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

[RIP-Felix]

Adding capacitance is what reduces low frequency ripple. A degraded cap would be a capacitor with low capacitance and lower capacitance means more low frequency ripple. So if you want to reduce the low frequency ripple then you must increase the total capacitance. In the 2500 series, I don't think the high ripple is because of degraded caps reducing capacitance but because the overall capacitance is low from the factory.

[

The PS3 vram is pretty budget. It's only 2 phase? So I think we'll hit the current limit very fast or start burning up our VRMs if we increase the current usage. I wouldn't raise the voltage beyond 1.1v for a daily driver, staying below 1.05v is even better.

Rather than correct you, I'll just link KEMET's K-Sim and let you discover the truth for yourself.

I lied...
Here's an example filter I designed to approximate the broadband performance of a NEC/Tokin proadlizer, in order to replace them with as close of a 1:1 drop-in as possible using a polymer/MLCC array. Spoiler, it's not possible, they were great, but the Tantalizer does work great. @Nascar1243 was able to raise his 40nm frankie all the way up to 1.6v without tripping voltage protection using them, Overclocking all the way to 1GHz core. That's an extreme use case and honestly, I didn't expect them to handle it, but they passed the test, proving the engineering principals are sound.

Impedance being lowered for this frequency range is what we want, because it's the range (100KHz to 10MHz) of switching ripple/harmonic noise the 2-phase (RSX) and 3-phase (CELL) Voltage regulation modules produce. Frequencies higher than that are decoupled by an array of 36x 0.1uF MLCCs.

Total Capacitance just affects how long the stored energy can last before the next "bang" (pulse of voltage) comes from the VRM. What I'm talking about is providing a low impedance path to ground for unwanted noise/ripple within this target band of frequencies. A well attenuated filter targets that frequency band by selecting the array of capacitors properly. Not just the total capacitance to smooth out the voltage drops between pulses from the buck converters. Once you have enough capacitance, more is just wasted, and if you go too high can cause issues, like increasing inrush current, potentially damaging components or tripping voltage protection designed to limit it. It it can lead to oscillations and increased instability, which I noticed when increasing it beyond 10mF. Among other issues.

As I said, a few extra capacitors isn't going to hurt, but it's not just about total capacitance. Their frequency response matters too and that can't be predicted or work efficiently if the existing caps have worn out and are not performing to specification anymore. They could be limiting the potential effect of your newly added caps would have had without the bad apple spoiling the bunch.

 

[ryeno]

Rather than correct you, I'll just link KEMET's K-Sim and let you discover the truth for yourself.

I lied...
Here's an example filter I designed to approximate the broadband performance of a NEC/Tokin proadlizer, in order to replace them with as close of a 1:1 drop-in as possible using a polymer/MLCC array. Spoiler, it's not possible, they were great, but the Tantalizer does work great. @Nascar1243 was able to raise his 40nm frankie all the way up to 1.6v without tripping voltage protection using them, Overclocking all the way to 1GHz core. That's an extreme use case and honestly, I didn't expect them to handle it, but they passed the test, proving the engineering principals are sound.

Impedance being lowered for this frequency range is what we want, because it's the range (100KHz to 10MHz) of switching ripple/harmonic noise the 2-phase (RSX) and 3-phase (CELL) Voltage regulation modules produce. Frequencies higher than that are decoupled by an array of 36x 0.1uF MLCCs.

Total Capacitance just affects how long the stored energy can last before the next "bang" (pulse of voltage) comes from the VRM. What I'm talking about is providing a low impedance path to ground for unwanted noise/ripple within this target band of frequencies. A well attenuated filter targets that frequency band by selecting the array of capacitors properly. Not just the total capacitance to smooth out the voltage drops between pulses from the buck converters. Once you have enough capacitance, more is just wasted, and if you go too high can cause issues, like increasing inrush current, potentially damaging components or tripping voltage protection designed to limit it. It it can lead to oscillations and increased instability, which I noticed when increasing it beyond 10mF. Among other issues.

As I said, a few extra capacitors isn't going to hurt, but it's not just about total capacitance. Their frequency response matters too and that can't be predicted or work efficiently if the existing caps have worn out and are not performing to specification anymore. They could be limiting the potential effect of your newly added caps would have had without the bad apple spoiling the bunch.

