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

[Razmann4k]

yes, that was something I was going to mention. the 8th spe seems to do nothing, and it's disabled by default, so even enabling it, it has no effect on games or anything. I'm not sure of the reasoning behind that. the point being that the ps3 is capable of so much more than sony allowed it to do, but maybe cooling was a factor in their decision? I don't have a cfw ps3, but from what I understand, stability with oc often relies on cooling the system to a certain level. maybe they didn't want a loud fan, so the average, legit user didn't think something was wrong with their system. I talked with al azi about this in regards to ps4 cooling, and we both think it's fine as is. it's often people coming from the ps3, knowing how hot the system actually gets, who worry.

As Starship2 said, the 8th SPU was disabled to improve yields, as if Sony made 1000 CELLs and 200 of them had a dead SPU, by disabling 1 of them, they now have 1000 working chips even if they're all slightly weaker in theory. However, games were never compiled with that 8th SPE in mind, therefore turning it on does nothing because it's never given any code to execute, it just sits there dormant.

 

[Starship2]

Sorry for talking about older posts, but i think this is important data.

All 20xx slims cannot pass the 650 barrier in core or 800 or 850 memory. They just can't. None of them can. The silicon won't allow it. It isn't efficient and matured enough. Don't know why people try absurdly high OCs from the get go without doing proper testing while only having 1 console. But, we all do learn the hard way I suppose.

i dont think thats right. maybe you tested some very unlucky ones. i got 2 20xx slims, one slim took 700/850 and i guess it reached its maximum and the other one was way stronger. it took 750/925 first try and runs without any problems. Games tested :GTA IV, FarCry3, Crysis 3, TLOU. pretty demanding games. didnt touch voltages on these slims, but i will do this weekend and see whats possible.
its always silicon lottery and yeah, some RSX's are prone to handle lower frequencies and responding worse than others. anyways i think we dont have enough data and its too early to point at some models to say "they cant handle 700 core in general". my slims proof youre wrong. and, for gods sake, that one slim with 750/925 is a very fucked up one. it has been running a long time with high temps, has still stock thermal paste on it. and i wont change it. i wanna test it under realistic conditions to create reproducibility.

I need more 65nm console data from their max overclocks. I couldnt make a good graph of their averages with the small sample size. 650 core seems to be about the average, but 800 on VRAM should be doable on all models. Even 90nm.

confirm. my C04 does 650/800 out of the box. but it seems like it is maximum i can do on that one. my slims with 65nm (20xx) can do way better. look above. i hope this helps collecting your data, at least 3 more consoles. if you need more details feel free to ask. i plan doing some further tests this weekend. tweaking voltages. talking about that, do you know stock voltages from 90nm/65nm fabs? couldnt find anything.

 

[RIP-Felix]

Hello i need some help guys to identify some Syscon errors that iam not understanding does anyone know what 2020 and 2120 error codes on Syscon means with someone explaining them with simple words?

HDMI error. Often occurs with another, more serious error and is not of concern. We dont have a clear understanding of it's specific triggers in code. We just know how they tend to pop up.

They can occur normally from AV/HDMI issues. IDK, often a video reset clears them up. If a bunch of them pop up in one power on and you dont get video, it can be the HDMI transmitter or Digital Video encoder.

 

[Starship2]

Hi everyone.
As i said before, i have done some testing over the weekend. i decided to take the "yard condition" slim (65nm, 2004B, dyn001) that was overclocked to 750/925 with stock voltage because i thought that the potential would be the biggest at this particular one.
Spoiler: it is not.

First i had to find out whats the stock voltage (dont have my multimeter here..), so i began to start with 0,96v. Console booted, but no picture. there is only one conclusion: its undervolted. i went higher and higher until the core stabilizes and found out that 1,1v has to be the stock voltage. so now i had a reference and knew from my experience, how far i can go as i do have some older experiences from my curie graphic cards. i decided to test core frequency first, so i built a 800/900 and flashed it. i raised voltage to 1,15v (took the value for 1,165v as there are some losses due to efficiency) and it stabilized with 800/900. tested Crysis 3 and some other games and everything was running fine, no artifacts, no crashes. i was pretty positive and sure it will take a lot, so i next tested 850/900.

