Mac OS 9 Lives
Mac OS 9 Discussion => Hardware => CPU Upgrades => Topic started by: Protools5LEGuy on August 13, 2014, 07:30:58 PM
-
Old article about unleashing the 166 bus on first MDDs
http://www.macbidouille.com/articles/89/page1 (http://www.macbidouille.com/articles/89/page1)
http://www.macbidouille.com/news/2002/10/08#3610 (http://www.macbidouille.com/news/2002/10/08#3610)
http://translate.google.com/translate?u=http%3A%2F%2Fwww.macbidouille.com%2Fniouzcontenu.php%3Fdate%3D2002-10-08%233610+&langpair=fr%7Cen&hl=en&ie=ISO-8859-1&prev=%2Flanguage_tools (http://translate.google.com/translate?u=http%3A%2F%2Fwww.macbidouille.com%2Fniouzcontenu.php%3Fdate%3D2002-10-08%233610+&langpair=fr%7Cen&hl=en&ie=ISO-8859-1&prev=%2Flanguage_tools)
Many thanks to Amber's eyes who compared the 2 computers under a magnifying glass for 2h30, Pimus for lending us a brand new G4 Bi-1GHz and Bakayo for his incredible welding skills.
Here are the modifications needed to change the G4 Bi-867's bus clock from 133 to 167MHz.
You'll note that changing the bus speed to 167MHz would overclock the CPU to over 1GHz. To prevent this, you'll have to change the CPU's multiplier coefficient from x6.5 to x6.
WARNING : for your computer to remain stable, you have to use DDR-SDRAM PC2700 modules. The usual PC2100 modules are not certified for working at these kinds of frequencies.
Note that to get the bus from 133 to 166 Mhz, overclockerait machine more than 1 Ghz. To avoid this, you must retake the multiplier of 6.5x to 6x (see this news )
Watch out !!!!!
To ensure the stability of the machine after changes, it is imperative to have only DDR-SDRAM PC 2700! The strips are not recommended 2100 !!!
Changing bus takes 2 steps:
1) changes on the motherboard (the Word is Amber)
the photo is the underside of the motherboard.
I surrounded the three resistors in question.
They are in order R677, R678 and R676
(http://files.macbidouille.com/mbv2/news/photonews11/867133167.jpg)
(http://files.macbidouille.com/mbv2/news/photonews11/res133166big.jpg)
a bus for 133mhz: there is a resistor at each location
for a bus to 167MHz: you must turn in the R676
(It's about 100% on because I had 2:30 to scan both cards under the microscope and that's the only difference)
So there must be other possibilities (I have not taken the time to try and ram no longer lends itself to such practices)
Apple has added a second protection on Bi 867 cards If you just change the settings on the motherboard only does it happen in 167 Mhz on the condition to map a 1Ghz or 1.25 bipro Ghz.
So I'll turn it back to Amber:
So on the map girl pictured, there is a series of three resistors
numbered as follows.
Bus 167 Mhz
R27 nothing
R22 nothing
Resistor R26
Bus 133 Mhz
R27 nothing
Resistor R22
Resistor R26
So it is a security or a lock over Apple
(http://files.macbidouille.com/mbv2/news/photonews11/bus167cartefille.jpg)
In this screenshot, you see the bus at 167 MHz. The processor board is that of a true dual 1Ghz, not Overclock.
(http://files.macbidouille.com/mbv2/news/photonews11/preuve167.jpg)
-
On Mac-Forums one guy make this...
http://www.mac-forums.com/forums/apple-desktops/164289-overclocking-dual-g4-mdd-867mhz-l3-cache-info.html (http://www.mac-forums.com/forums/apple-desktops/164289-overclocking-dual-g4-mdd-867mhz-l3-cache-info.html)
I have succesfully overclocked Dual G4 867mhz machine by changing the FSB from 133mhz to 166mhz, I removed the two resistors, R676 and R22. I have tested that the machine is stable in Xbench 1.3, Geekbench and in normal usage. G4 chips in my machine are RX933PC so they are rated for 933mhz and now working at 1080mhz.
But, I looked up the Samsung L3 cache chips on my processor card and as far as I know, they are rated for 250mhz (-HC25). L3 cache in MDD models with 1MB of it operate at 4:1 ratio, meaning that the cache in my machine is already operating out of spec, at 270mhz.
Is there any software which I can use to verify the L3 cache speed? Xbench displays incorrect speed, ~4Ghz. I wonder if there is a way to modify L3 cache ratio to 5:1 so I could possibly overclock even further, to 1.25 Ghz? Probably I could just try and see if it can take it, but I would like hear if anyone in here has ever done overclockin for their Dual MDD 867mhz machines and was it how succesful?
-
Here is the "bible" of MDD cpu modules overclocks
http://web.archive.org/web/20110722181011/http://bitsandpieces.info/Multipliers.htm (http://web.archive.org/web/20110722181011/http://bitsandpieces.info/Multipliers.htm)
(http://web.archive.org/web/20110722181011im_/http://bitsandpieces.info/images/Boardfront-small_001.jpg)
This section refers to 1420MHz modules only :-[
Below is the multiplier table for the 3-Phase LTC3732CG - Note the default voltage for the 1420MHz module is 1.60V
(http://web.archive.org/web/20110722181011im_/http://bitsandpieces.info/images/1420VIDtable.jpg)
-
http://xlr8yourmac.com/systems/g4_mirrored_drive_doors/G4_MDD_CPU_Module.html (http://xlr8yourmac.com/systems/g4_mirrored_drive_doors/G4_MDD_CPU_Module.html)
Front PLL_CFG array
(http://xlr8yourmac.com/systems/g4_mirrored_drive_doors/PLL_CFG_Top.gif)
Back PLL_CFG array
(http://xlr8yourmac.com/systems/g4_mirrored_drive_doors/PLL_CFG_Back.gif)
Voltage regulator
(http://www.xlr8yourmac.com/systems/g4_mirrored_drive_doors/VID.jpg)
-
http://translate.google.com/translate?sl=auto&tl=en&js=n&prev=_t&hl=en&ie=UTF-8&u=http%3A%2F%2Fwww.macbidouille.com%2Farticles%2F89%2Fpage2&act=url (http://translate.google.com/translate?sl=auto&tl=en&js=n&prev=_t&hl=en&ie=UTF-8&u=http%3A%2F%2Fwww.macbidouille.com%2Farticles%2F89%2Fpage2&act=url)
Anyone willing to take 133 bus MDD to a higher ground?