I think you're overcomplicating this. Based on your scope measurements, the 2500 series VRM has low frequency ripple. There was a little bit of high frequency noise but not too much so the MLCC were doing their job well enough.

https://www.digikey.com/en/articles...dc-dc-converters-to-minimize-power-rail-noise

To fix the low frequency ripple, you must add more capacitance. That's it. Adding 3 470uF caps brings the total capacitance back to PS3 FAT spec.

They could be limiting the potential effect of your newly added caps would have had without the bad apple spoiling the bunch.

Unless they are shorted, they won't hurt the performance of added caps because ESR and capacitance are combined.

 

[RIP-Felix]

The scope is only measuring what what it sees in the moment. In the video I didn't go into much detail about my triggering method, but what I was doing is increasing the trigger threshhole selecting for larger ripple. And I didnt perform an FFT to show the harmonic noise. That's not the point of the video and would have overcomplicated it. But since you brought it up...

I use the anlogy of a particularly large wave at the beach. You can sit there watching them for a good long time and see hundreds between 3 and 6 feet. Then every hour or so a 10 footer comes along.

Ripple and noise is like that. This is why people often report their bad tokins begining to show signs of wear when they're console randomly shuts down in games. It's not consistent and could be 5 mins or an hour after starting. It degrades over time becoming more and more frequent.

It's necessarily complicated because of the 2 major roles capacitors play. To deliver current in between buck pulses and to decouple unwanted frequencies, providing a low impeedance path to ground so they don't go through your processor and wreak havoc like a tsunami!

It's not that complicated. No need to oversimplify and ignore the 2nd major role capacitors have. You must take into account the combined frequency response of the entire filter. It's not optional.

The only reason to ignore this is if you want to feel safe in your overly simple all or nothing statment that "more capacitance reduces low frequency ripple." I guess there's psychological comfort in that, but it's not 100% true like you are stating it is. That's copium. It's partially true and can be 100% wrong if the caps you use cause an antiresonane peak, which cause an impeedance spkie right in the middle of your frequency response curve. At which point when the 10 foot wave hits, it blows right past your filter and...BEEP-BEEP-BEEP.

My point is that no one knows that capacitance is the only factor that has degraded. There are many other specifications that do too. ESR, leakage current, parasitic capacitance and impeedance. The frequency response will shift and if it shifts too far can let important frequencies past.

You can't dissmiss all of that and say they dont matter and capacitance is the only one that does.

Nuance matters.

 

[ryeno]

The scope is only measuring what what it sees in the moment. In the video I didn't go into much detail about my triggering method, but what I was doing is increasing the trigger threshhole selecting for larger ripple. And I didnt perform an FFT to show the harmonic noise. That's not the point of the video and would have overcomplicated it. But since you brought it up...

I use the anlogy of a particularly large wave at the beach. You can sit there watching them for a good long time and see hundreds between 3 and 6 feet. Then every hour or so a 10 footer comes along.

Ripple and noise is like that. This is why people often report their bad tokins begining to show signs of wear when they're console randomly shuts down in games. It's not consistent and could be 5 mins or an hour after starting. It degrades over time becoming more and more frequent.

It's necessarily complicated because of the 2 major roles capacitors play. To deliver current in between buck pulses and to decouple unwanted frequencies, providing a low impeedance path to ground so they don't go through your processor and wreak havoc like a tsunami!

It's not that complicated. No need to oversimplify and ignore the 2nd major role capacitors have. You must take into account the combined frequency response of the entire filter. It's not optional.

The only reason to ignore this is if you want to feel safe in your overly simple all or nothing statment that "more capacitance reduces low frequency ripple." I guess there's psychological comfort in that, but it's not 100% true like you are stating it is. That's copium. It's partially true and can be 100% wrong if the caps you use cause an antiresonane peak, which cause an impeedance spkie right in the middle of your frequency response curve. At which point when the 10 foot wave hits, it blows right past your filter and...BEEP-BEEP-BEEP.

My point is that no one knows that capacitance is the only factor that has degraded. There are many other specifications that do too. ESR, leakage current, parasitic capacitance and impeedance. The frequency response will shift and if it shifts too far can let important frequencies past.

You can't dissmiss all of that and say they dont matter and capacitance is the only one that does.

