Homebrew RELEASE sys-clk under/overclocking sysmodule

[krakotay]

Id genuinely love to see actual numbers instead of claims. Please give me some type of power draw comparison, id love to see it. I'm not trying to be sarcastic either I genuinely want to see this.

Is it actual? I can show you it

 

[Anxiety_timmy]

Is it actual? I can show you it

Sure

 

[krakotay]

Sure

Oh shit... i got a big problem. I don't know, how many W eats switch with default sys-clk. But i tested in doom eternal with grafical mode, that product of Cooler3D eats just 8-10W with 1300+ mhz OC GPU and 1400+ OC CPU. Regular switch rev. 2 eats 7w in stock (!), so...

 

[Anxiety_timmy]

Oh shit... i got a big problem. I don't know, how many W eats switch with default sys-clk. But i tested in doom eternal with grafical mode, that product of Cooler3D eats just 8-10W with 1300+ mhz OC GPU and 1400+ OC CPU. Regular switch rev. 2 eats 7w in stock (!), so...

These were on Mariko, Correct?
Also can you get Voltages for the CPU and GPU?

 

[krakotay]

These were on Mariko, Correct?
Also can you get Voltages for the CPU and GPU?

Yes, switch lite, its mariko
No problems. 810mv CPU and 810mv gpu.

Post automatically merged: Jan 6, 2023

810mv CPU

Without charging its 765 CPU, 810 gpu

 

[Cooler3D]

Id genuinely love to see actual numbers instead of claims. Please give me some type of power draw comparison, id love to see it. I'm not trying to be sarcastic either I genuinely want to see this.

As you wish:

vs

More than TRIPLE the performance per watt versus RetroNX SysCLK, even at minimal overclocking. Read more about the testing methodology at the link (sorry for the Russian, Google Translate will help).

Once again, not +3%, not +10%, not even +30%. A hilarious +300% superiority over garbage SysCLK by RetroNX team.

The Switch on 4IFIR, even with the stock TDP of portable mode (namely power consumption of OFW Switch Light without overclocking), has significantly more performance than the Switch overclocked beyond its limits of SysCLK by RetroNX.

 

[hippy dave]

As you wish:

View attachment 346528 vs View attachment 346529

More than TRIPLE the performance per watt versus RetroNX SysCLK, even at minimal overclocking. Read more about the testing methodology at the link (sorry for the Russian, Google Translate will help).

Once again, not +3%, not +10%, not even +30%. A hilarious +300% superiority over garbage SysCLK by RetroNX team.

The Switch on 4IFIR, even with the stock TDP of portable mode (namely power consumption of OFW Switch Light without overclocking), has significantly more performance than the Switch overclocked beyond its limits of SysCLK by RetroNX.

Idk man maybe don't be such an arsehole about people whose work you built off of?

 

[Cooler3D]

Idk man maybe don't be such an arsehole about people whose work you built off of?

I have very good reasons to consider them hypocritical scoundrels.
These people exhausted my credit of respect a year ago, at the announcement of 4IFIR, which they subjected to targeted, full-scale persecution:

Spoiler: More info about incident

For a whole month, I patiently listened to their humiliating, public insults. I tried to encourage them to discuss, at least to understand why they are lying inventing absurd accusations against me. I tried to understand what caused their hatred. When I got tired of it, I just left the English segment.

A year has passed since then, the number of downloads has long exceeded a million, and members of the RetroNH community still make up stories when someone mentions my project. Some members of RetroNX openly, not at all embarrassed, stalker after me on the Russian-language forum, for further discussion with their leaders in the community Discord.

Now that by its mere existence, 4IFIR has made all their false accusations untenable, becoming a living brand on their reputation, I have no reason left to be more polite to them than they were to me.

P.s. Less than 2% of the code remained from the original project in 4IFIR, but despite all of the above, I did not remove references to RetroNX from the project. Professional ethics.