-
http://translate.google.com/translate?sl=auto&tl=en&js=n&prev=_t&hl=en&ie=UTF-8&u=http%3A%2F%2Fwww.macbidouille.com%2Farticles%2F89%2Fpage2&act=url (http://translate.google.com/translate?sl=auto&tl=en&js=n&prev=_t&hl=en&ie=UTF-8&u=http%3A%2F%2Fwww.macbidouille.com%2Farticles%2F89%2Fpage2&act=url)
Anyone willing to take 133 bus MDD to a higher ground?
i have one.. but im afraid :o :o :o
after i destroyed my quicksilver 867mhz i am not very brave with hardware mods :D
-
It is more difficult to pass from 166 to 172 Mhz. I've read the instructions once a year since 2003, and I'm still don't know for sure what is the quartz oscillator that must be changed (there isn't even a picture ) and where I can buy it...
-
i have one.. but im afraid :o :o :o
after i destroyed my quicksilver 867mhz i am not very brave with hardware mods :D
To summarize all this post for ours MDD 866 there are 2 things to be done. The board is 133 bus and the cpu is 133bus...
On the motherboard back side
(http://files.macbidouille.com/mbv2/news/photonews11/867133167.jpg)
(http://files.macbidouille.com/mbv2/news/photonews11/res133166big.jpg)
for a bus to 167MHz: you must turn in the R676
And you have a 166 bus MDD board (heaven in OS9 I guess). This needs DDR RAM at 333 instead of 266 (?) so the original memory could not go 333.
You could buy a 166 MDD cpu or mod the same 866 to be working on 166 bus.
(http://files.macbidouille.com/mbv2/news/photonews11/bus167cartefille.jpg)
Bus 167 Mhz
R27 nothing, R22 nothing, Resistor R26
Bus 133 Mhz
R27 nothing,Resistor R22, Resistor R26
The thing that everybody talks about is that the 866 at 166 make fail the cache of the cpu,
I have succesfully overclocked Dual G4 867mhz machine by changing the FSB from 133mhz to 166mhz, I removed the two resistors, R676 and R22. I have tested that the machine is stable in Xbench 1.3, Geekbench and in normal usage. G4 chips in my machine are RX933PC so they are rated for 933mhz and now working at 1080mhz.
But, I looked up the Samsung L3 cache chips on my processor card and as far as I know, they are rated for 250mhz (-HC25). L3 cache in MDD models with 1MB of it operate at 4:1 ratio, meaning that the cache in my machine is already operating out of spec, at 270mhz.
and also is recommended to take down the multiplier from 6.5 to 6.
From bits and multipliers...
(http://web.archive.org/web/20110722181011im_/http://bitsandpieces.info/images/Boardfront-small_001.jpg)
(http://web.archive.org/web/20110722181011im_/http://bitsandpieces.info/images/Boardrearcopy-small_001.jpg)
Step 3. The table below outlines the PLL_EXT and PLL_CFG[0-3] settings for different bus and CPU speeds. For this example I am starting with a stock 1250MHz speed (7.5x multiplier) setting and upgrading to 1500MHz (9x multiplier) setting. The first number in the fields of the first column represents the PLL_EXT resistor and the following 4 numbers represent PLL_CFG[0], PLL_CFG[1], PLL_CFG[2] and PLL_CFG[3] resistors.
Example: 167MHz system bus and 1250MHz (7.5x multiplier) CPU speed is represented as:
PLL_EXT = 0 (resistor or bridged)
PLL_CFG[0] = 0 (resistor or bridged)
PLL_CFG[1] = 0 (resistor or bridged)
PLL_CFG[2] = 0 (resistor or bridged)
PLL_CFG[3] = 1 (no resistor or open)
Our target speed is 1500MHz (9x multiplier) - This is represented as:
PLL_EXT = 1 (no resistor or open)
PLL_CFG[0] = 0 (resistor or bridged)
PLL_CFG[1] = 1 (no resistor or open)
PLL_CFG[2] = 1 (no resistor or open)
PLL_CFG[3] = 1 (no resistor or open)
There in http://web.archive.org/web/20110722181011/http://bitsandpieces.info/Multipliers.htm (http://web.archive.org/web/20110722181011/http://bitsandpieces.info/Multipliers.htm) is a table that says that our 866 on 166 would be 1083Mhz with
Search for the "PLL Multiplier Config + Actual Speed on given bus" table.
Try to buy the smallest soldering iron you can buy. For the JBC solders here in Europe you can find very thin tips for.
We have to be brave to live this era ;)
-
It is more difficult to pass from 166 to 172 Mhz. I've read the instructions once a year since 2003, and I'm still don't know for sure what is the quartz oscillator that must be changed (there isn't even a picture ) and where I can buy it...
The instructions from bits and pieces, macbidoule, xlr8yourmac? What is the source of that info?
From bits... there are 3 zones to look out :
a)PPL CFG (0-3)+ PPL EXT on processor side
b)PPL CFG (0-3)+ PPL EXT on back side
c)VID (0-4) on back side
(http://web.archive.org/web/20110722181011im_/http://bitsandpieces.info/images/Boardfront-small_001.jpg)
(http://web.archive.org/web/20110722181011im_/http://bitsandpieces.info/images/Boardrearcopy-small_001.jpg)
From macbidoule there is also a zone to brake cpu to 133
(http://files.macbidouille.com/mbv2/news/photonews11/bus167cartefille.jpg)
-
It is more difficult to pass from 166 to 172 Mhz. I've read the instructions once a year since 2003, and I'm still don't know for sure what is the quartz oscillator that must be changed (there isn't even a picture ) and where I can buy it...
I guess the quartz oscillator is the same than the voltage generator (?), the chip on the VID (0-4). Maybe if you share your quartz oscillator instructions we can find out if it isn`t...
My references are here in this post ALL. Any more info, please, share here.
Is there any technician with skills to pull out the RAM of the CACHE of the cpu daughterboard and change it? Maybe is easier that what we thought to go 2 Mb cache.
