System Profiler Software for Overclocks

[cobalt60]

Just booted my 7300 for the first time yesterday, and the first time I booted a PPC604 Mac in like, 15 years. This machine is going to be used for bus overclock testing 3 different mainboards; a 7300, a 7600, and an 8600. I am using an XLR8 ZIF Carrier with a 350MHz G3 750L, 320MB 60ns EDO RAM non-interleaved, and Mac OS 8.6.

I set it to run a 45MHz bus and 4.5X multiplier (202.5 MHz); Apple System Profiler reported a CPU speed of 200MHz. Figured it was rounding. Next set 40MHz x 4.5 (180MHz); Apple System Profiler again reported 200MHz. Seems like it's just reporting the original speed of the machine.

What software is available that will accurately report the CPU speed, if any?
Is there software that will also report the bus speed and/or multiplier?
What should I use to stress test the bus over-clocks?

 

[cobalt60]

Found most of my answer, and I'm sure many are aware of it; MacBench. It shows actual CPU and bus speed.

 

[joshc]

Clockometer should be pretty accurate in showing the actual CPU clock speed.

 

[Phipli]

And Newer Gauge Pro, possibly Sonnet Metronome. Much prefer Gauge Pro.

 

[Phipli]

https://macintoshgarden.org/apps/gauge-pro

 

[cobalt60]

Awesome, I will try them all. Normally on my newer PCs, my stress test consists of encoding videos; any suggestions on what I can use here? I usually prefer real world CPU intensive apps over dedicated stress-test programs, but that may be more due to superstition (over fears of running chips unecessarily hot and hard.
And actually in the case of a bus overclock, I guess I'd want to focus on memory benchmarks and stress tests; is that true? MacBench 5 actually didn't seem to have a memory benchmark, any suggestions on how to test that?

 

[Phipli]

Awesome, I will try them all. Normally on my newer PCs, my stress test consists of encoding videos; any suggestions on what I can use here? I usually prefer real world CPU intensive apps over dedicated stress-test programs, but that may be more due to superstition (over fears of running chips unecessarily hot and hard.
And actually in the case of a bus overclock, I guess I'd want to focus on memory benchmarks and stress tests; is that true? MacBench 5 actually didn't seem to have a memory benchmark, any suggestions on how to test that?

I usually just run a few benchmarks back to back to be honest, that or play MP3s at a high bitrate. If you allocate the minimum of RAM to the MP3 software, it will force it to use the disk a lot too.

 

[Byrd]

Norton System Info is pretty accurate with CPU speeds, and the completed benchmark, CPU/FPU speeds will also tell you if your overclock compares to similar speeds.

I've also found NSI a decent stress test, it was the only program that would glitch out my SE/30 with Daystar Powercache clone that was giving me graphical issues. Another good one is Unstuffing a large file in the background ... while running the Graphing Calculator demo, for example. The Quake 1 timedemo another one. The lack of preemptive multitasking means "stressing" the CPU isn't as easy as other OSes.

 

[cobalt60]

60MHz froze on boot. I had what appeared to be a stable system at 58MHz. However, un-stuffing some small installers gave a "could not verify data integrity" or some such message. I suspected it could have been related to the bus overclock, and so tested lower speeds. 55MHz never gave the error, 57MHz gave it about 50% of the time. Am I correct to assume there is a decently high likelihood that the bus overclock is indeed causing this issue?

Also want to point out, MacBench 5 seems quite inaccurate when reporting bus and CPU speed. I mean no more than 1 or 2MHz on the bus, but thats still not great.

play MP3s at a high bitrate

If this causes failure, is it always going to be a system freeze? Or might there be a less obvious symptom like stuttering?

Norton System Info

Will download now, thanks.

 

[Phipli]

Norton System Info is pretty accurate with CPU speeds

I don't agree, it is often incorrect with accelerators.

 

[joshc]

I don't agree, it is often incorrect with accelerators.

This is my experience of it too. I think Norton is OK with reading the clock speed of stock systems.

 

[Phipli]

If this causes failure, is it always going to be a system freeze? Or might there be a less obvious symptom like stuttering?

Stuttering is normal behaviour for playing mp3s in the 90s

I was thinking getting it hot until it crashes.

 

[cheesestraws]

I usually prefer real world CPU intensive apps over dedicated stress-test programs, but that may be more due to superstition (over fears of running chips unecessarily hot and hard.

A computer really ought to be able to run at 100% indefinitely if it is to be considered stable. If it can’t it’s either overclocked too far or the cooling isn’t up to the job

 

[Phipli]

A computer really ought to be able to run at 100% indefinitely if it is to be considered stable. If it can’t it’s either overclocked too far or the cooling isn’t up to the job

 

[cobalt60]

I've had computers run fine at 100% reported CPU use for extended periods, but crash (and/or run uncomfortably hot, cant remember its been a while) after a short time using a dedicated stress test program. So I still always run the CPU at "100%" for testing.

 

[CC_333]

Not relevant here per se (except perhaps to later G4s, G5s and Intel Macs), but with multi core or multi processor systems, to max them out "at 100%, you actually need to run them at anywhere from 200% (for a dual core CPU a pare of single core ones), up to 800% or more (for quad core CPUs with hyperthreading).

But for 68k Macs, 100% is probably more than enough.

c

 

[cobalt60]

That's why I normally use encoding, it can use all 24 threads my CPU has. I also have a desktop resource monitor (conky) display at all times my resource use, including per thread use. Normally encoding would push all cores to 100%, but on my 12900K it only pushes to like maybe 85%. So I do have to figure out now a way to actually push to 100%; probably can just use different encoding parameters.
Compiling is another good way to push all cores to 100% CPU use

 

[cobalt60]

by "per thread use" I meant "per logical core". Was trying to account for the hyper-threading.

 

[bigmessowires]

Did you find that MacBench was able to detect when you changed the bus speed? I'm trying to understand why my 6214CD (that was apparently motherboard-swapped for a 6300) reports in System Profiler as 37.5 MHz bus, 2.5x multipler, 94 MHz CPU speed. I'm guessing the System Profiler doesn't actually know the bus speed, and guesses incorrectly based on the Gestalt ID.

 

[cobalt60]

Did you find that MacBench was able to detect when you changed the bus speed?

My current machine has a 52MHz bus and 10x multiplier.

Apple system profiler : 52 .. | 520
Gauge Pro ........... : 52.0 .| 519.9
MacBench 5 .......... : 53.25 | 532.5
CPU Director ........ : 52 .. | 519

So MacBench is quite a ways off. I think GaugePro is probably the most accurate.

I'm guessing the System Profiler doesn't actually know the bus speed, and guesses incorrectly based on the Gestalt ID.

I believe this happens on New World Macs, but didn't think it did on Old World.