First boot: freeze in xmb. i raised voltage to 1,2v. still freezes, but a little bit later. so i raised voltage again to 1,22v but still kept freezing in xmb. at this point i knew, that this is gonna be a tough one as that is way too much volts for that low frequency. i knew it wouldnt help to push voltage higher, but i did. raised it again to 1,25v. still freezing. at this point it was absolutely clear, that the rsx cant take that much voltage and stabilize. i set fanspeed at 100% in fantable and as everbody expected, it didnt freeze in xmb instantly anymore. i tried to start a game, but it froze. of course it did. i raised voltage to 1,28v, but nope, it turned of with 3 beeps when a game is started. errlog says 1002 but thats what i expected. Tokins are worn out and cant provide that much power any longer.

yep, thats a bummer. GPU started so insanely high with stock voltages but just lets me add another 50 mhz...
alright. i went back to 800 core as it seems to be the maximum that chip can handle.
but... what about vram? maybe i got some more luck there. i put a 800/950 on and lowered voltage to 1,25v and yep, it boots. tested crysis 3 again for 30 minutes with 50% fanspeed and gpu temperature was at 65 Celsius. Everything was fine. so i decided to push it further. 800/975 was next. and again it perfectly boots and runs stable in crysis 3. now i had a goal. i wanted a 4 digit vram clock and flashed 800/1000. nothing special to say, it runs.

so i tested crysis 3, farcry 3 and gta 4. it does a great job. performance is very decent and there is no flickering/texture warping or artifacts. temperatures seem fine and thats what they have to be. [notice the word "cell" in that crysis scene, maybe its a little foreshadowing whats going to be overclocked next?lol]

if temps exceed 70 C it starts with artifacts but it wont crash. i decided to stop here even though i am very sure i could raise vram to 1050. but at this point its not necessary as it is way more than some of you would expect from a 65nm rsx.

whats the conclusion and what do we learn from that?
there are many things i can confirm:

1) yes, chips do have to run on low temps. 70C seems to be a magical border at slim consoles. sweet spot is 60-65 C.
2) yes, 65nm seem to be not the best choice for overclocking as they only can handle poor core frequencies
3) no, its absolutely wrong that they cant go over 650 core. "none of them can", but mine do obviously. even with stock voltages. they dont go high, thats correct. but they can definitely do 700 or in some cases 750.
4) yep, it looks like at least some of them can reach crazy vram speeds, as mine does easily 1ghz.
5) and last thing: there has to be a difference between 65nm slim and 65nm fats, as i have never seen such crazy vram speeds on fat 65nm fabs. my 65nm fat does struggle with 650core and 850 vram, but i think we need to get more data here. maybe my fat is an unlucky one, but i dont think so.

GPU: rsx65 a06 800/1000 vpe:ff shd:3f [K4A023000:1:2:15:b:3:3:8:1][3d:0:2:0:1:2:0][2:1:0]

