PS3 Project RSX Boost: Overclock your Retail PS3 RSX Speeds (ps3 cfw only)

[raidriar]

Tested on my B01, 600/750 version. I am delidded and repasted, 19 blade fan, 226 PSU. There has been only a slight increase in temperature, nothing that dynamic fan control cannot handle. Definite bump in FPS. Tested Jak 2 and Red Faction Guerilla. Unfortunately TLOU is 30 locked, no gains there.

Has anybody tried to push clocks to 625/650/700 on the core?

 

[Checkity_CheckYoSelf]

Tested on my B01, 600/750 version. I am delidded and repasted, 19 blade fan, 226 PSU. There has been only a slight increase in temperature, nothing that dynamic fan control cannot handle. Definite bump in FPS. Tested Jak 2 and Red Faction Guerilla. Unfortunately TLOU is 30 locked, no gains there.

Has anybody tried to push clocks to 625/650/700 on the core?

Any improvements on ps2 gameplay by chance?

I've only tested on my 2501a slim model (4.84 Rebug Dex version) and i'm seeing about the same as you. Some games locked, some games pretty noticeable increase in framerate. No major temp changes for mine either. Fan has to work a touch harder but nothing substantial.

 

[bguerville]

If the idea is to collate reliable data in order to assert the impact of the various overclock (and underclock) settings then it would be better for everyone to use the same benchmarking setup and provide precise temps/data.

To be complete, a report for a game title test should ideally include
1. the exact overclock settings for both RSX core and memory, the console model's SKU reference, CFW type and version (and hardware mod if any).
2. the room temperature
3. the manual fan setting percentage (not dynamic fan control) used for the duration of the test
4. the temps when the console has been idle for 10mn
5. the temps after 15mn of game play with manual fan control without overclock
6. the temps after 15mn of game play with manual fan control with overclock
7. the game frame rate average without overclock
8. the game frame rate average with overclock

 

[sandungas]

If the idea is to collate reliable data in order to assert the impact of the various overclock (and underclock) settings then it would be better for everyone to use the same benchmarking setup and provide precise temps/data.

To be complete, a report for a game title test should ideally include
1. the exact overclock settings for both RSX core and memory and the console model's SKU reference.
2. the room temperature
3. the manual fan setting percentage (not dynamic fan control) used for the duration of the test
4. the temps when the console has been idle for 10mn
5. the temps after 15mn of game play with manual fan control without overclock
6. the temps after 15mn of game play with manual fan control with overclock
7. the game frame rate average without overclock
8. the game frame rate average with overclock

Some games locked

And most specially, we need a list of that games that are having problems, the first time someone mentioned it i thought it was a bit weird, but later it was confirmed by other user, and now you are confirming it even more... so... somone figures why is happening that ?

I was wondering if is caused by some game functions that are dependant of a specific access speed... lets say the game have a function that is trying to retrieve a texture for a specific lapse of 100-200 milisecons
If the texture is not retrieved in 200 miliseconds the function breaks (as expected), but if the texture is retrieved in lower than 100 miliseconds the function breaks too (this is not so expected)

You know... is the same problem that happens in emulators that are trying to emulate the disc drive but are not doing any adjustments of speed access
Lets say... if the game is an ISO stored in the internal hdd the emulator is loading it faster than the original disc inserted in the drive... so a few games are broken
In this case... overclocking the RSX RAM should give faster access too and we could fall into the same problem, but in PS3 games

-----
Anyway... to try to figure this the first thing we need to know is an accurate list of the games that are having this problem

Spoiler

And after identifying a few of them it would be needed to do some tests in them by using different clock speeds to find the "sweetspot" where the overclock becomes stable

 

[bguerville]

And most specially, we need a list of that games that are having problems, the first time someone mentioned it i thought it was a bit weird, but later it was confirmed by other user, and now you are confirming it even more... so... somone figures why is happening that ?

I was wondering if is caused by some game functions that are dependant of a specific access speed... lets say the game have a function that is trying to retrieve a texture for a specific lapse of 100-200 milisecons
If the texture is not retrieved in 200 miliseconds the function breaks (as expected), but if the texture is retrieved in lower than 100 miliseconds the function breaks too (this is not so expected)

You know... is the same problem that happens in emulators that are trying to emulate the disc drive but are not doing any adjustments of speed access
Lets say... if the game is an ISO stored in the internal hdd the emulator is loading it faster than the original disc inserted in the drive... so a few games are broken
In this case... overclocking the RSX RAM should give faster access too and we could fall into the same problem, but in PS3 games

-----
Anyway... to try to figure this the first thing we need to know is an accurate list of the games that are having this problem

Spoiler

And after identifying a few of them it would be needed to do some tests in them by using different clock speeds to find the "sweetspot" where the overclock becomes stable

Yes definitely..
So far, the reported data has not proven all that useful tbph, often too vague, with too little data or no point of reference to compare to etc...

