Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
"fastest" flash, and using a pencil to overclock.
- Thread starter Reflex88
- Start date
- Sort by reaction score
Boot into OF:
Code:
dev /cpus/@0 // Make the CPU the active package
spd // Check the HID1 to see what DFS mode we are in( full/half/quarter)
: set-dfs-real-low pvr@ 10 rshift 8004 <> if exit else hid1@ 1 1f 8 - lshift or hid1! 1 ms 4 1 gpio! then ; //We define the word and the function of set-dfs-real-low
set-dfs-real-low // We try to invoke setting HID1 bit 8 to 1 and wait 1 ms for the GPIO
spd // we check the status of HID1 agian, and if things work we will see that bit 8 is setNote: Everything after // is just my notes, don't type // or that which comes after.
Ok we would expect to se HID1 = 808xxxxx, so I've got the code wrong for our set-dfs-real-low function.This is on the mini:
It locks up at this point and doesn't do anything until I reboot it.
I think the powerbook is doing DFS through hardware because there is some other stuff going on with the voltage at the same time.
Let me ask around and see if anyone can help out.
Great things are going on here 
What I am wondering about is whether it is possible to add faster RAM too. I took a look at the firmware of my Mini 1.5GHz but I did not find where a faster RAM timing of say CL2 can be configured. Of course, the memory controller would have to support this as well.
If I'd have a description of the chipset ...
What I am wondering about is whether it is possible to add faster RAM too. I took a look at the firmware of my Mini 1.5GHz but I did not find where a faster RAM timing of say CL2 can be configured. Of course, the memory controller would have to support this as well.
If I'd have a description of the chipset ...
Speaking of memory, were 2 GB DDR1 333MHz sticks ever manufactured? I wonder if there's any serious limitation preventing the sole RAM slot from addressing more than one gigabyte.Great things are going on here
What I am wondering about is whether it is possible to add faster RAM too. I took a look at the firmware of my Mini 1.5GHz but I did not find where a faster RAM timing of say CL2 can be configured. Of course, the memory controller would have to support this as well.
If I'd have a description of the chipset ...
As far as I can tell, PC3200 would be the newest RAM that would be backwards compatible with the Mini, and they never made 2GB SODIMM, only some very rare ECC full sized desktop DIMMs.
they did make 2GB DDR1 UDIMMs but they are stupidly rare
the Later revision of the Xserve G5 was compatible with them
(and indeed I suspect PowerMac G5s, if you go through the About this Mac thread in the Photos section of this forum, someone posted a 2.7Ghz G5 with 16GB of RAM, and I know the later G5's use the same U3 heavy North bridge/memory controller that the Xserve G5's use)
I would love to find some to try in my PowerMac G5 and Mac Mini but I have not been able to find any
Kingston made a kit KTA-G5400E/4G for the Xserve which was 2x2GB ECC UDIMMs sticks for example
but I have litreally never seen any for sale on ebay etc, there where also some other 3rd party 2GB ECC and non ECC UDIMMs, but those are just as rare, and what few places I can find with them listed want like $70-140 a stick
(being UDIMMs the ECC part is not a problem thankfully, Uninorth macs will work fine with ECC UDIMMs, they will just ignore the ECC bit, its RDIMMs they dont work with)
Last edited:
I applaud this mod, but what is the point ?so the 7448 is cooler than the 7447A/B - what use will this do and how much faster is the 7448 compared to the 7447A/B ? I don’t mean to be negative, but what does all this mean for us PowerPC Mac users in 2021 ??
That's rather easy to answer: It's like any other hardware upgrade everyone does, with the usual returns. Faster boot times, faster-running apps, faster OS use (rarer hang-ups and the like), overall increased system smoothness etc.. For instance, I archive software and deal with archived software a lot, and so deal with a lot of compression and decompression. I cannot emphasize enough the difference of doing so with, say, a stock 1.25GHz PPC that comes stock with the high-end OS-9-capable MDD, or even the 1.5GHz PPC Super mini, against the likes of a 7448 running at 2.x speeds. It running cooler also allows for less loud fans, which is a very, very welcome quality of life improvement when using those systems (particularly the MDD, but minis occasionally get louder than necessary, too).
All this makes a lot more sense under OS 9: G4s are the only upgrade path, meaning no G5s. But machines like the mini can surprisingly run OS X very well, too, and with a 7448 upgrade, it would make all the difference in the above points. Being able to run higher-end OS X games in a deliciously tiny setup thanks to a 7448 upgrade is also very attractive, which is important considering there are no G5-based minis (and they would be very loud if they somehow existed). Make that mini crunch some World of Warcraft better, and join the Macintosh Garden's own private PPC server.