If for some reason you are interested in the details, you can trace this story from a frozen thread on GBATEMP. A brief retelling with conclusions is available on 4PDA.
If the toxic perpetrators of anti4IFIR campaign - suppress their pride and make a public apology, I am ready to listen to them and even forgive.

p.s. Something tells me they won't.

 

[hippy dave]

I have very good reasons to consider them hypocritical scoundrels.
These people exhausted my credit of respect a year ago, at the announcement of 4IFIR, which they subjected to targeted, full-scale persecution:

Spoiler: More info about incident

For a whole month, I patiently listened to their humiliating, public insults. I tried to encourage them to discuss, at least to understand why they are lying inventing absurd accusations against me. I tried to understand what caused their hatred. When I got tired of it, I just left the English segment.

A year has passed since then, the number of downloads has long exceeded a million, and members of the RetroNH community still make up stories when someone mentions my project. Some members of RetroNX openly, not at all embarrassed, stalker after me on the Russian-language forum, for further discussion with their leaders in the community Discord.

Now that by its mere existence, 4IFIR has made all their false accusations untenable, becoming a living brand on their reputation, I have no reason left to be more polite to them than they were to me.

P.s. Less than 2% of the code remained from the original project in 4IFIR, but despite all of the above, I did not remove references to RetroNX from the project. Professional ethics.

If for some reason you are interested in the details, you can trace this story from a frozen thread on GBATEMP. A brief retelling with conclusions is available on 4PDA.
If the toxic perpetrators of anti4IFIR campaign - suppress their pride and make a public apology, I am ready to listen to them and even forgive.

p.s. Something tells me they won't.

Probably just stay the fuck out of their thread then tbh.

 

[Cooler3D]

Probably just stay the fuck out of their thread then tbh.

"They" managed to mess up the 4IFIR thread so much that it turned out to be easier to freeze it than to continue to participate in it. Now "this" is a 4IFIR thread. Anyone interested in overclocking deserves to know the truth, and the right to choose their preferred overclocking tool. And I guess what will be preferred.

Unscrupulous liars can go to their Discord, they will not mislead anyone else on GBATemp.

 

[hippy dave]

"They" managed to mess up the 4IFIR thread so much that it turned out to be easier to freeze it than to continue to participate in it. Now "this" is a 4IFIR thread. Anyone interested in overclocking deserves to know the truth, and the right to choose their preferred overclocking tool. And I guess what will be preferred.

Unscrupulous liars can go to their Discord, they will not mislead anyone else on GBATemp.

I feel like claiming the moral high ground doesn't work so well when you're being such a dickhead about it, but I've exhausted all the interest I had in this, so whatevs, be a dickhead.

 

[Anxiety_timmy]

As you wish:

View attachment 346528 vs View attachment 346529

More than TRIPLE the performance per watt versus RetroNX SysCLK, even at minimal overclocking. Read more about the testing methodology at the link (sorry for the Russian, Google Translate will help).

Once again, not +3%, not +10%, not even +30%. A hilarious +300% superiority over garbage SysCLK by RetroNX team.

The Switch on 4IFIR, even with the stock TDP of portable mode (namely power consumption of OFW Switch Light without overclocking), has significantly more performance than the Switch overclocked beyond its limits of SysCLK by RetroNX.

Do you have voltage lines for cpu/gpu
and tbh alot of those benefits come from dram so thats neat.

 

[Cooler3D]

Do you have voltage lines for cpu/gpu
and tbh alot of those benefits come from dram so thats neat.

You are welcome, even more power efficient modes - coming soon.

About lines:
For Mariko, ST1-2-3-4+ etc means: "Power consumption equivalent to mode X steps lower."
ST4, for example = - 4 steps = power consumption of portable stock mode.

 

[guily6669]

As you wish:
..............
Once again, not +3%, not +10%, not even +30%. A hilarious +300% superiority over garbage SysCLK by RetroNX team.

The Switch on 4IFIR, even with the stock TDP of portable mode (namely power consumption of OFW Switch Light without overclocking), has significantly more performance than the Switch overclocked beyond its limits of SysCLK by RetroNX.

