Has anybody discovered how to undervolt Cell or RSX?

RIP-Felix said:
No, I asked @M4j0r about that ince and while I don't clain to understamd what he told ne it was something to the effect of 'there's alot that needs to happen before that can happen.' I get the gist that it's not worh the effort, but not impossable.




The VDDR Voltage reduction is different than the VDDC core voltage we're talking about here. RSC_VDDR is the refrence voltage for the FlexIO. I think it's the RSX's interface controller. It's independently powered vs the Cell/SB FlexIO refrence voltage (YC_RC_VDDIO). I think the reasonnis so the RSX's connection in the FlexIO can be sequenced as it's brought online in a distict stage after the Cell and SB are.
Click to expand...
Yes, I was wrong. I checked and found that the voltage provided by IC6200 does not participate in the VDDC we changed.
 
I like this idea of UVOC (Undervolted Overclock). Bumping up to a modest OC on an efficient 40nm will give you like 80% of the gains due to the diminishing returns of high OCs, but leave a little headroom for an Undervolt to reduce temps. Best of both worlds. Especially when combined with a CPU UV.
 
RIP-Felix said:
I like this idea of UVOC (Undervolted Overclock). Bumping up to a modest OC on an efficient 40nm will give you like 80% of the gains due to the diminishing returns of high OCs, but leave a little headroom for an Undervolt to reduce temps. Best of both worlds. Especially when combined with a CPU UV.
Click to expand...
I did this on my pc, since it's a SFF case and I can't install a huge heatsink, it seems you've found a new target.
 
Quite the development since I made this thread. Any tutorial on how to set the voltages? Any way to do it through CFW? This is likely critical to keeping the 90NM consoles alive without swapping RSX.
 
villahed94 said:
I personally don't own a Sherwood system (..yet) but I assume the addresses must be similar (RSX address -1 ).
Will be interesting to see how later systems react to the CELL undervolting, though this is huge for Phats because this will prevent senseless kills by failed delids (ask me how I know).
Click to expand...
As i promised i threw an eye on my slim and i found the Cell VID address. it is in fact 0x50.
First i took a look at the stock value, i expected it to be FF. and thats what i found.
IMG_0323.jpeg


so i played around with some values, first i decided to feed it a low value

IMG_0324.jpeg


i let it run for about 15 minutes until the temps reach its maximum and wont raise any further (fixed it at 76% fanspeed cause im doin overvolting later)
IMG_0325.jpeg


it plateaud at 37C CPU. as a proof of concept i chosed a ridiciously low voltage next to see if it crashes and throw some errors, especially cell errors. and that was the case. i put in a value something around 0.7v and tried to boot.
IMG_0328.jpeg


aaaaaand yep. it freezed on bootscreen and on top it ylods. that could be useful data, so leets see whats in the errlog:

IMG_0329.jpeg


say hello there, we got some Cell attention and livelock errors. ignore the 3 bottom errors, they are not recent as u can obviously see from the timestamp.

so next i wanted to validate. i took a high value, could be something like 1,25v.
IMG_0330.jpeg


we are able to boot again and ive let it sit for 15 minutes again... and yep, we got significantly higher temps on cell (of course it affects RSX a little due to dissipational reasons).
IMG_0333.jpeg


so i can a 100% confirm that 0x50 is Cell VID and 0x51 is RSX VID on slim consoles. VID Tables are in fact the same for both chips @RIP-Felix
 
raidriar said:
Quite the development since I made this thread. Any tutorial on how to set the voltages? Any way to do it through CFW? This is likely critical to keeping the 90NM consoles alive without swapping RSX.
Click to expand...
While it in theory could be done through CFW I don't recommend doing so because if you apply unstable settings you'll soft-brick your console.
Everything should be done via SYSCON.
 
Tried the undervolting process on a spare CECHG04 I had.

Results:
RSX went down to 0.96V where it's happily runnin' (90nm too. Original voltage 1.20v)
CPU went down to 1.00V.

Temperature drop:
Before: 72/68 on idle
Now: 65/62 on idle.

Couldn't test a game yet since I finished it way too late (almost 4 AM, I was barely able to see the screen lol) but I'll be doing that in the following days, since I also have a A01 and a C04 waiting to arrive.
 
I've got a frankie CECHA00 with 40nm RSX and I am thinking about trying this out with reducing voltages from both CELL and RSX. I was wondering, what is the RSX_VDDC in slims/super slims with 40nm RSX? If I understood correctly, the frankie voltage mod only reduces RSX RAM chips voltage (VDDR), but not core voltages. And by changing VDDC VID values, does the VDDR voltage also change or are they separate from each other?
 
Legetzu said:
I've got a frankie CECHA00 with 40nm RSX and I am thinking about trying this out with reducing voltages from both CELL and RSX. I was wondering, what is the RSX_VDDC in slims/super slims with 40nm RSX? If I understood correctly, the frankie voltage mod only reduces RSX RAM chips voltage (VDDR), but not core voltages. And by changing VDDC VID values, does the VDDR voltage also change or are they separate from each other?
Click to expand...
From what I can gather, a 40nm RSX will request 0.95V by default. However there have been tests where it can go below the minimum VID table for Mullion (0.8V)
 
IIRC @Sampsonay got a Frankie A01 as down as 0.82-0.83V I think.