 

[littlebalup]

If the idea is to collate reliable data in order to assert the impact of the various overclock (and underclock) settings then it would be better for everyone to use the same benchmarking setup and provide precise temps/data.

To be complete, a report for a game title test should ideally include
1. the exact overclock settings for both RSX core and memory and the console model's SKU reference.
2. the room temperature
3. the manual fan setting percentage (not dynamic fan control) used for the duration of the test
4. the temps when the console has been idle for 10mn
5. the temps after 15mn of game play with manual fan control without overclock
6. the temps after 15mn of game play with manual fan control with overclock
7. the game frame rate average without overclock
8. the game frame rate average with overclock

I would add CFW version/setup, PS3 model (and hardware mods if any), an idle delay between 5 and 6 to cool down the beast.

 

[Checkity_CheckYoSelf]

Yes definitely..
So far, the reported data has not proven all that useful tbph, often too vague, with too little data or no point of reference to compare to etc...

It might be a good idea to put together some sort of submission process and a means of recording this info in that case if we're going to try to set a standard of sorts. I was trying to make an excel sheet when I started testing mine, but the games I tested weren't seeing any dramatic improvements (at least by the FPS counters measures) so I didn't feel it worthy to report.

 

[bguerville]

It might be a good idea to put together some sort of submission process and a means of recording this info in that case if we're going to try to set a standard of sorts. I was trying to make an excel sheet when I started testing mine, but the games I tested weren't seeing any dramatic improvements (at least by the FPS counters measures) so I didn't feel it worthy to report.

I totally agree with you there, ideally it should have been added to the OP ages ago.
Regarding the frame rate, I think what we would really need is an average rate over a set period of time like say 10 or 15mn rather than in "real time".

I would add CFW version/setup, PS3 model (and hardware mods if any)

I edited my post to add that information as well [emoji6]

an idle delay between 5 and 6 to cool down the beast.

Yes, in theory, going back to idle temps between steps 5 and 6 is the way to go but I didn't mention it because I figured that if you start with standard settings, play 15mn then proceed to apply the overclock, by the time you are ready to begin playing another 15mn, the temps will have gotten back down close to idle temps, am I wrong in that assumption?

 

[bguerville]

@aldostools
You added the FPS counter into webman-mod, didn't you?
If so, could you by any chance add a Web command that would capture the minimum, maximum and average values for both temps and FPS over a set period of time?
The command would take 2 arguments, the duration in seconds and the target log file path?
That way people could trigger the capture from a remote browser or the custom pad combo, when the target path is omitted the command could maybe just return the data to be displayed in the remote browser (after displaying a please wait message for the duration of the operation).
Something along those lines...
Of course, the fan would need to be set to manual or possibly syscon for the min/max/avg temps measurements to be worth getting.

It's just an idea to help with testing though, I haven't really checked the source so I dunno how easy it would be to make this happen, no pressure if you don't have time or cannot be bothered.. [emoji6] I would totally understand, I have been busy and unable to write a single line of code for the past 3 or 4 weeks..

 

[raidriar]

There needs to be a "warm up" period too or else the temperature readings will be wrong. Has anybody figured out a way to manipulate clock from XMB/reset lv1? Would suck to have to keep flashing firmwares to modify the clocks. In addition, are voltage controls exposed? It would be really cool to try and drop the voltage on cell and RSX especially for 90nm models.


As for PS2 improvement, I would think none, at least in the case of A/B models since it relies on actual PS2 hardware. Maybe netemu can see some improvements

 

[aldostools]

@aldostools
You added the FPS counter into webman-mod, didn't you?
If so, could you by any chance add a Web command that would capture the minimum, maximum and average values for both temps and FPS over a set period of time?
The command would take 2 arguments, the duration in seconds and the target log file path?
That way people could trigger the capture from a remote browser or the custom pad combo, when the target path is omitted the command could maybe just return the data to be displayed in the remote browser (after displaying a please wait message for the duration of the operation).
Something along those lines...
Of course, the fan would need to be set to manual or possibly syscon for the min/max/avg temps measurements to be worth getting.

It's just an idea to help with testing though, I haven't really checked the source so I dunno how easy it would be to make this happen, no pressure if you don't have time or cannot be bothered.. [emoji6] I would totally understand, I have been busy and unable to write a single line of code for the past 3 or 4 weeks..

Yes, I added the FPS plugin 2.2. I skipped the version 2.3 but it can be used replacing the sprx in /dev_hdd0/tmp/wm_res with the binaries from TheRouLetteBoi's github

webMAN MOD basically adds the shortcut L3+R3 to launch the plugin already built by TheRouLetteBoi.
I didn't touch the FPS plugin code. To capture the FPS, that plugin must be modified.
I don't have the development environment required to compile the changes.