Does you mod work on Erista V1 too and does it make that much difference in it compared to latest OC suite? (Now they support Erista out-of-the-box and don't even need to add any line to Hekate settings for the memory OC).

Also what exactly consist your mod in terms of the improvements, ram trimmings + voltage optimisations?

 

[Anxiety_timmy]

I have very good reasons to consider them hypocritical scoundrels.
These people exhausted my credit of respect a year ago, at the announcement of 4IFIR, which they subjected to targeted, full-scale persecution:

Spoiler: More info about incident

For a whole month, I patiently listened to their humiliating, public insults. I tried to encourage them to discuss, at least to understand why they are lying inventing absurd accusations against me. I tried to understand what caused their hatred. When I got tired of it, I just left the English segment.

A year has passed since then, the number of downloads has long exceeded a million, and members of the RetroNH community still make up stories when someone mentions my project. Some members of RetroNX openly, not at all embarrassed, stalker after me on the Russian-language forum, for further discussion with their leaders in the community Discord.

Now that by its mere existence, 4IFIR has made all their false accusations untenable, becoming a living brand on their reputation, I have no reason left to be more polite to them than they were to me.

P.s. Less than 2% of the code remained from the original project in 4IFIR, but despite all of the above, I did not remove references to RetroNX from the project. Professional ethics.

If for some reason you are interested in the details, you can trace this story from a frozen thread on GBATEMP. A brief retelling with conclusions is available on 4PDA.
If the toxic perpetrators of anti4IFIR campaign - suppress their pride and make a public apology, I am ready to listen to them and even forgive.

p.s. Something tells me they won't.

Issue stems from you claiming that things that aren't safe were safe on V1 models. It just left a sour taste in everyone's mouth over there.

 

[Cooler3D]

Issue stems from you claiming that things that aren't safe were safe on V1 models. It just left a sour taste in everyone's mouth over there.

Spoiler: The only issue

First, I never claimed this. It's one of those tales they made up to smear me. On the contrary, I described the project as even more dangerous than it really was.

Secondly, how does this even justify a dastardly conspiracy to deliberately denigrate my reputation through their own tales? Massive harassment and insults against me, attempts to humiliate, slander and slander as much as possible. Purposeful, organized, full-scale bullying. This is actually a crime to which these dishonest crooks did not have and could not have any right.

I mentioned that my calculations did not reveal any reasons that indicate the danger of a console failure. Consider the project dangerous. I asked them, instead of insulting me, to constructively substantiate the accusations leveled against me. It's a pity that I did not immediately understand that they were making up these claims, and their real goal was to denigrate me and cancel my project.

I also mentioned that at that time there was not one known case of a console failure due to 4IFIR. Even at that time, there were already thousands of people willing to test the experimental alpha at their own risk. Now that the install base has long jumped over a million, and the project has been around for over a year, guess how many console failures have been recorded? Zero. It is a fact.

According to my offenders, their own experiments with overclocking led to the destruction of a couple of consoles of members of the Discord community. Even assuming they weren't lying, I can only sympathize with those who dare to test experimental patches from incompetent authors. Only now, their shit code has nothing to do with 4IFIR.

Fortunately, time put everything in its place, justice prevailed, my reputation was restored long ago, which cannot be said about the reputation of these scoundrels, which deservedly received irreparable damage. None of these unscrupulous crooks had the guts to simply apologize in advance. Where did the courage and pry with which they once persecuted me all over the Internet go? Now that their guilt has long been obvious and proven, these cowards even ran away from their own thread. What a pitiful sight.

Post automatically merged: Jan 13, 2023

Does you mod work on Erista V1 too and does it make that much difference in it compared to latest OC suite? (Now they support Erista out-of-the-box and don't even need to add any line to Hekate settings for the memory OC).

Also what exactly consist your mod in terms of the improvements, ram trimmings + voltage optimisations?

