.2 wattage vs portable stock)".

 

[Foxi4]

I did this over a week ago. No matter what forum I post on, a link to the repository will certainly be there. I would love to go back to GBA, but I feel like you guys really dislike me, I feel like I'm annoying the two of you, and as a result, your attitude towards me feels humiliating. I don't like annoying people, I don't like asking for favors. It makes me uncomfortable, I feel like an outcast here. This is the only thing that stops me, I have no ulterior motives.

Nobody is hostile to you - you haven’t been warned, you haven’t faced any disciplinary action, you’re allowed to post to your heart’s content and if anything, the community gathered around you to give you tips on how to make your project better. Your problem is interpreting every piece of advice as hostility, then retaliating and finally crying wolf when someone loses their temper with you. I’m glad that you’ve put in the effort to start a proper GIT for one of your projects as was suggested to you - I visited it, it’s heaps better compared to your previous attempts and very easy to browse. If you were a little bit more amenable to taking criticism on board, you wouldn’t be having nearly as many issues. When you’re posting figures that are verifiably bollocks and only “work” when using your convoluted, non-standard method of measurement, you’re going to get called out on it. That’s not a negative statement about you, that’s a clarification on what can or cannot be achieved.

When you say “x12 performance” the average person who has seen an electronic device before will assume “huh, if I’m running at 30 frames now and my performance is going to increase x12, I can run at 360 frames!” which is idiotic. What’s actually happening is a modest increase in performance (measurement #1) and a decrease in power consumption (measurement #2).

You are doing *yourself* a disservice by posting figures that are *outrageous* and I am helping you correct that so that your documentation doesn’t look ridiculous to anyone who’s ever overclocked anything before.

Wait a second. I'm confused, or you're wrong. Let's go first:

We have a game that renders at 640*360 (230400 pixels, 360p) on the stock console. The frame rate drops below 20 fps.
In order to render this game at 1280*720 (921600 pixels, 720p) with a frame rate not dropping below 60 fps, we need:
(921600 / 230400) * (60 / 20)
We get 12. We need at least x12 to the original performance for this game to work at 720p at a stable 60fps.
Right?
In my marketing crap, it is said that 4IFIR, with a number of reservations, in ideal conditions is able to give up to x8 in performance, compared to stock in portable mode.
This means that in an ideal scenario, testing should show something like:
720p at 40fps = 480p at 60fps.
So? You were right, I made a mistake when counting not in my favor, lol. Thank you for noticing.

Jesus wept. Fine, I’ll walk you through it. We’re going to assume that the native resolution is stable-ish because we need a baseline of some kind. It’s not stable, but it’s stable enough for our purposes at this time.

@30-ish FPS
640x360 = 230400
1280x720 = 921600

Pixel count difference = 921600 - 230400 = 691200

That’s how many more pixels per frame you need to fill, how big of a % increase is that?

691200/230400x100 = 300%

That’s 3x the work required, so you would need 3x compute - at the same framerate, forget about 60 FPS. If you wanted 60 with dips, you’d need to double that, if you wanted 60 constant, you’d need triple that (since 20 FPS is your lowest dip). At no point are you anywhere near 12, you need 3x compute at the same framerate, 6x compute at double the framerate and 9x compute at three times the framerate.

If you want to save yourself the hassle of counting percentages, you can just use a calculator. No harm in doing that, but do it right.



https://percentagecalculator.net/

It’s a super simple equation anyway - Increase/OriginalFigure x 100 = Increase %

For the record, all this time we’re dancing around the fact that the Switch runs *underclocked* by default compared to the NVidia spec for the purposes of saving battery life and keeping thermals under control. Realistically you’re only “overclocking” it by a small fraction since the die itself was always intended to run faster than it’s running in the Switch, by a *huge* margin.

 

[Cooler3D]

@30-ish FPS
640x360 = 230400
1280x720 = 921600

Pixel count difference = 921600 - 230400 = 691200

That’s how many more pixels per frame you need to fill, how big of a % increase is that?

691200/230400x100 = 300%

x4 ≠ 100% +400% = 500%
x4 = 100% * 4 = 400%

 

[hippy dave]

Putting code on a git still doesn't count if you actively avoid answering people who ask where it is. Stop wiping your arse with software licences, and stop spouting paragraph after paragraph of whiney feels, and just do the things you're supposed to do.

 

[linuxares]

Putting code on a git still doesn't count if you actively avoid answering people who ask where it is. Stop wiping your arse with software licences, and stop spouting paragraph after paragraph of whiney feels, and just do the things you're supposed to do.

He probably have blocked you and why he can't see your posts.

