PS3 PS3 Undervolting Thread

[TriForcer]

Hmm okay. I'm still in the early stages of collecting data on this but I'm reasonably certain that the VID values found in the LV1 dump can't be directly used to measure the actual voltage provided to the RSX, or at least they don't actually match the values found in @RIP-Felix 's voltage chart. I'm not sure how many people paid careful attention to that video but he says that the Voltage/hex values listed in his spreadsheet are just taken from the closest public equivalent since the controllers seem to be a Sony specific part.
Just as an example I can point to some of the ones I've measured on my consoles.
eg. I took measurements on 2 consoles of mine (a CECHA and CECHE), one measured at 1.185V and the other at 1.190V, but both have an LV1 VID of 39 which should provide 1.2250V

That was good to know. I'll measure it better on my machines in the future. Thanks for the information

My latest undervolted PS3. Haven't tried turning it on after being turned off for a few days, so I'll probably need to adjust the RSX, but it seems to be stable like this right now.
APS-226, Nidec D14F-12BS1 01H1, 12V/2,65A, 1TB HGST 7200rpm
…
Now I have tried to start up my PS3 (CE132226453-CECHA01) again after it was completely turned off for 4 days. It started up on the first try and plays TLOU without any problems, so these new value will be the final results for this machine

But I have a question about the Q6200 (the voltage regulator for RSX). It seems to be intact on this machine? Is this a problem? My other machines have this one modified with a jumper.

---------------
Now I have undervolted another machine (08-27430423-0383465-CECHC04) and this is its data.
Tested with TLOU, GT6 (loop demo mode) and GoW - Ascension.

 

[DoublesAdvocate]

That was good to know. I'll measure it better on my machines in the future. Thanks for the information

My latest undervolted PS3. Haven't tried turning it on after being turned off for a few days, so I'll probably need to adjust the RSX, but it seems to be stable like this right now.
APS-226, Nidec D14F-12BS1 01H1, 12V/2,65A, 1TB HGST 7200rpm
…
Now I have tried to start up my PS3 (CE132226453-CECHA01) again after it was completely turned off for 4 days. It started up on the first try and plays TLOU without any problems, so these new value will be the final results for this machine

But I have a question about the Q6200 (the voltage regulator for RSX). It seems to be intact on this machine? Is this a problem? My other machines have this one modified with a jumper.

---------------
Now I have undervolted another machine (08-27430423-0383465-CECHC04) and this is its data.
Tested with TLOU, GT6 (loop demo mode) and GoW - Ascension.

Interesting, your results now have me wondering if a number of the other voltages posted here were accurate or if people were just using the number in the LV1 dump and assuming it was accurate. Is that the same CECHC04 you posted about earlier in the thread running at 1.3125V?
Also nice that you added if they were under any load when you took the measurement, and the location where you measured them in the circuit, this could explain some of the small differences as well.
Based on the data so far it seems like 90nm chips with a VID of 39 (bin 2) are set to 1.2250V but after going through the mosfets and filter circuits it comes out to about 1.185-1.190V, a drop of ~0.035-0.04V
90nm chips with a VID of 16 (bin 3) are set to 1.3125V but after going through the mosfets and filter circuits it comes out to about 1.275-1.289V, a drop of ~0.0235-0.0375V
90nm chips with a VID of 1c (bin 1) are quite rare but they should be set to 1.1625V and measure out to about 1.1225-1.1275V
Based on my tests there can be a drop of about 0.001V measuring before the tokins vs after the tokins so a number of components between the regulator and the RSX could add up to explain the drops.
Part of the reason I've been curious about documenting this stuff is so people aren't confused when they're modding their console and wondering why the VID value they set through the syscon is different when they actually measure them.

 

[TriForcer]

Interesting, your results now have me wondering if a number of the other voltages posted here were accurate or if people were just using the number in the LV1 dump and assuming it was accurate. Is that the same CECHC04 you posted about earlier in the thread running at 1.3125V?