Soon, I will translate and publish a post here, with all the information on 4IFIR, description of features, etc.
In short, yes, the improvements over Kazushima's project are quite significant for Erista as well. Most of them, one way or another, are associated with an increase in performance or an increase in power consumption. After the upcoming update, when the latest optimizations from Mariko will be ported to Erista (including complex optimization of GPU modes, including the new 4IFIR Optimized and 4IFIR Maximized modes), the benefits will become even more obvious.

 

[Anxiety_timmy]

Spoiler: The only issue

First, I never claimed this. It's one of those tales they made up to smear me. On the contrary, I described the project as even more dangerous than it really was.

Secondly, how does this even justify a dastardly conspiracy to deliberately denigrate my reputation through their own tales? Massive harassment and insults against me, attempts to humiliate, slander and slander as much as possible. Purposeful, organized, full-scale bullying. This is actually a crime to which these dishonest crooks did not have and could not have any right.

I mentioned that my calculations did not reveal any reasons that indicate the danger of a console failure. Consider the project dangerous. I asked them, instead of insulting me, to constructively substantiate the accusations leveled against me. It's a pity that I did not immediately understand that they were making up these claims, and their real goal was to denigrate me and cancel my project.

I also mentioned that at that time there was not one known case of a console failure due to 4IFIR. Even at that time, there were already thousands of people willing to test the experimental alpha at their own risk. Now that the install base has long jumped over a million, and the project has been around for over a year, guess how many console failures have been recorded? Zero. It is a fact.

According to my offenders, their own experiments with overclocking led to the destruction of a couple of consoles of members of the Discord community. Even assuming they weren't lying, I can only sympathize with those who dare to test experimental patches from incompetent authors. Only now, their shit code has nothing to do with 4IFIR.

Fortunately, time put everything in its place, justice prevailed, my reputation was restored long ago, which cannot be said about the reputation of these scoundrels, which deservedly received irreparable damage. None of these unscrupulous crooks had the guts to simply apologize in advance. Where did the courage and pry with which they once persecuted me all over the Internet go? Now that their guilt has long been obvious and proven, these cowards even ran away from their own thread. What a pitiful sight.

Post automatically merged: Jan 13, 2023

Soon, I will translate and publish a post here, with all the information on 4IFIR, description of features, etc.
In short, yes, the improvements over Kazushima's project are quite significant for Erista as well. Most of them, one way or another, are associated with an increase in performance or an increase in power consumption. After the upcoming update, when the latest optimizations from Mariko will be ported to Erista (including complex optimization of GPU modes, including the new 4IFIR Optimized and 4IFIR Maximized modes), the benefits will become even more obvious.

Some of your edits on 4IFIR aren't exactly stable and you only have to ask quite a few of the switches I've come across. Any code be it yours or switch oc suite or zatchy or my garbage spaghetti code will have the same results if you push dram too high. DRAM will corrupt storage if it can't handle frequencies and timings, and its not a all size fits all solution either. I know for one of my consoles it was so poorly binned that it couldn't handle 1.8ghz on DRAM even with 1.2v. While the Erista unit I use is perfectly fine at 1.8ghz with stock voltages. With 4IFIR's 2.133ghz option on DRAM, constant crashes and at one point even SDFS corruption happened. For some even greater variations, look no further than speedo values. My own Erista I use cannot handle above 2.008GHZ-2.013GHZ on cpu, while my friends own Erista handles above 2.091GHZ. Granted this could be fixed with voltage tweaks but again some switches can just barely do 1.897GHZ, with some even failing to reach that. Status monitor reports the values as you set them because PCV reports those values back, but checking clockspeeds using other methods reveals the real clockspeed. I have yet to see any Erista units actually benefit from above 921 on GPU but id love to be proven wrong on that front.

Post automatically merged: Jan 13, 2023

Im of the mind of let people burn their hardware if they want, but some just see higher performance and don't stop to question if its even good for their units, and you have to remember most of us aren't on Mariko.

 

[guily6669]

Spoiler: The only issue

First, I never claimed this. It's one of those tales they made up to smear me. On the contrary, I described the project as even more dangerous than it really was.

