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

No i meant core. The highest core is 900 stable. And I got fantastic news. One of my January's was crashing at 900 core...
I was specifically asking about the 2501A. I have recorded 3x 2501B models from you already. The A is the only one I couldn't find the core/VRAM frequencies of. Now you say that was core. But didn't say what VRAM is. Can you see why this is frustrating?

It's actually quite hard to find a post from you or anyone else that shows the LV1 string of the console you're currently talking about.
Ok yeah. Those won't ever overclock high. The very process of putting a 40nm RSX on a fat is stressful to the chip. It losses overclockablity due to the heat and stress. But your OC at 700/850 is still very good regardless. That's more than you could ask for from a fat PS3.
My E01 Franky with a NOS 40nm CXD5301 does 800/975 in what I'm calling Murder Crysis, which is Crysis3 with hacked Cvars pushing everything to the max.

I'm using crysis 3 because it's late in the PS3 lifecycle and uses the CELL SPUs to the full. The PS3 hardware if fully leveraged when the hacked config and the specific test location I used has every element on screen. Shading, lighting effects, water physics, particle effects, explosions, NPCs and AI behaviors, grass and foliage animations stretching off into the distance. A long alley way for loading textures off in the distance to load into VRAM, leveraging the maxed out draw distance. Plenty of 3D models and high resolutions assets. Action occurring in the scene keeping everything in motion.

This one spot is perfect for leveraging everything the game can do to push the PS3 HW.
Stand here.jpg Look here.jpg
At stock 500/650 it'll only do 12FPS and at 800/975 it does about 19FPS, which is actually playable, surprisingly! This level of detail is absurd and I didn't think it would even come close to approximating stock performance in this game at normal settings. This game without my Murder Config struggles to maintain 20FPS.

The OverPowerd heatsink keeps the 40nm nice and cool. I think actually makes the OC's more stable. I set a static 35% Fan and recorded temps and FPS to characterize it's potential. This is fairly accurate and you can use it to predict the relative performance of any OC. In my case, 800/975 took 12FPS to 19FPS. A 37% increase in true performance. And I know this because the games settings can make full use any performance increase. There's nothing unused in this game/config/scene.
upload_2024-5-26_7-33-48.png

CoreVRAMFPSCPUGPUAmbientGPUΔTFan%
45090012.25625425.32935
45095012.45635525.33035
50065012.06635425.12935
55070013.226254252935
60075014.31635624.83135
65080015.02635624.63135
70080015.8635724.63235
75080016.65645825.13335
80080017.296358253335
80085017.75635824.73335
80090018.15645925.63335
80092518.42645925.33435
80095018.75635924.83435
80097518.98635924.63435
About temps. I measured the ambient room temperature of using a thermocouple taped to the intake of the PS3. Not touching the plastic, where it could be transferring heat from the console, but suspended in the air in front of the intake. This is the true temperature of the air entering the console. If you take the difference in ambient and GPU you get the change in temperature, or Delta T. That is an even more accurate representation of the actual GPU temperatures between runs. This helps to see if there was a significant difference between them.

I wanted to see if down clocking the Core 50MHz to 450MHz, but overclocking the VRAM to as high as was stable (950) would match performance of stock clocks but reduce overall temps. While I was able to match and exceed FPS, the temps were not significantly different. Perhaps even slighly higher. However, I believe 450/850 would match 500/650 in FPS for this game. That might reduce temps. It should be noted that this approach is going to sacrifice some Core performance and tasks that are shader/processing heavy, such as lighting/shaddows, water simulation, explosions, particle effects...etc will ne negativly affected. Whereas high resolutions texture loading and streaming data to/from the Core will improve. These could balance out and be unniticble to the end user. And if that comes with a reduction in temps, especially for a 90nm RSX with defective underfill where small reductions in temps are a net positive, then this may not be such a bad idea.

And this is where it gets really REALLY weird. @Nascar1243 found that 300/900 and 300/750 punched a hole in spacetime and initiated Core Fusion...
Core Fusion.jpg
He jumped from about 10FPS at 300/800 to 18FPS at 300/900! And 15FPS at 300/750. And they accomplished this physics defying, mind bending, wacko non-sense with 7C cooler temps. IT'S COMPLETELY BONKERS!