In the On Mac-Forums one guy make this...
http://www.mac-forums.com/forums/apple-desktops/164289-overclocking-dual-g4-mdd-867mhz-l3-cache-info.html (http://www.mac-forums.com/forums/apple-desktops/164289-overclocking-dual-g4-mdd-867mhz-l3-cache-info.html)
I have succesfully overclocked Dual G4 867mhz machine by changing the FSB from 133mhz to 166mhz, I removed the two resistors, R676 and R22. I have tested that the machine is stable in Xbench 1.3, Geekbench and in normal usage. G4 chips in my machine are RX933PC so they are rated for 933mhz and now working at 1080mhz.
But, I looked up the Samsung L3 cache chips on my processor card and as far as I know, they are rated for 250mhz (-HC25). L3 cache in MDD models with 1MB of it operate at 4:1 ratio, meaning that the cache in my machine is already operating out of spec, at 270mhz.
Is there any software which I can use to verify the L3 cache speed? Xbench displays incorrect speed, ~4Ghz. I wonder if there is a way to modify L3 cache ratio to 5:1 so I could possibly overclock even further, to 1.25 Ghz? Probably I could just try and see if it can take it, but I would like hear if anyone in here has ever done overclockin for their Dual MDD 867mhz machines and was it how succesful?
there are references about the 1Mb cache of the 866 dual, limited to 250 MHz operating at 4:1 ratio working out of specs... What Samsung cache chips take the 1.42, 1.33 and 1.25 "golden duals". Is out there ANY reference of swaping the cache chips?
-
from what i have read about pc overclocking..
this is totally possible
that the cache memory cannot reach the proper speed
this is oftenthe biggest factor in overclocking
which is why pc overclockers always always want the craziest fast
ram so they can achieve higher multipler
so it makes sense that the cache ram is whats holding it back
plus.. i also read that the quicksilver 733mhz cpu is very popular for overclocking
because it lacks the cache ram entirely.. and therefore it can be overclocked to 933mhz
very easily!
https://translate.google.com/translate?sl=fr&tl=en&js=y&prev=_t&hl=en&ie=UTF-8&u=https%3A%2F%2Fsites.google.com%2Fsite%2Foverclockmac%2F&edit-text=&act=url
https://translate.google.com/translate?sl=fr&tl=en&js=y&prev=_t&hl=en&ie=UTF-8&u=https%3A%2F%2Fsites.google.com%2Fsite%2Foverclockmac%2F&edit-text=&act=url
as we can see in the attach document changing a 733 qs cpu is the only overclock for mac
that u can do by cutting a single trace to a resistor going from 733 to 800
or by moving the resistor u can make it a 867 or a 1067
to make it a 933 or 1000 u need an extra resistor from some place to add
this seems like the best option to get a 1ghz single cpu quicksilver!!
-
from what i have read about pc overclocking..
this is totally possible
https://translate.google.com/translate?sl=fr&tl=en&js=y&prev=_t&hl=en&ie=UTF-8&u=https%3A%2F%2Fsites.google.com%2Fsite%2Foverclockmac%2F&edit-text=&act=url
Thanks for the link!
Someone tried too to put in order the bits article
https://translate.googleusercontent.com/translate_c?act=url&depth=1&hl=en&ie=UTF8&prev=_t&rurl=translate.google.com&sl=fr&tl=en&u=https://sites.google.com/site/overclockmac/g4mdd&usg=ALkJrhj7-0rGMVdKifn347AFqLocMO3unQ (https://translate.googleusercontent.com/translate_c?act=url&depth=1&hl=en&ie=UTF8&prev=_t&rurl=translate.google.com&sl=fr&tl=en&u=https://sites.google.com/site/overclockmac/g4mdd&usg=ALkJrhj7-0rGMVdKifn347AFqLocMO3unQ)
At least I can copy that table from there...
(https://sites.google.com/site/overclockmac/tablemddpll.jpg)
!!! Warning !!!
The presence of a resistance equivalent to a "0" in this table and the lack of resistance to "1
-
im reading here.. that its possible to take a sawtooth to 133mhz bus by openfirmware only???
am i reading right?
http://www.xlr8yourmac.com/G4ZONE/sawtooth/SawtoothCPUdesign.html
as long as u have pc133 ram..and mac os 9.1+
+ openfirmware hack
= 600mhz g4 from a 450mhz
:o
-
https://translate.google.com/translate?sl=fr&tl=en&js=y&prev=_t&hl=en&ie=UTF-8&u=https%3A%2F%2Fsites.google.com%2Fsite%2Foverclockmac%2F&edit-text=&act=url
https://translate.google.com/translate?sl=fr&tl=en&js=y&prev=_t&hl=en&ie=UTF-8&u=https%3A%2F%2Fsites.google.com%2Fsite%2Foverclockmac%2F&edit-text=&act=url
as we can see in the attach document changing a 733 qs cpu is the only overclock for mac
that u can do by cutting a single trace to a resistor going from 733 to 800
or by moving the resistor u can make it a 867 or a 1067
to make it a 933 or 1000 u need an extra resistor from some place to add
this seems like the best option to get a 1ghz single cpu quicksilver!!
Taking the MDD866 dual to 10xxMHz is just pull off two resistors too...
The QS 733 chip even overclocked to 1067 should be "slower" than a DA 733 due to the 2Mb cache...
What about to "exchange" the cache chips? It is micro tech, not nano tech. ;D :D Any TV Service repair shop should have the electric soldering iron to desolder the cache...What is unknown is if the "new" cache would be a waste of time and money. Maybe not all cpu daugtherboard can take advantage of the chips installed. Some cpus work despite having the cache out of specs/broken
-
im reading here.. that its possible to take a sawtooth to 133mhz bus by openfirmware only???
no.
am i reading right?
http://www.xlr8yourmac.com/G4ZONE/sawtooth/SawtoothCPUdesign.html
as long as u have pc133 ram..and mac os 9.1+
+ openfirmware hack
= 600mhz g4 from a 450mhz
:o
this CPU at 600MHz will not even boot.
-
Of course the cache chips can be changed, even the G4 chips... it cost between 100 and 200 EUR. But why? I have not ever found a synthetic or real test that shows some benefice of 2MB of L3 against 1MB of L3. My last try was with the xServe CPU...
-
I've pushed a [email protected] dual module up to 1.750 in to a MDD. It was the last time I felt a computer FAST: Pure speed.
-
I've overclocked some PC also, the basis are the same, but very different CPUs and motherboards, so it is different technic.
