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

Mitsu™ said:
I don't know which version of the game you have, so :
US :
https://www.mediafire.com/file/lr6raa0aopo4cgf/NPUB30302C2SAVEENCBIN16.zip/file
EU :
https://www.mediafire.com/file/q604bdf7n3h1x0r/NPEB00575C2SAVEENCBIN16.zip/file


Even with that, you will likely hit the limits of the GPU well before 1200MHz.
But 1000MHz... why not :)
Click to expand...
Well doubling the speed of the RSX is not nothing at all:tranquillity:
If the CELL ever gets overclocked I imagine it will also be able to be pushed quite far.
 
hey.. anyone got the latest firmware with 750/1000 please..? im doing some testing with my ps3 right now

(EDIT: im testing Sonic the Hedgehog (2006), a game nobody likes but still lags a lot!!)
 
Last edited:
EumelEumel said:
tyvm.
my 0D 2504 pulls the white flag on 900 Core it seems, when it gets hotter than 52° :biggrin2:
Click to expand...
Perhaps when we increase voltages you can stabilise 900!

marcoselball said:
hey.. anyone got the latest firmware with 750/1000 please..? im doing some testing with my ps3 right now

(EDIT: im testing Sonic the Hedgehog (2006), a game nobody likes but still lags a lot!!)
Click to expand...
1000 memory is pretty dangerous, 950 is the max for most 25xxs
 
Last edited by a moderator:
EumelEumel said:
tyvm.
my 0D 2504 pulls the white flag on 900 Core it seems, when it gets hotter than 52° :biggrin2:
Click to expand...
So 850MHz core is your limit, a great OC ! :)
52°C would be hard to maintain x)
marcoselball said:
oh well, I guess I'll just stay with 750/950, I like these setting too so it's fine, TYSM for the help too
Click to expand...
Razmann4k is right, 1000MHz is risky. Plus, we don't know which RSX revision you have, so if you try 1000MHz on a CXD5300A1GB, you'll likely brick it.
(Btw, does Sonic 06 runs better ?)

On my side, running Crysis HD for almost 3 hours with a target of 68°C (the default WebmanMod target) at 850/1000 and it's perfectly stable.
No freeze, no crash, zero artifacts.
So, 70°C is a guaranteed freeze but 67~68°C is fine.
Okay.
IMG_20240612_003724.jpg
 
Mitsu™ said:
So 850MHz core is your limit, a great OC ! :)
52°C would be hard to maintain x)

Razmann4k is right, 1000MHz is risky. Plus, we don't know which RSX revision you have, so if you try 1000MHz on a CXD5300A1GB, you'll likely brick it.
(Btw, does Sonic 06 runs better ?)

On my side, running Crysis HD for almost 3 hours with a target of 68°C (the default WebmanMod target) at 850/1000 and it's perfectly stable.
No freeze, no crash, zero artifacts.
So, 70°C is a guaranteed freeze but 67~68°C is fine.
Okay.
View attachment 43405
Click to expand...
wait, I did tried 1000 an hour ago, everything seems fine
how do I find that CXD.. number? I know its on IHS, but is there a software way?
 
Mitsu™ said:
So 850MHz core is your limit, a great OC ! :)
52°C would be hard to maintain x)

Razmann4k is right, 1000MHz is risky. Plus, we don't know which RSX revision you have, so if you try 1000MHz on a CXD5300A1GB, you'll likely brick it.
(Btw, does Sonic 06 runs better ?)

On my side, running Crysis HD for almost 3 hours with a target of 68°C (the default WebmanMod target) at 850/1000 and it's perfectly stable.
No freeze, no crash, zero artifacts.
So, 70°C is a guaranteed freeze but 67~68°C is fine.
Okay.
View attachment 43405
Click to expand...


I could really say Sonic 06 runs better


Here in the video, at 1:01:10, there is a huge lag spike, but with the overclock the lag spike was reduced a lot (the video it's just an example, I didn't record any gameplay of that btw)

(EDIT: the tornado part without the overclock is a bit laggy too, after the overclock I could say It's a little less laggy)

(EDIT 2: and I have the 25XX model so I guess 1000MHz it's dangerous, Thanks for warning me!)
 
Last edited:
Mitsu™ said:
So 850MHz core is your limit, a great OC ! :)
52°C would be hard to maintain x)

Razmann4k is right, 1000MHz is risky. Plus, we don't know which RSX revision you have, so if you try 1000MHz on a CXD5300A1GB, you'll likely brick it.
(Btw, does Sonic 06 runs better ?)