Now temper your excitement. XMB ran sluggish and you wouldn't want to actually use the console like this. It's just a strange novelty and a curiosity from an architectural standpoint. As for why the actual eff this happened we still don't know.
  • Hypothesis 1 (refuted): We thought it could be a quark in the way FPS calculates, but he said it actually does feel smoother in game. Like it really was running at that FPS, not just giving a false reading.
  • Hypothesis 2 (strained): I speculate it could be because 900 VRAM is a perfect 3x integer of 300 CORE and that there might be an architectural explanation for why 3 VRAM cycles per Core cycle might synchronize and improve efficiency or bandwidth of the shader piplines. Allowing for less idle time and improved latency for coordination between VRAM and core. And if that were the case, then the same might be true of any 2x or 3x integer. But I was not able to replicate it using a 2x 450/900. That was the highest 2x integer I can achieve. 500/1000 wouldn't be stable for me. Other integers may reveal more context if other were to try them, but in our tests it has only occurred at 300/900 and 350/750. @nascar was able to replicate it on another console, so this isn't a fluke. It appears to only work with lower cores. HOWEVER, you'll notice 300/750 isn't an integer. So why would it work? That's a chink in the hypothesis.
  • Hypothesis 3: It could be a quark in the way this game in particular uses CORE/VRAM and the available bandwidth. Or the settings I've enabled in the custom config.
  • Hypothesis 4: It's possible that something broke at these low of cores. Like some visual effect or processor heavy task that isn't obvious while looking at the screen. If it literally cannot run and isn't, then it won't eat up compute cycles and the FPS would jump up. It doesn't cause the system to freeze or anything, but then we don't want to loose the save by running around, so maybe we would find an issue if we attempted to progress through the game. Like loading a new area.
Anyway, that was weird. If you have any ideas as to why that could happen, let me know. Maybe a quark of the Curie architecture?
 
Last edited:
Probably a 700MHz one.

@RIP-Felix

This 2504A can do 850 on core if cooled enough, personally I can't consider this stable. What are your thoughts on this ? Can we consider a heat-dependent OC stable ?
At 70°C, after ~30min on Crysis (850MHz on core) : freeze.

My last fully stable OC on it is 800/950 (I didn't tried 975 on memory, I'm scared to approach 1000MHz.)

Where did you pull up the code to get the rsx revision? I tried dumping my syscon error log to see if I can find it to provide more info to RIP-Felix but I did not find it.
 
I added it myself, it's not in any dump.
You can find this info on the IHS.
View attachment 43213

im shooting in the dark for this but im seeing if the seller is willing to send a picture of the donor's IHS. Hopefully they comply if they do have it.

Edit: yea they said they don't keep up with rex revisions but they do know what the 90nm rsx revision is which isn't really helpful. It was worth a shot
 
Last edited:
And that's why I only buy January 2501Bs. Every single one of them of my 4 have been able to do 850 easy. One of them doing 900 stable fully. So that's the highest the silicon lottery can get you. 950 doesn't seem to be remotely possible.

Edit: so been testing for an hour now. 900/975 on the one that does 850/975 stable without temp compensation. So I put the fan to 80%. Yes very loud. It was hugging 55c and 56c. It finally settled at 56c. Zero crashing or artifacts. 56c and below is the magic number to keep it stable. 57 and 58c. Might. But I'll have to turn down the fan speed and check. So another hour of testing it is.

Edit 2: ok so 57c and above it doesn't like it. 55 and 56 is the highest It can do for a whole hour no issue. Once I turned down the fan speed to 70 from 80 it went up to 57, and after a few minutes on 57, crash. But 56c for a whole hour was good. So 57 and up is a nope. So this is good for those that love doing delids and improve cooling in every way. I don't delid, and never have. But after I get enough January's. I'll delid this one and try to optimize cooling as much as possible to try and reduce temps the best I can. This is great data though. That extra 50 extra MHz that would usually cause crashes, stabilized with cooler temps. But that's only if you don't artifact at all, but only crash. Once I avoid the crashes due to temp, there were zero artifacting and instability. Man I wrote a Bible.
Hi mitsu i had a question does slim 2504B can handle 800/900? Just asking because iam thinking to buy one
 
All overclocking is possible in theory, just have to work out how everything is linked which is the tricky part you don't want to be doing something like overclocking the bus for the storage controller at the same time for instance, that would be really, really bad. You'd corrupt storage devices left and right.

