150 MHz QS Bus Speed!

[FBz]

Overclocking Apple Quicksilver System Bus to 150 MHz?

Searching for instructions on overclocking the MDD 1.25 GHz & 1.42 GHz CPU’s
stumbled upon MacTron’s post http://macos9lives.com/smforum/index.php/topic,2903.msg18351.html#msg18351
Mentions increasing Quicksilvers’ bus speed to 150 MHz & found this bit of how-to instruction.
https://translate.google.com/translate?sl=auto&tl=en&js=y&prev=_t&hl=en&ie=UTF-8&u=https%3A%2F%2Fsites.google.com%2Fsite%2Fjonathanrenaudg%2Fhome&edit-text=&act=url

Might give this a whirl.
Seems easy enough and reportedly… with little resultant heat increase?
I’ve plenty of QS “A” MOBOs to test this on before same on “B” board.

Although I do wonder about this:
“Also take note that OSX will not recognize in these analyzes,
the overclock of the bus to 150MHz nor the real frequency of your processor.
Regardless, the important thing is that it works!”

Anyone here attempted this, in this manner / fashion?

[FBz]

150 MHz System Bus hack on a Quicksilver 820-1276 “A” board. Used a 22Ω .25W resistor and a soldered in-place jumper. Used same RAM and nVidia from Quicksilver that had a bootable “B board” to start… with a solo Inland 128 GB SSD (via a trusty RXD-629A7-7 bridge adapter) and only 1 GB RAM installed. QS 867 MHz CPU.

No massive differences noted with QuickBench 2.0  - but maybe QB isn’t the best method to measure the touted 12% System Bus speed increase against original config. It seems faster. [QuickBench results from the original are top results, over the results from the hack (bottom).] Click attachments.

Original 22Ω resistor appears different - but the newer resistor has same values.

Example per instructions:


Step 1:


Completed:


Unable to source a 1 GHz DP for the QS but did get a 933 and I thought this hack might help it a bit towards that 1 GHz performance goal. Next I’ll do same hack on Quicksilver “B board” max the RAM and install that 933 CPU.

[DieHard]

Wow, this is so nicely posted... you get another "DieHard Santa" package this year !

I will try to include small items this time as the ADC monitors provide further "clutter" to your lab

Oh, and let me know if you need a few dozen more "Bribge" SATA to ATA adapters 'cause I am starting to run out of experiments with them

[Bolkonskij]

I never cease to be amazed at what folks are still able to achieve on our older Macs - well done, FBz! Does it run stable?

[FBz]

Thanks Bolkonskij und Diehard
(aka: Lemmy).

Again however, credit where credit is due goes to Jonathan Renaud.
I just followed his instructions from a post back from around 2007(?).

Yes, it’s been running stable since the hack. Haven’t done the “B board”
with that 933MHz CPU yet but it’s on my list. (Been a bit preoccupied
with a “white whale”.) Still thinking that if possible on a QS, maybe
something similar on an MDD… to improve DTR with SSDs?

And as far as “clutter” goes here…
How can you “see” when I’ve taped over the camera on this iMac?

[FBz]

Well, only took a little over two years to finally get back to testing with a Quicksilver “B” board.
Confirmation now http://macos9lives.com/smforum/index.php/topic,6079.msg45423.html#msg45423

Newly acquired 933 MHz Quicksilver was best candidate for swapping “B” board and testing hack.
And no, I did not overclock it. Just added that jumper and resistor as first shown in the link above.

AND with MAChSpeed _3.5.0_OS9.sit from Macintosh Gardens…
https://macintoshgarden.org/apps/machspeed-control-combo
Now it doesn’t just seem faster, it is confirmed faster!!!
(Big thanks to @evanboonie over at MacRumors.com)
https://forums.macrumors.com/threads/g4-7448-and-7457-swap.2233532/page-10?post=32608426#post-32608426



Not quite the targeted 150 MHz bus speed, but pretty damned close.
I’ll max the RAM and this will become my QS base test machine.
(Wonder how a 1GHz DP Quicksilver might also perform / report?)

Next, will try to overclock an 867 MHz QS CPU to 933
- because it is already so similar to an actual 933 MHz.
https://thehouseofmoth.com/ppc-overclocking-station/

Never anything to do around here.

[mrhappy]

YeeHaa FBz… Ride em cowboy!!LOVE this!!!

[aBc]

Do not encourage him. 