On my side, running Crysis HD for almost 3 hours with a target of 68°C (the default WebmanMod target) at 850/1000 and it's perfectly stable.
No freeze, no crash, zero artifacts.
So, 70°C is a guaranteed freeze but 67~68°C is fine.
Okay.
View attachment 43405
Click to expand...
A locked 30! What does this section run at on stock 500/650? Thanks for the save file for this section btw!
33% fan speed isn't bad at all, my 2100 isn't quite as cool but that's because the paste I bought was dry and I didn't have any other paste, I should repaste with something better soon lol especially before going above its current 600/750 clocks.
 
luhtoe said:
hi, I was considering overclocking my fat cechl 65nm to 600/750 but I have heard that even a small OC is unstable on the fat units due to the NEC tokins and the increased heat generation. is it a good idea or should I keep it stock?
Click to expand...

I can't tell you if it's a good idea or not but I will chime in about what I have been able to do with my CECHL01. I have been testing mostly with a 60fps patched eboot of warhawk. I've been incrementing MFW firmware builds by lowest MHz available. I found for my system, 700MHz core was unstable in game and produced artifacting and then crashed the system. So I backed my core clock down to 650Mhz and having been pushing memory 25MHz at a time. So far 800 has been stable but given what info I have read in this thread from others I think I will stop here for this system.

I am interested in voltage control if that becomes available. Even if it is only available by syscon I would like to mess with it. If we can get more voltage control, I would like to make a watercooled ps3 slim to push clocks.

Here is info about my system:
CFW 4.91.2 Evilnat Cobra 8.5 [D-PEX] [BETA 9]
rsx65 a06 650/800 vpe:ff shd:6f [AP0008827:1:1:15:17:e:2:5:2][1d:0:3:0:1:2:0][2:1:0]

Syscon log info:
Firmware Version: 4.91 (build 50754)
Platform ID: CokF10
Product Code: 00 82
Product Sub Code: 00 08
Hardware Config: 000000000307BCBF
Syscon Firmware Version: 065D.0000000000000000 (EEPROM: 0000000000000000)
Bringup Count: 1001, Shutdown Count: 975
Runtime: 167 Days, 1 Hours, 24 Minutes, 38 Seconds
 
EumelEumel said:
wait, I did tried 1000 an hour ago, everything seems fine
how do I find that CXD.. number? I know its on IHS, but is there a software way?
Click to expand...
Well you can check the datecode of your PS3 to get some idea, if it's 0C (July-September 2010) then it's more likely the CXD5300A1GB, if it's 0D (October - December 2010) or 1A (January - March 2011) then there's a chance of a CXD5300CGB or CXD5300GGB which can handle higher clocks than the previously mentioned revision. Even if it's a 0D or 1A that still doesn't give you assurance of course, the only way to know for sure from my knowledge is to check the IHS unless someone else knows if it's detectable in software?
 
Razmann4k said:
Well you can check the datecode of your PS3 to get some idea, if it's 0C (July-September 2010) then it's more likely the CXD5300A1GB, if it's 0D (October - December 2010) or 1A (January - March 2011) then there's a chance of a CXD5300CGB or CXD5300GGB which can handle higher clocks than the previously mentioned revision. Even if it's a 0D or 1A that still doesn't give you assurance of course, the only way to know for sure from my knowledge is to check the IHS unless someone else knows if it's detectable in software?
Click to expand...
I actually teared it apart, cleaned and re-pasted it today. And made pictures which already came in handy.. I have CXD5300CGB 0D Datecode

So I shouldn't try to go higher on VRAM? haha
currently on 850/1000
 
Razmann4k said:
A locked 30! What does this section run at on stock 500/650? Thanks for the save file for this section btw!
33% fan speed isn't bad at all, my 2100 isn't quite as cool but that's because the paste I bought was dry and I didn't have any other paste, I should repaste with something better soon lol especially before going above its current 600/750 clocks.
Click to expand...
To be honest, I don't even remember how it runs at 500/650 :') I need to check.

EumelEumel said:
I actually teared it apart, cleaned and re-pasted it today. And made pictures which already came in handy.. I have CXD5300CGB 0D Datecode

So I shouldn't try to go higher on VRAM? haha
currently on 850/1000
Click to expand...
My PS3 also have a CXD5300CGB RSX and is at 850/1000. If you can maintain your PS3 below 70°C (which is easy), it should stay stable.

Above 1000MHz is really dangerous. If you don't have a hardware flasher, then don't try it.
 
EumelEumel said:
I actually teared it apart, cleaned and re-pasted it today. And made pictures which already came in handy.. I have CXD5300CGB 0D Datecode

So I shouldn't try to go higher on VRAM? haha
currently on 850/1000
Click to expand...
CXD5300CGB is better of course than than CXD5300A1GB, I think you're pretty much at the limit for your chip, maybe you can go a bit higher on the core to 900 if you're lucky, but I honestly wouldn't touch the VRAM any more than that because if you don't have a hardware flasher you might brick your PS3. When voltage mods are available, maybe you can increase your core to 900 or maybe more but we don't know yet.
 
