Investigating Titanium CPU Throttling

[Doq]

tl;dr:​

Titanium PowerBooks (at least the 3,4 and 3,5, possibly others) will throttle the CPU and disable the L3 cache if they detect that either a bad battery or no battery is present (specifically, it will throttle if it detects that the battery is below 5%). You can also set this state manually (i.e. to preserve life if on battery) by selecting Slower Processor Speed in the Control Strip, at least on OS 9.

---

We now resume our original post:

Here's the bit of context for why this thread exists:
Earlier today and yesterday I posted a picture in the What have you done with a PowerPC today? thread showing a 1 GHz Titanium PowerBook G4 next to a 1 GHz Aluminium PowerBook G4 and some Geekbench 2 scores. I and some others had pointed out that the Titanium score was rather suspect and not quite in line with what the average score for this machine actually is-- according to EveryMac, this should be about 556, near the 542 that the Aluminium got.

This thread is to figure out why.

When I did the tests as shown in the picture, I ran the Titanium's test using the Aluminium's boot volume in Target Disk Mode. That gave the shown result of 303. This led to a few replies inquiring about login items and hidden energy hogs, which, while possible, I deemed wasn't relevant since if this was the case, the Aluminium's score would have been hampered as well.

Following that, I was replied to by Amethyst1, suggesting that Spotlight might have been doing some shenanigans. Sure enough, when I rebooted in, Spotlight was doing some shenanigans. After turning that off, I reran Geekbench, twice, and got a score of 349. An improvement, yes, but still pretty suspect compared to what it should be.

I pointed out that in OS 9's System Profiler, the CPU speed reported both 667 and 1000 MHz. What makes this more suspect is that OS X does not report this anywhere in the system-- not in System Profiler, About This Mac, or even in the Terminal.

Now at a dead end, I did some research, and from what little I could find, I found an answer that stated that the CPU will throttle itself down if there is no battery or if the battery is bad. I don't think this is correct, because if it was, wouldn't the Aluminium-- which had its battery go bad a while ago-- also throttle down? I can't pull up the System Profiler from OS 9 because OS 9 doesn't run natively on the Aluminium, and Classic gimps results.

Should I consider getting or rebuilding a battery to see if this is actually true?
---
As I was writing this thread, I got another reply suggesting that it's thermal. I'll get my hands on some Arctic MX4 sometime this month to see if it helps.

 

[B S Magnet]

Now at a dead end, I did some research, and from what little I could find, I found an answer that stated that the CPU will throttle itself down if there is no battery or if the battery is bad. I don't think this is correct, because if it was, wouldn't the Aluminium-- which had its battery go bad a while ago-- also throttle down? I can't pull up the System Profiler from OS 9 because OS 9 doesn't run natively on the Aluminium, and Classic gimps results.

Unlike the Intel-based Mac laptops, the absence of a battery (or the presence of a dead battery) has no impact on CPU speed with PowerPC laptops. To wit, the PowerBook I use for testing Snow Leopard has never had a working battery.

Should I consider getting or rebuilding a battery to see if this is actually true?
---
As I was writing this thread, I got another reply suggesting that it's thermal. I'll get my hands on some Arctic MX4 sometime this month to see if it helps.

The one thing I might try is to install OS X from scratch from the install DVD/iso on the titanium. Fresh start for it, and it further assures that any initial configurations handled by installation scripts are fine-tuning the install for the titanium environment.

Of course, running iStat menus on both laptops should give you a sense of thermals (though there will likely be fewer sensors on the titanium), and fresh thermal paste on any computer is always a good praxis to follow.

But I think there may be something worth exploring in a fresh installation of OS X itself specific to the titanium’s hardware environment. If you want to use the same HDD as the aluminium, perhaps split it into two even partitions and, whilst inside the titanium, run the installer to install to the new partition. That way, the aluminium can still use the original partition to boot and the titanium can use the new partition for its boot.

 

[Amethyst1]

the absence of a battery (or the presence of a dead battery) has no impact on CPU speed with PowerPC laptops.

IIRC @LightBulbFun discovered that the 867MHz/1GHz Ti does downclock and disable L3 when the battery is gone.

