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

[xXFREDBERTXx]

I use my Phat L04 with a CXD2991GB 65nm RSX for that OC and my results are rock stable 650/900Mhz now ☝
Sadly 700Mhz Core is not stable it drops artifacts in games so yeah to less stock voltage sadly in these L04 Series…
But hey a nice free performance boost and why not take these now and GT6 benefit so far the most from that OC and it is now with constantly 60FPS playable

 

[LuanTeles]

About my 251XX, I finally experienced a crash caused by the VRAM. I overclocked it again,why not?

1050/1275 MHz

The system was freezing during the coldboot animation. Luckily, webMAN mod was already running, so I just lowered the temps to 65°C by setting the fan speed to 80%, and I was able to boot again XD

CELL games are working fine, but I can't test RSX games since 900 core and 950 core MHz is already very unstable here, let alone 1050 MHz. I think it might work if my PS3 could stay below 60°C, but since I've never changed the thermal paste and it already runs quite hot on the XMB, testing isn't possible at the moment. But at least I finally hit a clock speed where my 251XX crashed.

 

[Charles_n_town]

About my 251XX, I finally experienced a crash caused by the VRAM. I overclocked it again,why not?

1050/1275 MHz

The system was freezing during the coldboot animation. Luckily, webMAN mod was already running, so I just lowered the temps to 65°C by setting the fan speed to 80%, and I was able to boot again XD

CELL games are working fine, but I can't test RSX games since 900/950 MHz is already very unstable here, let alone 1050 MHz. I think it might work if my PS3 could stay below 60°C, but since I've never changed the thermal paste and it already runs quite hot on the XMB, testing isn't possible at the moment. But at least I finally hit a clock speed where my 251XX crashed.

that's a pretty insane overclock speed. good thing it didn't brick, but at least you pushed your system to its limits.

 

[Flirti]

I use my Phat L04 with a CXD2991GB 65nm RSX for that OC and my results are rock stable 650/900Mhz now ☝
Sadly 700Mhz Core is not stable it drops artifacts in games so yeah to less stock voltage sadly in these L04 Series…
But hey a nice free performance boost and why not take these now and GT6 benefit so far the most from that OC and it is now with constantly 60FPS playable

Another confirmation that on FAT with 65 RSX the memory takes 900 without any problems.
With disgusting core overclocking, 600-650.

 

[Flirti]

Hello again guys.
I found a game that will burn out your RSX.
This is Risen 3.
No other game loads my RSX like this game does.
Real death for FAT 90nm.

 

[Lazard]

Hello i need some help guys to identify some Syscon errors that iam not understanding does anyone know what 2020 and 2120 error codes on Syscon means with someone explaining them with simple words?

 

[richardjohn]

Hey guys, this is my very first post, I hope everyone is well. I've been following for quite sometime, I have a CHECH 2504B and currently running 4.91.2 OC 700/1000 and haven't experienced anything weird so far. I am currently testing gt6

what manufactured is it bro?

what manufactured is it bro?

what time is it?

 

[fpoker102]

If you have sound on XMB, you can try to blindly update your FW.

I don't think your VRAM died. In our tests and with the logs provided by Haxxxen, it seems like there isn't enough power supply for it. Try to provide us with a syscon log.

Personally, i think your console can be recovered as you can still hear that the XMB is working.




I'll make one today.

Update

@JoeSapecudo's system runs 900/1000 better than mine, but on Crysis, it randomly crashes after some minutes (It's rare, but it is happening). Therefore, we should assume that the latest stable clocks for 2511X are 850/1000. Although it works fine for most of the game, Crysis and GOW are experiencing these issues, so it's better to keep everything 100% stable.

I don't think it's worth pushing to 950 core, at least on our consoles. It may become stable in the future on Super Slim models, but I do believe we can still push the memory further. We haven't observed any artifacts in our systems, and overclocking the memory is providing a greater FPS boost than overclocking the core.

I'll create an updated table soon.

Important notes:

Keep in mind to observe the following:

If your console is reaching the memory clock limit, it will start artifacting. If you notice this happening, DO NOT push it further, as it will 100% brick your PS3.

If your console is reaching the core clock limit, it will crash in RSX-extensive games such as Crysis and GOWs. We haven't had any consoles bricked due to the core.

And, of course, test only incrementing by 50MHz each time.


One question:

If you've tested, let's say, 700/1000 and it worked, could we assume that the 1000MHz memory will work in all core speeds, right? As I could see, if a memory clock is supported, it will work just fine at every core speed, so there is no need to start from scratch in the next core increment.

what is the model number of your ps3 with date Code plz.

 

[fpoker102]

So are 2501B and 2501A Slims safe with OC when it comes to Nec tokens?

Brother has anyone checked 1025 and 1050 mhz on PS3 GGB model?

How is it compared to a 900+/900+? I'm curious if it'd be more beneficial to use a modest core with a higher memory than trying to push both evenly.