Alright thanks guys
@Razmann4k 900 is not possible on mine. Not entirely sure but it seems like I have to maintain temp below 52c to be able to push it that high, which obviously is not possible unless you can live with a jet inside your living room.
 
I can only do intermittent testing in my spare time but I'm comfortable in saying that once you hit 650MHz core frequency further increases don't really matter all that much for the GPU the core isn't especially frequency starved. Before you'd even need to really consider going above 650MHz vRAM is going to need to be at 850MHz, every +100MHz vRAM should be accompanied by +50MHz core increase to keep bottlenecking minimal.

For vRAM I would say 800MHz is what should be aimed for to iron out the worst of the performance issues in GPU limited titles. As far as I can find out all PS3s are fitted with GDDR rated for 667\700 (evidence more heavily supports the latter) so every PS3 should be capable of 800MHz vRAM speeds you'd have to be very, very unlucky to not be able to reach that. In fairly bad cases you'll likely go from 17FPS to 24-26FPS @ 650\800. Some games are just a lost cause though like Assassin's Creed those games appear to be very CPU bottlenecked (likely down to pure and simple bad\lazy coding).

Using what data I could get access to you'd likely need to increase the Cell frequency to 4GHz to remove most of the CPU bottleneck which on paper should be possible with every Cell as it appears like the Cell originally was designed for at least 4GHz but ended up being 3.2GHz due to thermal issues. Keeping other bus frequencies locked to their stock values might be difficult though it depends what Sony did with the retail units if they laser cut the Cell it's game over but if they just softlocked it (which in theory makes more sense but is common to laser cut and usually what happens) then keeping internal bus frequencies locked while increasing Cell core frequency should be possible. The PS3 isn't something the people I talk to specialise in, none of us do, but with what we do we are very familiar with the standard process of production for these kinda of things and fusing certain elements so have hammered out a few of the basics and concluded that a CPU OC should be possible with work if the Cell isn't laser cut but if there isn't a way to increase frequency with just a multiplier change then PLL dividers will be needed to keep internal bus frequencies around what they should be and that's when things get complicated, likely impossible, to accomplish an OC.
 
Ketxxx said:
I can only do intermittent testing in my spare time but I'm comfortable in saying that once you hit 650MHz core frequency further increases don't really matter all that much for the GPU the core isn't especially frequency starved. Before you'd even need to really consider going above 650MHz vRAM is going to need to be at 850MHz, every +100MHz vRAM should be accompanied by +50MHz core increase to keep bottlenecking minimal.

For vRAM I would say 800MHz is what should be aimed for to iron out the worst of the performance issues in GPU limited titles. As far as I can find out all PS3s are fitted with GDDR rated for 667\700 (evidence more heavily supports the latter) so every PS3 should be capable of 800MHz vRAM speeds you'd have to be very, very unlucky to not be able to reach that. In fairly bad cases you'll likely go from 17FPS to 24-26FPS @ 650\800. Some games are just a lost cause though like Assassin's Creed those games appear to be very CPU bottlenecked (likely down to pure and simple bad\lazy coding).

Using what data I could get access to you'd likely need to increase the Cell frequency to 4GHz to remove most of the CPU bottleneck which on paper should be possible with every Cell as it appears like the Cell originally was designed for at least 4GHz but ended up being 3.2GHz due to thermal issues. Keeping other bus frequencies locked to their stock values might be difficult though it depends what Sony did with the retail units if they laser cut the Cell it's game over but if they just softlocked it (which in theory makes more sense but is common to laser cut and usually what happens) then keeping internal bus frequencies locked while increasing Cell core frequency should be possible. The PS3 isn't something the people I talk to specialise in, none of us do, but with what we do we are very familiar with the standard process of production for these kinda of things and fusing certain elements so have hammered out a few of the basics and concluded that a CPU OC should be possible with work if the Cell isn't laser cut but if there isn't a way to increase frequency with just a multiplier change then PLL dividers will be needed to keep internal bus frequencies around what they should be and that's when things get complicated, likely impossible, to accomplish an OC.
Click to expand...
A 3.8 to 4ghz cpu clock would 100% fix tons of annoying cpu bottlenecked games, also along with the RSX OC, we could say good bye to dog shit framerates and stutters accross the board. Only the most cursed lazy port jobs will not be saved, which honestly aren't actually that many.
 
Tanzu15 said:
A 3.8 to 4ghz cpu clock would 100% fix tons of annoying cpu bottlenecked games, also along with the RSX OC, we could say good bye to dog shit framerates and stutters accross the board. Only the most cursed lazy port jobs will not be saved, which honestly aren't actually that many.
Click to expand...