-
Of course the cache chips can be changed, even the G4 chips... it cost between 100 and 200 EUR. But why? I have not ever found a synthetic or real test that shows some benefice of 2MB of L3 against 1MB of L3. My last try was with the xServe CPU...
BUT Real World tests and DAW uses show that a 2Mb of L3 make the CPU to have more inertia, more power. I guess this is mainly because in a CPU cycle the CPU can see 1 or 2 Mb. On x86 computer is clear that a CPU with more cache is more "powerful". Maybe the synthetic tests you have made are simply not aware of this. I thought cinebench would...
The geekbench test is well know that has nothing to see with the real world tests.
Final Cut Pro 3.0 test the system in the first run to know how many video layers/ rtime effects can be run at the same time. (Yes, I know it can be hacked) ... http://www.xlr8yourmac.com/tips/final_cut_pro_3_tweaks/final_cut_pro_3_tweaks.html (http://www.xlr8yourmac.com/tips/final_cut_pro_3_tweaks/final_cut_pro_3_tweaks.html)
Logic >4 take advantage of the 2nd processor and the Altivec if found.
Chris had come with the perfect example. Compare the DA 733 with the QS 733 performance with cinebench or geekbench or OS9benchwhateveriscalled.
-
in most of your mdd overclocks u are probably using pc3200 ddr 400 ram right?
the machines were originally made for pc2100 ddr266 or pc2700 ddr-333 ram
(2100 for the first mdds i beleive)
which is why using pc3200 ram allows the cpu to go higher..
u have this room for overclocking to be successfull ..
because the ram can match the increased frequency
on a g4 if u use pc100 ram.. theres no way overclock is gonna work
people who have shitty ram and try overclocking will fail
well really its not that the ram is shit.. it just isnt rated high enough to match the territory
that u push teh cpu to.. its "off the map" of what it can do
-
I've pushed a [email protected] dual module up to 1.750 in to a MDD. It was the last time I felt a computer FAST: Pure speed.
:o
You say that just to create ENVY. Are you sure M.A.R.L. do not accept Paypal? I see the same business that Apple is the master: To show a incredible product that every user has to say "I want one".
Did you use a fridge as a box for the MDD@1750? I think the case would melt down... ;D
-
in most of your mdd overclocks u are probably using pc3200 ddr 400 ram right?
the machines were originally made for pc2100 ddr266 or pc2700 ddr-333 ram
(2100 for the first mdds i beleive)
which is why using pc3200 ram allows the cpu to go higher..
u have this room for overclocking to be successfull ..
because the ram can match the increased frequency
on a g4 if u use pc100 ram.. theres no way overclock is gonna work
people who have shitty ram and try overclocking will fail
well really its not that the ram is shit.. it just isnt rated high enough to match the territory
that u push teh cpu to.. its "off the map" of what it can do
No at all. RAM speed only has to be taken in to account if you change the main buss speed, not the CPU speed.
-
I've pushed a [email protected] dual module up to 1.750 in to a MDD. It was the last time I felt a computer FAST: Pure speed.
:o
You say that just to create ENVY. Are you sure M.A.R.L. do not accept Paypal? I see the same business that Apple is the master: To show a incredible product that every user has to say "I want one".
Did you use a fridge as a box for the MDD@1750? I think the case would melt down... ;D
The main MDD fan was connected directly at 12v. Really.
LOL
-
No at all. RAM speed only has to be taken in to account if you change the main buss speed, not the CPU speed.
bus speed affects system ram yes u are correct
and cpu increase affect cache ram
Of course the cache chips can be changed, even the G4 chips... it cost between 100 and 200 EUR. But why? I have not ever found a synthetic or real test that shows some benefice of 2MB of L3 against 1MB of L3. My last try was with the xServe CPU...
this is why the man in the article ptleguy was looking at
was talking about taking off his cache ram + replacing them
with higher quality chips.. not larger sized chips.
i cant find the link but i remember reading something about 270mhz?
and the quality of his cache chips.. and how he looked up the spec of the ram chips
-
haha, resistors under the microscope ... you guys are so anal with your hardware ... just like i am with my DSP code. :)
-
Chris had come with the perfect example. Compare the DA 733 with the QS 733 performance with cinebench or geekbench or OS9benchwhateveriscalled.
isnt the 733mhz DA alot better then the QS?
the QS has no cache... but 733 DA does have it , i think?
diehard was the one who told me the 733DA kicked the QS 733's ass
-
I'm talking of 1MB of L3 against 2MB of L3, I'm *Not* talking about having L3 against not having it. In the G4 7450 the pipelining was increased from 4 to 7 stages, so L3 caché is very important here, but 1 or 2 MB? I never could find a difference.
The QS 733 was a marketing downgraded hardware as the MDDs with 133Mhz Bus.
-
I'm talking of 1MB of L3 against 2MB of L3, I'm *Not* talking about having L3 against not having it. In the G4 7450 the pipelining was increased from 4 to 7 stages, so L3 caché is very important here, but 1 or 2 MB? I never could find a difference.
The QS 733 was a marketing downgraded hardware as the MDDs with 133Mhz Bus.
http://www.everymac.com/systems/apple/powermac_g4/specs/powermac_g4_733.html (http://www.everymac.com/systems/apple/powermac_g4/specs/powermac_g4_733.html)DA
Introduced January 9, 2001 disc July 18, 2001 cost US$3499 with superdrive and 3000 with CD
http://www.everymac.com/systems/apple/powermac_g4/specs/powermac_g4_733_qs.html (http://www.everymac.com/systems/apple/powermac_g4/specs/powermac_g4_733_qs.html)QS
Introduced July 18, 2001 at US$1699
You pay in feb 2001 3500$ for a 733 and in august 2001 1700 for a 733 HALF!
But the cpu of both 733 cant be more night and day!
1Mb of L3 cache vs no L3 cache on QS
733 DA Geekbench 2 =430-450 http://browser.primatelabs.com/geekbench2/2446079 (http://browser.primatelabs.com/geekbench2/2446079)
733 QS Geekbench 2 =400-405 http://browser.primatelabs.com/geekbench2/1360228 (http://browser.primatelabs.com/geekbench2/1360228)
Dawn, it is difficult to find test on same OS...