Nuance matters.

Again, you're overcomplicating this. 1) We want to reduce low frequency ripple. The only way to reduce low frequency ripple is to add more capacitance. The purpose of bulk capacitance in the vrm circuit is to remove the low frequency ripple caused by the switching. 2) Adding more capacitors will NOT increase the impedance because each extra capacitor decreases it because the capacitors are in parallel. 3) Inrush current might be an issue if we increase the capacitance too high but that doesn't seem to have happened by only adding 3 extra 470uF caps.

 

[RIP-Felix]

Impeedance and ESR combines when there are more than 1. And the calculation is hairy when you combine caps with different values. It can cause antiresonance that increases impeedance in the important range. At this point, I don't have much more to add.

If the caps are identical, then those factors cancel out and it simply divides by the number of caps. That's why I reccomended you use 470uF Aluminum polymer or Tantalum Polymer panasonic caps of the same kind as your motherboard has.

When combining different values, like a 560uF instead of 470uF, or and old and new one those factors do not cancel out and affect the resulting frequency response curve.

It's why what I did isn't the reccomendation, just a hack. And the result is a good example of why it matters. It didn't work well. It did a little bit. Sure. But I mean long lasting actual repair should apply a more encumpassing understanding of the engineering principals. I would replace all the caps with new ones if I were going to sell this console to a customer, since I measured wear indicating they are not performing to specification anymore. Leaving them on there, knowing they've worn out is irresponsable and lazy.

Just replace the old with new and be done with it. That's repair, when a proper diagnosis is made.

Changing the design should only be considered when you know the outcome.

As stated before, you can increase capacitance to a point. Adding 3x 470uF caps isn't going to make it worse. My results showed it was better, but not as effective as it should have been. Or could have been had I replaced all of them with new caps.

At this point I've said all that I need to. I think you have enough to mull it over and research further, if you're willing. I'm just trying to help you not fall victum to the same trap I see many people fall for.

 

[ryeno]

I would replace all the caps with new ones if I were going to sell this console to a customer, since I measured wear indicating they are not performing to specification anymore. Leaving them on there, knowing they've worn out is irresponsable and lazy.

Just replace the old with new and be done with it. That's repair, when a proper diagnosis is made.

Changing the design should only be considered when you know the outcome.

As stated before, you can increase capacitance to a point. Adding 3x 470uF caps isn't going to make it worse. My results showed it was better, but not as effective as it should have been. Or could have been had I replaced all of them with new caps.

At this point I've said all that I need to. I think you have enough to mull it over and research further, if you're willing. I'm just trying to help you not fall victum to the same trap I see many people fall for.

Remove the 7 factory polycaps and replace them with new poly caps and check the ripple. I guarantee you that you wouldn't see any improvement because the problem is low total capacitance.

 

[ryeno]

Yeah, that's why I'm asking people to post their SYSCON Errorlog dump as well. It has the Platform ID listed and is right there in evilnat's CFW tools now. Easy to correlate to the MB rev using the Devwiki.

I am also curious to see what errors may be triggered by pushing the RSX clocks this high. There are some errors that could silently trigger, like...

• 2020/2120, 2022/2122, or 2024/2124 that could indicate AV instability (DVE and HDMI transmission).
• 1002 could be due to RSX core voltage instability, that would trigger a YLOD.
• 1001s and 1004s is very likely to be in the log as they occur from unexpected shutdowns, which will happen when the console freezes and is forced off. But it can also indicate CPU core voltage instability, but I doubt that would be happening this case, since we're not OC the Cell (yet).
• FBVDDQ might cause 10x 2120/ 1x 3013 error combos from a short, but insufficient voltage/filtering I'm not sure what error it might cause. I'd just like to have the logs in case something interesting pops up.

I edited the config file directly using the Cvar commands after reading down the list and how they work.

So start a new game and skip the movie. Then leave the controller alone. When Sarah stands up she'll be facing a mirror. That's been a standard test spot in TLOU for the last few years and most of us that test temps and FPS us it as out baseline. So pleas use it as well or we won't be talking about the same consistent setup/workload.

