Really, the best way to do it is just to download the datasheets/Application Guides to the processors, and key in on the part called "Electical Specifications." It's usually a table. Intel tends to have it's own way of doing things, but their method may actually make more sense to someone who's not used to looking at electrical specifications all day.
They tend to have "typical," "maximum," and "absolute maximum" ratings. A friend of mine was involved in a project that tried to break a P4 by running strings of high-load instructions to get it past the abs max. He ended up pulling it off by getting the current flux so high that it magnetized some transisitors somewhere and created calculation errors. But unless you're trying to do something ridiculously synthetic, the "typical" metric is the number that counts.
But, otherwise, if you're interested in mobile chips (XScales, ARM9s, PPC405, etc), I can probably type you up a quick comparo.
Yes, you can be sure I have already done some searching around on Google and various manufacturer datasheet pdf's. I guess the way it is specified did not give me great confidence that I found what I was looking for. From what I gather, the PPC 750FX is somewhere around the 1.4-1.5 v range and 6.5 amps under max load for 800 Mhz operation. What I really wanted to know is what it runs at under 900 Mhz operation (consistent to the way Powerlogix runs their Z900 product). I suppose I could just plot out the data points to assess linearity and extrapolate for 900 Mhz, if I weren't so lazy. :P I guess its the 1.4-1.5 v that just sounds weird to me. I thought processors ran on stuff like 3.3 and 5 v taps, but this has probably become less so now that processors have really been miniaturized with high clockrates and minimized power dissipation.
So that is the data I have unveiled for an 800 Mhz G3. What I want to compare to (though it would have been nice to see how all kinds of common processors these days compare) is the much older 350 Mhz G3 (have no idea what "PPC 750 x" series it was), used in BW Powermacs. That is the one that has been difficult to dig up data on.
"What for", you may ask? I was just wondering if the stock 200 W power supply might be running too close to its limits with a 900 Mhz G3 upgrade. This is more academic study than design plausibility, at this point for me. So forgive me if this sounds like a bizarre request. Any additional input on the topic would be appreciated.
"What for", you may ask? I was just wondering if the stock 200 W power supply might be running too close to its limits with a 900 Mhz G3 upgrade. This is more academic study than design plausibility, at this point for me. So forgive me if this sounds like a bizarre request. Any additional input on the topic would be appreciated.
6.5amps (6500mA . . . puts things into better perspective) sounds a bit steep for the 750. That would be 9.1W at 1.4V. As far as the 3.3v issue, your chip has a very precise DC-DC converter on the daughtercard that is about 85% efficient. Some of the very newest ones will be 90% efficient, and I know Powerlogix makes a big deal in their literature about having well engineered DC-DC converters. So don't expect your CPU to draw more than 10W (12W if you're overclocking) of 3.3V power.
Anyway, I bet your 200W supply will have no trouble with the top-of-the-line G3, given also the fact that your 3.3V power is largely untapped unless you have a bunch of PCI cards. If worse comes to worse you could do like I did with an old 7500, and modify the power supply to dump out more 3.3V current.
The Samsung S3C2410 maxes out around 150mA (80mA typ) at 203Mhz! We use it in a lot of our products, but it's also worthwhile to note that, at high loads, you'll be hard pressed to find a chip of any kind that offers better performance per watt than does the 750fx.
Very good info you present, Splinemodel! This is exactly the sort of discussion I was needing.
The power supply modifying thing for more current- is that a complicated thing to do or can it be done by a novice that has like "soldered a wire or 2"? Are there any downsides to doing so? Reduced stability or something? Shortened service life?
So, I'm gathering that regardless of the CPU generation, it is the 3.3 v line that ultimately feeds the CPU, not the 5 v line (which is typically for motherboard and associated control circuits)?
I've noted the specs on the stock power supply. It gives 20 amps for either of the 3.3 v and 5 v lines, but no more that 100 W max for the both of them together. So I guess 20 amps would cover the consumption of a 750fx under pretty much any worst case condition you can think of...