DA 733 was top of the line while QS 733 was low end of line. Same speed 6 month later. I guess some buyer of the DA 733 could have rage over Apple... They paid 3000-3500$ for a machine that was 1700 later...
-
when the QS 733 came out, the germans keyboards magazine came out with a test of reverb plug-ins and the QS 733 was not faster than the 533 for most reverbs tested. for the steinberg reverb-32 the 733 was even slower compared to the 533 and dual 500 G4s.
you can find noticable differences between having cache vs. none also for soft GL (diablo 2) and G4 optimized fast fourier transform (such as fftw, some max objects or vst plug-ins), and also for the one or other image editing process.
-
The main MDD fan was connected directly at 12v. Really.
LOL
is it depending on which pins u draw power from on a 4pin molex that determines the voltage to be 12v vs 5v ?
pic attached is of a 12v connection to illustrator the 4pin molex connector is using only the right most 2 pins
could i use this technique to up the voltage of my fans?
-
is it depending on which pins u draw power from on a 4pin molex that determines the voltage to be 12v vs 5v ?
Yes it is.
pic attached is of a 12v connection to illustrator the 4pin molex connector is using only the right most 2 pins
could i use this technique to up the voltage of my fans?
If the fans have variable speed usually receives variable voltage. To increase the voltage of this kind of fans 5v usually is summed (ie: in parallel= variable voltage + 5 volt ) ... but be careful! ...
-
If the fans have variable speed usually receives variable voltage. To increase the voltage of this kind of fans 5v usually is summed (ie: in parallel= variable voltage + 5 volt ) ... but be careful! ...
From a Molex connector
(http://www.overclock.net/content/type/61/id/1234872/width/350/height/700/flags/LL)
You can take (without a potentiometer) +5 (red black), +12 (yellow black) and +7 (12-5 red yellow)
-
Just found some more pictures about the 166 bus unlock
http://forums.macrumors.com/showpost.php?p=18789514&postcount=1 (http://forums.macrumors.com/showpost.php?p=18789514&postcount=1)
(http://forums.macrumors.com/attachment.php?attachmentid=461614&stc=1&thumb=1&d=1392831515)
(http://forums.macrumors.com/attachment.php?attachmentid=461615&stc=1&thumb=1&d=1392831515)
Full post http://forums.macrumors.com/showthread.php?p=18789514#post18789514 (http://forums.macrumors.com/showthread.php?p=18789514#post18789514)
-
I'm talking of 1MB of L3 against 2MB of L3, I'm *Not* talking about having L3 against not having it. In the G4 7450 the pipelining was increased from 4 to 7 stages, so L3 caché is very important here, but 1 or 2 MB? I never could find a difference.
I agree with Mactron… 1 MB vs 2 MB did not yield as big of a difference as other factors; unfortunately, when my company was testing all of these G4 systems in a real world DAW environments, we did not have an exact platform (G4 model) that only differed by cache. We standardized on the QS933 with 2 MB cache for some client's needs and the single MDD 1.25 with 1 MB for others. The MDD was clearly the performance winner in the VST arena, but there were other factors like faster PATA and memory architecture, I think the 2 MB on the 933 probably help a smidgin, but again we chose the MDD for the powerhouse and the QS 933 was used in environments that needed to be quieter. We gave up on the Dual processor stuff altogether at the time to avoid compatibility issues. After picking the 933 QS or MDD 1.25 as the base system, for VST systems we added cards like the UAD-1 and the power core PCI to get pro results without overtaxing the CPU… this made a VST DAW similar to a ProTools rig… DSP cards helped offset the CPU usage.
Over-clocking sounds great…it definitely adds another level to maximize a DAW. We just focused on what worked at the time and produced stable and consistent results. Since the over-clocking also adds a layer of trouble-shooting, we avoided the information that was available at the time (which was very little). I think of the Over-clocking as real art form, and not for the average users, but I value all of the efforts that have yielded amazing results and made G4s go faster than thought possible
-
honestly ptle5guy--
the 167 mhz mod isnt worth the trouble
u can go out and find a used g4 with the factory 167 bus for less then 50 bucks here
theres one sitting at the shop down the street right now for 40$ from my house
why would i risk messing around with the motherboard when i can just check the classified ads;D
-
honestly ptle5guy--
the 167 mhz mod isnt worth the trouble
u can go out and find a used g4 with the factory 167 bus for less then 50 bucks here
theres one sitting at the shop down the street right now for 40$ from my house
why would i risk messing around with the motherboard when i can just check the classified ads;D
Chris, I would not pay 50 bucks for a motherboard with one resistor less... ;D
I have changed motherboards on PC boxes since 1998... It is only to find a good guide to take off the motherboard from MDDs.
To unsolder SMD components is not easy, but also is not hard if you practice. I have been using a soldering iron to repair my guitar cables since 1990... I am not an expert, but the practice would lead to do it right as Harrymatic says on the post.
It is confirmed that the only difference in motherboards is the dawn R676 resistor. I wish apple would use a jumper, but that is not the case.
Locate and remove the resistor marked R676 on the underside of the board – it’s near the heatsink under the CPU area. (see photos) You’ll need a fine pencil-tip soldering iron. Make sure that the removed resistor doesn’t get lost somewhere on the board! I didn’t find this too difficult as I do a lot of work with electronics, but if you don’t have experience with SMD components, practice loads on some old PCBs first.
Imagine you can get 2 MDD for 50 bucks, one working at 866 and other 1.25 but with the PSU dead. What would you choose? Ah, you would choose BOTH, because you have the money...That is not my case...
Anyway, here in Spain a working MDD still cost 100 Euros (120-130 USA dollars). Even some original Sawtooth are trying to be sold at 100. And MDD PSU also 100. I wish we have those offers here in Europe, but is not the case...
The iFixIt guides for MDD are not deep, but at least we have the best graphic guide I have found for them:
https://www.ifixit.com/Teardown/Powermac+G4+MDD+%28Mirrored+Drive+Doors%29+Teardown/1624 (https://www.ifixit.com/Teardown/Powermac+G4+MDD+%28Mirrored+Drive+Doors%29+Teardown/1624)
It is up to the user if it worth to pull out the motherboard from the box. Eventually, if you buy a MDD 166 bus motherboard, you will have to pull out the old-slow motherboard...
Please, if someone swap out a MDD 133 board, and have no interest on it, PM me. I would pay shipping cost in Europe.