1. Rip Crysis3 to PS3 using multiman. DO NOT update.
2. FTP to /dev_hdd0/GAMES/BLUS30976-[Crysis3]/PS3_GAME/USRDIR
3. View attachment 43011
   
4. Copy system.cfg and keep it safe. Or rename original.
5. Save the following modified system.cfg to
   /dev_hdd0/GAMES/BLUS30976-[Crysis3]/PS3_GAME/USRDIR
   

   ; crysis3
   
   sys_game_folder=c3
   
   sys_localization_folder=localization
   
   release_install=1
   
   sys_languages=english,french,spanish
   g_language=english
   
   con_restricted=0
   sv_cheatprotection=0
   r_displayinfo=1
   pl_health.normal_regeneration_rateSP=99999
   pl_health.normal_threshold_time_to_regenerateSP=0.1
   g_godmode=1
   g_infiniteammo=1
   g_infinitesuitenergy=1
   g_inventoryNoLimits=1
   v_goliathMode=1
   g_infiniteCloak=1
   sys_MaxFps=0
   e_ParticlesObjectCollisions=Dynamic
   e_ParticlesMotionBlur=1
   e_Tessellation=255
   r_WaterTessellationHW=1
   r_SilhouettePOM=1
   e_GI=0
   r_FogShadows=0
   r_FogShadowsWater=0
   r_SSReflections=0
   r_Sharpening=0
   r_ChromaticAberration=0
   r_ssdo
   e_GsmCache=0
   e_ShadowsResScale=40
   e_ShadowsPoolSize=4096
   e_ShadowsMaxTexRes=1024
   e_GICache=0
   e_TessellationMaxDistance=255
   e_MergedMeshesInstanceDist=255
   e_MergedMeshesViewDistRatio=255
   e_LodRatio=255
   e_ViewDistRatio=255
   e_ViewDistRatioDetail=255
   e_ViewDistRatioVegetation=255

• Delete any game data if it exists.
  
  Note: If you DL Crysis3 then it'll be the updated 1.04 version, which was patched to prevent this from working.

Are you still using this config file? I get visual errors when I try using

e_ShadowsMaxTexRes=1024

The rain on the floor becomes corrupted

I'm using

; crysis3
sys_game_folder=c3
sys_localization_folder=localization
release_install=1
sys_languages=english,french,spanish
g_language=english
sys_MaxFps=0
; Real-time Global Illumination
e_GI = 0
; Volumetric Water Shadows
r_FogShadowsWater = 0
; Depth of Field
r_DepthOfField = 0
; Color Grading
r_ColorGrading = 0
; Explosion Blur
g_radialBlur = 0
; Particle Collisions
e_ParticlesObjectCollisions = 1
; Shadow Casting Light Scale
e_ShadowsResScale = 1.0
; Tessellation Distance Limit
e_TessellationMaxDistance = 5.00
; Grass Instance Animation Distance
e_MergedMeshesInstanceDist = 2.0
; LOD distance ratio
e_LodRatio = 4.0
; View distance ratio (Vegetation)
e_ViewDistRatioVegetation = 21
; Chromatic Aberration
r_ChromaticAberration = 0.0
; Bloom amount
r_HDRBloomRatio = 0.000
; Ambient Occlusion Method
r_ssao = 0
r_ssdo = 0
; Sun Shadow Map Size
e_ShadowsMaxTexRes = 256
; Grass Instance Max Distance
e_MergedMeshesViewDistRatio = 10.0
; View distance ratio (General)
e_ViewDistRatio = 25
; View distance ratio (Detail)
e_ViewDistRatioDetail = 19

https://tweakguides.dmegaming.com/Crysis3_15.html

I finished OCing one of my PS3s. The results are kinda disappointing.
factory voltage was 1.031v. 700 core stable.
after doing the cap mod and increasing voltage to 1.05v, 750 core was stable in Crysis 3.
1.081v 800 was not stable. Textures were flickering. 1.1v may or may not have been stable but I didn't like the temp.
So I'm running 1.05v 750 core. I might raise it 1.056v to have some extra headroom.


NOTICE: SSD UPDATE FASTER THAN HDDS. So It's a good idea to use an SSD when you're increasing clockspeeds because if it's unstable then you can update the system faster.

 

[Dhmhtrhs]

If the ps3 reach 1ghz on overvold no problem do you think 2ghz is even possible? Someone better try it

 

[LuanTeles]

Which issue is that exactly?