Now considering that you wouldn't rely on power supply specs to hold rock steady under "max" conditions, maybe a more conservative output rating (if you want things to run predictably and reliably) is 50-60 W cont for the 3.3 and 5 v lines together? Also consider that this stock power supply is getting on in age (not quite 10 years of use, but getting there). So maybe it isn't exactly capable of the ultimate output/stability it once was rated for (though it seems to work just fine, as far as powering all original equipment in the computer). Is it possible that it is now getting juuuuust close enough to cause difficulty when you drop in a 750fx upgrade clocked at 900 Mhz? The only other things I have installed is an internal 80 GB HD, 384 MB of RAM, and the original ATI Rage128 videocard. Do you suppose that plus the motherboard is enough to take an appreciable number of watts from that conservative 50 W power supply baseline, such that the CPU power line may run a bit lean under demanding conditions?
On one hand, it still seems farfetched that this 200 W power supply is overmatched, but on the other, it actually seems apreciably cutting it close once you consider all the extra "details", losses, and safety factors, no?
I have yet to do a voltage sag test on the power supply upon start up, but I suppose that will be coming soon for grins and giggles...
I wouldn't worry too much. Just make sure your power supply isn't dusty. In the end it's a collection of soild state, macroscopic devices. The only enemy of solid state devices is conductor creep (aka metallic drift). That is, individual particles actually move very slowly in an electric current, but eventually they can cause the material to drift, which causes problems. Fortunately for you, this takes a lot longer than 10 years in a power supply.
As for modifying a power supply, well, there's not much documentation on that, and I wouldn't recommend making a drastic change to your output current. I got another amp out of the 5V line by changing some resistor values, and in retrospect it definitely wasn't worth the time. I assume you have a B&W G3. Those have the 200W PS and the ZIFF. . . I wouldn't worry about the power situation at all. I have a Sawtooth G4 (220W PS) with Dual G4 500s, 512MB of RAM (4x 128MB modules), one 120GB ATA drive, 2x 68GB Seagae Cheetahs, an ATTO PCIExpress Dual SCSI 160 card, a GeForce 3, and an Adaptec 2930. No problem. The thing runs 24/7 as a server. The dual G4's alone probably make up the 20W difference.
Well I guess that really puts things in perspective!
So now my "CPU voltage and current" topic drifts to a "why doesn't my Powerlogix Z900 work" topic. \ I don't have too many avenues left other than something in the motherboard just cannot stay in sync with this upgrade or there is something flakey with this CPU chip (that the Powerlogix people would rather burn in hell before admitting that their product may be faulty and worthy of replacement).
Comments
Generally that information requires deep digging in a companies PDF files.
Google search should get you a fair bit though.
I have a few numbers of Freescale's processors sitting around, was there anything specific or were you running a comparison?
They tend to have "typical," "maximum," and "absolute maximum" ratings. A friend of mine was involved in a project that tried to break a P4 by running strings of high-load instructions to get it past the abs max. He ended up pulling it off by getting the current flux so high that it magnetized some transisitors somewhere and created calculation errors. But unless you're trying to do something ridiculously synthetic, the "typical" metric is the number that counts.
But, otherwise, if you're interested in mobile chips (XScales, ARM9s, PPC405, etc), I can probably type you up a quick comparo.
So that is the data I have unveiled for an 800 Mhz G3. What I want to compare to (though it would have been nice to see how all kinds of common processors these days compare) is the much older 350 Mhz G3 (have no idea what "PPC 750 x" series it was), used in BW Powermacs. That is the one that has been difficult to dig up data on.
"What for", you may ask? I was just wondering if the stock 200 W power supply might be running too close to its limits with a 900 Mhz G3 upgrade. This is more academic study than design plausibility, at this point for me. So forgive me if this sounds like a bizarre request. Any additional input on the topic would be appreciated.