Is ps3 CXD5300GGB will support 1100 VRAM?

2511A seem to be the highest overclockers here. But yeah try in 50 increments. Do not do higher than 950 on the memory. Unless you want to risk a brick.

what is the date code of 2511A

Wait, what? Do you now have 8 PS3s? That's insane!

BTW, I remember another guy here using 950 core, am I right? It's still a shame that I mine system can only go up to 850Mhz fully stable. I can play for a few minutes at 900MHz before experiencing a random crash.

I just installed the 900 core again to see if the temperature affects when the system crashes.

what is the model number & date code of your ps3 brother luan?

 

[Starship2]

Edit:
dont misunderstand that 2000models have 40nm's, they dont. sorry for my misformulation here. I just wanted to refer to Flirti's theory

I can confirm. These 40nm's can pack a punch.
My 2004B model took 750/925 easily on first try, whereas my fat C04 struggled with 650/800. on 650 Core my c04 has a little artefacting in some demanding games (TLOU). Ive written fantables to both consoles that keep rsx under 60C. I will push my slim to the limits at the weekend, my guess is 750/950-975 should barely work without playing with VID. For stock voltages this is a pretty good result for such an environment (such as tiny power delivery at the chips (compared to pc), much heat in a small case, inferior cooling)

 

[shnyaps]

I can confirm. These 40nm's can pack a punch. My 2004B model took 750/925 easily on first try, whereas my fat C04 struggled with 650/800. on 650 Core my c04 has a little artefacting in some demanding games (TLOU). Ive written fantables to both consoles that keep rsx under 60C. I will push my slim to the limits at the weekend, my guess is 750/950-975 should barely work without playing with VID. For stock voltages this is a pretty good result for such an environment (such as tiny power delivery at the chips (compared to pc), much heat in a small case, inferior cooling)

2000 has 65nm

 

[Starship2]

2000 has 65nm

Right! my context was a little bit messed up, i didnt mention that the 2000 models have a 40nm. just saying that these are sick and confirming in the next sentence the answer from Flirti, that 65nm can easily take 900 on Memory. Sorry for my misformulation, i shouldve read it twice before posting ...

Another confirmation that on FAT with 65 RSX the memory takes 900 without any problems.
With disgusting core overclocking, 600-650.

So it looks like 65nm in Slim models can take more as we see 65nm in fat models just support poor core frequencies. Maybe some others can also post their results of fat 65 and slim 65 models as a proof of concept. the question is why can slims handle higher core frequencies better? is that advantage chip related or do slim mainboards provide better/stable current for rsx? pretty interesting for building an overkill system with record overclocking if we can point out whats the most stable/powerful board/gpu combination.
Results from LuanTeles are insane and its crazy to see in which regions we are able to push the rsx. we are talking about a 100% gain in clockspeeds.
makes me thinking about what cell could do... too bad its not possible yet.

 

[Charles_n_town]

So it looks like 65nm in Slim models can take more as we see 65nm in fat models just support poor core frequencies. Maybe some others can also post their results of fat 65 and slim 65 models as a proof of concept. the question is why can slims handle higher core frequencies better? is that advantage chip related or do slim mainboards provide better/stable current for rsx? pretty interesting for building an overkill system with record overclocking if we can point out whats the most stable/powerful board/gpu combination.
Results from LuanTeles are insane and its crazy to see in which regions we are able to push the rsx. we are talking about a 100% gain in clockspeeds.
makes me thinking about what cell could do... too bad its not possible yet.

Afaik, cell is impossible without modifying Syscon. The only reason RSX works is that there's a way to override its Syscon values in lv1.

 

[Starship2]

Afaik, cell is impossible without modifying Syscon. The only reason RSX works is that there's a way to override its Syscon values in lv1.

of course we have to do it via config ring in syscon. thats what we have already done with unlocking 8th SPE. i bet its actually not possible because we dont know anything about its multipliers yet. we only know REF_CLK but not how its multiplied and how its synced with XDRam. it will be hard to find out because of its complex architecture, but im pretty sure one day we will find out how everything relates.

 

[Charles_n_town]

of course we have to do it via config ring in syscon. thats what we have already done with unlocking 8th SPE. i bet its actually not possible because we dont know anything about its multipliers yet. we only know REF_CLK but not how its multiplied and how its synced with XDRam. it will be hard to find out because of its complex architecture, but im pretty sure one day we will find out how everything relates.

what I mean is how would you override the values of the cell in a firmware update? maybe sony was going to try optimizing the rsx, and that's the reason its values are present? I don't know much about hardware, but I think I remember reading that sony was originally going to have the cell do the processes of a gpu as well. however, they couldn't get it to work, so they used a cheap rsx chip to do it instead, so perhaps they were going to boost performance somehow and never did?

