Today I will try to upload an update with full support for HOS16. After that, I will publish here.
This notification means that you have crossed the power consumption threshold of ~15 watts from battery. For ethical reasons, I did not disable this mechanism. Instead, I tried to optimize the power consumption of the console components to fit as much performance as possible into this threshold. On MARIKO, this made it possible to achieve phenomenal performance (up to CPU 2.5GHz GPU 1.4GHz RAM 2.5GHz) available from the battery (~10 watts). On ERISTA, the power consumption of components is also reduced, but I have not yet managed to port all the mechanisms for reducing power consumption to it, and combinations of maximum frequencies for the CPU and GPU from the battery are not yet available to you.
However, if you overclock only one of the two components, for example 2.3GHz for a 3DS emulator, you will be well within the ~15 watt threshold.
It is also worth understanding that thanks to low-level optimizations aimed at increasing performance per watt, including due to low-level optimization of timings and memory bandwidth, at the same frequencies with unlocked SysCLK - on 4IFIR you will have significantly more performance.
As a result, on the 4IFIR-e ST6+, at 1.2GHz CPU 0.69GHz GPU 2.1GHz RAM in graphics-heavy scenarios (60fps mod), you get more performance than an unlocked SysCLK would give at maximum frequencies (1.7GHz / 0.9GHz / 1.6GHz), but at the same time, on 4IFIR, you remain almost at the stock level of power consumption of the console without overclocking (at least not go beyond stock wattage too much).
Actually, this is how MARIKO achieved "up to x5 performance per watt" (in some scenarios even higher, but I was too shy to state it). On ERISTA, the results are not "up to x5" yet, but also quite impressive, and I'm working towards further optimizing the performance per watt.
Conclusion:
Pick up the values of frequencies that fit into the threshold of 15 watts. Disable CPU frequency autoboost (when running on battery power, it hinders rather than helps, leading to short-term power consumption jumps, which, in combination with GPU overclocking, are enough to trigger protection). 1.2GHz / 0.84GHz / 2.1GHz fit well into the 15 watt threshold on the ST6+.
Last edited by Cooler3D,
friend we are talking about the same thing. Perhaps the translator is mistranslating. You are right! but I supplement the fact that everyone everywhere writes like that. there I bought a vacuum cleaner, it says the power is twice as much in the motor, and in fact the suction is only 20% stronger than in the previous model. my analogy with the motor of a car is also about this. you're right, but everyone writes like that, I'm just used to it. I read about overclocking laptops and frames, GPUs, everywhere they confuse nominal power with
maybe because generally this topic is mentioned in proportion to the power with the consumption, if something is 20% more powerful, but the consumption is half, it is considered that it doubles the power in proportion to the other device, this is like Cooler calculates this I think, performance divided by watts, and make the comparison
P.S. It can be confusing for the user, I'm not saying it's wrong, but maybe not the best way to put it, hahaha
If everyone writes like that then everyone’s wrong and needs to be corrected. Just because the number of people who are wrong is large doesn’t make them right. If you wrote that kind of claim in a brochure for your car tuning service, your customers would sue for false advertising, and they’d be right. You can have a ridiculous increase of RPM, brag about it to the entire world and everyone will be very impressed until you tell them that you measured it while you were in neutral, so the engine wasn’t actually doing any work. Words mean things.
Just two years ago a lot of overclockers pointed out a bug in Alder Lake CPU’s. Without enabling FLL OC Mode in the BIOS they’d report all sorts of nonsense in CPU-Z, including ridiculous overclocking results like reaching 8+GHz frequencies which were obviously fake. Did that stop GIGABYTE from using said results in their marketing materials? No. Did they get *lit up* online for it? Heck yes, because it’s not a real result, and everybody (including GIGABYTE) knew that from the start.
Don’t confuse my post for some kind of assassination of character - I’m just saying that the figures presented are not true. That doesn’t mean that performance hasn’t increased, it means that the increase wasn’t measured correctly or wasn’t represented correctly, or both. If this is the best solution available to you right now then you should keep on using it - god speed. That being said, claims of performance increase should be realistic and accurate, otherwise they’re misleading to users who don’t know any better.
Who cares really what the number really is. if its even 1% better then sysclk im using it lol
By all means, do it! Everyone should use whatever works for them. That being said, misleading claims of performance benefits are not fair to the end user. You can’t tell someone that their performance will triple and expect them to not be upset when they complain that their performance only increased by a quarter - that’s fair criticism. If that kind of thing is happening, it’s an indication that the methodology for testing needs to be changed. Personally, I’d be very reserved about posting *any* performance increase claims because they’re highly variable - some people will see a big uplift, other people’s results might be more modest. Once you put a number out there, the expectations are set, and that’ll bite you every time.
EDIT: If we’re talking about a figure like performance per watt (like someone above mentioned) then that needs to be disclosed at the time of writing so it’s clear to whoever’s reading the result and no false expectations are set. Performance per watt and “performance” in general are not the same, a higher PPW just means that you did X amount of work using less power than normal. It doesn’t even mean that overall performance has increased - you might end up running a test *slower* than at stock, but because your power consumption has decreased significantly, the machine is performing more work *per watt*. It’s a different category. Still very relevant, especially on a battery powered system, but you’re measuring a “different kind of performance”, so to speak.
