0.13µm AthlonXP 866MHz vs 0.18µm Pentium III 866MHz

Mikael · Post by **Mikael** » Wed Apr 12, 2006 7:14 am

I'm planning to build a quiet Linux box and I have both of the CPUs described below:

AthlonXP 1700+ "Thoroughbred B" 0.13µm
Pentium III 866MHz "Coppermine" 0.18µm

Whatever CPU I end up choosing, I plan to run it at 866MHz, since I really don't need more. I would like to know which of these CPUs that will run coolest at the following settings:

AthlonXP @ 866MHz (1.5V)
Pentium III @ 866MHz (1.65V)

As you can see, the Athlon is slightly undervolted. I wanted to undervolt more, but any voltage below 1.5V produces huge amounts of errors in Memtest86, resulting in general stability issues. I have no idea why the CPU can't operate correctly on 1.4V at 866MHz, but it may just be an extremely bad sample.

Anyway, I don't have a PIII mobo available yet, so I don't know if I may be able to undervolt. The comparison is therefore between the undervolted Athlon and the stock PIII. Which will run cooler?

jaganath · Post by **jaganath** » Wed Apr 12, 2006 7:36 am

The comparison is therefore between the undervolted Athlon and the stock PIII. Which will run cooler?

The P-III, by a loooong way. P-III's TDP at stock clock and voltage is less than half the Athlon at stock clock and volts; frequency has a linear effect on heat output, volts has a quadratic effect.

Mikael · Post by **Mikael** » Wed Apr 12, 2006 7:49 am

jaganath wrote:The P-III, by a loooong way. P-III's TDP at stock clock and voltage is less than half the Athlon at stock clock and volts; frequency has a linear effect on heat output, volts has a quadratic effect.

If I remember correctly, the TDP for the PIII is ~30W. The TDP for a 1.5V 1700+ (1.47GHz) is 49W. If a linear relationship to frequency is assumed, the AthlonXP ends up at the same power consumption as the PIII. Remember that this is not a Palomino 0.18µm core, but the significantly cooler running Thoroughbred B core.

The reason I posted this question in the first place is that it isn't all that clear cut. The specifications indicate that they would be equal in terms of power output, but I wanted to hear from people with first hand experience of the two.

jaganath · Post by **jaganath** » Wed Apr 12, 2006 8:15 am

If I remember correctly, the TDP for the PIII is ~30W.

More like 23W:

cpu elec specs

Mikael · Post by **Mikael** » Wed Apr 12, 2006 8:37 am

jaganath wrote:
If I remember correctly, the TDP for the PIII is ~30W.
More like 23W:

cpu elec specs

Yep, it would seem so. I probably got my data from the 1.7V and 1.75V parts:

27.71W (1.7V core)
30.8W (1.75V core)

Then it all depends on if the numbers for either of these CPUs are conservative or not... I'm of course hoping for the Athlon to be pretty close, since I don't have a motherboard for the PIII. That would save some trouble. That said, the AthlonXP seems to produce very little heat with the settings I run it on, so it probably won't be much of a problem anyway. If only it would operate correctly at 1.15V like my previous Athlons could...

QuietOC · Post by **QuietOC** » Wed Apr 12, 2006 9:09 am

It sounds like you 1700+ may be a Thoroughbred-A.

A good Thoroughbred-B is a good match for a Tualatin Celeron. There's no contest between it and a Coppermine. Having looked at used Tbred-B prices, I'd just eBay it and get S754 Sempron system. I have a brand new "E6" Sempron 64 2500+ that runs 1.4GHz at 0.88V if you're interested.

Mikael · Post by **Mikael** » Wed Apr 12, 2006 9:17 am

QuietOC wrote:It sounds like you 1700+ may be a Thoroughbred-A.

No, I'm positive that it is a "B". The stepping is 0251 JIUHB.

QuietOC wrote:A good Thoroughbred-B is a good match for a Tualatin Celeron. There's no contest between it and a Coppermine. Having looked at used Tbred-B prices, I'd just eBay it and get S754 Sempron system. I have a brand new "E6" Sempron 64 2500+ that runs 1.4GHz at 0.88V if you're interested.

Well, I think this will do for the moment. The CPU isn't a total catastrophe and it will do perfectly well in a low-end Linux box. I may buy a cheap AM2 Sempron system later this year, though, just for the fun of playing around with it.