-
i would choose the 867 and leave it as an 867 ;)
which is what i have.. and its fast enough as it is..! really ! i dont notice that its slower then the single 1.25
in fact it has a higher benchmark then the 1.25 single
naturally 2 against 1 cpu
-
i would choose the 867 and leave it as an 867 ;)
which is what i have.. and its fast enough as it is..! really ! i don't notice that its slower then the single 1.25
in fact it has a higher benchmark then the 1.25 single
naturally 2 against 1 cpu
I would be happy with the 866 dual if my last PowerMac were the DA 466, but since I bought the GHz dual cpu module from QS02, I feel it slower... Once you test speed, you get addicted...
Yeah, the 866 is a good machine compared to Digital Audios, but only on OSX. I feel the MDD with 2 gigs of ram as fast as the QS02 with 1 gig on Tiger and Leo.
But for OS9 the QS02 is faster/more powerful.
I do not force users to do the mod. Just wanted to get all the info in one place. I like stock machines too, but the benefits of pulling out the R676 is too much to ignore it...
-
167 Mhz is the "native" bus speed of MDDs. (by the way, it can be overclocked to 172 with motherboard oscillator change) although some "low end" MDD's were working at 133 Mhz for marketing reasons (this is easy recoverable via motherboard multiplier resistors settings). Well I never have done it, it is easiest try to acquire the good ones (167 Mhz) ;)
Sorry to bring back an old thread, but I have seen you reference the 172 MHz bus clock a couple time but not sure where you got the info. Very curious if you have URL/doc/etc laying around yet with this info?
The MDD bus can be changed from 133 to 166, just changing some resistors in the Motherboard. But If you already have an 166 Mhz Bus System, The only way to go further is to change the base quartz oscillator of the motherboard.
MacTron: Let's be accurate. It is 1 resistor to be pull of for the motherboard to get 166.
-
Old article about unleashing the 166 bus on first MDDs
http://www.macbidouille.com/articles/89/page1 (http://www.macbidouille.com/articles/89/page1)
Can you tell us if the MacBidole article is well translated? http://macos9lives.com/smforum/index.php?topic=1652.0 (http://macos9lives.com/smforum/index.php?topic=1652.0)
Well, for starters, it's MacBidouille!
automated translations always make me cringe...
at least the first bit is correct. the following two should read:
----------------------------------------------------------------------------
Note that changing the bus from 133 to 166 Mhz would overclock the machine to more than 1 Ghz. To avoid this, you must reduce the multiplier from 6.5x to 6x
Watch out !!!!!
To ensure the stability of the machine after changes, it is imperative to have only DDR-SDRAM PC 2700! 2100 simms are to be avoided !!!
Changing the bus takes 2 steps:
1) changes to the motherboard
the photo is the underside of the motherboard.
I surrounded the three resistors in question.
They are in order R677, R678 and R676
For a bus of 133mhz: there is a resistor at each location
for a bus to 167MHz: you must remove R676
(It's about 100% on because I had 2:30 to scan both cards under the microscope and that's the only difference)
So there must be more to it (I have not taken the time to try and ram no longer lends itself to such practices)
Apple has added a second protection on Bi 867 cards: If you change the settings just on the motherboard, it will go to 167 Mhz only if you install a 1Ghz or 1.25GHz bipro card.
(So I'll turn it back to Amber:)
So on the daughterboard pictured, there is a series of three resistors,
numbered as follows.
167 Mhz Bus:
R27 nothing
R22 nothing
R26 Resistor
133 Mhz Bus:
R27 nothing
R22 Resistor
R26 Resistor
So it is a security or a lock from Apple
----------------------------------------------------------------
I'll leave it to you to edit the original posts: I have no way to do this...
-
i wonder what the benefit of a higher system bus is when you do not clock the processor higher at the same time.
-
i wonder what the benefit of a higher system bus is when you do not clock the processor higher at the same time.
Do the Math :)
Calculate the RAM bandwidth (PC-2100 or PC-2700) and the System Bus at 133 or 166 Mhz.
You'll see a bottleneck in the main system bus, this is worst in the 133 Mhz case than in the 166 Mhz case.
DDR 133, PC-2100 2128 MB/s (2,1 GB/s)
DDR 166, PC-2700 2656 MB/s (2,6 GB/s)
-
Hello,
What about the M8570 (EMC1914) with firewire 800? I have 2 machines, one is 133Mhz, the other is 167Mhz.
How can I mod the 133Mhz machine to become a 167Mhz?
Thank you
-
DDR 133, PC-2100 2128 MB/s (2,1 GB/s)
DDR 166, PC-2700 2656 MB/s (2,6 GB/s)
...which is about ten times more what you need in any of these machines, because thre more you can write and read from memory the more you will also like to have processing power to process all that data.
-
... which is about ten times more what you need in any of these machines,
OK. I'll put some extra data:
The Double Data Rate memory bus bandwidth (from RAM to U2 northbridge):
DDR 133, PC-2100 2128 MB/s (2,1 GB/s)
DDR 166, PC2700 2656 MB/s (2,6 GB/s)
The System Bus bandwidth (from U2 northbridge to CPU moule)
MDX 133 Mhz, 1064 MB/s (1.0 GB/s)
MDX 166 Mhz, 1296 MB/s (1,3 GB/s)
The PPC G4 (7455) in the MDDs usually needs around 10 GB/s to run without wait states.
In the Dual CPU modules this is even worst, as this small bandwidth is shared between the two CPUs!
-
i guess i dont understand the theory. but can you explain it with an example of an application?
say if i want to play audio from a sampler. on a 133 mhz machine i can play some 200 mb/s from RAM - in practice. on a 167 mhz machine i could play some 250, in a machine with DDR3 RAM i could play some 300mb/s.
but in all cases i will not have a computer fast enough to also have one effect plug-in on every channel anyway. so why would i care about the bus throughput?
maybe i fail to see that there are also nonrealtime processing jobs in this world. :)
-
Hello,
What about the M8570 (EMC1914) with firewire 800? I have 2 machines, one is 133Mhz, the other is 167Mhz.
How can I mod the 133Mhz machine to become a 167Mhz?
Thank you
AFAIK to mod the motherboard to run 166 instead of OEM133 you have to take 1 resistor of the backside of the motherboard. But I did not ever do that MOD.
i wonder what the benefit of a higher system bus is when you do not clock the processor higher at the same time.