…and that is also suspect, but that’s besides the point.
On the bright side that kind of claim is actually fairly easy to validate.you're right, but you're too steamed on . You must be an engineer or a lawyer. I read a lot of nonsense , but I analyze it myself, the author here is a cool cooler, the only one in the world who does this in principle, even if he writes that the power is growing by 10,000%, I will not care. I just read it and I will continue to analyze it myself, and I will read what people write, someone has rolled out a worthy review above. you, too, create a super review it would be interesting from you with your thinking and mind
I’m not “steamed up”. Cooler and I are on relatively good terms, I think - this isn’t exactly our first encounter. If anything, I’m giving him tips on how to better represent his results to the community so that he doesn’t have as hard of a time as he tends to in other places.
I saw the GPU voltage on this 4ifir thingy is absurdly low. If 1305mhz (displayed) with 820mV. Does it mean The GPU only runs at 1075mhz - 1095mhz in real world application? So what's the point then?
I tried benchmarking regular OC Suite at 1075mhz with your pack that runs at 1305mhz with a pretty similar voltage And it runs pretty much the same. Shit even 998mhz on the regular oc suite runs better compared with 4ifir. 1305mhz is supposedly runs with 980mV.
Horizon OS Undervolting is pretty much non-existent at this point.
I mean, just don't mislead everyone else and claiming about power efficiency and stuff. Im not against this crap, i just don't like the way you informed your own pack falsely.
I tried benchmarking regular OC Suite at 1075mhz with your pack that runs at 1305mhz with a pretty similar voltage And it runs pretty much the same. Shit even 998mhz on the regular oc suite runs better compared with 4ifir. 1305mhz is supposedly runs with 980mV.
Horizon OS Undervolting is pretty much non-existent at this point.
I mean, just don't mislead everyone else and claiming about power efficiency and stuff. Im not against this crap, i just don't like the way you informed your own pack falsely.
Moderation asked not to respond to trolling, so I will refrain from commenting.
Last edited by Foxi4,
If he is not lying on purpose, then I think I know what misled him (other than a lack of understanding of the principles of overclocking and the operation of computational pipelines). He tried to assemble the OC Suite himself, changing the maximum GPU frequency in a configuration with remote parameters "for modification by those who do not know how to program".
He lowered the maximum value for the GPU there to 11XX (he couldn’t lower it lower), and at the output I got a kip in which there is 13XXMHz, but they have a voltage of 11XX. And the performance is 11XX.
This prompted him to a bright idea that the performance decreases proportionally with the voltage, lol, and there can be no increase in 4IFIR. In 4IFIR-e, which tears to pieces OC Suite at the same frequencies in terms of performance.
By the way, yes, many users who are tired of looking into the development diary because of the Russian language, even to see comparative videos and screenshots, asked me to upload comparative tests for the GBA. Apparently it will be necessary to force the implementation of their request.
P.s. I don't take offense at all to those who think it's impossible. On the contrary, I take it as a compliment, it means that I managed to do something so incredible that it's hard to believe. Moreover, sooner or later you will be convinced of this personally, and will return to report your surprise to me.
Another thing is when someone, realizing that it works, tries to sabotage the process of distribution and development of my work in order to preserve the remnants of the reputation of the leader of their discord community, and a primitive utility for basic overclocking.
Not entirely related to 4IFIR but related to overclocking so why not.
I did these tests a while back and I might retry them with a few other things.
Main issue is that I only really recorded peak power, although the average was about 4-10% lower.
These are CPU only, with 1-3 thread load under HOS, with handheld 0% brightness, airplane mode on and joycons connected.
Power in Watts, (Erista)
Context:
The board TDP is 18W
CPU regulators can support about 2.295ghz before hitting their limits on Erista.
I did these tests a while back and I might retry them with a few other things.
Main issue is that I only really recorded peak power, although the average was about 4-10% lower.
These are CPU only, with 1-3 thread load under HOS, with handheld 0% brightness, airplane mode on and joycons connected.
Power in Watts, (Erista)
| 2091 | 11.45 | 10.58 | 8.84 |
| 1963 | 10.28 | 9.68 | 8.06 |
| 1887 | 9.88 | 9.01 | 7.33 |
| 1785 | 9.35 | 8.29 | 6.94 |
| 1683 | 8.78 | 8.01 | 6.77 |
| 1581 | 8.29 | 7.55 | 6.49 |
| 1428 | 7.57 | 6.77 | 6.15 |
| 1326 | 7.22 | 6.38 | 6.04 |
| 1224 | 6.96 | 6.15 | 5.88 |
| 1122 | 6.56 | 5.99 | 5.62 |
| 1020 | 6.28 | 5.82 | 5.48 |
| 918 | 6.14 | 5.54 | 5.09 |
| 816 | 5.96 | 5.31 | 4.98 |
| 714 | 5.78 | 5.2 | 4.71 |
| 612 | 5.56 | 4.98 | 4.53 |
| 510 | 5.29 | 4.7 | 4.36 |
| 408 | 4.98 | 4.59 | 4.31 |
Post automatically merged:
Context:
The board TDP is 18W
CPU regulators can support about 2.295ghz before hitting their limits on Erista.
- Status
what a shame, to have to go back to an OG size battery soon.