Has anybody discovered how to undervolt Cell or RSX?

[NightLess]

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?

Have you measured the VDDC voltage? Is the voltage setting too high/too low for the 3012 error code?

 

[Awjs]

I have not but I made a discovery that should've been obvious after you said, "Mullion VID table". I just assumed that every board that uses the SW syscon is rocking the ISL6326 buck controller for both cpu and gpu it makes sense from a cost perspective... But the CECHL Ver 001 uses an ONsemi ncp5399 for the CELL and it's hard too see without a complete disassembly a ISL65XX? for the RSX.

I'll see if I can get a pic of the RSX controller from a board seller if they have the time. Breaking a fresh repaste of MX-6 and LTT PTM7950 feels bad.. =(

Here's the Datasheet for the CELL
digikey.hk/htmldatasheets/production/880020/0/0/1/NCP5399.pdf

Since this board is using 2 different buck controllers, It's safe to assume the address values of 50/51 are different?

Just confirmed the RSX on the Ver-001 uses a ISL6568

Okay, after looking at both datasheets for ISL6568 and NCP5399 and cross referencing with mullions table it's a match. So both use the same address to X voltage value. Now I need to find the correct address 51 or 3111 and maybe stop being lazy and hook up some probes to the vcores...

 

[Mitsu™]

Has anyone here tried undervolting a 90nm FAT PS3 with a small underclock? Logically, it should provide more headroom for undervolting.
Of course, you'd need to find a balance to avoid impacting performance too much, playing with the core/VRAM ratio.
This might be interesting for preservation purposes.
Sacrificing a bit of performance for some more degrees lower doesn't seem too bad, especially compared to some of the crazy/risky/hazardous modifications I've seen in the PS3 scene (for poor results most of the time).
I don't have a FAT to experiment on, so I don't know. Just throwing the idea out there.

 

[NightLess]

Okay, after looking at both datasheets for ISL6568 and NCP5399 and cross referencing with mullions table it's a match. So both use the same address to X voltage value. Now I need to find the correct address 51 or 3111 and maybe stop being lazy and hook up some probes to the vcores...

If it is Sherwood then the rsx address would not be 3111, right? This is the address of Mullion. You found that Cell and rsx use different buck converter controllers. Maybe you need the VID table of the corresponding model. Don't be lazy. It is always safe to measure it, unless your seal is still there.

 

[Awjs]

Sorry, when I said, "same address," I mean VID tables, both buck converters use the same tables and after cross referencing both tables and converting from binary to hex it's a complete match to the Mullion table provided by RIP-Felix.

The command to program the Sherwood Fats are as follows..

w 50/51 with any Mullion Vcore VID table by Felix, then fix your checksum like you would do normally with a Sherwood and you're done.

I didn't check Vcore for the CELL but it's quite noticeable/obvious but I did check voltage changes for the RSX because I'm overclocked, by Default "FF" is 1.1 "3E" for my 65nm RSX.

The voltage scaling as expected for this old proccess is quite bad, for 600mhz I needed 0.9625 where as with 650mhz I need 1.0500.

Thank for all resources and thank you Nightless for the idea that maybe programing Sherwood phats are slightly different from slims due to the 6 vs 8 digit binary register VRM table difference.

I hope this helps someone else with a 65nm phat or at least a VER-001 board like mine.

Cheers,
Awjs

 

[polyscracker]

I was watching @RIP-Felix overclock tutorial on youtube and it got me interested. Not on the overclock, but the underclock, and I found this thread. I went to the github page for the VID tables and read something interesting.

Note: The default value is FF on retail consoles. Writing FF to these addresss will revert RSX Core to stock voltage.

My question is: does it also applies to the CELL? If I screw it and it doesn't boot, could I just write FF back and get to the default value it was previously there?

I'm asking it because I've got a CECHB01 console, and I would like to make it run cooler without having to resort to deliding the CELL which is very risky and I'm totally against it, unless I were getting 80ºC at very high fan speed for no reason.

 

[Awjs]

I was watching @RIP-Felix overclock tutorial on youtube and it got me interested. Not on the overclock, but the underclock, and I found this thread. I went to the github page for the VID tables and read something interesting.



My question is: does it also applies to the CELL? If I screw it and it doesn't boot, could I just write FF back and get to the default value it was previously there?

I'm asking it because I've got a CECHB01 console, and I would like to make it run cooler without having to resort to deliding the CELL which is very risky and I'm totally against it, unless I were getting 80ºC at very high fan speed for no reason.

Yes

 

[polyscracker]

Yes

Thanks. While I was waiting for a response, I reread the entire thread and found the answer on pages 3 and 4.
But confirmation is always helpful, so thank you!