Benifit? A dual 866. 866/133= (Aprox) 6.5 multiplier.
If the CPU module were able to work the same in 166...
6.5*166=1079 Hz, so it could be a GHz machine but it seems the cache RAM of the Dual 866 can't go that far.
Here are the modifications needed to change the G4 Bi-867's bus clock from 133 to 167MHz.
You'll note that changing the bus speed to 167MHz would overclock the CPU to over 1GHz. To prevent this, you'll have to change the CPU's multiplier coefficient from x6.5 to x6.
So, 6x166=996, 4 Hz less than a GHz... but...
Should it be quicker than a 6.5*133 machine? Sure it is.
-
or u could just buy a better mdd for les then 60$ anywhere in north america
or online
-
or u could just buy a better mdd for les then 60$ anywhere in north america
or online
In the future, with the success of forums likes this, the number of cheap PPC will be reduced... :'(
Just look at how much it cost now a 2006 MacBook...
Ppc boards are becoming in collector items, or antiquities.
The cost of a 2nd hand G4 cpu module has grown since 2010, and it will be worst.
Chris, it is good for the environment to MOD hardware instead of trashing it.
Anyway, I can't have more machines in my house. 3 PowerMacs and 4 Intels is way too much machines for anyone but a geek-collector. And I will never send to trash ANY PowerPC.
-
the difference between 133mhz + 166mhz is an improvement yes.. but imo its nothing too crazy to worry about.. i use machines that have 133mhz bus all the time.. and they are still capable of getting the job done just fine.. g3, sawtooth, quicksilver. 70% of the mac g3/g4s dont have this extreme bus..... its ok;0 its only for those obsessed with having the best ;)
i think its better to focus on using the software !
-
its only for those obsessed with having the best ;)
I like to have the best , exactly as most of us. But I'm not obsessed with having it. But I always recommend to buy the best and powerful Mac Hardware Mac Os 9 capable ( the FW400, 166 Mhz Bus MDDs) because nowadays the price is very similar between a G3 or a Yikes with a very bad motherboard or a MDD with one of the best motherboards ever made for a Mac Os 9 machine.
Anyway, each one have its own work flow and each work flow have its hardware bottlenecks. If your workflow implies using 16 or more audio tracks, each one with its effects line, you will need a very fast hard drive and probably Powerful DSP system like Pro Tools. If your work flow rely on 16 tracks of virtual native instruments as mine does, you will need a faster CPU and a fast system bus.
-
i use machines that have 133mhz bus all the time.. and they are still capable of getting the job done just fine.. g3, sawtooth, quicksilver. 70% of the mac g3/g4s dont have this extreme bus..... its ok;0 its only for those obsessed with having the best ;)
i think its better to focus on using the software !
G3 beige run a 66MHz bus and G3B&W a 100 bus as the PCI G4 and Sawtooth (first AGP). Since Digital Audio or Gigabit Ethernet there is 133 bus(I think it started on DA).
-
Another important point is that the CPU speed must be around 4.5 x the bus speed. So for a MDD at 1.25Ghz the optimum system bus should be around 275 Mhz. The presence of the L3 and L3 cache and the 7 stage pipelining, helps to reduce the dramatic impact of the slow 166 Mhz bus, ... but forget about a 133 Mhz system bus ...
The background of this, was the problems that Motorola had to improve the G4, because it was more interested in selling cellulars at its time ...
-
Another important point is that the CPU speed must be around 4.5 x the bus speed. So for a MDD at 1.25Ghz the optimum system bus should be around 275 Mhz. The presence of the L3 and L3 cache and the 7 stage pipelining, helps to reduce the dramatic impact of the slow 166 Mhz bus, ... but forget about a 133 Mhz system bus ...
The background of this, was the problems that Motorola had to improve the G4, because it was more interested in selling cellulars at its time ...
[youtube]w9VWRB07yqc[/youtube]
-
i agree for sure. get the best u can get.. and the budget now for having a 167 is not so different from buying any other mac.. my point is for alot of people that these mods can lead to a nonfunctional computer!! better to just buy the right one.. or use the 867 mdd as it is..
thats just my opinion. u guys go off + mod your motherboards to your hearts content;) i have a mdd 867 sitting downstairs (without a power supply) im not worried about modding it thats for sure.. because i can just get another one for an affordable price if i need.
alot easier to just buy one ... at least where i live.
perhaps im afraid to solder ;D i think its easier to trust the standard config of the machine.. and be happy.. rather then to create instability + wierd problems that can take over your life. video cards i would be much more liekly to try to mod because u can easily replace.. a system board.. im chicken ;D
-
BTW: upto Digital Audio 533 all G4s CPUs had an even shorter pipelining!
A 4 stages pipelining!
So this was a real impressive CPU at its time!
... and even today
-
protools guy = ultimate steve jobs fanboy
8) 8) 8)
-
protools guy = ultimate steve jobs fanboy
8) 8) 8)
Steve NeXTStepped the Macs... We could have CopLand instead of MacOS9 if Steve didn't came out with the NeXT Step OS
http://en.wikipedia.org/wiki/Copland_%28operating_system%29 (http://en.wikipedia.org/wiki/Copland_%28operating_system%29)System 7 was ahead of its time, and Ellen Hancock was wrong on that .
She decided it was best to cancel the project outright and try to find a suitable third-party system to replace it. Development officially ended in August 1996, and after a short search they announced that Apple was buying NeXT in order to use their NeXTSTEP operating system as the basis of a new Mac OS.
Apple should have bought Also BeOS while they finished CopLand and OSX. The first 10.0, Puma and Jaguar are Jokes as OSes, but BeOS wasn't.
Yes, I am a fan on the work he did while the transition from OS9 to OSX making these wonderful machines we use. But he was pushing really hard everyone in the industry to develop for OSX Cocoa not carbonized apps instead of OS9 compatible carbonized apps... The OS9 in the coffin was totally unnecessary to my taste, and even a Bad Strategy for a business. They could be selling OS9 machines longer as they did with Apple II machines on the last 80's and first 90's. Apple had a lot of series of machines pre Steve Jobs, and that "Zen" lineup on products in Steve'Age damaged the life on OS9...
Also the Steve work on G5 to Intel was awesome. iBooks, PowerBook Ti, iPod, iTunes, iPhone, iPad... Those are wonderful devices, but G4s has an special place for me.