@RIP-Felix @Mitsu™ @LuanTeles mother of god. 1ghz on both. I got brave and finally pushed it to 1ghz. It's perfect in xmb but in crysis it's not stable. And crashes with black flickering. So with voltage this is an easy 1ghz core. And without voltage at all 950/1000 stable all day.

@Tanzu15, I successfully clocked mine to 1000/1225 MHz. However, when I tried 950 and 1000 MHz on the core, it resulted in an instant crash on CRYSIS, GOW, and other RSX-intensive games. The game crashes right after the first in-game frame. I can reach the main screen and loading screens just fine, but as soon as the game starts loading the graphics, the screen freezes while the sound continues. I can still access the in-game XMB and the impose screen to exit the game without any issues.

Instant Crash

• GOW
• CRYSIS

Working (Cell.be games)

• Detuned
• Bayonetta
• Heavenly Sword
• All homebrew apps

I don't have any more games on my HDD to test.

EDIT:
I rolled back my system to 850/1000 MHz to see if the HDMI output issue I was having sometimes is caused by the VRAM clock. I haven't had much time lately as I'm working in another city and living there, so I only come home on weekends. Therefore, it might take me a bit longer to report back on this.

 

[LuanTeles]

NOTICE: SSD UPDATE FASTER THAN HDDS. So It's a good idea to use an SSD when you're increasing clockspeeds because if it's unstable then you can update the system faster. Setting fan to 95% while updating is also a good idea.

A good idea is to update the system with a PS3PATCH.PUP as a last resort since it installs much faster. Of course, we will lose the jailbreak, but we can use the bgtoolset right after it. At least this way, we can save our console.

Another idea is to use the FAN TEST boot mode to cool down the console, as it increases the fan speed to 100%.

 

[Tanzu15]

@Tanzu15, I successfully clocked mine to 1000/1225 MHz. However, when I tried 950 and 1000 MHz on the core, it resulted in an instant crash on CRYSIS, GOW, and other RSX-intensive games. The game crashes right after the first in-game frame. I can reach the main screen and loading screens just fine, but as soon as the game starts loading the graphics, the screen freezes while the sound continues. I can still access the in-game XMB and the impose screen to exit the game without any issues.

Instant Crash

• GOW
• CRYSIS

Working (Cell.be games)

• Detuned
• Bayonetta
• Heavenly Sword
• All homebrew apps

I don't have any more games on my HDD to test.

EDIT:
I rolled back my system to 850/1000 MHz to see if the HDMI output issue I was having sometimes is caused by the VRAM clock. I haven't had much time lately as I'm working in another city and living there, so I only come home on weekends. Therefore, it might take me a bit longer to report back on this.

1000 memory honestly is the most stable memory clock for the memory that supports it. Mine does 950/1000 easily zero diff. And zero bugs. Crashes or issues. Yours would need more voltage for core stability. At 1000 core I'm fully stable in xmb and in game menus but with crysis, it starts black shadow flickering in then in 5 to 8 seconds freezes. So some more voltage would grant me stability. But that's so pointless. 950/1000 fully Stable 24/7 is a dream come true. It's the fastest ps3 recorded. And it's fully stable no draw backs when it comes to texture flicker etc etc. the silicon must be one of the best manufactured at that point in time. I have 2 other 900/1000 consoles too. And a bunch of other 850/975s.

 

[RIP-Felix]

Are you still using this config file? I get visual errors when I try using

e_ShadowsMaxTexRes=1024

The rain on the floor becomes corrupted

I'm using

; crysis3
sys_game_folder=c3
sys_localization_folder=localization
release_install=1
sys_languages=english,french,spanish
g_language=english
sys_MaxFps=0
; Real-time Global Illumination
e_GI = 0
; Volumetric Water Shadows
r_FogShadowsWater = 0
; Depth of Field
r_DepthOfField = 0
; Color Grading
r_ColorGrading = 0
; Explosion Blur
g_radialBlur = 0
; Particle Collisions
e_ParticlesObjectCollisions = 1
; Shadow Casting Light Scale
e_ShadowsResScale = 1.0
; Tessellation Distance Limit
e_TessellationMaxDistance = 5.00
; Grass Instance Animation Distance
e_MergedMeshesInstanceDist = 2.0
; LOD distance ratio
e_LodRatio = 4.0
; View distance ratio (Vegetation)
e_ViewDistRatioVegetation = 21
; Chromatic Aberration
r_ChromaticAberration = 0.0
; Bloom amount
r_HDRBloomRatio = 0.000
; Ambient Occlusion Method
r_ssao = 0
r_ssdo = 0
; Sun Shadow Map Size
e_ShadowsMaxTexRes = 256
; Grass Instance Max Distance
e_MergedMeshesViewDistRatio = 10.0
; View distance ratio (General)
e_ViewDistRatio = 25
; View distance ratio (Detail)
e_ViewDistRatioDetail = 19