The first one I reported is not measured in the same way as the last two, but I can check the first one again when I get home (for the weekend).
I have more PS3s, but they will come as they come...

 

[DoublesAdvocate]

The first one I reported is not measured in the same way as the last two, but I can check the first one again when I get home (for the weekend).
I have more PS3s, but they will come as they come...

Can't hurt, but just confirming you took it from the LV1 is good enough for me

 

[oiamlaffin]

I've got a bunch of CECHA00 and CECHB00 that I have started to mess around with undervolting and one is really kicking my ass.

Japanese CECHA00, original 90nm RSX and CELL, back tokins replaced with tantalizers. I did not take initial voltages from the board, from the others I have messed with 1.125v (3D) has been my starting point for testing which always seems stable, just work down from there.

Had the CELL at 1.1v and the RSX at 1.0875, was playing GT6 to stress test and the console would lock up after about 1.5hrs of running. No beeps, no YLOD, no reported errors in syscon, temps stay around or under 68c webman fan curve.

Have tried 1E, 3D, and 1D, but all seem to exhibit the exact same crash.

Also tried to bump RSX voltage up and down, which makes zero difference, still hard system lock after about an hour and a half.

Guessing the CELL just needs somewhere north of 1.15v to be stable, but was curious if you guys have any insight.

 

[DoublesAdvocate]

I've got a bunch of CECHA00 and CECHB00 that I have started to mess around with undervolting and one is really kicking my ass.

Japanese CECHA00, original 90nm RSX and CELL, back tokins replaced with tantalizers. I did not take initial voltages from the board, from the others I have messed with 1.125v (3D) has been my starting point for testing which always seems stable, just work down from there.

Had the CELL at 1.1v and the RSX at 1.0875, was playing GT6 to stress test and the console would lock up after about 1.5hrs of running. No beeps, no YLOD, no reported errors in syscon, temps stay around or under 68c webman fan curve.

Have tried 1E, 3D, and 1D, but all seem to exhibit the exact same crash.

Also tried to bump RSX voltage up and down, which makes zero difference, still hard system lock after about an hour and a half.

Guessing the CELL just needs somewhere north of 1.15v to be stable, but was curious if you guys have any insight.

You can always just get the default voltages by setting a value of FF rather than something in the chart like 1E, 3D, or 1D, that way you can take measurements.
If you dump the LV1 of the RSX it will also tell you the default voltage of the chip which will either be VID 39: 1.2250V, VID 16: 1.3125V, or VID 1c: 1.1625V.
It varies from chip to chip, some just come out of the factory running better than others and there's nothing you can do to change that.
It's worth mentioning that the temperature of the console also has some influence on the ability to lower voltages, a console with a delid and full repaste running at 100% fan speed may undervolt better than one running at stock voltages with the fan speed set to something low.

 

[kraevd.pl]

Had the CELL at 1.1v and the RSX at 1.0875, was playing GT6 to stress test and the console would lock up after about 1.5hrs of running. No beeps, no YLOD, no reported errors in syscon, temps stay around or under 68c webman fan curve.

Just for statistics: CECHC04, 101 days of use, APS-227, RSX & CELL delided & repasted
CELL CXD2964GB
RSX CXD2971DGB 0711HJL, VRAM Qimonda HYB18H512322AF-14
RSX LV1 string: [4K0303201:1:3:5:6:e:4:f:2][39:2:1:1:1:3:1][0:0:0]
CELL stock voltage measured: 1.248V
CELL undervolted to 1.150V (3D). If lower - screen freezes in Last of Us long session play.
CELL real voltage measured after undervolting: 1.121V

RSX stock voltage measured: 1.195V (should be 1.225V according to VID 39)
RSX undervolted to 1.050V (21). If lower - artifacts in Gran Turismo 6 & Heavy Rain long session play.
RSX real voltage measured after undervolting: 1.029V

CELL is still significantly hotter than RSX

 

[DoublesAdvocate]

I'll add that to my database. By the way, do you happen to have any photos of the RSX in your console? I'm wondering about which model it is exactly and what VRAM it uses.

