Amiga.org
Amiga computer related discussion => Amiga Hardware Issues and discussion => Topic started by: Hyperspeed on September 02, 2004, 02:07:48 AM
-
I was just wondering...
If your Motorolla CPU got too hot, would it switch into a slower mode
or disable caches?
I'm thinking in particular of the '040 but I've performed a
SysSpeed2.6 test on a Blizzard 1260 and occasionaly the module will
register it as a 49Mhz 68060, sometimes a 50Mhz.
This could be the Blizzard slowing down when 2x SIMMs are installed,
but I wondered if any 68k chips had intelligient features like this.
No ATX style features that can be accessed through software, such as
temperature reading?
I suppose if there were we'd have seen something on Aminet by now
though...
-
I know for sure when you do a real-time cpu clocking check on a PC ,it does tend to drop some or even higher some CPU speed, but not enough for it to be noticeable by a megahertz. But who knows, try to keep taking the tests and see if by chance it wasnt checking it at a bad time? :-)
-
I had an A4000 at one time. I'd have to say yes, that the 040 CPU has some kind of thermal throttle. It had a heat problem. The case would get hot. Hotter than I though it should be. After about a half hour to an hour, depending on the ambient temp, I noticed that it would slow down. Especially on hot summer days when it would come to a complete hault. After I install a 80CF/M case fan, the CPU had a HSF, I had no more heat problems. The case always felt cool after that.
-
@Argo
Installing fans can make something be cool and suck at the same time. :-D
-
As I understand it, all CPUs slow down a little when they get hot, it's something to do with Electrons not being able to travel quite as fast once the medium is hot, or something. It's a vauge memory of something I was once told in a Microprocessor lesson. The CPU slow-down is of course quite small.
I have a total of 6 fans running in my A1200 system. Everything runs cool as a cucumber in there, and it's not actually that loud. You need to shop around for decent fans :)
Brian
-
But who knows, try to keep taking the tests and see if by chance it wasnt checking it at a bad time?
LOL :) CPUs will be calling in sick next and having time off work.
Kev
-
According to the Motorola 68060 datasheet, there's nothing to indicate it does anything different when the temperature gets high. It does have a built in thermistor which you can use with external circuitry to sense the junction (die) temperature. If you got a bit clever with an op-amp, you could connect it to the analogue to digital converter in the joystick port, and write a program to display the junction temperature on workbench.
Don't forget you can disable parts of the CPU in software, like caches and things which will reduce it's heat dissipation, but execution speed will suffer.
Maximum junction temperature is typically around 110°C, so the device should operate normally up to that point. If you exceed that rating, it's likely to do bad things.
If you're worried about CPU cooling, there is an entire section in the datasheet about it, along with some simple calculations and tables for heat dissipation, thermal resistance of the case, etc.
As for the clock speed reported by your program, it probably measures the timing of a certain clock signal to get that info. The crystal oscillator's output (which clock signals are derived from) is a fixed frequency, and there's definately no way it's going to change according to CPU temperature.
Most likely your oscillator is about 49.9MHz or something, so depending on the way the program measures and reports whatever clock source, it may sometimes get 49MHz, sometimes get 50MHz.
Use a real frequency counter if you really want to remove all doubt.
-
I had that problem with a revision 3.0 A3640. The machine would get slower and slower until it would lock up. It wasn't overheating but as it warmed up, the cpu board wasn't reading a zorro signal properly. More heat == more resistance. The problem would show up in summer but the machine was ok in winter. If I removed all zorro cards, the problem would go away. It would also stop if I removed the case cover and put a fan blowing on the cpu board. This was on an A4000D.
-
When I bought my first A4000, back in 1994, I bought it from a rather dodgy character of a certain ethnic minority that does not have a reputation for honesty. I tested it in his house, it ran OK, but once I got it home, I found that after an hour of running with the case on, it used to slow down and finally stop.
There was an animation on some PD disk of a dog running. If I left this anim running, it would show the slowdown effect quite dramatically. With some freeze spray, I traced it to one of the PGAs under the CPU card. It cost less than $10 to have it replaced with a later version, and I've never had any trouble in the ten years since then.
tony
-
tonyw wrote:
With some freeze spray, I traced it to one of the PGAs under the CPU card.
Could you explain what this procedure is all about? Freeze spray :-?. It sounds this is knowledge I could use in feature endeavours with my A4k :-).
-
I don't think they have Intel style 'clock throttling' built in. Even AMD didn't introduce that until recently AFAIK.
Old references (http://www4.tomshardware.com/column/20011029/counterstrike-01.html)
-
BrianJHoskins wrote:
As I understand it, all CPUs slow down a little when they get hot, it's something to do with Electrons not being able to travel quite as fast once the medium is hot, or something. It's a vauge memory of something I was once told in a Microprocessor lesson. The CPU slow-down is of course quite small.
I have a total of 6 fans running in my A1200 system. Everything runs cool as a cucumber in there, and it's not actually that loud. You need to shop around for decent fans :)
Brian
6 fans?!?
It seems that you don`t have optimal casing for your A1.2k. Oh, otherwise you may have PPC+BVision:) and veeery hot hdd:)). My system (A1240T) uses ONE FAN (driven by a thermistor), fitted on a CPU. I measured total power consumption and noticed that it`s about 60-70Watts, so I removed noisy fan from PSU. I checked how long it takes the 040 to overheat and noticed that it must work for at least 5 hours to overheat (DigiBoosterPro with 10CH mod on HiFi stereo 14bit: 33kHz, tested on August this year, with 28C outdoor). Before it crashed due to overheat, I haven`t noticed any reasonable slowdown. Maybe it`s like you say about those Electrons, only. So I decided to leave only one fan, controlled by the CPU temperature. The computer is so quiet I can`t hear it when I stand in the door of my room (2m away:), especially after replacing rather noisy Caviar with Seagate Barracuda...
-
GadgetMaster wrote:
I don't think they have Intel style 'clock throttling' built in. Even AMD didn't introduce that until recently AFAIK.
Old references (http://www4.tomshardware.com/column/20011029/counterstrike-01.html)
It has been around since 486 atleast in one form or the other.
The 030 in some Powerbookmodels could be switched down to a 16Mhz powersaving mode!
-
GadgetMaster wrote:
I don't think they have Intel style 'clock throttling' built in. Even AMD didn't introduce that until recently AFAIK.
Old references (http://www4.tomshardware.com/column/20011029/counterstrike-01.html)
AMD PowerNow has been included in mobile edition btw.
-
@Hyperspeed
as already pointed by Castellen, thers no way to determinate the exact clock via software even using the good WhichAmiga by Piru.
WhichAmiga is the most precise tool for the 68k side, but sometimes fails to check the exact clock (and FSB also) of a PPC (just a little).
About the heating, not only the CPU is involved.
I have followed the optimal suggestions of Dave's page (a little fan on the ram controller) CSPPC related but useful even for other enviroments.
EDIT- btw . . .nice Michal heatsink (http://www.amiga.org/gallery/index.php?n=689=13)