When they went to Intel, there was no need to buy Apple Machines to run Tiger, Leo, SL...
I recognize that have seen every MacWorld since the Steve Jobs back to Apple, some from the NeXT Cube, and the 1984 Macintosh one. What Apple owner didn't do that? Every time I play ..
[youtube]6KJZLiic2F8[/youtube]
makes my DA GHz dual and my MDD 867 run faster and prouder.
-
Seeing G4 kicking the ass of intel proc. is fun
[youtube]KLWom93pEHY[/youtube]
Steve Jobs running OS9 get nervous... ;D
BTW: upto Digital Audio 533 all G4s CPUs had an even shorter pipelining!
A 4 stages pipelining!
So this was a real impressive CPU at its time!
... and even today
What was the fastest G4 7400? All 7450 G4 are 7 stage pipeline?
-
What was the fastest G4 7400? All 7450 G4 are 7 stage pipeline?
The G4 500 DP (Gigabit)
... and the Digital Audio G4 533, Mac Server G4 533 (Digital Audio) and Mac Server G4 533 DP (Digital Audio) 533 with the g4 7410 (4 stages too)
All 7450 G4 are 7 stage pipeline?
Yes it is.
Take a look:
Apple Power Macintosh G4 533 (Digital Audio) Geekbench 2 (32): 356
Apple Power Macintosh G4 733 (Quicksilver) Geekbench 2 (32): 372
Increasing a 38 % the processor speed on the 7450 (quicksilver 733) and only a 4% of Geekbench improvement compared with the 7410 (Digital Audio 533 ) .
Well, the reduced cache memory on the quicksilver also damage its performance...
-
Seeing G4 kicking the ass of intel proc. is fun
well enjoy - because u will never see it again lol
-
This is an easy mod. 1.0ghz/167mhz is a huge benefit in OS9. I've done it on two 867's. Regarding fans... Scrap and kiss good-bye to that power connector to the MOBO, as it's that which causes the awful RF in the builtin audio and the whole wind-tunnel problem. I run just one fan atop the heatsink at 7 volts directly from the drive source (+5 & +12 = 7 volts), and the power supply has one of the fans removed for better air flow. No other fans. No need to use a soldering iron. You can snip the resisters off, which pulverizes the resisters, but who cares? The resisters for the 133 and 167 busses are reversed on Macbidule's translated page. You need to remove, not add, one to get a 167 bus. This was a French > English translation mistake. To go from 867/133 to 1000/167 you need only remove resistors.
-
interesting. i had an 867 that i got rid of. i wish i hadnt now!
is there any place where this is more clearly documented?
it seems like a very careful modification only for the expert + brave people
-
I can see this topic hasn't been updated for ages, but if anyone ever seeks this info:
I can confirm that the 1.0ghz single-processor PowerMac 3,6 (FW800) also supports the r676 removal to overclock the FSB from 133 to 166Mhz mod. You should just be even more careful for the fact that the resistor layout differs slightly from the 2002 MDD.
I have done it with my FW800 MDD and as soon as you put a multiprocessor daughtercard, the FSB is recognized as a 166Mhz one.
-
Yes very true - I installed a dual 1.42ghz dual CPU daughter card onto this board and is what I am using to write this message here. Runs beautifully all the time.
One thing I do use for cooling this machine now that I never did earlier was the altering of the info plist file of the AppleFan.kext to alter when the main MDD fan is to kick in - this keeps the CPU temp around the 49ºC mark under heavy loads. I understand that by doing this it may cut fan life down but the machine never gets beyond the 50ºC mark ever. This seems to only work for TIGER - macOSX10.4.11 PPC.
-
I presume the same bus speed mod is also possible on the Xserve Rev.1 motherboard - anyone have details for that?
-
I'm not convinced the Xserve can have its FSB overclocked.
As far as i know, the original Xserve was introduced a couple months before the 2002 MDD, and the OEM hardware configurations ( SP and MP ) all had 133Mhz FSB.
-
I'm not convinced the Xserve can have its FSB overclocked.
It's not an overclocked bus.
There is two models of DDR xServe G4:
A) A 133 Mhz System bus and ATA 100 Hard disk Interfaces and FW400.
B) A 166 Mhz System bus and ATA 133 Hard disk Interfaces and FW800.
But the FSB can be overclocked from 166Mhz to 172 Mhz.
Probably both procedures are similar to the ones performed in MDDs
-
I'm not convinced the Xserve can have its FSB overclocked.
It's not an overclocked bus.
I meant overclocked as in "Overclocked from Apple's stock config", not pushing beyond the 7455's specs.
To be honest, I never saw an Xserve G4 (or even a RackMac) in front of me; even though I know they use the same board design as the MDD's, I'm completely unfamiliar with these.
I just noticed that there were machines that had onboard ATA-7 :)
-
Last sunday I went 30 Km away from my city to pick a broken MDD. It was a Uni 1.25 GHz MDD with SAMSUNG PSU (broken I guess) and a Radeon 9000 and a Sonnet Allegro FW400 ( probably with onboard broken firewire ).
I went to everymac and compared geekbench 2 results of 867 Dual Vs Uni 1.25 and were the same! For Mac OS X there is no advantage one machine vs the other. But on Mac OS 9 M.A.R.L. and other cultist would choose the 1.25.
So I decided to put the 166 and 1.25 on the 867 dual box that has a working AcBel PSU.
Notice the colour greener on 133 and bluer on 166
-
The boards are different. So I decided to take use on my new phone and take some pictures of the R676 on the back side of the motherboard to kill the subject at last.
-
On the CPU Daughtherboard side:
I could find R22, R26 and R27 on the front of the Dual 867, and I couldnt find R22 on the front or back of the Uni 1.25.
I could find R26 and R27 on the back side of the Uni 1.25
-
Just in case someone want to find R22 (maybe is not needed)
-
Question : I have a working 866 DP , and a broken 1.2 DP, may I swap the Cpu? Can it cause problems with the bus speed? In theory the cpu will be underclocked due to the slower bus clock, other downsides?
-
Ok the CPU swap worked! Now I have a dual processor @ 1ghz . The slower bus resulted in a underclocked CPU (1Ghz vs 1.25) but for me is now enough, not in the mood to risk damaging the hw overclocking!