https://tweakguides.dmegaming.com/Crysis3_15.html

I finished OCing one of my PS3s. The results are kinda disappointing.
factory voltage was 1.031v. 700 core stable.
after doing the cap mod and increasing voltage to 1.05v, 750 core was stable in Crysis 3.
1.081v 800 was not stable. Textures were flickering. 1.1v may or may not have been stable but I didn't like the temp.
So I'm running 1.05v 750 core. I might raise it 1.056v to have some extra headroom.


NOTICE: SSD UPDATE FASTER THAN HDDS. So It's a good idea to use an SSD when you're increasing clockspeeds because if it's unstable then you can update the system faster.

I only used it to maximize the console usage. The idea was to load the VRAM full of textures, and leave no performance on the table while I was collecting data to gauge performance gains. I wouldn't play the game at those settings. And yeah, there are graphical issues with those settings and if you try to play it, it crashes loading new areas. So it's just for testing max performace at that test location.

But it's interesting to see how tweaking various setting affect the game.

 

[Ketxxx]

I had a little more time recently to put the couple PS3s I have through some more tests, both of them are identical using the same thermal materials, PSU, etc. The only difference is that the 2103 is using the "upgraded" heatsink compared to the 2503 and for experiments sake the 2103 hasn't undergone any de-lidding while the 2501 has had a CPU de-lid. What follows are my observations and general thoughts and notes;

2103: 66c @ stock in DMC4 snowy mountain castle overpass with 35-50% fan set, auto @ 65c.
2503: 60c with CPU delid @ 650\800 in DMC4 snowy mountain castle overpass with 35-50% fan set, auto @ 63c.

Observations: 2503 on average runs with a 10% lower fan speed, a significant difference for sure. It can also be inferred by these results that a GPU de-lid could potentially drop fan speed another 7-15% and temps another 4-8c depending on workload. The GPU is a totally different animal and likely isn't getting the workout it should be having to spend so much time waiting on the dog slow Cell CPU to feed it the data it wants so more accurate figures are hard to lock down.

2103 runs with an average of 45-47% fan speed somewhat suggesting the beefier heatsink is either A: not the benefit it has been considered to be over the thermodynamically weaker 2503 heatsink or B: any advantages the 2103 heatsink has is being totally negated by the fact the IHS at least on the CPU needs removing.

Regardless of if it is the CPU or GPU that hits your defined fan speed either one will trigger the fan to ramp up.

Now, these results are quite interesting and certainly show why at this stage a de-lid is strongly recommended but I wouldn't go as far as to give the whole de-lidding thing a strong essential necessity tag because even as someone very experienced removing CPU and GPU IHS's I don't recommend this for the PS3 lightheartedly because the CPU and particularly RSX are so easy to kill with this process. I've never come across more fragile GPUs or CPUs than what the PS3 uses.

At the moment I am working on a much safer way to remove the CPU and GPU IHS and my working theory is that once the PS3 is jailbroken and some custom perameters set you can strip the console down, take a infrared temperature gun and point it at the CPU and GPU to measure how hot they are getting and wait for one of them to reach about 90c, hopefully the PS3s temperature protection won't kick in where it usually does using a CFW so 90c can be reached. The heat will soften the rubber seal and epoxy on the RSX so you can gently lever both off. In the case of the RSX don't try to flip it all at once, work one corner at a time then either just gently hold the corner up with the tool you are using or as I will likely do place something under the corner and move to the next to hopefully get 2 corners done in one thermal cycle. Once the IHS cools too much just repeat the thermal cycle for the other 2 corners. Even doing this I wonder if it will reduce RSX deaths from de-lid attempts as it really is that fragile I've had RSX IHS's willingly pop off using other methods that really didn't do anything excessive and the RSX has still died.