I'm waiting for my UART to Serial USB converter to arrive (maybe tomorrow), and I will proceed with undervolting both the RSX and CELL.
I have a fair amount of experience with overclocking and undervolting, but only in the PC realm. However, as long as I can revert to the default values if something goes wrong, it should be fine.

If I can, I will report back the results on this thread.

 

[NightLess]

Thanks. While I was waiting for a response, I reread the entire thread and found the answer on pages 3 and 4.
But confirmation is always helpful, so thank you!

I'm waiting for my UART to Serial USB converter to arrive (maybe tomorrow), and I will proceed with undervolting both the RSX and CELL.
I have a fair amount of experience with overclocking and undervolting, but only in the PC realm. However, as long as I can revert to the default values if something goes wrong, it should be fine.

If I can, I will report back the results on this thread.

You need to correspond to the Mullion table, the Cell address is 3110, RSX is 3111, I usually measure the VDDC voltage to make sure there is no abnormal voltage.

 

[polyscracker]

You need to correspond to the Mullion table, the Cell address is 3110, RSX is 3111, I usually measure the VDDC voltage to make sure there is no abnormal voltage.

Yes, I already had the Mullion table and the addresses for both Cell and RSX saved in advance. I wrote a tutorial for myself with all the steps I've gathered from this thread, Reddit, YouTube, and the PS3dev wiki to make things easier. Thanks for confirming the information needed for my particular system. I'm hoping for the best.

-- Merged Post---

I did it! It works! No more noisy jet-like fan. My CECHB01 is finally quiet.
I'm too tired because this week the hot weather is breaking consecutive records everyday in my city. And I can't do nothing about it. I feel like I'm boiling in my room. 35ºC but the feeling of 45%! I'm not used to it. Anyway...

I modified the SYSCON to undervolt both Cell and RSX, and I got pretty good results, even with this super hot weather. I can choose between 10 to 15 cooler temps or up to 8% less noise from the fan.

I would post some pictures to show my results, but I'm having issues with the forum not openning pages or going back to the home page for no reason, and javascript errors back and fourth. I thought it was some manteinance going on, but it's has been like this for 2 days. I can't even edit my posts.

I changed the Cell voltage to 1.0v and it booted and even was gaming on it for 7 minutes and the XMB crashed, but the game kept running with the fan doing it's job. So after that, I decided to set the voltage to 1.125v and test it, and I got the results I had mentioned before. The RSX I simply set the voltage to 1.1v and got great results aswell.

I played God of War 3 for almost an hour and got no issues at all. Before the undervolt, my target temp was 75ºC on WebMAN and the cooler would ramp up to about 38% to keep it that way. It's the most demanding game I own. Now after the undervolt, I can keep the target temp at 70ºC and the fan never goes beyond 31%! It's amazing!

Also while idle on XMB, I have the same 70ºC target temp, but the fan sits at 25% to 27% after the game session ends and the temperature stabilizes.

Also also while playing a PS2 game, Shadow of the Colossus, in software mode (ps2emu checked on WebMAN so I can see the temperatures in real-time), I changed from the recommended 40% and set it at 33%. This games is very taxing and the FPS drops to 15 frames in alot of moments. Well, I've been running the game for almost an hour and the temperatures are stable at 69ºC for Cell and 59ºC for RSX. It's unbelivable.

Last but not least, I also got a 40watts reduction in power consumption (I have a kill-a-watt to measure it).

I would like to thank everyone involved in this undervolt thing because the noise of my B01 was driving me crazy. I will try to post some pictures to prove it, but other people in previous pages had similar results, and the forum's issues prevent me from using it while logged in, I don't know why. I tried another browsers and got the same results, so if pictures don't shows up, you're gonna have to take my word for it.

Spoiler: VIEW IMAGES

EDIT: it seems I solved my issues with the forum not working for me when logged in.

I have a question about an issue that I don't know if I should take precaution or not.
After I edited the SYSCON address and did all this, everytime I turn on or reset/restart my PS3, the fan spins really fast (maybe at 80%, It's really loud!) and it quickly starts to slowdown because the temperature is not high at all. I thought about editing the SYSCON fan curve, but I would like to know if this is maybe a bug with WebMAN Auto Fan Controller.

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.

Somehow it's the same thing happening to me. It's not WebMAN that went crazy with the fan. It's SYSCON. I've only changed the bytes to modify the voltage and didn't touched anything else. Could someone please shed a light on this, please? The system is working very cool, but SYSCON makes the fan spin at 80% until WebMAN takes control of the FAN, then it works as it should.

EDIT: OK. I fixed by rewriting the default COK-001/2 fan curve. Job done.

EDIT2: Adding benchmarks for PS2 game in Hardware Mode. It get's less hot compared to PS2 Software mode, as expected. I've read the opposite on another thread about the matter, so decided to test it myself. The game is Shadow of the Colossus and this spot runs at 15FPS. I left it running for 50 minutes while monitoring the temps with the SYSCON comamnds tmp 0 and tmp 1. Also you can see that the fan is set to a fixed 33% value.
This game on software mode used to make my Cell get close to 78ºC before the undervolting, and that was with a 40% fixed value for the fan.