 

[AphoticD]

IIRC @LightBulbFun discovered that the 867MHz/1GHz Ti does downclock and disable L3 when the battery is gone.

I can confirm this is the case.

I just booted a Titanium 867MHz into OS9 and can see it steps down to 666MHz confirmed as “Slower Processor Speed” in the Energy Saver control strip (which won’t accept a mode change) if there is no battery installed.

Once I install the battery I can switch to “Faster Processor Speed” and ASP reports 866MHz.

Gauge Pro incorrectly reports it is a 601 @ 827.8MHz with no L2/L3 cache on Faster and 638.3MHz on Slower mode, however Memory Performance is higher on Slower Mode (216.9MB/sec) vs Faster Mode (185.3MB/sec) which is interesting.

Both modes report the presence of a “1MB L2 cache” in ASP, with no mention of L3, but booting into Tiger shows the 256K L2 + 1MB L3 in the Processor Prefpane from CHUD Tools.

I am not sure if the L3 is being disabled without the battery though.

 

[AphoticD]

In line with the OP’s experience, I just ran geekbench in Tiger on the Ti 867 and got the following results...

With battery: 386
Without battery: 248

Both with the same “Highest” performance setting in Energy Saver prefs.

 

[ervus]

I put a 7457 in a TiBook and can confirm there are some shenanigans. The 7447 based PowerBooks will change speed with DFS (and voltage too), but the Ti with a 7455 has two separate sets of PLL settings that it uses to change speed. I also tried putting 2MB of L3 in it and saw some weirdness there too. I could force it on, but the firmware is doing some stuff that I've not seen on other apples. If you put a volt meter on it you can clearly watch it changing speeds, but I didn't notice that temperatures had anything to do with it.

 

[Doq]

Mmm.... yeah.

Guess I'm grabbing a battery as well. I do have a proper 65W adaptor though so least that's in check.

I do have a battery, but it's aftermarket, and completely shot. Since it is aftermarket, I might consider replacing the cells instead without feeling bad about ripping in to it.

 

[B S Magnet]

IIRC @LightBulbFun discovered that the 867MHz/1GHz Ti does downclock and disable L3 when the battery is gone.

Oh wow. I did not know this! Thank you for the link reference!

Until this, I was entirely unaware of any iBook/PowerBook which underclocked depending on power source. Are there others which behave similarly or was it limited to that one PowerBook revision?

 

[LightBulbFun]

IIRC @LightBulbFun discovered that the 867MHz/1GHz Ti does downclock and disable L3 when the battery is gone.

indeed (and I think the 667Mhz/800Mhz DVI models do the same as well), and this is true for any later Mac too, even intel Macs will down clock their lowest CPU speed if you dont have a working battery installed

on PowerPC macs, you can use Sonnet Metronome to see what speed your CPU is running at and what caches are working (it also worth double checking using the diagnostic tab in system profiler that your L3 cache is actually working, it notorious for falling on these machines and the 1Ghz AlBooks)

 

[B S Magnet]

indeed (and I think the 667Mhz/800Mhz DVI models do the same as well), and this is true for any later Mac too, even intel Macs will down clock their lowest CPU speed if you dont have a working battery installed

I was previously aware of Intel-based MB/MBP models stepping down effective clock speed by half when running only from the MagSafe adapter. I am also familiar with the Dynamic Frequency Switching of the 7447s, but my understanding of that feature is it downsteps/upsteps the CPU on the fly (and mostly transparent to the user) depending on usage demands and not, say, whether the system is running on battery only, mains only, or both.

 

[LightBulbFun]

In line with the OP’s experience, I just ran geekbench in Tiger on the Ti 867 and got the following results...