Another viable option could be to use a variable temperature soldering iron, heat it to 100c then place it on one corner at a time to gently lift each RSX corner once the thermal eloxy has softened. Again you'd need a temperature gun to ensure you don't heat the corner excessively you don't want to kill the memory chip with heat.

Hopefully someone found this useful and if anyone tries my de-lid suggestions I'd be interested to know how it goes I'll be doing this myself when I get a chance but don't know when that chance might be.

@ryeno & @RIP-Felix; You're both overcomplicating it. Just filter out your ripple\noise\transient spikes at the source. Sure, you might have to throw in a few upgraded caps here and there further along in the circuit or add a bit extra filtering if you really want to optimise but you don't need to radically change things, in fact that's a bad idea, a VERY BAD idea, particularly when it comes to capacitance.

 

[Panagiotis5373]

Hi guys does anyone have any proper tutorial for ps3 liquid metal apply? I heard it can actually damage some copper that is connect to the gpu and rsx to move the air but that thing is very sensitive to metal and i don't understand how should i do it any help please?

 

[ryeno]

I finished overclocking my 2 consoles.

rsx40 a01 700/925 vpe:ff shd:3f [NN6051-11:0:4:b:15:4:6:0:1][25:0:e:0:1:0:1][1:1:0]
179d 01:35:38 • 2,930 ON • 2,894 OFF (36)
700MHz @ stock voltage
750MHz @ stock voltage & cap mod.
800MHz @ 1.056v
925MHz VRAM

950MHz VRAM was game stable but resulted in occasional crashes when loading games from the XMB. Currently running it at 800MHz @ 1.062v but it tested stable at 1.056v. 850MHz was not stable at 1.125v and I didn't want to raise the voltage any further.

rsx40 a01 700/925 vpe:ff shd:3f [NN8036-12:0:4:c:2:a:6:0:1][1e:0:a:0:1:0:1][1:1:0]
342d 20:21:00 • 2,781 ON • 2,727 OFF (54)
700MHz @ stock voltage
750MHz @ 1.025v & cap mod
800MHz @ 1.115v & cap mod
875MHz VRAM

I was experiencing hard crashes with this console so I lowered the VRAM from 900 to 875 and it's been stable so far. It throws out a 1601 or 1701 error code when it crashes. This console runs hotter than the other console and probably needs a delid. Currently running it at 750MHz @ 1.031v but it tested stable at 1.025v.

I ran into a problem with auth. I couldn't get syscon.ps1 or Windows64_gui_ps3_syscon_uart_script.exe to auth so I used SysconReader.exe auth and then switched to syscon.ps1 to write.

@ryeno & @RIP-Felix; You're both overcomplicating it. Just filter out your ripple\noise\transient spikes at the source. Sure, you might have to throw in a few upgraded caps here and there further along in the circuit or add a bit extra filtering if you really want to optimise but you don't need to radically change things, in fact that's a bad idea, a VERY BAD idea, particularly when it comes to capacitance.

Rip thinks that the capacitance is fine but degraded. I think the 2501 VRAM has insufficient capacitance because the vram is only 2 phase and they only have ~3300uF while the NEC TOKIN consoles had 4000 or 4800.

You should do the capmod. I had 1 console that went from 700 to 750 on the core from the cap mod. If you're reducing ripple by 10 or 20mv then that's the same as increasing voltage by 2 or 3 bins.

Hi guys does anyone have any proper tutorial for ps3 liquid metal apply? I heard it can actually damage some copper that is connect to the gpu and rsx to move the air but that thing is very sensitive to metal and i don't understand how should i do it any help please?

Don't use liquid metal. It's not worth the risk.