Meanwhile I'm preparing one of my freshly arrived BCs, a C04, for a downvolt on the CELL and RSX. I replaced the original 15 blade fan (which was rattling, for some strange reason) with a modified Delta KFB1412H 17 blade fan. Hope this makes it as quiet as the 15 blade Nidec and as good as the 19 blade Minebea fan.

EDIT: And CECHC04 is in tests as we speak. Lowered CELL from a blistering 1.22V to 1.13V, and the RSX from a alarmingly 1.30V (yikes!!) to 1.09V. Aiming for lower on the RSX.
 
Last edited:
I undervolted my CELL using the Mullion Vcore VID table, and for some reason my SYSCON fan control went crazy. For some reason it keeps fans at 100%. Anyone have idea why that is?

EDIT: Before anyone asks, I didn't accidentally increase voltage that the temps would skyrocket. Using Webman fan control set to 75c, the PS3 stays much cooler than before.

EDIT 2: Fixed, somehow I accidentally changed thermal config values from syscon.

I managed to undervolt frankie 40nm RSX to lowest mullion VID value (0.8375V) and my CELL is running at 1.1125V VID at the moment. I tried using lower VID for CELL and I'm pretty sure I tried 1.1 VID before and it crashed, so 1.125V might be the lowest I can go if it continues to stay stable. Also my PS3 seems to be one of the first batches ever made, as the minimum version goes to 1.00, so that might mean that the CELL fab process wasn't as good as later 90nm ones and voltage can't go as low as maybe others CECHA/Bs can. I haven't checked the real voltages with multimeter, so I am not sure what's the real voltage that's going to the chips.

Here's Uncharted 2 running with 28% fan, which is usually the speed my frankie goes up after a few minutes:

IMG_2619.JPEG



And here's 25% fan, which is much quieter than 28% and pleasant for playing games and watching movies:

IMG_2618.JPEG




For someone who wants to try undervolting and wants to use SYSCON while the console is fully reassembled you could do it like this:
IMG_2604.JPG


IMG_2608.JPEG



One thing I found out that my console makes coil whine especially when playing games. In example while playing GTA 5 the coil whine is there, but after pausing the game it goes away. Is this most likely coming from power supply or somewhere else?
 
Last edited by a moderator:
Legetzu said:
One thing I found out that my console makes coil whine especially when playing games. In example while playing GTA 5 the coil whine is there, but after pausing the game it goes away. Is this most likely coming from power supply or somewhere else?
Click to expand...
It's definetly the power supply. I've got a couple of ZSSR PSUs and one of them has especially bad coil whine. It's annoying and I had to change it. I find APS226s don't have this issue.
I'm wondering if using an APS231 on undervolted consoles is a viable alternative now.

Edit: I've also permagrounded DIAG for easy Syscon access. Are there any negative effects to this?
 
villahed94 said:
It's definetly the power supply. I've got a couple of ZSSR PSUs and one of them has especially bad coil whine. It's annoying and I had to change it. I find APS226s don't have this issue.
I'm wondering if using an APS231 on undervolted consoles is a viable alternative now.

Edit: I've also permagrounded DIAG for easy Syscon access. Are there any negative effects to this?
Click to expand...

Funny thing is that I'm quite sure I have APS-226, but not 100% sure, as I didn't check the model. It's getting 230V from the wall, if that makes any difference for it to coil whine.
 
Hi, could this be done on a SuperSlim console or is CFW a must? I get that we are changing stuff on syscon and fixing its checksum so I guess it should work, but I'm no expert on the matter.
I know that doing this on a SuperSlim is kinda overkill as those are already low power and not that hot but any thermal advantage is welcome.
 
I decided to make a account to report my findings with my CECHL it seems w 50 or 51 doesn't to anything besides make it unable to boot. Interestingly 5f works but anything else under or over say 60 or 5e will not post. Maybe I missed something here? It seems pretty straight forward and simple, auth, into w 50/51 xxValue, eepcsum into, w 7fe flipping bits, will result into a A0223012 error.

Just made a 650/875 overclock to my RSX so far stable with GTA5 and Crysis HD play through but I wanted to undervolt too... :(
 
Awjs said:
I decided to make a account to report my findings with my CECHL it seems w 50 or 51 doesn't to anything besides make it unable to boot. Interestingly 5f works but anything else under or over say 60 or 5e will not post. Maybe I missed something here? It seems pretty straight forward and simple, auth, into w 50/51 xxValue, eepcsum into, w 7fe flipping bits, will result into a A0223012 error.

Just made a 650/875 overclock to my RSX so far stable with GTA5 and Crysis HD play through but I wanted to undervolt too... :(
Click to expand...
Did you use Mullion's vid form?
 
Sherwood's reference at github.com/RIP-Felix/PS3-GPU-Overclock I haven't tried mullion's because my console was the first gen with a sherwood chip installed no?
 
You must log in or register to reply here.

Similar threads

Back
Top