With battery: 386
Without battery: 248

Both with the same “Highest” performance setting in Energy Saver prefs.

was just looking back at this as I finally got a working (NOS ) battery for my TiBook a week ago

and so I was finally able to add it to my geekbench library as now the processor ran at full speed etc

but I noticed im getting much higher scores then you so I wonder why that is

how much RAM does your TiBook have and what GB version are you using? (and you are using a 65W charger right?)

in tiger under GeekBench 2.2.0 I got a nice round 500


heres my TiBook in tiger with GB 2.2.0

PowerBook G4 (867 MHz) GB2.2.0 - Geekbench

Benchmark results for a PowerBook G4 (1 GHz/867 MHz) with a PowerPC G4 (7445/7455) processor.

browser.geekbench.com

and here it is in Leopard with GB 2.2.7

PowerBook G4 (867 MHz) GB2.2.7 - Geekbench

Benchmark results for a PowerBook G4 (1 GHz/867 MHz) with a PowerPC G4 (7445/7455) processor.

browser.geekbench.com

 

[B S Magnet]

and here it is in Leopard with GB 2.2.7

PowerBook G4 (867 MHz) GB2.2.7 - Geekbench

Benchmark results for a PowerBook G4 (1 GHz/867 MHz) with a PowerPC G4 (7445/7455) processor.

browser.geekbench.com

It’s only when one evaluates the 547 score in 2.2.7 for the 867MHz setup by comparing it with the highest 2.2.7 score I’ve managed with a 1670MHz A1138 of just 984, does one begin to observe the moderate but measurable bump in speed when having an L3 cache on board. It basically amounts, if I did my math correctly, to a 7 or 8 per cent speed bump.

Does this figure pan out? Anyone want to check the math?

 

[ervus]

We were talking about L3 in the chip swapping thread too. From what I've seen it depends a lot on what you're doing. It can make either no difference at all, or up to 80% improvement. In general it helps more on a slower bus and with the 7455 (vs 7457). The biggest improvement I've seen was on a Cube with a 7455: so it was a slow bus and small L2 cache. The software I was running was processing terrain squares into a triangle mesh, so the software was jumping around a lot in a big chunk of memory to interpolate elevation points, not running a little benchmark loop. Processing video files seems to benefit quite a bit from L3 as well. Some data compression software will show the benefit when larger libraries are used.

 

[AphoticD]

how much RAM does your TiBook have and what GB version are you using? (and you are using a 65W charger right?)

Hello @LightBulbFun ,

My Titanium 867MHz has 1GB RAM, GB 2.2.0 on 10.4.11 and 65W power.

I’ll make some time to run the benchmarks again and also on my 667MHz and 550MHz Titaniums to compare.

Hope you are doing well mate.

 

[ervus]

I think the earlier TiBooks used different CPUs. I'm guessing the 550MHz uses a 7410? The 667MHz probably has a 7450 chip, but maybe no L3 installed? I'm trying to remember pictures I've seen of TiBook logic boards... My only experience is with the 1GHz version.

 

[Macbookprodude]

I can confirm this is the case.

I just booted a Titanium 867MHz into OS9 and can see it steps down to 666MHz confirmed as “Slower Processor Speed” in the Energy Saver control strip (which won’t accept a mode change) if there is no battery installed.

Once I install the battery I can switch to “Faster Processor Speed” and ASP reports 866MHz.

Gauge Pro incorrectly reports it is a 601 @ 827.8MHz with no L2/L3 cache on Faster and 638.3MHz on Slower mode, however Memory Performance is higher on Slower Mode (216.9MB/sec) vs Faster Mode (185.3MB/sec) which is interesting.

Both modes report the presence of a “1MB L2 cache” in ASP, with no mention of L3, but booting into Tiger shows the 256K L2 + 1MB L3 in the Processor Prefpane from CHUD Tools.

I am not sure if the L3 is being disabled without the battery though.

Its there. OS 9's system profiler shows the L3 under L2 cache. It could be an error in the system profiler. But Tiger and Leopard and maybe Linux show L3 is working at 1MB.

 

[Sebasboy]

I put a 7457 in a TiBook and can confirm there are some shenanigans. The 7447 based PowerBooks will change speed with DFS (and voltage too), but the Ti with a 7455 has two separate sets of PLL settings that it uses to change speed. I also tried putting 2MB of L3 in it and saw some weirdness there too. I could force it on, but the firmware is doing some stuff that I've not seen on other apples. If you put a volt meter on it you can clearly watch it changing speeds, but I didn't notice that temperatures had anything to do with it.

damn you did what i was wanting to do..

upgrade 1mb to 2mb cache and cpu swap on the ti book..

overclocked my 1ghz to 1.27 ghz and reaching 701 cinebench points..

sadly i broke my battery trying to recell it (it bricked itself) and now i cant seem to get my hands on a decently priced working 3rd party battery...

any ideas?