 

[kraevd.pl]

I'm wondering about which model it is exactly and what VRAM it uses.

Sure, RSX CXD2971DGB 0711HJL, VRAM Qimonda HYB18H512322AF-14

 

[Gou]

Anyone had any experience with COK-002W boards?
I just run into one, did not even know they existed until now, and apparently it's RSX runs at 1.13v stock.
Will report back with undervolt results when i have more time, just thought it was strange. Its definitely a 90nm Sony die though, with binning of 1, which also did not realize was a thing.
rsx: b03 500/650 vpe:ff shd:6f [G8B685700:1:2:19:d:d:3:f:2] [1c:1:0:0:1:3:1][0:0:0]

 

[DoublesAdvocate]

Anyone had any experience with COK-002W boards?
I just run into one, did not even know they existed until now, and apparently it's RSX runs at 1.13v stock.
Will report back with undervolt results when i have more time, just thought it was strange. Its definitely a 90nm Sony die though, with binning of 1, which also did not realize was a thing.
rsx: b03 500/650 vpe:ff shd:6f [G8B685700:1:2:19:d:d:3:f:2] [1c:1:0:0:1:3:1][0:0:0]

Holy cow you not only have the rarest COK board but also the rarest 90nm RSX!
I'm not joking when I say I only know of a 2 photos of the COK-002W motherboard.
This is also the first time I've seen a bin 1 90nm RSX in a retail console and just like I predicted a few posts up it's running at 1.13V.
Would you be able to get some high quality photos of the motherboard, RSX, and other chips on the board and send them to me?

 

[kraevd.pl]

Anyone had any experience with COK-002W boards?
I just run into one, did not even know they existed until now, and apparently it's RSX runs at 1.13v stock.

I have COK-002W 1-873-513-11 RSX running on stock voltage: 16 (1.3125V)
b08 500/650 vpe:ff shd:7b [AP0005713:1:1:11:a:12:2:f:1][16:3:1:1:1:3:1][0:0:0]
RSX CXD2971DGB 0709HJL VRAM Qimonda HYB18H512322AF-14
Is that something special?

 

[Gou]

Holy cow you not only have the rarest COK board but also the rarest 90nm RSX!
I'm not joking when I say I only know of a 2 photos of the COK-002W motherboard.
This is also the first time I've seen a bin 1 90nm RSX in a retail console and just like I predicted a few posts up it's running at 1.13V.
Would you be able to get some high quality photos of the motherboard, RSX, and other chips on the board and send them to me?

Sure i can do that. I delided without gluing the IHS back on, so it will be a bit of pain but no worries.
Perfect time to reglue them i guess

I have COK-002W 1-873-513-11 RSX running on stock voltage: 16 (1.3125V)
b08 500/650 vpe:ff shd:7b [AP0005713:1:1:11:a:12:2:f:1][16:3:1:1:1:3:1][0:0:0]
RSX CXD2971DGB 0709HJL VRAM Qimonda HYB18H512322AF-14
Is that something special?

Same board, but different RSX. Yours is bin 3 toshiba i think. I guess the fact that my console won silicon lottery is unrelated to the W in the board model then

Edit: Photos attached. Hope this helps. I did not glue the IHS because i am running low on thermal paste and i wouldnt have enough to reassemble it, but if i recall correctly the RSX part number is just a normal CXD2971DGB

 

[DoublesAdvocate]

I have COK-002W 1-873-513-11 RSX running on stock voltage: 16 (1.3125V)
b08 500/650 vpe:ff shd:7b [AP0005713:1:1:11:a:12:2:f:1][16:3:1:1:1:3:1][0:0:0]
RSX CXD2971DGB 0709HJL VRAM Qimonda HYB18H512322AF-14
Is that something special?