 

[Starship2]

what I mean is how would you override the values of the cell in a firmware update? maybe sony was going to try optimizing the rsx, and that's the reason its values are present? I don't know much about hardware, but I think I remember reading that sony was originally going to have the cell do the processes of a gpu as well. however, they couldn't get it to work, so they used a cheap rsx chip to do it instead, so perhaps they were going to boost performance somehow and never did?

not sure about that. only thing we can tell is that only 4 firmwareupdates contained syscon patches. there are no clues so far that sony planned to boost performance later on. id rather say they modified fan curves and reduced clocks a litte as it turned out rsx being defective.
what do you mean by how would we override the values? since we are able to connect via uart we can override for ex. config ring values. we just need to know the address. we dont have to do it over a firmware update, even if we can. the question is not how we physically do it, but how the frequencies are built and how it correlates with Rambus. this is unknown battlegrounds at this moment.

 

[Charles_n_town]

not sure about that. only thing we can tell is that only 3 firmwareupdates contained syscon patches. there are no clues so far that sony planned to boost performance later. id rather say they modified fan curves and reduced clocks a litte as it turned out rsx being defective.
what do you mean by how would we override the values? since we are able to connect via uart we can override for ex. config ring values. we just need to know the address. we dont have to do it over a firmware update, even if we can. the question is not how we physically do it, but how the frequencies are built and how it correlates with Rambus. this is unknown battlegrounds at this moment.

what I mean is that the only reason a lot of people are interested in this is that it just requires installing a firmware update. if you have to solder or even open the system, most people wouldn't bother, especially with dubious results that may or may not cause a brick.

 

[Starship2]

what I mean is that the only reason a lot of people are interested in this is that it just requires installing a firmware update. if you have to solder or even open the system, most people wouldn't bother, especially with dubious results that may or may not cause a brick.

alright, im with you. however, for adjusting any clock speeds you always have to change values in syscon. thats what the firmware update does. there are just 2 ways to do, either we use uart and write it direclty or we do a syscon patch implemented in firmware. last one is more attractive for most people, thats right. but for us enthusiasts it doesnt matter. thats not the problem we are pointing at. we have to increase our knowledge of how cell operates and how it multiplies reference clock in relation to every single SPE.
lets hope for the best that some day we gain the possibility to do so. im seeing 4Ghz there
also we need to find a way to make use of the 8th spe

 

[Charles_n_town]

alright, im with you. however, for adjusting any clock speeds you always have to change values in syscon. thats what the firmware update does. there are just 2 ways to do, either we use uart and write it direclty or we do a syscon patch implemented in firmware. last one is more attractive for most people, thats right. but for us enthusiasts it doesnt matter. thats not the problem we are pointing at. we have to increase our knowledge of how cell operates and how it multiplies reference clock in relation to every single SPE.
lets hope for the best that some day we gain the possibility to do so. im seeing 4Ghz there
also we need to find a way to make use of the 8th spe

yes, that was something I was going to mention. the 8th spe seems to do nothing, and it's disabled by default, so even enabling it, it has no effect on games or anything. I'm not sure of the reasoning behind that. the point being that the ps3 is capable of so much more than sony allowed it to do, but maybe cooling was a factor in their decision? I don't have a cfw ps3, but from what I understand, stability with oc often relies on cooling the system to a certain level. maybe they didn't want a loud fan, so the average, legit user didn't think something was wrong with their system. I talked with al azi about this in regards to ps4 cooling, and we both think it's fine as is. it's often people coming from the ps3, knowing how hot the system actually gets, who worry.

 

[Starship2]

yes, that was something I was going to mention. the 8th spe seems to do nothing, and it's disabled by default, so even enabling it, it has no effect on games or anything. I'm not sure of the reasoning behind that. the point being that the ps3 is capable of so much more than sony allowed it to do, but maybe cooling was a factor in their decision? I don't have a cfw ps3, but from what I understand, stability with oc often relies on cooling the system to a certain level. maybe they didn't want a loud fan, so the average, legit user didn't think something was wrong with their system. I talked with al azi about this in regards to ps4 cooling, and we both think it's fine as is. it's often people coming from the ps3, knowing how hot the system actually gets, who worry.

No! Cell BE is not only used in PS3, it was also used in some Servers. To increase yield rate and make it more profitable they put cells with 7 guaranteed functional spe's in the ps3 because in first instance it wasnt necessary to have 8 spe's in PS3. that way they didnt have to put expensive testing procedure for 8 working spe's to the ps3 cpu's and they could make use of cpu's with 1 defective spe.
our problem why we cant use 8th spe is because hypervisor doesnt know how to handle a further spe. hypervisor doesnt even know that there is another spe. and we are not able to modify the hypervisor, probably never will. one thing using that spe could have been adding another task, for example running a music player or browser in background while gaming. or if we install otherOS, we could make use of it. But its a bummer if we think that PS3 games could make profit from enabling that spe, thats not gonna happen...
good point, of course it wasnt even in sonys interest to let it run with all spe's for given reasons. the cooling system wouldnt be able to handle that at acceptable fan speeds. would have made all things more worse