BTW, what mobo did you use to undervolt that Sempron? It seems as if most of them have an artificial lock that prevents undervolting on non CnQ parts. I just got a Sempron 2800+ system with an Asrock nForce 410/GeForce 6100 mobo for the family company and there are no voltage options in BIOS. Guess I'm screwed on undervolting, right?

QuietOC · Post by **QuietOC** » Wed Apr 12, 2006 9:32 am

Mikael wrote: BTW, what mobo did you use to undervolt that Sempron? It seems as if most of them have an artificial lock that prevents undervolting on non CnQ parts. I just got a Sempron 2800+ system with an Asrock nForce 410/GeForce 6100 mobo for the family company and there are no voltage options in BIOS. Guess I'm screwed on undervolting, right?

Biostar Tforce 6100 S754.

Well, there is some discussion about those limits. My Tforce won't undervolt my E6 Sempron 64 3100+ under 1.06V (1.1V setting). I believe it has something to do with the CPU not neccessarily the motherboard. (Maybe some of the Semprons are actually Turions...)

No, it is not impossible to undervolt. You can jumper and/or pull CPU pins to change the default VID. I haven't done that on a S754, but I am sure it is possible. It might even be fairly easy.

jaganath · Post by **jaganath** » Wed Apr 12, 2006 9:36 am

No, it is not impossible to undervolt. You can jumper and/or pull CPU pins to change the default VID.

The usual caveats apply, ie you can destroy your CPU if you do this wrong, SPCR bears no responsibility if your CPU goes up in flames, stocks may go down as well as up, etc....

Mats · Post by **Mats** » Wed Apr 12, 2006 10:25 am

Reasons for AXP:

- You underclock it, which will give you a TDP of 29 V, and that's before undervolting (which you can't do).

- S462 lasted much longer, with the latest Sempron CPU's (which most of them are Tbred B with higher FSB, and a few the quite similar Barton. Your 1700+ is pretty much a Sempron 2200+ running 33 MHz slower.) stopped being manufactured just half a year ago, being on the market for over 3 years is a long time and not a bad thing. Should be pretty easy to find a mobo even with SATA if you want that. P3 mobos disappeared a long time ago, possibly more than 3 years for that CPU.

- The TDP's are not comparable, the two companies have different methods for measuring it. An AMD TDP corresponds to a higher Intel TDP. So if both have the same TDP, the AMD uses less power than the Intel.

- Tbred B is more modern than the Coppermine. You also have the ability to raise the speed back up with the AMD if you want.

Mikael · Post by **Mikael** » Thu Apr 13, 2006 10:27 am

NEWS!

I've done some quite extensive testing the past few hours. As I said before, the CPU does not produce errors at 1.5V with a 133MHz FSB. I now also noticed that a pretty significant voltage increase was required to get something half stable at 166MHz FSB. What I then did was to decrease the bus speed to 100MHz. Not only was the CPU now stable at 1.4V, I've been able to go all the way down to 1.2V. I haven't tried lower and I've only tested Memtest so far (no Prime or anything), but it looks promising.

I also took the liberty of increasing the CPU frequency to 1GHz, to offset the lower bus speed somewhat. So, the CPU is now happily Memtesting at 1GHz (10*100) at 1.2V and 2-2-2-6-1T timings.

What's even better is that it should still be slightly faster than the PIII and now slightly cooler! Let's just hope it manages 24h Prime95 too...

EDIT: Okay, WinXP i installed and running. I did get a problem at the install, due to the low voltage. It has now been running Prime95 for a while at 1.3V and things are fine so far. I hope to be able to take the voltage down to 1.25V.

Anyway, this box is running extremely cool at the moment. The only case ventilation is done through the PSU. The CPU is cooled by some pretty small GlacialTech heatsink, with a fan at 2200RPM (that's going to be taken care of) and the heatsink isn't even luke warm despite the CPU running at full load! It's easily cooler than my body temperature. The chipset is passively cooled and luke warm. The PSU casing hasn't heated up at all... This is looking great so far!

mattthemuppet · Post by **mattthemuppet** » Mon Apr 17, 2006 3:11 am

I think you may be limited to 1.25V anyway as IIRC there's some limit or other that let T'red Bs boot windows at less than that voltage. I think I managed to get my Sempron [email protected] down to 1.2V using software, but it wouldn't boot when it was set at that in BIOS (had to up it to 1.25 or 1.275V, can't quite remember). Still, it's running at stock (1.67GHz) @1.375V perfectly happily.