It would address a good number of issues, most notably open world games should show a significant improvement (Sacred 2, GTA & AC3 to name a few examples). One of the worst performing games I have is Naughty Bear the areas the game used to have significant slowdown isn't completely gone but it is the difference of about 16FPS and 22-24FPS. That change is with quite modest clock increases - 650\750. I've just nudged up to 775 for testing as my focus is on seeing if I can get to 850MHz on the vRAM which is a significant bandwidth boost for the GPU while having a sensible GPU frequency which is plenty high enough right up to 1000MHz vRAM frequency really bottlenecking from the GPU core still wouldn't be significant at that stage.

What we really need is a tool that can read what memory modules the GPU is using, similar to GPU-Z, that way people could find out what the particular GDDR used with their GPU should be capable of by it's part number so guesswork could be eliminated. The whole reason I made a bunch of custom firmware nudging vRAM up in the smallest possible steps is exactly because of the current guesswork involved so patient people can avoid bricking their console. I'll test Lego Jurassic World later it's a bit more easily repeatable the problem areas than Naughty Bear is. In fact for testing Lego JP is especially good at visually showing if the GPU is core frequency starved as when those situations arise the game "judders" where it's exhibiting the problem as a frame pacing issue.

Temps are still perfectly good with my testing, 40% fan, maximum GPU temp I've seen in Sacred 2 was 62\63c. Focus on bandwidth, that should net you the most gain while being able to keep temps down with lower core frequencies and units quieter.

On a side note, is there a tool that allows for community game patches? There's an awful lot of games that would see massive performance improvements from nothing but one or two simple configuration changes. Eg; Sacred 2 runs like shit because the developers decided in their infinitely stupid "wisdom" that the game should render at 1080p internally which means even on a 720p display you won't see any better performance because the image is just being downscaled to the TV. Because of this Sacred 2 has VERY frequent frame rate drops to the mid teens, a simple resolution change in a config file to a true 720p rather than 1080p would see the performance be 2x-2.5x better meaning it would hold a perfect 30FPS at all times at a minimum. If you wanted to keep as high of a resolution as possible while gaining enough frames to just maintain 30FPS then something close to 900p should do that.
 
Ketxxx said:
It would address a good number of issues, most notably open world games should show a significant improvement (Sacred 2, GTA & AC3 to name a few examples). One of the worst performing games I have is Naughty Bear the areas the game used to have significant slowdown isn't completely gone but it is the difference of about 16FPS and 22-24FPS. That change is with quite modest clock increases - 650\750. I've just nudged up to 775 for testing as my focus is on seeing if I can get to 850MHz on the vRAM which is a significant bandwidth boost for the GPU while having a sensible GPU frequency which is plenty high enough right up to 1000MHz vRAM frequency really bottlenecking from the GPU core still wouldn't be significant at that stage.

What we really need is a tool that can read what memory modules the GPU is using, similar to GPU-Z, that way people could find out what the particular GDDR used with their GPU should be capable of by it's part number so guesswork could be eliminated. The whole reason I made a bunch of custom firmware nudging vRAM up in the smallest possible steps is exactly because of the current guesswork involved so patient people can avoid bricking their console. I'll test Lego Jurassic World later it's a bit more easily repeatable the problem areas than Naughty Bear is. In fact for testing Lego JP is especially good at visually showing if the GPU is core frequency starved as when those situations arise the game "judders" where it's exhibiting the problem as a frame pacing issue.

Temps are still perfectly good with my testing, 40% fan, maximum GPU temp I've seen in Sacred 2 was 62\63c. Focus on bandwidth, that should net you the most gain while being able to keep temps down with lower core frequencies and units quieter.

On a side note, is there a tool that allows for community game patches? There's an awful lot of games that would see massive performance improvements from nothing but one or two simple configuration changes. Eg; Sacred 2 runs like shit because the developers decided in their infinitely stupid "wisdom" that the game should render at 1080p internally which means even on a 720p display you won't see any better performance because the image is just being downscaled to the TV. Because of this Sacred 2 has VERY frequent frame rate drops to the mid teens, a simple resolution change in a config file to a true 720p rather than 1080p would see the performance be 2x-2.5x better meaning it would hold a perfect 30FPS at all times at a minimum. If you wanted to keep as high of a resolution as possible while gaining enough frames to just maintain 30FPS then something close to 900p should do that.
Click to expand...

There is an application for patching eboots, check out the Game Patches thread for more info https://www.psx-place.com/threads/game-patches.43706/page-3#post-389436
 
You must log in or register to reply here.

Similar threads

Back
Top