Errlog:
Firmware Version: 4.90 (build 50747)
Platform ID: ÿÿÿÿÿÿþ›
Product Code: 00 85
Product Sub Code: 00 09
Hardware Config: 000000000203BC3C
Syscon Firmware Version: 0832.00000000083E0832 (EEPROM: 00000000083E0832)
Bringup Count: 5766, Shutdown Count: 5025
Runtime: 365 Days, 15 Hours, 31 Minutes, 14 Seconds
Error Log
01: FFFFFFFF Sat Jan 1 01:59:59 2000
02: FFFFFFFF Sat Jan 1 01:59:59 2000
03: FFFFFFFF Sat Jan 1 01:59:59 2000
04: FFFFFFFF Sat Jan 1 01:59:59 2000
05: FFFFFFFF Sat Jan 1 01:59:59 2000
06: FFFFFFFF Sat Jan 1 01:59:59 2000
07: FFFFFFFF Sat Jan 1 01:59:59 2000
08: FFFFFFFF Sat Jan 1 01:59:59 2000
09: FFFFFFFF Sat Jan 1 01:59:59 2000
10: FFFFFFFF Sat Jan 1 01:59:59 2000
11: FFFFFFFF Sat Jan 1 01:59:59 2000
12: FFFFFFFF Sat Jan 1 01:59:59 2000
13: FFFFFFFF Sat Jan 1 01:59:59 2000
14: FFFFFFFF Sat Jan 1 01:59:59 2000
15: FFFFFFFF Sat Jan 1 01:59:59 2000
16: FFFFFFFF Sat Jan 1 01:59:59 2000
17: FFFFFFFF Sat Jan 1 01:59:59 2000
18: FFFFFFFF Sat Jan 1 01:59:59 2000
19: FFFFFFFF Sat Jan 1 01:59:59 2000
20: FFFFFFFF Sat Jan 1 01:59:59 2000
21: FFFFFFFF Sat Jan 1 01:59:59 2000
22: FFFFFFFF Sat Jan 1 01:59:59 2000
23: FFFFFFFF Sat Jan 1 01:59:59 2000
24: FFFFFFFF Sat Jan 1 01:59:59 2000
25: FFFFFFFF Sat Jan 1 01:59:59 2000
26: FFFFFFFF Sat Jan 1 01:59:59 2000
27: FFFFFFFF Sat Jan 1 01:59:59 2000
28: FFFFFFFF Sat Jan 1 01:59:59 2000
29: FFFFFFFF Sat Jan 1 01:59:59 2000
30: FFFFFFFF Sat Jan 1 01:59:59 2000
31: FFFFFFFF Sat Jan 1 01:59:59 2000
32: FFFFFFFF Sat Jan 1 01:59:59 2000

i am thinking about desoldering the tokins and start a new test with tantalizers. i dont have much hope there as voltage is already too high for such poor frequency. im pretty sure i shouldnt go over 1,3v Vcore...
thanks for reading and i hope i could help.

to be continued...

//Part 2
I was hustlin' to achieve more. i wanted to squeeze every last mhz out of my slim, so i had to focus on the problem with the occuring 801002 error when putting load on the system next. i was pretty confident that my nec/tokins have aged significantly due to nearly 9000 hours of runtime. I decided to remove the bottomside tokins and throw some original sony tantalizers on (from a donor board). First i soldered some mixed 470µF & 330µF caps on it ( yeah i know mixing capacitors doesnt work out well). Just had these on my hands and i was only supposed to see if it instantly crashes again at 1.28v.

Of course, it did. i guess this is my personal pitfall. but coming to pitfalls, i remembered RIP-Felix's words about capacitance so i gave it another shot. i was thinking about the best configuration and decided to put a 2x3 330µF pattern on it. i bridged them together to make impedance as low as possible and came to 1980µF at total whats exactly the value of one tokin. so capacitance isnt the issue anymore.

now it has to work. tried 1.28v again and it runs fine in xmb. starting a game works either, but after about 20mins it shuts off and flashes red. cool. at least i could make it last longer, so i was pointing at the right problem. after this im sure that i have to get rid of every tokin. i have to do the upper side. i am so close to get it running with 850 core. its not just achieving a goal, it is troubleshooting at its best. it would be another confirmation that the whole VRM assembly seems to be a huge bottleneck. so my journey isnt over yet.

 

[kursatd24]

guys, i broke my ps3 at july 15. @Starship2 helped me so right now i can see the recovery menu but my rsx is unstable. can i change my rsx voltage from syscon via uart connection?

SM-S911B cihazımdan Tapatalk kullanılarak gönderildi

 

[LuanTeles]

My system is currently running at 1000/1275 MHz.

It's actually quite interesting to let the GPU reach 70°C. At first, a dark spot appears on the screen, then it slowly starts to spread, almost like liquid crystal leaking from an old TV.

When it hits 71°C, the icons start to become distorted and more black spots and "leaking liquid crystal" appears

Just a fan increase to keep it under 70ºC + reboot fixes everything

 

[Starship2]

My system is currently running at 1000/1275 MHz.