Or pick whatever FPS game, RPG or what have you. Maybe edit things in Photoshop faster, or have webpages render faster and more smoothly under TenFourFox.Also, friendly reminder: a lot of Mac OS X (and ALL of Mac OS) software require PPC machines, with no way around it. All emus are stuck with low FPS, inaccuracies and other issues, and while QEMU is by far the best among them, it's sadly not even remotely close to being good enough to be able to replace any real hardware, and the forecast seems to be it never really will, certainly not even in the next 10 years (at the very least). Maybe we will want to re-check this 30 years later, but better not expect much even then, to avoid disappointment.
Now, if only we somehow developed ourselves to make a community that developed tools that facilitated extending/expanding Mac OS and Mac OS X support to newer PPC hardware, graphics cards and the like, that would be ideal. But until then, getting our trusty hardware accelerated with 7448s will do.
Strategia,
my idea was not about the actual size limit of the RAM but about the speed.
AFAIK, the Mini is configured to fixed CL3 timing of the SDRAM module. I tried to use ultra-fast "gamer" CL2 SDRAM chips, but it seems that the Mini does not read this timings from the embedded SPD configuration, so it uses the CL2 capable modules with the standard "slow" CL3 timing.
We could gain speed without any modification to the processor if we would just enable the chipset's memory controller to perform CL2 timing when CL2 modules are being used.
But I do not have any access to the chipset's specs nor to the firmware (besides reengineering of the binary, of course).
Having a 1.5x faster RAM memory I'd guess that you will recognize a difference. And you would not likely be able to increase the CPU core frequency by the same factor. Given the limited size of the cache memory the RAM speed is one of the key factors in regards to overall performance. Why should gamers have paid lots of money for this kind of memory modules in the past?
However, currently I do not have any comparison benchmark result for this.
What I found is:
arstechnica.com
forums.anandtech.com
However, currently I do not have any comparison benchmark result for this.
What I found is:
SDRAM CL2 CL3
Is the difference between CL2 and CL3 gains enough to shell out a few more bucks for CL2?What do you consider to be good reliable memory (micron, corsair, nec, samsung, infineon, siemens, etc..)? View image: /infopop/emoticons/icon_rolleyes.gif
arstechnica.com
How much faster is CL2 vs CL2.5 vs CL3?
I'm looking at setting up a new box based on the Epox 8RDA+ with 1GB of PC2700 DDR and a Barton chip (not sure of which one yet). I've been lurking around PC's for a long time but don't know the answer to this question: In the real world (not synthetic benchmarks) how much performance...
forums.anandtech.com
I’m pretty certain that’s what I pulled out of my June 2004 G5 (following contortion-level troubleshooting behind the probable cause(s) of KPs due to overheating in my nominally hotter-than-average tower. I pulled them out because all the other slots were filled with PC-3200/333MHz 1GB sticks, and these two were the oddballs (and were only recognized as 1GB DIMMs by the G5).
Was this what you were looking for?
Attachments
By using this thread’s graphite pencil method in lieu of the resistors, I finally got past my hesitations and went ahead to try to overclock my key lime clamshell iBook the other day to the next step (from 66MHz bus clock to 82.5MHz, using this guide, which would yield a 577MHz CPU clock speed).
The good news is I can attest the graphite method works. It works flawlessly.
The bad news is I can get to Open Firmware and can execute a few basic commands or a partial verbose boot before the system freezes (with the latter producing an odd link/load error associated with IONDRVSupport.kext which does not occur when the bus speed is set to OEM 66MHz).
So the overclock didn’t work, but it afforded me a chance to replace the CPU and memory controller’s OEM thermal pads (which were oily and crumbling) with some Noctua thermal paste, so the experience was still an overall good outcome.
Next up (possibly off-topic?):
I hope to source a 600MHz or 700MHz PPC750CXe chip from a junked “Summer 2001” iMac — the CXe variants designed to run on 100MHz system buses — and to try to transplant that to the iBook’s logic board to configure with either 82.5MHz or 99MHz. I will, of course, need to hire one of you BGA-soldering whiz kids for the transplant.
The good news is I can attest the graphite method works. It works flawlessly.
The bad news is I can get to Open Firmware and can execute a few basic commands or a partial verbose boot before the system freezes (with the latter producing an odd link/load error associated with IONDRVSupport.kext which does not occur when the bus speed is set to OEM 66MHz).
So the overclock didn’t work, but it afforded me a chance to replace the CPU and memory controller’s OEM thermal pads (which were oily and crumbling) with some Noctua thermal paste, so the experience was still an overall good outcome.
Next up (possibly off-topic?):
I hope to source a 600MHz or 700MHz PPC750CXe chip from a junked “Summer 2001” iMac — the CXe variants designed to run on 100MHz system buses — and to try to transplant that to the iBook’s logic board to configure with either 82.5MHz or 99MHz. I will, of course, need to hire one of you BGA-soldering whiz kids for the transplant.
