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

I would just let webMan Do it's dynamic thing and keep the console at 68c. Best comprimise of fan noise and reliability.

The reason for the 90nm relibility issue is because the underfill expands just before it softens between 70-75C. The expansion causes a tensile pressure spike on the bumps, over time cracking them. That's each time the temp crosses that 70c line. Once above that temp, normal expansion forces bend and crack the bumps over time, because the underfill is more tacky than hard. It doesn't provide support to the bumps anymore.

The rule with underfill is to keep the temps significantly below that transition zone (Thermal glassification temperature, Tg). Personally I think 68C is best comprimize with fan noise, but I worry about local hotspots achiving temps higher than that. Which is the reason for the safe margin. Somthing like 65C or even 60C might be better, but achieving that is hard to do without turning the fan into a jet engine. It would seriously ruin the experiance unless you wear headphones. So a balance betwrrn noise and temps have to be struck. I think 68C is a good one.

Note. This only applies to the 90nm RSX. 65 and 40nm do not have bad underfill. They can operate upwards 75C no problem. However, when overclocking I've noticed that they really don't like temps above 65C. Above that they start artifacting, especially with higher core clocks. I'm not sure what the mechanism is there. Something to do with how temps affect parasitc impeedance and capacitance in the transitors. I need to do more research on that.
It's the same as with PCs. Temps will always affect clocks. I got another January that does 950 core 1000 memory, it will crash at 65c. But below that it stays very stable without any crashes at all. So that means without any voltage 900 is it's fully stable clock which can hit up to 70c or 75c without crashing. Typically you want your PS3 not to crash at 70c. That is the true stability test after passing the no artifacts. But with a good 45% fan speed or 50, you can keep it nice and cool. 45% to me ain't loud at all.
 
Picked up a cheap October 2010 2501A Slim which seems to have a SONY fab'd RSX

rsx40 a01 500/650 vpe:ff shd:3f [AP0026812:0:2:a:10:17:2:0:1][20:0:a:0:1:0:1][1:1:0]

Overclocked to 700/750 @ stock voltage and doing some testing in Sonic Unleashed rn, no artifacts but this system definitely needs new thermal paste aha, so soon I will delid and replace all the paste (and the tantalum caps probably) to see how far I can push this thing lol

edit: seems stable at up to 900/950 at stock voltage, 950/950 caused a crash and showed artifacting in game, going to experiment with increasing voltage once i repaste :)

Did you measure the capacitance of the original caps?
I still have the old ones, going to measure soon
 
Last edited:
@Mitsu™ @Charles_n_town


dNBgrxi.png

i just build my first own CFW with a low OC, can someone verify they got the same md5 as me?
you can use something like https://md5file.com/calculator

It's based on the:
CFW Evilnat 4.90 Cobra 8.4 [PEX] [CEX]_PS3UPDAT.PUP
With a custom OC of just:
550/650
Output name: test.pup, but it became: test.pup_4.90.CEX-PS3UPDAT.PUP
MD5 hash: 245799963D350B6384F4493F85A4FCD8



Well I personally never test this oc mod, I'm going to try only that hw side of 512 ram swap. Atm I'm only one getting rsx ram decap test and only that could build one rsx tester board without reball rsx (test with pogo pins).
Some of my test are in this link
https://s.go.ro/8lcdk9cq
Once ready I'll ask Felix to write a complete understanding for all tutorial. Is just hard to explain in each case particular how to build it.
This is just reference build, tech's with experience will understand ideas for sure.

upgrading 512MB vram or system ram? any further progress? awesome work!
 
Last edited:
@Mitsu™ @Charles_n_town


dNBgrxi.png

i just build my first own CFW with a low OC, can someone verify they got the same md5 as me?
you can use something like https://md5file.com/calculator

It's based on the:
CFW Evilnat 4.90 Cobra 8.4 [PEX] [CEX]_PS3UPDAT.PUP
With a custom OC of just:
550/650
Output name: test.pup, but it became: test.pup_4.90.CEX-PS3UPDAT.PUP
MD5 hash: 245799963D350B6384F4493F85A4FCD8





upgrading 512MB vram or system ram? any further progress? awesome work!
Those are the right settings, yes.
 
.. can someone verify they got the same md5 as me?
No, it changes between builds. Build another and it'll have a different MD5. I tried to do this when I 1st made one too, I wanted to be sure I didn't screw anything up. But they didn't match. I just had to trust the process. If you followed the steps correctly and unchecked the rebuilder options, it said it built sucessfully, then you'll be good.
 
No, it changes between builds. Build another and it'll have a different MD5. I tried to do this when I 1st made one too, I wanted to be sure I didn't screw anything up. But they didn't match. I just had to trust the process. If you followed the steps correctly and unchecked the rebuilder options, it said it built successfully, then you'll be good.

I see, thanks. I find it strange since I thought it would just swap out the clock values, like the earlier patch did. I wonder what else it's doing to make the files different on every build, @haxxxen?

I had the same error pop up as you did, but deleting Settings.xml while it was running solved it.
 
I see, thanks. I find it strange since I thought it would just swap out the clock values, like the earlier patch did. I wonder what else it's doing to make the files different on every build, @haxxxen?

I had the same error pop up as you did, but deleting Settings.xml while it was running solved it.
encryption and compression will change values every time.
 
effect of ripple/noise with higher clocks. I suspect that increasing them increases noise and ripple that makes any OC less stable, especially near the limit. FBVDDQ (VRAM voltage) on a COK-00X MB has 2 unpopulated 7343 pads for polymer bulk filtering caps (tantalum). By default there's 1. I have been wondering if populating those would improve long term reliability of the RSX. OTOH, populating those "may" allow you to clock higher too. A 90nm RSX it's not exactly reccomended, given it's thermal limitation (70c) and related reliability concerns. But for a frankie with a 65 or 40nm, the heatsink, VRM, and additional filtering pads have the potential to really push the limits.