You can't even make an estimated prediction on what kind of performance uplift there would be as I don't think XDR efficiency is easily available information. I would try and put some VERY rough numbers together making some general assumptions but I'm way too tired for that right now. Suffice to say a 15-30% bandwidth improvement variable factors considered wouldn't be out of the realms of possibility depending how high the OC is. I'll try to work some numbers out when I'm not knackered but if we assume 25.6GB\s is the theoretical maximum with a real-world maximum being 21GB\s say then you could say operational efficiency is 82% on average. Taking poor coding in mind though it could be as low as about 60% in some cases which would be about 15.3GB\s.
@Tanzu15 it's on my to-do list but first I need to figure out why MFW Builder throws the error below at me when I click RSX Overclocking and try to build a CFW. Googling is not helping. Decent amount of others getting this error though. I don't like building MFWs without knowing what all potential errors are it flies against my very reputation in the PC arena and my nature of not getting things wrong.
move mfw to path without spaces
 
Can you please make 800/900 evilnat v9 beta
Done : https://www.mediafire.com/file/03tcxym0x747rzk/800_900_BETA9_PEX.PUP_4.91.CEX-PS3UPDAT.PUP/file
Hi mitsu i had a question does slim 2504B can handle 800/900? Just asking because iam thinking to buy one
Yes, but not all. If you have to deal with date codes, then search for a "0D" or "1A" (1A is still CFW-compatible IIRC).
"0D" = October / November / December 2010
"1A" = January / February / March 2011 (After January, I'm really not sure about the CFW compatibility.)

These are the dates to search and the most susceptible to support high OC.
January's models are the best you can buy, but good luck finding one (especially if you have to deal with date codes ^^').

Unrelated edit :

I tried 975MHz VRAM and it's perfectly stable, no artifacts nor crash/freeze !
(I've updated the SYSCON log posted here).
So 800/975 is my highest stable OC on this 2504A (0D - CXD5300CGB) ! :D
IMG_20240527_010316.jpg
 
Last edited:
All my 2501As can only do 700 core and 950 or 975 on the memory. Depends on the memory lottery. Any PS3 from November and up, should be able to do 975 on memory. Just my data I can gathered. None of my 2501As can do even 750 stable at all. Yes 700 core fully stable only. @RIP-Felix


Done : https://www.mediafire.com/file/03tcxym0x747rzk/800_900_BETA9_PEX.PUP_4.91.CEX-PS3UPDAT.PUP/file

Yes, but not all. If you have to deal with date codes, then search for a "0D" or "1A" (1A is still CFW-compatible IIRC).
"0D" = October / November / December 2010
"1A" = January / February / March 2011 (After January, I'm really not sure about the CFW compatibility.)

These are the dates to search and the most susceptible to support high OC.
January's models are the best you can buy, but good luck finding one (especially if you have to deal with date codes ^^').

Unrelated edit :

I tried 975MHz VRAM and it's perfectly stable, no artifacts nor crash/freeze !
(I've updated the SYSCON log posted here).
So 800/975 is my highest stable OC on this 2504A (0D - CXD5300CGB) ! :D
View attachment 43218

yeah the consoles that do 800 or 850 aka December -January or February or March as reported by some that 3.56 minimum can do cfw, can do 975 memory stable. All 4 of my January 2501Bs do 975 fully stable. 1000 having issues. Only tested 1000 on one. And the issue was bad. So yeah. Safest highest with chance of fast recovery is 975. 1000 and up it's GG for the console.

.
 
Last edited by a moderator:
All my 2501As can only do 700 core and 950 or 975 on the memory. Depends on the memory lottery. Any PS3 from November and up, should be able to do 975 on memory. Just my data I can gathered. None of my 2501As can do even 750 stable at all. Yes 700 core fully stable only. @RIP-Felix

How much of a difference in fps between 700 core and 800? I have one of my 2501a running 700 - 850 and it runs great. Worth trying 750?

yeah the consoles that do 800 or 850 aka December -January or February or March as reported by some that 3.56 minimum can do cfw, can do 975 memory stable. All 4 of my January 2501Bs do 975 fully stable. 1000 having issues. Only tested 1000 on one. And the issue was bad. So yeah. Safest highest with chance of fast recovery is 975. 1000 and up it's GG for the console.
 
I was specifically asking about the 2501A. I have recorded 3x 2501B models from you already. The A is the only one I couldn't find the core/VRAM frequencies of. Now you say that was core. But didn't say what VRAM is. Can you see why this is frustrating?

It's actually quite hard to find a post from you or anyone else that shows the LV1 string of the console you're currently talking about.

My E01 Franky with a NOS 40nm CXD5301 does 800/975 in what I'm calling Murder Crysis, which is Crysis3 with hacked Cvars pushing everything to the max.