Attachments
Make sure you're running PC100 or faster RAM, as I think these BUS overclocks also OC the RAM, Old PC66 may not cut 82.5Mhz and surely won't do 100Mhz.By using this thread’s graphite pencil method in lieu of the resistors, I finally got past my hesitations and went ahead to try to overclock my key lime clamshell iBook the other day to the next step (from 66MHz bus clock to 82.5MHz, using this guide, which would yield a 577MHz CPU clock speed).
The good news is I can attest the graphite method works. It works flawlessly.
The bad news is I can get to Open Firmware and can execute a few basic commands or a partial verbose boot before the system freezes (with the latter producing an odd link/load error associated with IONDRVSupport.kext which does not occur when the bus speed is set to OEM 66MHz).
So the overclock didn’t work, but it afforded me a chance to replace the CPU and memory controller’s OEM thermal pads (which were oily and crumbling) with some Noctua thermal paste, so the experience was still an overall good outcome.
Next up (possibly off-topic?):
I hope to source a 600MHz or 700MHz PPC750CXe chip from a junked “Summer 2001” iMac — the CXe variants designed to run on 100MHz system buses — and to try to transplant that to the iBook’s logic board to configure with either 82.5MHz or 99MHz. I will, of course, need to hire one of you BGA-soldering whiz kids for the transplant.
Trust me, I have no PC-66 RAM anywhere in my home, and this 512MB stick I’m using consistently tests the quickest with both Xbench and GeekBench (of all the 512MB PC-100/PC-133 SO-DIMMs I have handy). It’s a Samsumg unit with 16 discrete RAM modules (instead of the usual 8) on it, and I think it was sourced from OEM memory on my long-dead PB Ti G4/400.
EDIT: Though, now that I think of it, the on-board 64MB is probably 66MHz RAM. If so, then I‘d be perplexed how any past overclocking successes with other clamshell iBooks managed to ever happen at all.
Last edited:
I dont think so, I suspect those where sold together as a 2GB kit and so the sticks where labeled with the total capacity of the kit
Kingston are known for doing this as well which is very annoying/frustrating at times
id double check the actual speed of the onboard RAM chips you may find they are 100Mhz rated
these G3 CPUs dont generally tend to overclock very far, so im not surprised it fell over at 577Mhz
On the first, I think you’re right. I found this in web archive which reveals those DIMMs were indeed probably 2x1GB and marketed all sly, but it is noteworthy that they were also the only pair in my G5 which were rated at 400MHz and not 333MHz:
http://web.archive.org/web/20060107...e/details_en.asp?prodtechmate=SXP-2G/PC3200K2
As for the onboard 64MB on the iBook, do I know of a way to verify their rated clock speed? System Profiler, my go-to, simply shows the speed as “Built-in” — and given how the logic board is set to 66MHz, I doubt Apple would have gone out of their way to put in faster-rated (and at the time, undoubtedly more spendy) RAM for on-board.
Via firmware, is it possible to deactivate onboard RAM temporarily and to force the system to run only on the added SO-DIMM? If so, then I might try to move the bus clock speed back to 82.5MHz and test-boot without on-board memory.
By the way, back on the topic (subtopic?) of Mac mini and firmware flashing, do we know how to edit back the machine serial number back into the ROM, before flashing?
I'm willing to flash my 1.5GHz PowerMac10,2 Mac mini with the 7448-patched PowerMac10,1 ROM, but I'd like to keep my existing serial to prevent incorrect accusations from the MorphOS staff that my serial is changed because of some license-bypassing reason or some other similar nonsense. (Their license morphos.key file is bound to a machine's serial number.)
I'm willing to flash my 1.5GHz PowerMac10,2 Mac mini with the 7448-patched PowerMac10,1 ROM, but I'd like to keep my existing serial to prevent incorrect accusations from the MorphOS staff that my serial is changed because of some license-bypassing reason or some other similar nonsense. (Their license morphos.key file is bound to a machine's serial number.)
I would hold off on that, I'll be looking into it further and I'll see if I can figure out how to make the 10,2 actually boot an OS with the 7448.
Decided to try some of this overclocking on my 1.42GHz G4 Mini and it runs great at 1.67GHz. I used a tiny blob of solder instead of a pencil though. Not because I wanted to, I just couldn't actually find a pencil in the house.
Edit: Well, it made it through a ~30 minute Cinebench run so I’m calling it stable.
Edit: Well, it made it through a ~30 minute Cinebench run so I’m calling it stable.
Last edited:
If it was a daily driver or mission critical system, sure, but for a computer I turn on once or twice a month to screw around with it's fine.