The CPU/RSX is already captured checking the option CPU/RSX/FAN Chart in /setup.ps3

The captured CPU/RSX is shown in a graphical html chart in /dev_hdd0/xmlhost/game_plugin/cpursx.html

 

[bguerville]

Yes, I added the FPS plugin 2.2. I skipped the version 2.3 but it can be used replacing the sprx in /dev_hdd0/tmp/wm_res with the binaries from TheRouLetteBoi's github

webMAN MOD basically adds the shortcut L3+R3 to launch the plugin already built by TheRouLetteBoi.
I didn't touch the FPS plugin code. To capture the FPS, that plugin must be modified.
I don't have the development environment required to compile the changes.

The CPU/RSX is already captured checking the option CPU/RSX/FAN Chart in /setup.ps3
View attachment 37019

The captured CPU/RSX is shown in a graphical html chart in /dev_hdd0/xmlhost/game_plugin/cpursx.html

Gotcha, no problem, I didn't know whether you integrated the code into wMM or just relied on loading the sprx file without messing with it.
Well, if someone wishes to help, the plugin source is open..

 

[PostalDude__]

If the idea is to collate reliable data in order to assert the impact of the various overclock (and underclock) settings then it would be better for everyone to use the same benchmarking setup and provide precise temps/data.

To be complete, a report for a game title test should ideally include
1. the exact overclock settings for both RSX core and memory, the console model's SKU reference, CFW type and version (and hardware mod if any).
2. the room temperature
3. the manual fan setting percentage (not dynamic fan control) used for the duration of the test
4. the temps when the console has been idle for 10mn
5. the temps after 15mn of game play with manual fan control without overclock
6. the temps after 15mn of game play with manual fan control with overclock
7. the game frame rate average without overclock
8. the game frame rate average with overclock

Might be an idea to measure power consumption (if possible) as power consumption is directly tied to extra heat made by the console

 

[bguerville]

Might be an idea to measure power consumption (if possible) as power consumption is directly tied to extra heat made by the console

Sure, let's put it this way, the more data the better! Those who can get precise power consumption measurements should of course add that too.
Someone needs to make a proper template so testers can use it for their reports, the template can always be improved as things move along.

It would also be interesting to double check that there aren't any errors at lv1, lv2, rsx & syscon levels popping up at any point, game crashes or not. It might be good to check the various log sources whenever possible and report any anomaly.

 

[PostalDude__]

Sure, let's put it this way, the more data the better! Those who can get precise power consumption measurements should of course add that too.
Someone needs to make a proper template so testers can use it for their reports, the template can always be improved as things move along.

Indeed, the hardest thing with all of this testing though, we are somehow going to need a static benchmark that is the same every time, like PC benchmarks, other wise results may not be super accurate, Idk if anyone has made anything similar to this for the ps3 as of yet?

 

[bguerville]

Indeed, the hardest thing with all of this testing though, we are somehow going to need a static benchmark that is the same every time, like PC benchmarks, other wise results may not be super accurate, Idk if anyone has made anything similar to this for the ps3 as of yet?

Ideally yes you would have a series of standard benchmarks but personally I dunno of any such thing being available to the public on the PS3. Maybe someone does?
OC results seem to vary greatly from one game to the next, some testers reported insignificant FPS and temp changes in some games while in others they reported visible FPS changes and temps rise of up to 10C.
Of course, all this testing being non standard with a variety of setups and we cannot draw any conclusions from it.

 

[PostalDude__]

Ideally yes you would have a series of standard benchmarks but personally I dunno of any such thing being available to the public on the PS3. Maybe someone does?
OC results seem to vary greatly from one game to the next, some testers reported insignificant FPS and temp changes in some games while in others they reported visible FPS changes and temps rise of up to 10C.
Of course, all this testing being non standard with a variety of setups and we cannot draw any conclusions from it.

the next best thing might be to share around save games, of games that have an unlocked FPS/push the system to its limit, a few of these come to mind, like GOW 3, although its not uncapped it rarely reaches its 60FPS cap and it also cooks the RSX within an inch of its life.

 

[bguerville]

the next best thing might be to share around save games, of games that have an unlocked FPS/push the system to its limit, a few of these come to mind, like GOW 3, although its not uncapped it rarely reaches its 60FPS cap and it also cooks the RSX within an inch of its life.

That's not a bad idea actually, when testing on Rebug REX 4.84.2, the fake save data owner makes this trivial too.

We would need to make a list of the games that are the most demanding and create appropriate saves for test runs, we know them all really.

The other problem being that like sandy pointed out, we also need a list of games with which testers had problems, those may not be the most demanding or impact temps greatly.. in other words as many games as possible need to be tested with a range of overclock settings, independently of the data collation about temps.

 

[raidriar]

GT6, GOW3, GTA4, GTA5 are all great test games. Uncharted 1-3, TLOU for maintaining 30FPS as much as possible

 

[PostalDude__]

GT6, GOW3, GTA4, GTA5 are all great test games. Uncharted 1-3, TLOU for maintaining 30FPS as much as possible

the naughty dog games can be "improved" as its possible to disable the 30FPS frame rate cap via the dev menu that can easily be applied using an EBOOT, which would turn them into good tests.