Spoiler

EDIT3: adding one screenshot for PS2 software mode on the same conditions (1 hour with fixed fan at 33%)

Due to forums javascript errors, I can not edit my posts to save my life, so if any mod see this, the double post it's not my fault. The forums doesn't work when I'm signed in. It works when I'm not signed in though.

Anyway... I managed to find the lowest sweet spot for my B01 90NM Cell and RSX voltages.

Tested for many hours in God of War 3 and also in Shadow of the Colussus in PS2 Hardware mode, and I've got Cell at 1,1125v and RSX at 1,1v. Anything lower, and I get crashes or artifacts. The target temp at Webman is the same: 70ºC. The fan however got a little more quiet in PS3 games, it never crosses 30% compared to the 31% from my last report. And in PS2 hardware mode, I got a decrease of 2ºC after measuring for the same ammount of time than before. Really nice. I didn't test the PS2 software mode because I'm really tired for today. The ambient temperature is the same as before. HOT!

I would've upload pictures, but due to the errors with the forum, all I can do is report via text only. I hope it's OK.

Thanks mods for merging my posts. If you could merge this last one too, I would be grateful. It's becasue I went ahead and did the last test, even though I'm tired due to the hot weather.

Here is PS2 software mode with a more agressive undervolt. Just compare to before and you'll notice a 2ºC decrease in temperature at the same conditions. If by any chance the picture doesn't shows up, it says: CELL 67ºC / 57ºC / FPS 15

 

[vldru]

Hello everyone. I want to share my experience of setting custom voltages for my consoles. For cechc08 I installed cell 1E rsx 1E, cechh08 cell 24 rsx 20, with such voltages the consoles began to emit significantly less heat, they work stably in different modes. I want to note that all the consoles were revived by me, after the GPU failed, by the reflow method on a homemade soldering station.

Hello. I need help finding a voltage table for super slim. The ones in the table are not suitable, they are too big for SS.

 

[Estox]

Wow that are awesome finds. I'll test these to my fat systems today. But I got one question. What VID table you talking about? I can't find it anywhere in here i read the whole thread. Am I blind?

 

[vldru]

[QUOTE="Wow that are awesome finds. I'll test these to my fat systems today. But I got one question. What VID table you talking about? I can't find it anywhere in here i read the whole thread. Am I blind?[/QUOTE]
https://github.com/RIP-Felix/PS3-GPU-Overclock

 

[Estox]

I undervolted my CECHL04 it now bearly reaches 50°C. It maxes at 48/48 in TLOU.

to undervolt a CECHL it is a SW Syscon you need adresses 0x50 and 0x51 and use the MULLION VID table despite its a SW system.

to fix the checksum missmatch type in "eepcsum" after that type that value given endian byte swaped in adress 0x7FE so like "w 7fe xx xx". then your ps3 boots up again.

F.ex. eepcsum results in csum = 0xB883 you have to type in " w 7fe 83 b8". reboot console, done. ps3 boots again. this only counts for CECHL VER-001 boards. I don't know if the csum adress is the same with other SW revision. I'll have to find out. and do not forget to use the mullin VID table despite it is a SW ps3. it depands on the used buck controller not on the syscon revision wich table is to be used.

Update 13.04.:

I finally undervolted my super slim 28nm version. All I said for CECHL counts for her to BUT you have to use the SW VID table.

So I can sum up all fat ps3's despite it has sw or mullion syscon need to be downvolted with the hex values out of the mullion table. Starting with the slim 2500 and later you habe to use the sw vid table.

 

[gmanarte]

I undervolted my CECHL04 it now bearly reaches 50°C. It maxes at 48/48 in TLOU.

Was there any noticeable performance issues? Meaning, was there a fps drop at certain point? Any strange noticeable hardware behavior?

Now a general question, Is it possible to do a minimun undervolt with a high OC FW and have it stable? With almost same performance but better temperatures?

 

[Estox]

Does anybody know if undervolting the cell or rsx in a CECHA/B effects the voltage the EE+GS or the GS in a C/E receives?

@gmanarte

No performance differences at all same fps like before. Just running more efficiently. That is the awesomen thing about undervolting. 0 downsides only awesome xD

 

[Starship2]

Does anybody know if undervolting the cell or rsx in a CECHA/B effects the voltage the EE+GS or the GS in a C/E receives?

@gmanarte

No performance differences at all same fps like before. Just running more efficiently. That is the awesomen thing about undervolting. 0 downsides only awesome xD

It doesnt effect them as they are not linked in any ways. They have their pretty own voltage and frequency references.