Secondly, how does this even justify a dastardly conspiracy to deliberately denigrate my reputation through their own tales? Massive harassment and insults against me, attempts to humiliate, slander and slander as much as possible. Purposeful, organized, full-scale bullying. This is actually a crime to which these dishonest crooks did not have and could not have any right.

I mentioned that my calculations did not reveal any reasons that indicate the danger of a console failure. Consider the project dangerous. I asked them, instead of insulting me, to constructively substantiate the accusations leveled against me. It's a pity that I did not immediately understand that they were making up these claims, and their real goal was to denigrate me and cancel my project.

I also mentioned that at that time there was not one known case of a console failure due to 4IFIR. Even at that time, there were already thousands of people willing to test the experimental alpha at their own risk. Now that the install base has long jumped over a million, and the project has been around for over a year, guess how many console failures have been recorded? Zero. It is a fact.

According to my offenders, their own experiments with overclocking led to the destruction of a couple of consoles of members of the Discord community. Even assuming they weren't lying, I can only sympathize with those who dare to test experimental patches from incompetent authors. Only now, their shit code has nothing to do with 4IFIR.

Fortunately, time put everything in its place, justice prevailed, my reputation was restored long ago, which cannot be said about the reputation of these scoundrels, which deservedly received irreparable damage. None of these unscrupulous crooks had the guts to simply apologize in advance. Where did the courage and pry with which they once persecuted me all over the Internet go? Now that their guilt has long been obvious and proven, these cowards even ran away from their own thread. What a pitiful sight.

Post automatically merged: Jan 13, 2023

Soon, I will translate and publish a post here, with all the information on 4IFIR, description of features, etc.
In short, yes, the improvements over Kazushima's project are quite significant for Erista as well. Most of them, one way or another, are associated with an increase in performance or an increase in power consumption. After the upcoming update, when the latest optimizations from Mariko will be ported to Erista (including complex optimization of GPU modes, including the new 4IFIR Optimized and 4IFIR Maximized modes), the benefits will become even more obvious.

I will be waiting for news for Erista

Some of your edits on 4IFIR aren't exactly stable and you only have to ask quite a few of the switches I've come across. Any code be it yours or switch oc suite or zatchy or my garbage spaghetti code will have the same results if you push dram too high. DRAM will corrupt storage if it can't handle frequencies and timings, and its not a all size fits all solution either. I know for one of my consoles it was so poorly binned that it couldn't handle 1.8ghz on DRAM even with 1.2v. While the Erista unit I use is perfectly fine at 1.8ghz with stock voltages. With 4IFIR's 2.133ghz option on DRAM, constant crashes and at one point even SDFS corruption happened....................

Well anyone going into OC should always check things, I my self am not planning to go beyond the actual full Nvidia tegra SOC specs on my Erista (other than the memory cause I'm using 1863mhz when needed)...

Many people for example talk about SOC degradation, the SOC is usually the last important thing in those kind of devices unless going overvoltage a lot, the biggest problem will be the actual motherboard power design and those things are made to be as cheap as possible for the specified specs.

The SOC wearing faster is not even any kind of concern, I mean I had my intel I7 2600K from 2011 (3.8ghz max turbo) at 4.9ghz on high voltage and on a air cooler for all this time and only manage to burn 1 core half a year ago, 11 years in use with many thousands of hours.

Portable things made on a tight budget will have many other major failures than the SOC, usually power delivery might go puff on OC.

Erista I have seen a few years ago and seem to be kinda locked into 18W max stock and that is when charging both the console + joy-cons + dock USB\hdmi while playing as it will reduce charge power to not go above it's USB-C power design despite the charger being above 30W...