It's actually quite interesting to let the GPU reach 70°C. At first, a dark spot appears on the screen, then it slowly starts to spread, almost like liquid crystal leaking from an old TV.

When it hits 71°C, the icons start to become distorted and more black spots and "leaking liquid crystal" appears

Just a fan increase to keep it under 70ºC + reboot fixes everything

How much Vcore are you running? I guess your gpu is responding very quick to small changes in Vcore. If im right it would make sense to create some graphs where you can see how the average chips scale with Vcore according to Mhz. This could give us orientation whether you have a "good" chip or not. In my case it was like my gpu didnt respond to any Vcore increases in a valuable way. Your speeds are mindblowing, great result. Also very interesting how it reacts to temperature.

 

[LuanTeles]

How much Vcore are you running? I guess your gpu is responding very quick to small changes in Vcore. If im right it would make sense to create some graphs where you can see how the average chips scale with Vcore according to Mhz. This could give us orientation whether you have a "good" chip or not. In my case it was like my gpu didnt respond to any Vcore increases in a valuable way. Your speeds are mindblowing, great result. Also very interesting how it reacts to temperature.

My Vcore is default, @RIP-Felix once told me based on my LV2 string, but i don't remember now

But it's not stable; I'm just doing some tests. My stable clocks are 850/1200, which generally run well at any temperature (80ºC+ or so). However, I'm not entirely sure about the 1200 VRAM clockk. I noticed some occasional screen flickering (occoured just 3 times), which a reboot fixed. Now, though, the LEDs on my TV have burned out, so I'm not sure if the issue was with the VRAM clock or if my TV was simply on its last legs.

 

[Starship2]

My Vcore is default, @RIP-Felix once told me based on my LV2 string, but i don't remember now

But it's not stable; I'm just doing some tests. My stable clocks are 850/1200, which generally run well at any temperature (80ºC+ or so). However, I'm not entirely sure about the 1200 VRAM clockk. I noticed some occasional screen flickering (occoured just 3 times), which a reboot fixed. Now, though, the LEDs on my TV have burned out, so I'm not sure if the issue was with the VRAM clock or if my TV was simply on its last legs.

you achieved these clocks by default vcore? wow. i would never have expected that this is even possible. seems like you have an 1% sample.

 

[villahed94]

Hello again guys.
I found a game that will burn out your RSX.
This is Risen 3.
No other game loads my RSX like this game does.
Real death for FAT 90nm.

Any particular section that loads the RSX real bad? I'm trying to test the thermals and stability of an undervolted 90nm RSX.

 

[Raidmax]

Hey, it's been a while since I've been here, and I just recently came upon my 2504B, the one that I lucked out with a 5300GGB on.

What's the base OC this can do? I recall it being a high value, around 900 or so but I'd like to ask first.

Now, though, the LEDs on my TV have burned out, so I'm not sure if the issue was with the VRAM clock or if my TV was simply on its last legs.

I'd doubt a VRAM clock can burn out LEDs , likely the LEDs have given in. Although it might be offtopic, here's a hint - if it's a LG or Samsung, try reducing the backlight value to circa 35 if you plan to replace the LEDs. You'll have them last A LOT LONGER than even the intended lifespan they claim to be rated for.

 

[shnyaps]

Hey, it's been a while since I've been here, and I just recently came upon my 2504B, the one that I lucked out with a 5300GGB on.

What's the base OC this can do? I recall it being a high value, around 900 or so but I'd like to ask first.

Step by step

 

[Starship2]

Hey, it's been a while since I've been here, and I just recently came upon my 2504B, the one that I lucked out with a 5300GGB on.

What's the base OC this can do? I recall it being a high value, around 900 or so but I'd like to ask first.



I'd doubt a VRAM clock can burn out LEDs , likely the LEDs have given in. Although it might be offtopic, here's a hint - if it's a LG or Samsung, try reducing the backlight value to circa 35 if you plan to replace the LEDs. You'll have them last A LOT LONGER than even the intended lifespan they claim to be rated for.

as the other ones announced many times, start with 600/750 and go up step by step. even if you have the possibly "best sample" on paper, its not guaranteed that your particular sample can go that high.
raise core frequency first and try to find its limitations. then do vram. be advised, u should have a flasher. otherwise youre not able to unbrick, if things go bad.