I'm using crysis 3 because it's late in the PS3 lifecycle and uses the CELL SPUs to the full. The PS3 hardware if fully leveraged when the hacked config and the specific test location I used has every element on screen. Shading, lighting effects, water physics, particle effects, explosions, NPCs and AI behaviors, grass and foliage animations stretching off into the distance. A long alley way for loading textures off in the distance to load into VRAM, leveraging the maxed out draw distance. Plenty of 3D models and high resolutions assets. Action occurring in the scene keeping everything in motion.

This one spot is perfect for leveraging everything the game can do to push the PS3 HW.
View attachment 43210 View attachment 43211
At stock 500/650 it'll only do 12FPS and at 800/975 it does about 19FPS, which is actually playable, surprisingly! This level of detail is absurd and I didn't think it would even come close to approximating stock performance in this game at normal settings. This game without my Murder Config struggles to maintain 20FPS.

The OverPowerd heatsink keeps the 40nm nice and cool. I think actually makes the OC's more stable. I set a static 35% Fan and recorded temps and FPS to characterize it's potential. This is fairly accurate and you can use it to predict the relative performance of any OC. In my case, 800/975 took 12FPS to 19FPS. A 37% increase in true performance. And I know this because the games settings can make full use any performance increase. There's nothing unused in this game/config/scene.
View attachment 43209
CoreVRAMFPSCPUGPUAmbientGPUΔTFan%
45090012.25625425.32935
45095012.45635525.33035
50065012.06635425.12935
55070013.226254252935
60075014.31635624.83135
65080015.02635624.63135
70080015.8635724.63235
75080016.65645825.13335
80080017.296358253335
80085017.75635824.73335
80090018.15645925.63335
80092518.42645925.33435
80095018.75635924.83435
80097518.98635924.63435
About temps. I measured the ambient room temperature of using a thermocouple taped to the intake of the PS3. Not touching the plastic, where it could be transferring heat from the console, but suspended in the air in front of the intake. This is the true temperature of the air entering the console. If you take the difference in ambient and GPU you get the change in temperature, or Delta T. That is an even more accurate representation of the actual GPU temperatures between runs. This helps to see if there was a significant difference between them.

I wanted to see if down clocking the Core 50MHz to 450MHz, but overclocking the VRAM to as high as was stable (950) would match performance of stock clocks but reduce overall temps. While I was able to match and exceed FPS, the temps were not significantly different. Perhaps even slighly higher. However, I believe 450/850 would match 500/650 in FPS for this game. That might reduce temps. It should be noted that this approach is going to sacrifice some Core performance and tasks that are shader/processing heavy, such as lighting/shaddows, water simulation, explosions, particle effects...etc will ne negativly affected. Whereas high resolutions texture loading and streaming data to/from the Core will improve. These could balance out and be unniticble to the end user. And if that comes with a reduction in temps, especially for a 90nm RSX with defective underfill where small reductions in temps are a net positive, then this may not be such a bad idea.

And this is where it gets really REALLY weird. @Nascar1243 found that 300/900 and 350/750 punched a hole in spacetime and initiated Core Fusion...
View attachment 43212
He jumped from about 10FPS at 300/800 to 18FPS at 300/900! And 15FPS at 350/750. And they accomplished this physics defying, mind bending, wacko non-sense with 7C cooler temps. IT'S COMPLETELY BONKERS!

Now temper your excitement. XMB ran sluggish and you wouldn't want to actually use the console like this. It's just a strange novelty and a curiosity from an architectural standpoint. As for why the actual eff this happened we still don't know.
  • Hypothesis 1 (refuted): We thought it could be a quark in the way FPS calculates, but he said it actually does feel smoother in game. Like it really was running at that FPS, not just giving a false reading.
  • Hypothesis 2 (strained): I speculate it could be because 900 VRAM is a perfect 3x integer of 300 CORE and that there might be an architectural explanation for why 3 VRAM cycles per Core cycle might synchronize and improve efficiency or bandwidth of the shader piplines. Allowing for less idle time and improved latency for coordination between VRAM and core. And if that were the case, then the same might be true of any 2x or 3x integer. But I was not able to replicate it using a 2x 450/900. That was the highest 2x integer I can achieve. 500/1000 wouldn't be stable for me. Other integers may reveal more context if other were to try them, but in our tests it has only occurred at 300/900 and 350/750. @nascar was able to replicate it on another console, so this isn't a fluke. It appears to only work with lower cores. HOWEVER, you'll notice 350/750 isn't an integer. So why would it work? That's a chink in the hypothesis.
  • Hypothesis 3: It could be a quark in the way this game in particular uses CORE/VRAM and the available bandwidth. Or the settings I've enabled in the custom config.
  • Hypothesis 4: It's possible that something broke at these low of cores. Like some visual effect or processor heavy task that isn't obvious while looking at the screen. If it literally cannot run and isn't, then it won't eat up compute cycles and the FPS would jump up. It doesn't cause the system to freeze or anything, but then we don't want to loose the save by running around, so maybe we would find an issue if we attempted to progress through the game. Like loading a new area.
Anyway, that was weird. If you have any ideas as to why that could happen, let me know. Maybe a quark of the Curie architecture?
which level were you testing? I only tested the prologue level and the wet surfaces were purple also noticed the higher res stuff going on. it would crash in the elevator in the 1st chapter.
 