Digital foundry when tested quite a few time ago the Switch Erista can jump to 20W on OC (I think it was actually just full Tegra X1 official clocks and not even going above it's specs), that's already above Switch internal power usage for how the board is designed as stock specs not to mention if the switch had to charge the battery + Joy-cons at the same time though it probably wouldn't even charge almost anything...

ps: On desktop PC it's already tricky to get actual full extreme OC when power delivery sometimes is not even any problem on a well engineered board + GPU PCB design made for extreme OC, it's mostly luck to get good silicon than can OC a lot with lower voltages than the majority... Usually RAM OC is the worse and takes a HUGE amount of time on a PC if you want to get the most out of it, need to increase voltage on memories, CPU memory controller and spend many countless hours tweaking the timings (sometimes also seem stable but might fail like after 2 hours of heavy testing, BSOD's and all sorts of errors)...

 

[Anxiety_timmy]

I will be waiting for news for Erista

Well anyone going into OC should always check things, I my self am not planning to go beyond the actual full Nvidia tegra SOC specs on my Erista (other than the memory cause I'm using 1863mhz when needed)...

Many people for example talk about SOC degradation, the SOC is usually the last important thing in those kind of devices unless going overvoltage a lot, the biggest problem will be the actual motherboard power design and those things are made to be as cheap as possible for the specified specs.

The SOC wearing faster is not even any kind of concern, I mean I had my intel I7 2600K from 2011 (3.8ghz max turbo) at 4.9ghz on high voltage and on a air cooler for all this time and only manage to burn 1 core half a year ago, 11 years in use with many thousands of hours.

Portable things made on a tight budget will have many other major failures than the SOC, usually power delivery might go puff on OC.

Erista I have seen a few years ago and seem to be kinda locked into 18W max stock and that is when charging both the console + joy-cons + dock USB\hdmi while playing as it will reduce charge power to not go above it's USB-C power design despite the charger being above 30W...

Digital foundry when tested quite a few time ago the Switch Erista can jump to 20W on OC (I think it was actually just full Tegra X1 official clocks and not even going above it's specs), that's already above Switch internal power usage for how the board is designed as stock specs not to mention if the switch had to charge the battery + Joy-cons at the same time though it probably wouldn't even charge almost anything...

ps: On desktop PC it's already tricky to get actual full extreme OC when power delivery sometimes is not even any problem on a well engineered board + GPU PCB design made for extreme OC, it's mostly luck to get good silicon than can OC a lot with lower voltages than the majority... Usually RAM OC is the worse and takes a HUGE amount of time on a PC if you want to get the most out of it, need to increase voltage on memories, CPU memory controller and spend many countless hours tweaking the timings (sometimes also seem stable but might fail like after 2 hours of heavy testing, BSOD's and all sorts of errors)...

If the SOC degradation thing comes from me, then no I don't mean a consoles SOC is going to die after overclocking for a long time. What can happen is a clockspeed is set and its running at that speed, but its throwing so many errors that the performance is worse than at lower clockspeeds. I have seen Eristas drawing more power than 18 watts, and sometimes they'll just pull from the battery whenever it can't get the power it needs from the charger. Either way, overclock until you get the performance you want, and then leave it there.

 

[guily6669]

The SOC degradation was more a general thing, not exactly directed for you, but while searching for OC stuff I see many ppl all over the place concerned mostly with it which is pointless...

I totally get what you are saying but in the PC I also had many similar problems to get even lower performance on high OC and doesn't even involve much software behind it like Switch which I know if it the code is not well optimized could lead to everything happening and not even having to do with the hardware OC of clocks+voltages but rather the software screwing it all...

I have very low experience in Switch OC as I just started, so far latest OC suite seems to be working ok on my Erista, but I never pushed much high clocks.

So far I have been having crashes only on Red Faction Guerrilla, but most likely was the inf ammo\jetpack\charges but still finished the game well. I'm almost sure it has nothing to do with the low SOC clocks but the memory at 1862mhz might be causing it too cause I never memtested it for hours, only like 30min OK, my memories could be heating a lot too cause I have beyond thick thermal pads in them directly to the backplate which probably isn't doing anything to cool them and I also removed the EMI shielding all around the SOC though I have always played with BT\Wifi OFF...