 

[Panagiotis5373]

I finished overclocking my 2 consoles.

rsx40 a01 700/925 vpe:ff shd:3f [NN6051-11:0:4:b:15:4:6:0:1][25:0:e:0:1:0:1][1:1:0]
179d 01:35:38 • 2,930 ON • 2,894 OFF (36)
700MHz @ stock voltage
750MHz @ stock voltage & cap mod.
800MHz @ 1.056v
925MHz VRAM

950MHz VRAM was game stable but resulted in occasional crashes when loading games from the XMB. Currently running it at 800MHz @ 1.062v but it tested stable at 1.056v. 850MHz was not stable at 1.125v and I didn't want to raise the voltage any further.

rsx40 a01 700/925 vpe:ff shd:3f [NN8036-12:0:4:c:2:a:6:0:1][1e:0:a:0:1:0:1][1:1:0]
342d 20:21:00 • 2,781 ON • 2,727 OFF (54)
700MHz @ stock voltage
750MHz @ 1.025v & cap mod
800MHz @ 1.115v & cap mod
875MHz VRAM

I was experiencing hard crashes with this console so I lowered the VRAM from 900 to 875 and it's been stable so far. It throws out a 1601 or 1701 error code when it crashes. This console runs hotter than the other console and probably needs a delid. Currently running it at 750MHz @ 1.031v but it tested stable at 1.025v.

I ran into a problem with auth. I couldn't get syscon.ps1 or Windows64_gui_ps3_syscon_uart_script.exe to auth so I used SysconReader.exe auth and then switched to syscon.ps1 to write.


Rip thinks that the capacitance is fine but degraded. I think the 2501 VRAM has insufficient capacitance because the vram is only 2 phase and they only have ~3300uF while the NEC TOKIN consoles had 4000 or 4800.

You should do the capmod. I had 1 console that went from 700 to 750 on the core from the cap mod. If you're reducing ripple by 10 or 20mv then that's the same as increasing voltage by 2 or 3 bins.


Don't use liquid metal. It's not worth the risk.

Already did the metal was so weird that when i try to spread it that thing just flied exactly outside the gpu and gg not turning on

 

[ryeno]

You must use MFWBuilder. With it you can apply OC to Evilnat Beta 7 CFW.
Or ask to someone if you can't manage to use it.


Didn't see this soon enough. 2004B model stable limits seems to be 700MHz on core / 900MHz VRAM.
950MHz VRAM really is too much for it.
I can't guide you through E3 usage though, I never used one myself.
But you will probably find help here.

Could you make a 800/925 profile please?

FATAL ERROR: patchtool.exe failed to execute! : couldn't execute "patchtool.exe": invalid argument
FATAL ERROR: Unable to patch self lv1.self.elf : patchtool.exe failed to execute! : couldn't execute "patchtool.exe": invalid argument
*****************
Error running script: Unable to patch self lv1.self.elf : patchtool.exe failed to execute! : couldn't execute "patchtool.exe": invalid argument

 

[WailkyS5M5]

Managed to successfully overlclock my Cech-2004A Slim without facing any bricks/artifacts whatsoever.
I went step by step, starting at 600/750 and applying 50Mhz increments each to the Core and testing Crysis 3 test spot like brother RipFelix did in his video.
I stopped at 650 Core, then did the same to the VRAM staying at 850. These are the values i saw are recommended to cap on for my model in brother Mitsu's video description.

Unrelated but i had to face a 1601/1701 YLOD once per applied overclock when playing either Crysis or Red dead. Aldostools told me that this is some issue with the FPS Plugin and the PS3 Software that causes this error. So that's the only "issue" i stumbled while doing this so far.

Temps did not really increase for me, maybe 1-2°C plus on the RSX at least which is good. Crysis 3 had huge improvements, GTA 5 too. For the early Rockstar games like GTA 4 and RDR not much has changed which is expected.
I might hook the ps3 up to a Capture card and make comparison videos to some undocumented games? Let me know if you would like to see that.

Wailky

 

[RoboKing's Cosmos]

@LuanTeles what is the recommended option for overclocking EvilNat CFW on a CECHL01 with fresh thermal paste? as it is recommended by your PS3 4K Pro. also my game Katamari Forever has lag too so hoping it'd improve that.

 

[Charles_n_town]

Could you make a 800/925 profile please?

FATAL ERROR: patchtool.exe failed to execute! : couldn't execute "patchtool.exe": invalid argument
FATAL ERROR: Unable to patch self lv1.self.elf : patchtool.exe failed to execute! : couldn't execute "patchtool.exe": invalid argument
*****************
Error running script: Unable to patch self lv1.self.elf : patchtool.exe failed to execute! : couldn't execute "patchtool.exe": invalid argument

Did you run mfw builder as administrator?