The board is quite uncommon, but the RSX on it is just a normal bin 3.

Sure i can do that. I delided without gluing the IHS back on, so it will be a bit of pain but no worries.
Perfect time to reglue them i guess

Same board, but different RSX. Yours is bin 3 toshiba i think. I guess the fact that my console won silicon lottery is unrelated to the W in the board model then

Edit: Photos attached. Hope this helps. I did not glue the IHS because i am running low on thermal paste and i wouldnt have enough to reassemble it, but if i recall correctly the RSX part number is just a normal CXD2971DGB

Interesting, I never noticed the 'W' is only present on one side. As for the RSX I was actually more curious about the datecode markings on it so I can try to narrow down the production date.

 

[Gou]

I will be sure to snap a photo next time i take it apart.
I have been running TLOU for 14 hours now and so far so good.
I have ended up at 1.05 on the RSX and 1.1375 on the CELL. Keep in mind that those values are from the voltage table.
I did not take it apart to measure the actual voltage. 1v on the RSX brought some light artifacts, 1.1 or lower on the CELL and the console YLODs after about 30mins.
Console is CECHC04 with datecode 7A and 218 days of operation.
It has settled at 64C on the CELL and 67C on the RSX with 29% fan speed. Depends on the room temp of course, it was going strong at 26% fan speed last night

Edit: Turns out i came to the conclusion too quick. It was very stable in TLOU, however it froze 5 minutes into GT6 demo mode. I bumped both by two/three steps just to have peace of mind. Running 1.1675 on CELL and 1.0875 on the RSX. So far so good but only time will tell

 

[DoublesAdvocate]

I will be sure to snap a photo next time i take it apart.
I have been running TLOU for 14 hours now and so far so good.
I have ended up at 1.05 on the RSX and 1.1375 on the CELL. Keep in mind that those values are from the voltage table.
I did not take it apart to measure the actual voltage. 1v on the RSX brought some light artifacts, 1.1 or lower on the CELL and the console YLODs after about 30mins.
Console is CECHC04 with datecode 7A and 218 days of operation.
It has settled at 64C on the CELL and 67C on the RSX with 29% fan speed. Depends on the room temp of course, it was going strong at 26% fan speed last night

Edit: Turns out i came to the conclusion too quick. It was very stable in TLOU, however it froze 5 minutes into GT6 demo mode. I bumped both by two/three steps just to have peace of mind. Running 1.1675 on CELL and 1.0875 on the RSX. So far so good but only time will tell

I'm guessing this is a CECHC04 right? Did you happen to get a measurement on the original CELL voltage btw?
Those temps seem quite high considering the console has been delided, repasted, and undervolted, but then again you're in Greece and it's summer so I'm guessing your room temp is probably quite high. My CECHE barely went above 50C when it was set to 35% fan speed in a 15-20C room.

 

[Gou]

Yes it's a C model. I think the difference would be because of the fan speed. I use the stock settings on webman except for minimum fan speed which is set to 27%. Room was about 27-28C when I posted those results.
The console is near silent and I don't think it went past 30%, so seems normal to me. Although the last time I messed with ps3 consoles was about 7 years ago and this was my first delis so I might be wrong.
Cell measured at 1.22v before doing anything to the console

 

[DoublesAdvocate]

Yes it's a C model. I think the difference would be because of the fan speed. I use the stock settings on webman except for minimum fan speed which is set to 27%. Room was about 27-28C when I posted those results.
The console is near silent and I don't think it went past 30%, so seems normal to me. Although the last time I messed with ps3 consoles was about 7 years ago and this was my first delis so I might be wrong.
Cell measured at 1.22v before doing anything to the console

Ah okay, a bit of extra room temperature and a bit lower fan speed would make sense for the higher temps.
Got it, I'll add that voltage to the list.