 

[Foxi4]

x4 ≠ 100% +400% = 500%
x4 = 100% * 4 = 400%

I don’t even know what you’re trying to say here. Do you know how percentages work? 100% is your baseline. That’s what you’re doing anyway. If you have a workload that is 300% larger than the baseline, you need to perform 300% of work. You’re not “adding” anything.

X = 100
3X = 3 x 100 = 300

You’re calculating an increase from workload A to workload B.

 

[Cooler3D]

X = 100
3X = 3 x 100 = 300

x1 = 100%, portable stock performance
x4 = 400% = portable stock performance multiplied by 4.

 

[hippy dave]

He probably have blocked you and why he can't see your posts.

He replied to me when it suited him.

 

[Foxi4]

x1 = 100%, portable stock performance
x4 = 400% = portable stock performance multiplied on 4.

View attachment 362735

I think my head is going to explode.

Cooler, for the love of Jesus, and Mary, and the Holy Spirit, and all else that is Holy. Which of the two things are you trying to say?

A) How much compute is required to render the scene?
B) How much your performance has to increase from stock?

Those two are not the same number. I will happily explain the difference, but you have to pick one or the other.

 

[Bricked]

@Foxi4 You will not get anywhere, people have been giving this Cooler3D character good advice since 2021, he hasn't acted upon any of it. He actively ignores it and prefers to engage in and instigate conflict.

The source code promise is also an empty one.

Here's what I found when I started to look into the history of this:

Now you might look at this screenshot and think "Oh Happy New Year! What's the big deal that was only 3 months ago."

Guess what, this is from January 2022.

So at this point he's had effectively the two pillars of the entire scene: SciresM and CTCaer, ask that he provide modified source. Nothing.

 

[Cooler3D]

I think my head is going to explode.

Cooler, for the love of Jesus, and Mary, and the Holy Spirit, and all else that is Holy. Which of the two things are you trying to say?

A) How much compute is required to render the scene?
B) How much your performance has to increase from stock?

Those two are not the same number. I will happily explain the difference, but you have to pick one or the other.

I'm confused too. It would probably be better if I just check what 4IFIR can demonstrate in a given game, and you tell me what it equivalent to, and whether it's cool or not.
And if the result does not seem very impressive to you, then I have been embellishing reality all this time. Ok?

 

[Foxi4]

I'm confused too. It would probably be better if I just check what 4IFIR can demonstrate in a given game, and you tell me what it equivalent to, and whether it's cool or not.
If the result is not "very impressive", then I have been embellishing reality all this time. Ok?

For the purposes of informing the audience and yourself, I will explain. A lot of this stuff is about phrasing your results correctly.

Let’s say we’re rendering *1* pixel at *1* frame per second. That’s *stock* when operating at 100%. It is our baseline and a given, that’s happening already.

Now, let’s say we want to render 1 pixel at *4* frames per second. That’s an additional 3 frames that we must render, so in addition to our stock value we also need to do an additional 300% of work (if we consider stock to be 100%)

Let’s say we succeed. We went from operating at 100% to 400%, since you now have 4 frames. That’s our performance.

1 frame = 100%
4 frames = X
4*100/1 = 400%

…but our *performance increase* is 300% compared to baseline because we were already rendering 1 frame anyway. We can’t treat our baseline *as part of the increase*, that’s a frame that was already there to begin with.

So it’s correct to say that we’re operating *at 4x* because we need 4x compute, it is correct to say that we’ve increased performance by 3 frames compared to stock, but it is incorrect to say that we’ve *increased performance by 4x* because we haven’t - we’re only rendering 3 additional frames compared to stock, and stock was 1.

All of that is admittedly splitting hairs, I’m being pedantic, but I just don’t like seeing imaginary figures based on nothing.

Realistically we have to run the game at stock, run the game with an overclock, see how much additional work you’ve managed to do and represent that as a percentage. Normally it runs at 20 FPS, if we get it to run at 60, that’s an *increase* of 40 additional frames, or +200% compared to baseline.

If I were a betting man, I *suspect* that you’ll be able to make the framerate more stable, but you won’t achieve a significant resolution boost. You just can’t the Tegra doesn’t overclock that well, we’re operating with silly-looking numbers only by the virtue of it being *underclocked* in the Switch to begin with. If it was running at NVidia’s values rather than Nintendo’s things would be significantly less impressive - most of what we’re doing is unlocking performance that’s already baked in.

I *think* I understand what you’re saying, 60 FPS is 3x the frames compared to 20 FPS, and that’s a valid way to present it, but you have to decide if you’re measuring performance or performance increase, as they are not the same.

 

[Cooler3D]

I *think* I understand what you’re saying, 60 FPS is 3x the frames compared to 20 FPS, and that’s a valid way to present it, but you have to decide if you’re measuring performance or performance increase, as they are not the same.