 

[shnyaps]

as the other ones announced many times, start with 600/750 and go up step by step. even if you have the possibly "best sample" on paper, its not guaranteed that your particular sample can go that high.
raise core frequency first and try to find its limitations. then do vram. be advised, u should have a flasher. otherwise youre not able to unbrick, if things go bad.

flasher and checked dump)

 

[Raidmax]

Guess I'll be sticking with stock then. I don't have the means to get a flasher.

@LuanTeles - are those screencaps off a HDMI capture card? I've been meaning to get a cheap one recently but held out on the purchase.

 

[batuervan]

HDMI error. Often occurs with another, more serious error and is not of concern. We dont have a clear understanding of it's specific triggers in code. We just know how they tend to pop up.

They can occur normally from AV/HDMI issues. IDK, often a video reset clears them up. If a bunch of them pop up in one power on and you dont get video, it can be the HDMI transmitter or Digital Video encoder.

I'm sorry, but I remember you every time I see this cat food. Thank you for all your efforts for Psx place.

https://ibb.co/PjX4wq9

 

[RIP-Felix]

Guess I'll be sticking with stock then. I don't have the means to get a flasher.

@LuanTeles - are those screencaps off a HDMI capture card? I've been meaning to get a cheap one recently but held out on the purchase.

1. Leave temps at stock. Disable webman while OCing. The higher temps will warn you of instability sooner.
2. Starting at 600/800 is pretty safe.
3. Then step by step thereafter, testing in crysis HD each time. I found like others did, its opening jungle crawl scene in the dark is actually really good for spotting early signs of instability. Play to the first sunrise and view of the ocean.
4. At the first signs of instability, back off one and your're done.

If it bricks or isn't able to update...

1. You may be able to log into webman mod over IP, on a browser on the same network. Even if you can't see anything onscreen it often is booted and running in the background. Boost fans to 95% using that.
2. If webman is unreachable, use SYSCON UART to crank fans to 100% changing the first fan steps fan percentage to FF. The cooler temps at boot will keep it stable in safemode longer, usually allow you to recover in time before it becomes too unstable and freezes. You might have to retry a few times.
3. You can try cooling the console outside early in the morning. Colder air might help too. Like in front of an open window. A freezer is likely to cause condensation, so I dont reccomend that.

Following this procedure makes it very unlikly you'll brick. But it's not guarinteed.

 

[Raidmax]

1. You can try cooling the console outside early in the morning. Colder air might help too. Like in front of an open window. A freezer is likely to cause condensation, so I dont reccomend that.

Getting it cooled down in emergency cases won't be an issue. I have a few 4pin Delta fans out of a dead Fujitsu-Siemens server that blow upwards of 4000RPM, and are between 8 and 12cm in size.

Worst case scenario, my balcony is close to a freezer in the early morning.

I started around 800 core however, given the SYSCON log was pretty barren (2x 2022 errors and a lone 21xx error that god knows where it came from.) which tells me this wasn't much of a used console (despite the case being in DIRE need of a painting session) so far nothing has crashed and I'm sitting around 68 on the GPU. I did use a GTA5 disc however, due to my PC HDD's space being almost full and not being able to get Crysis HD in there.

I'm around 900/950 at the moment. The last OC'd slim I had sold, a 650/800 I think, fared pretty well.

Later edit: Passed on 900/950 with flying colors. I don't think I've seen GTA 5 this responsive and smooth on the PS3.

 

[cha0shacker]

Anyone able to produce an update file from the current EvilNat Beta with 700/925?

 

[Charles_n_town]

Anyone able to produce an update file from the current EvilNat Beta with 700/925?

you can do it yourself by using mfw. use the bat to add in the keys for 4.91. then, open the mfw exe as administrator, select the 4.91 pup, output can be named whatever, select the tarbells as 4.xx cex, deselect both rebuilder options, select rsx, choose what you want, then build.