 

[Doq]

damn you did what i was wanting to do..

upgrade 1mb to 2mb cache and cpu swap on the ti book..

overclocked my 1ghz to 1.27 ghz and reaching 701 cinebench points..

sadly i broke my battery trying to recell it (it bricked itself) and now i cant seem to get my hands on a decently priced working 3rd party battery...

any ideas?

You might be able to repair it yourself with soldering and new 18650 batteries. I'm thinking about doing that to my Titanium's battery since I have an aftermarket battery-- I wouldn't feel comfortable doing this to an OEM battery, even if it was totally shot.

I missed the whole part about you tried to do this but it was too late I already posted. My b.

I'm not sure what you can do other than look for a replacement on the markets? Not sure if you'd have much luck though. The only one I found was untested used-- the untested part being the sketchy part.

 

[Macbookprodude]

damn you did what i was wanting to do..

upgrade 1mb to 2mb cache and cpu swap on the ti book..

overclocked my 1ghz to 1.27 ghz and reaching 701 cinebench points..

sadly i broke my battery trying to recell it (it bricked itself) and now i cant seem to get my hands on a decently priced working 3rd party battery...

any ideas?

What is the point since this can't be noticed ? We are talking maybe 5 percent to 10 percent increase. How does one upgrade the L3 1MB to 2MB ? That would be nice.

 

[Macbookprodude]

Titanium 1ghz batteries are all over the internet - ebay had some on there used but working great - I bought 3 of them for 20 bucks.

 

[Sebasboy]

What is the point since this can't be noticed ? We are talking maybe 5 percent to 10 percent increase. How does one upgrade the L3 1MB to 2MB ? That would be nice.

eh the l2 cache is soldered on the motherboard..

512k on each side of the board.

if you compare a same mhz g4 cpu without any l3 cache you can notice a massive loss in performance.

i saw a few vids from hrutkaymods where he overclocked a few desktop g4's and compared the score between the dual 1.5 ghz g4's vs the dual 2 ghz g4's..

the dual 1.5 had more l3 and was rather close to the 2ghz g4's with (iirc) half the l3 cache

 

[Doq]

The shocking conclusion is here. After taking a chance on an in-box OEM battery on eBay, it arrived, and I popped it in. I saw that it started charging and got all excited, but it was charging awfully slowly so I let it run for a bit. Came back, checked Metronome, saw that it said 1000 MHz with L3 cache and I got unbelievably excited.

The Geekbench results are in, Titanium PowerBook with a good battery:

The battery is still charging as I'm posting this, so while I was in Tiger I pulled up coconutBattery for it to tell me it only has 5 (five!) cycles on it. Super win on that one, might make this Titanium the new Companion after this series of events.

 

[Macbookprodude]

The shocking conclusion is here. After taking a chance on an in-box OEM battery on eBay, it arrived, and I popped it in. I saw that it started charging and got all excited, but it was charging awfully slowly so I let it run for a bit. Came back, checked Metronome, saw that it said 1000 MHz with L3 cache and I got unbelievably excited.

The Geekbench results are in, Titanium PowerBook with a good battery:
View attachment 1834693

The battery is still charging as I'm posting this, so while I was in Tiger I pulled up coconutBattery for it to tell me it only has 5 (five!) cycles on it. Super win on that one, might make this Titanium the new Companion after this series of events.

I could have told you that ! For some reason, titanium batteries are plenty on ebay !!!

 

[Doq]

I could have told you that ! For some reason, titanium batteries are plenty on ebay !!!

You have told me that.

That said, I found one (1). Of all of the 15" PowerBook G4 batteries on my eBay marketplace stateside, there was exactly one for the Titanium, sold as is, and I picked it up on a leap of faith.

It did work out though, so, nice.

 

[ervus]

Thanks for reporting your findings. So did you figure out what the various operating states are? On battery it runs at 667MHz, and on charger it runs at 1000 without L3, and with a good battery and plugged in you get 1000 with L3?

Or is this from changing the "Processor Performance:" setting under "Energy Saver"? Or does that option not appear for the TiBook?