Can someone make a 725mhz/900vram?
725MHz on core is impossible. It's either 700 or 750.
Mitsu™can you please make evilnat 850*900 betta9, thanks in advance.
Uploaded !
850/900
https://www.mediafire.com/file/znsa65tw2fepu8z/850_900_beta9_PEX.PUP_4.91.CEX-PS3UPDAT.PUP/file
How much of a difference in fps between 700 core and 800? I have one of my 2501a running 700 - 850 and it runs great. Worth trying 750?
According to RIP-Felix experiments on (Murder) Crysis 3, it's a 9,43 % improvement between 700 and 800 core.
700/800 : 15,8 FPS
800/800 : 17,29 FPS

If 700MHz core is completely stable, then you can probably try 750MHz core without risks. At worst you should notice instabilities.
 
Last edited:
Well when it comes to New Vegas or Bethesda games even OC has no help. Problem is the save data itself. If only we could find a way to port the pc mod of clearing unnecessary cache on the save data the game would run perfect. After I complete GTA V I am going to run Fallout 3 elite edition and see how the OC holds out.

I personally started a new save file for New Vegas which I'll play later with less frequent saves and auto saves. I have plenty of mods on it and ran well until the save file got big. Only games I was able to complete without struggle is Oblivion on 750/975

I also noticed something but I might be wrong, when I run games on standard OC'D pup some stuff are unstable but the minute i install Ps3 4k pro mod most things become stable on higher Overclocks. May someone please check this out. Flash a pup that's unstable(50mhz and 25mhz) higher than your stable OC and install ps3 pro mod and please get back to me! :)

I don't want to come across as rude but I haven't seen anything that has really been researched apart from a few tips etc on reddit and I've spent hours looking across google. (Please link me to any posts with evidence or code such as github etc so I can look and poke around :))

I'm also not trying to say that your wrong or invalid, I have just have not found anything at all online that being said I'm looking into this as well as a few other games to potentially make a patch or mod to help with the issue depends on my spare time at the moment I have some free time hence the look into fall out new vegas and skim of overclocking, I've done so much investigation with fallout new vegas only because it's in my best interests.

Again if anyone has anything such as mods and code I take a peek at I'm in your debt.

As for overclocking if someone can point me to like a repo with firmwares listed for slim so I can do more testing that would be great, I don't wanna flash a pex firmware as one of those bricked my other slim even though it should have been fine but to be fair that console was old and I had since I was a child and was the worst one I owned so It was just an excuse to the girlfriend to get another one.

Thanks in advanced!
 
I don't want to come across as rude but I haven't seen anything that has really been researched apart from a few tips etc on reddit and I've spent hours looking across google. (Please link me to any posts with evidence or code such as github etc so I can look and poke around :))

I'm also not trying to say that your wrong or invalid, I have just have not found anything at all online that being said I'm looking into this as well as a few other games to potentially make a patch or mod to help with the issue depends on my spare time at the moment I have some free time hence the look into fall out new vegas and skim of overclocking, I've done so much investigation with fallout new vegas only because it's in my best interests.

Again if anyone has anything such as mods and code I take a peek at I'm in your debt.

As for overclocking if someone can point me to like a repo with firmwares listed for slim so I can do more testing that would be great, I don't wanna flash a pex firmware as one of those bricked my other slim even though it should have been fine but to be fair that console was old and I had since I was a child and was the worst one I owned so It was just an excuse to the girlfriend to get another one.

Thanks in advanced!
What's your slim model and manufacture date?
 
New slim is 2503b, old one was a 2003a RIP :( I've still got spares but I don't have the models on hand as there in the loft at the moment.

Soz Date code is 0C
 
Last edited by a moderator:
Are the update files posted here okay to use on a ps3 without a working wifi chip? In particular I was looking at the files posted by Mitsu TM. (I am not allowed to link)

For context, I have a 2504A from 0D, so I want to try 700/850 to start. From what I have read on this forum, that should be safe
 
Back
Top