Originally posted by Randycat99
"What for", you may ask? I was just wondering if the stock 200 W power supply might be running too close to its limits with a 900 Mhz G3 upgrade. This is more academic study than design plausibility, at this point for me. So forgive me if this sounds like a bizarre request. Any additional input on the topic would be appreciated.
6.5amps (6500mA . . . puts things into better perspective) sounds a bit steep for the 750. That would be 9.1W at 1.4V. As far as the 3.3v issue, your chip has a very precise DC-DC converter on the daughtercard that is about 85% efficient. Some of the very newest ones will be 90% efficient, and I know Powerlogix makes a big deal in their literature about having well engineered DC-DC converters. So don't expect your CPU to draw more than 10W (12W if you're overclocking) of 3.3V power.
Anyway, I bet your 200W supply will have no trouble with the top-of-the-line G3, given also the fact that your 3.3V power is largely untapped unless you have a bunch of PCI cards. If worse comes to worse you could do like I did with an old 7500, and modify the power supply to dump out more 3.3V current.
The Samsung S3C2410 maxes out around 150mA (80mA typ) at 203Mhz! We use it in a lot of our products, but it's also worthwhile to note that, at high loads, you'll be hard pressed to find a chip of any kind that offers better performance per watt than does the 750fx.
The power supply modifying thing for more current- is that a complicated thing to do or can it be done by a novice that has like "soldered a wire or 2"?
So, I'm gathering that regardless of the CPU generation, it is the 3.3 v line that ultimately feeds the CPU, not the 5 v line (which is typically for motherboard and associated control circuits)?
I've noted the specs on the stock power supply. It gives 20 amps for either of the 3.3 v and 5 v lines, but no more that 100 W max for the both of them together. So I guess 20 amps would cover the consumption of a 750fx under pretty much any worst case condition you can think of...
Now considering that you wouldn't rely on power supply specs to hold rock steady under "max" conditions, maybe a more conservative output rating (if you want things to run predictably and reliably) is 50-60 W cont for the 3.3 and 5 v lines together? Also consider that this stock power supply is getting on in age (not quite 10 years of use, but getting there). So maybe it isn't exactly capable of the ultimate output/stability it once was rated for (though it seems to work just fine, as far as powering all original equipment in the computer). Is it possible that it is now getting juuuuust close enough to cause difficulty when you drop in a 750fx upgrade clocked at 900 Mhz? The only other things I have installed is an internal 80 GB HD, 384 MB of RAM, and the original ATI Rage128 videocard. Do you suppose that plus the motherboard is enough to take an appreciable number of watts from that conservative 50 W power supply baseline, such that the CPU power line may run a bit lean under demanding conditions?
On one hand, it still seems farfetched that this 200 W power supply is overmatched, but on the other, it actually seems apreciably cutting it close once you consider all the extra "details", losses, and safety factors, no?
I have yet to do a voltage sag test on the power supply upon start up, but I suppose that will be coming soon for grins and giggles...
As for modifying a power supply, well, there's not much documentation on that, and I wouldn't recommend making a drastic change to your output current. I got another amp out of the 5V line by changing some resistor values, and in retrospect it definitely wasn't worth the time. I assume you have a B&W G3. Those have the 200W PS and the ZIFF. . . I wouldn't worry about the power situation at all. I have a Sawtooth G4 (220W PS) with Dual G4 500s, 512MB of RAM (4x 128MB modules), one 120GB ATA drive, 2x 68GB Seagae Cheetahs, an ATTO PCIExpress Dual SCSI 160 card, a GeForce 3, and an Adaptec 2930. No problem. The thing runs 24/7 as a server. The dual G4's alone probably make up the 20W difference.
So now my "CPU voltage and current" topic drifts to a "why doesn't my Powerlogix Z900 work" topic.
Any suggestions?