The difference is in our understanding of what x2 is.
In my understanding, this is a stock multiplied by two.
In your understanding, this is a stock to which the stock multiplied by two is added.

Apparently the problem is in my English, but I understand x2 not as "stock + stock * 2", but just as "stock * 2".

It doesn't matter to me how to count. If you think that your interpretation is more correct, out of respect for you, I am ready to continue to consider your way. The essence of this will not change much. In both cases, the result is breakthrough and unprecedented. If, of course, this result corresponds to the real possibilities of 4IFIR. We will find out very soon. I went to check.

 

[Foxi4]

The difference is in our understanding of what x2 is.
In my understanding, this is a stock multiplied by two.
In your understanding, this is a stock to which the stock multiplied by two is added.

Apparently the problem is in my English, but I understand x2 not as "stock + stock * 2", but just as "stock * 2".

It doesn't matter to me how to count. If you think that your interpretation is more correct, out of respect for you, I am ready to continue to consider your way. The essence of this will not change much. In both cases, the result is breakthrough and unprecedented. If, of course, this result corresponds to the real possibilities of 4IFIR. We will find out very soon. I went to check.

It’s not an interpretation thing, it’s a matter of what you’re measuring. If you had 20 frames and now you have 60 then you’re +200% ahead in frames. If you were running at 100MHz and now you’re running at 200MHz then you’re running at 200%, or double the stock.

Like I said, all of this is moot as the majority of the wiggle room in Switch “overclocking” is just allowing the chip to run at clocks it was always designed to run at. Nintendo underclocked it heavily at stock to preserve battery life, so it’s really “overclocking” with a big asterix.

Don’t take this as me saying there’s not much to do here, there’s a lot of potential to unlock, solely by the virtue of stock values being so poor compared to what the chip can do.

 

[Cooler3D]

It’s not an interpretation thing, it’s a matter of what you’re measuring. If you had 20 frames and now you have 60 then you’re +200% ahead in frames. If you were running at 100MHz and now you’re running at 200MHz then you’re running at 200%, or double the stock.

Like I said, all of this is moot as the majority of the wiggle room in Switch “overclocking” is just allowing the chip to run at clocks it was always designed to run at. Nintendo underclocked it heavily at stock to preserve battery life, so it’s really “overclocking” with a big asterix.

What is your "realistic" forecast?

To ensure the operation of stock performance, the console of the second revision consumes approximately 5 watts. The game drops below 20fps, rendering in 360p. This is reliable information, it is easy to verify.

What do you think, at what maximum frame resolution will I be able to achieve a stable 60fps at 10 watts with 4IFIR? What result do you find impressive?

 

[Foxi4]

What is your "realistic" forecast?

To ensure the operation of stock performance, the console of the second revision consumes approximately 5 watts. The game drops below 20fps, rendering in 360p. This is reliable information, it is easy to verify.

What do you think, at what maximum frame resolution will I be able to achieve a stable 60fps at 10 watts with 4IFIR?

My realistic forecast really depends on what settings the particular game is running at and whether it’s CPU dependant or GPU dependant. You have to locate the bottleneck.

With games like Jurassic World the bottleneck is often times the CPU, not the GPU, so you have a lot of wiggle room on the Switch. They’re logic-heavy with a lot of AI/object instances running in the background, and those are primarily CPU tasks.

If the bottleneck is GPU, you *do not* have a lot of wiggle room on the Switch, particularly not on Erista.

Your increase in performance will always be dependant on whether the game is CPU bound or GPU bound, and by how much you can increase the performance of either.

Does that make sense?

For the record, this prediction is based on nothing - I don’t own JW2. I’m simply comparing to similar enough games in my head.

 

[Lamcza]

You don't know life if you don't drink kefir

God damn, you post the best one at least from polish ones... that Krasnystaw... sweet Jesus! It can cure the hangover in 2h and it can prevent the hangover if you drink it before! Freking magick potion! But some ppl like more "Kwaśne mleko" idk how they call it in Russia but i bet they have it.

 

[Cooler3D]

If the bottleneck is GPU

This terribly resource-hungry game, on the stock console, judging by the telemetry readings, bottlenecked simultaneously on the processor, memory and GPU.
Just think of any number and name it. You will not lose anything.

 

[Foxi4]

god damm you post the best one that krasnystaw... sweet Jesus! But some ppl like more "Kwaśne mleko" idk how they call it in Russia but i bet they have it.

Krasnystaw is the best kefir bar none, you are correct. Kwaśne mleko? Can’t say I’ve ever heard of that, do you mean maślanka? Surely not. Where can I get this kwaśne mleko?