(Already his hat does not fit.)

[IIO]

Oh, and let me know if you need a few dozen more "Bribge" SATA to ATA adapters 'cause I am starting to run out of experiments with them

have you yet tried to build a 5 foot tall model of the eiffel tower with bribge adapters?

[aBc]

In other news of the odd…

Swapped in one fan from an MDD’s PSU for a stock QS CPU cooling fan.
[In preparation for possible tweaking of that bus speed in a 1.0 GHz DP QS.]
So now it’s a half-a-wind-tunnel  Quicksilver. HD temp sensor recorded temp of 79˚F
- after one hour’s runtime (under load). Haven’t dB checked, but not for the DAW crew.
The 2410ML-04W-B60-B00 fan specs note a constant 5300 RPMs. Louder but 25 CFM.




In meantime, possible homes for at least five more Bribges have just become available here.


Hiding in the back there is a shy 450 DP Gigabit - now missing its’ Zip drive trim bezel & PSU.

Now go look under your bed IIO. (Krampus awaits you there.) You off-topic-shifting hooligan.

[Ncc74656]

my photos were way too large. here is an album of bench scores: https://imgur.com/gallery/tjHh5dn

here is the testing: https://imgur.com/gallery/VsEbYJj

thought i would add some to this.  G4 b board.

my single 800mhz cpu overclocked to 1ghz - default voltage of 1.85 and this boosted to 1.9 to reach. 475 geek bench score
https://i.imgur.com/bhFSetB.jpeg

i then swapped in a dual 1 - tested this factory at 861.
https://i.imgur.com/jlkJdCF.jpeg

stand alone i can get this to boot and start running benchmarks at 1.2 but no post higher than this. lost the silicon lottery here i think but im wondering. i have some single cpu boards that clock to 1.25, its the same chip. so maybe i desolder the 7455 cpu on the boards that require the least power at that frequency and ball them onto this dual board - see if i can go higher then?  i have not put a scope on this yet, i want to see just how much jitter or droop im getting in power delivery across everything. it could be as simple as a power stabilization issue. the tech sheet lists 1.87v max for these 7455's so apple runs them hot anyway.

many people say they can only get 1 step up on these cpu's so thats what i did. 1.07ghz at 909
https://i.imgur.com/0IZ1Uk5.jpeg

i stepped things back - 150mhz bus now
https://i.imgur.com/Q5VhexT.jpeg

dual 1ghz ƒ 150bus = 1.12Ghz  968
https://i.imgur.com/L8G2adN.jpeg

i then booted into 10.4 instead of 10.5 and reran this one test. 951   so 10.4.12 is SLOWER than 10.5.8
https://i.imgur.com/GNVuOoe.jpeg

next i will install sorbet and try that.

there was and is an issue with the higher bus speeds. to run 1.12 i am at 2v, so that is lots of thermal.  1.95 may be enough for stability, ill have to try that still BUT i kept at 2 as i was and am trying to push for 1.25ghz.

thermals are an issue for stability - the CPU sink gets hot but thats not the issue. the FSB clock controller is 11.5C higher at 150mhz vs 133 capping out at around 62C under bench testing.
on the CPU daughter card however the power fets under the heatsink in the corner of the board get HOT. clearing 95C before hard lock. there is not nearly enough air on them


the inductors get much warmer too. 19c hotter
https://i.imgur.com/AhBm51H.png

10 minutes after shutdown
https://i.imgur.com/SugXMPB.png

i added a large fan to blow under the heatsink and temps became stable.
https://i.imgur.com/h9m1hdk.png

i will need to work out a cooling solution here, tbd.  i am trying to keep everything period correct.

my next challenge will be teh X800 radeon but ill post that elseware. 

[aBc]

…mouth agape
and speechless.

Best image overview
see: https://imgur.com/gallery/VsEbYJj

[DieHard]

…mouth agape
and speechless.

DITTO

Incredibly cool (or should I say HOT) images !  Best ever

[redstudio]

I have a quicksilver with dual Gigadesign 1.6 GHz processors, updating the bus to 150 MHz doesn't work for me.. it starts, it freezes.. strange behavior.. it's due to the fact that Gigadesign expects a fixed frequency of 133 MHz? does anyone know how to get around obstacle? it would be nice to speed up the processor a bit.. among other things I could push it up to 1.8 ghz from the factory and with a little more speed on the bus maybe get closer to 2 ghz..