Are all stock rsx gpu voltages the same?
Whats the default voltage for different ps3 versions? Do they vary per each unit, or model revision or die shrink?
Have you tried overvolting without increasing frequency to check stability just from increased voltage?
and chart the power draw increase, temperature increase and ripple?

Have you tried undervolting to decrease temperature? Possibly 90nm longevity increase?

Does increasing fan speed to max affect the gpu voltage stability/droop?


  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

What's the difference in artifacting, temperature and power consumption between your heavy config and stock defaults?
 
Last edited:
I bought a 2000 heatsink. It doesn't fit on the 2500. There are 2 different spacings: the heatsink to the metal chassis and the heatsink to the motherboard. These 2 spacings are offset on the 2500 compared to the 2000 so the heatsink doesn't fit. I'd be pretty hard to try and retrofit it. I'd probably be easier to swap a 3000 rsx onto a 2000 motherboard.

Are all stock rsx gpu voltages the same?
Whats the default voltage for different ps3 versions? Do they vary per each unit, or model revision or die shrink?
Have you tried overvolting without increasing frequency to check stability just from increased voltage?
and chart the power draw increase, temperature increase and ripple?

Have you tried undervolting to decrease temperature? Possibly 90nm longevity increase?

Does increasing fan speed to max affect the gpu voltage stability/droop?




What's the difference in artifacting, temperature and power consumption between your heavy config and stock defaults?
https://www.psx-place.com/threads/p...peeds-ps3-cfw-only.36801/page-133#post-394237

The core voltage varies. For overclocking, you should start at 1.10v and then go from there, usually I'll reduce it after I find the max stable frequency,
 
Last edited:
I don't see memory speeds of 775, 825, 875, 925, 975, 1025 available in the drop-down of MFW Builder.
Is it possible to add those to the list of combinations, or am I missing something?
@haxxxen

The core voltage varies. For overclocking, you should start at 1.10v and then go from there, usually I'll reduce it after I find the max stable frequency,
Can that only be obtained by physical probing + syscon + sherwood VID table?

is the voltage info in the lvl1 dump? (if not, what does VIDbin mean?)

This slim is a cech-2001a September 2009
rsx65 a06 500/650 vpe:ff shd:3f [K4A067500:1:2:d:d:4:3:8:2][3d:0:2:0:1:2:0][2:1:0]

so according to
https://www.psx-place.com/threads/p...speeds-ps3-cfw-only.36801/page-79#post-389047

this is a toshiba DS 3 VIDbin 0?
 
I don't see memory speeds of 775, 825, 875, 925, 975, 1025 available in the drop-down of MFW Builder.
Is it possible to add those to the list of combinations, or am I missing something?
@haxxxen


Can that only be obtained by physical probing + syscon + sherwood VID table?

is the voltage info in the lvl1 dump? (if not, what does VIDbin mean?)

This slim is a cech-2001a September 2009
rsx65 a06 500/650 vpe:ff shd:3f [K4A067500:1:2:d:d:4:3:8:2][3d:0:2:0:1:2:0][2:1:0]

so according to
https://www.psx-place.com/threads/p...speeds-ps3-cfw-only.36801/page-79#post-389047

this is a toshiba DS 3 VIDbin 0?
There's a build with those 25 values somewhere in this thread. I think it might've been done by @Mitsu™ .
 
I don't see memory speeds of 775, 825, 875, 925, 975, 1025 available in the drop-down of MFW Builder.
Is it possible to add those to the list of combinations, or am I missing something?
@haxxxen


Can that only be obtained by physical probing + syscon + sherwood VID table?

is the voltage info in the lvl1 dump? (if not, what does VIDbin mean?)

This slim is a cech-2001a September 2009
rsx65 a06 500/650 vpe:ff shd:3f [K4A067500:1:2:d:d:4:3:8:2][3d:0:2:0:1:2:0][2:1:0]

so according to
https://www.psx-place.com/threads/p...speeds-ps3-cfw-only.36801/page-79#post-389047

this is a toshiba DS 3 VIDbin 0?
you could try to calculate the vid from the checksum or probe the vrm but this is pointless because there is no reason to run the GPU below 1v. Set the voltage to 1.1v, find your max stable OC, then reduce the voltage to lower heat and fan noise. I think most folks would end up running between 1.03-.1.08v for a daily system.
 
There's a build with those 25 values somewhere in this thread. I think it might've been done by @Mitsu™ .
Sweet!
found link in their video description


https://www.dropbox.com/scl/fi/s6ee...1&rlkey=2y5soj4zvstrjadwd50183wae&st=w6mifwz9


my biggest hurdle right now is VshFpsCounter crashing my system. It's so unstable that I highly suggest all stability testing and benchmarking to be done separately.

is VshFpsCounter less stable on a system without an SSD?
 
Last edited:
Hello friends, are these overclocking files for evilnat? For example, can I install one of these like normal evilnat? Is there a need for a different step? My device is ps slim 30xxx. I want to go up to the limit, I guess it won't exceed 900. Can you send me the link to these files?

Is a speed like 750/900 enough to prevent my Playstation from bricking? If there is no risk at these speeds, I will leave it like this. I would be very happy if you have the evilnat 4.91 750/850 file.
 
Last edited by a moderator:
Back
Top