[aBc]

(Is it)… due to the fact that Gigadesign expects a fixed frequency of 133 MHz? -redstudio

That’s possible. There’s a 1.2 GHz Sonnet CPU in a QS here, that I’ve yet to attempt the simple System Bus conversion to 150 MHz. So that might yield a similar result when I test that?

Have you also tested yours with any stock Apple QS CPU to see if that does work? (To test your conversion.) The theory being that if it works with a stock CPU, then the Gigadesign likely may rely upon 133 MHz.

And if so, I have seen references of alternate methods to increase the System Bus speed (other than the very simple jumper & resistor approach). It could also be a voltage / current concern too.

Much speedier demons than me - might need to respond here. @Ncc74656?

Best of luck and do keep us all posted.

[Ncc74656]

when you OC hte bus your going to be jumping your CPU speed and the voltage required by a %. at 150 its like 12.57 or such.

so take your cpu at a 133 bus and overclock it directly, figure its stable voltages with a 13 or so % oc. IF it wont run stable then you will need to downclock it to overclock the bus.  one step at a time.

[Ncc74656]

i just benched on 10.5.9 and got 979. so a 70 or so point bump just from the os

[redstudio]

.. probably with gigadesign processors the trick of increasing the bus doesn't work.. in fact the standard pin setting differentiates whether you have the 100 or 133 mhz bus.. you probably expect either one or the other not higher. Tried to change the speed on the pins but nothing... at 1.8ghz the Mac can't start... unfortunately I don't have the original CPU to make a comparison... once I removed the overclocking change I immediately started again with the dual 1.8's (with the computer open for avoid overheating..I have to study a way to circulate more air inside the house..) it works perfectly! I'm satisfied in the meantime...

[Ncc74656]

a few more. spent some time last night.
this is a dual 1ghz on a 133 bus - same board. os 9.2.2  sorry for poor picture quality.  8500 PC radEon, 1gb ram. IDE to sata adapter for a intel X25 SSD 128gb. (i bought a stack of 25 of these and use them everywhere. great SLC drives)
CPU: 1744
floating point: 1598
disc: 4167
graphics 2496
https://i.imgur.com/LFvjXj1.jpeg

here is the same system with a dual 1.0 with a 150 bus.
CPU: 2334
floating point: 1662
disk: 5007
graphics: 1972
https://i.imgur.com/TvkM9gG.jpeg


i also took this system (same from the tests in the post above) and tried to OC it to 1.19ghz but can not get a stable bench. i boosted voltage to 2.05v at core and still no joy. it could be a L3 cache limit as i hear they dont handle OC well? still. im not sure i understand why the single vs dual GPU boards have such a huge difference in OC potential. i have three single CPU boards at 800 or below mhz that are running at 1.25ghz now. stable.

installing sorbet 10.5.9 gave me a serious bump in performance that really surprised me. just as surprising as 10.4.12 running slower than 10.5.8 in my testing.
1000 in geekbench on all period correct hardware...  no fancy aftermarket stuff. B board wtih dual 1Ghz and old ram.
https://i.imgur.com/ad5w9wP.jpeg


finially i just so happened to have one CPU disabled on the daughter card while i booted OS9 and it gave me an error that one cpu has failed and to contact my support provider on bootup. that surprised me. didnt seem to effect bench marks but just an interesting tidbit.

next is my video card. going to be a bit out on that though as im replacing a bad memory chip on it with a samsung for better OC ability.

[Ncc74656]

there is another aspect (some what obvious) that i have not seen addressed here or anywhere really. on modern system we have full control of timings, voltages, and so fourth but not here.

I am using "Remember" program to run full tests (3 passes each) on the sticks.  to ensure stability.

the system will run and bench mark just fine with some memory combinations but when running newer games - instabilities creep in. Doom 3 is an example of this, when playing that game on my quicksilver i get lockups or stutters on sticks that were totally stable for days while web browsing or light multimedia.

overclocking the bus is going to bump ram speeds - we have 3 ram slots and i do not think apple engineered much "safety margin" in available amperage/voltage to their jedec spec.

i have 17 stick of SD ram, all 133. the Hynix chips are so far proving to be the most stable. NEC only works in single rank, one slot populated.  so far i have not found any sticks that will run all 3 banks with zero errors.

two 512mb sticks of crucial ram with hynix chips are stable in the system but none of the other sticks of 512 will work with them.