As for the original Q - I'd guess from my own limited experience that you should be able to get an underclocked and undervolted T'bred B down to the level where you can almost passively cool it (depending on HS, airflow and ambient temp). I'm sure mine could if I underclocked it and I ignored the temp readout during summer

Le_Gritche · Post by **Le_Gritche** » Mon Apr 17, 2006 5:54 am

Mikael wrote:What I then did was to decrease the bus speed to 100MHz. Not only was the CPU now stable at 1.4V, I've been able to go all the way down to 1.2V.

Did you keep the same CPU frequency by upping the multiplier or is that 1.5V @ 866MHz vs 1.2V @ 650MHz ?
I'm interrested to know, because I'm planning to take my 1GHz Thunderbird from 10*100 to 7.5*133. I assumed it would be able to undervolt the same, but you seem to report otherwise.

The CPU is cooled by some pretty small GlacialTech heatsink, with a fan at 2200RPM (that's going to be taken care of) and the heatsink isn't even luke warm despite the CPU running at full load!

Yes you can probably do something about the fan, maybe not go passive but still...
As I said I have a 1 Ghz atlhon, it's running at 1.65v from the stock voltage of 1.75v. It's cooled with a 80mm 2200rpm fan that I slowed down by soldering resistors of around 170 ohms. The motherboard can't report the fan speed (it's too slow) but I would say it's in the 800s. With a case temp of 30°C, the CPU is at 52°C idle and 60°C under load (CPUburner).
According to cpu-world, a 1700+ can sustain 90°C. Even when taking into account the difference between the reported temperature and the real temp, you can get quite hot without a problem.

If only it would operate correctly at 1.15V like my previous Athlons could...

If you had experience with Thunderbird Athlons, could you share your undervolting results ? I bought a mobo on ebay last week, it should be able to undervolt more than -0.1V, contrary to my current mobo.

Mikael · Post by **Mikael** » Mon Apr 17, 2006 7:08 am

Le_Gritche wrote:Did you keep the same CPU frequency by upping the multiplier or is that 1.5V @ 866MHz vs 1.2V @ 650MHz ?
I'm interrested to know, because I'm planning to take my 1GHz Thunderbird from 10*100 to 7.5*133. I assumed it would be able to undervolt the same, but you seem to report otherwise.

I actually increased the CPU frequency, so that's 1.5V @ 866MHz vs 1.25V @ 1GHz. I've had experience with several Socket A processors and this is the first one that has had this obvious connection between minimum usable CPU voltage and FSB speed. I'd be surprised if your Thunderbird behaved the same way as my Thoroughbred.

Le_Gritche wrote:Yes you can probably do something about the fan, maybe not go passive but still...

Yep, getting the fan down to less than 1000RPM should take care of the noise and easily be able to cool the CPU. I don't need to go passive, since the Fortron/Source (FSP) PSU will drown the CPU fan at that speed anyway. The PSU is still far from loud, though.

Le_Gritche wrote:If you had experience with Thunderbird Athlons, could you share your undervolting results ? I bought a mobo on ebay last week, it should be able to undervolt more than -0.1V, contrary to my current mobo.

I do have experience with Thunderbirds. I owned a 1GHz (10*100) Thunderbird with the "AXIA" stepping (legendary for overclocking well). I was more into overclocking than undervolting and silence back then, though. However, I did undervolt my particular CPU to 1.7V, while overclocking it to 1.33GHz. I never tried to lower the voltage to less than 1.65V. If I were to guess, I'd say that a good sample probably could operate correctly on 1.5V at 1GHz.

mattthemuppet: I think I remember my previous (and very overclockable) 1700+ booting at 1.15V on 1GHz (or close to it). It could have been my even more overclockable XP-M Barton, though. It's hard to remember all these experiments with different CPUs...

Either way, stability issues appear very sudden with this CPU. 1.225V is enough to run Prime95 and other tests for hours without errors, but 1.2V will go as far as to immediately corrupt a WinXP install. Somewhat of a mystery...

Le_Gritche · Post by **Le_Gritche** » Mon Apr 17, 2006 12:45 pm

Mikael wrote:I do have experience with Thunderbirds. I owned a 1GHz (10*100) Thunderbird with the "AXIA" stepping (legendary for overclocking well). I was more into overclocking than undervolting and silence back then, though. However, I did undervolt my particular CPU to 1.7V, while overclocking it to 1.33GHz. I never tried to lower the voltage to less than 1.65V. If I were to guess, I'd say that a good sample probably could operate correctly on 1.5V at 1GHz.

Mine is an ARGA stepping sadly. I couldn't find more than a handful of those in OC database online, not enough to make for a meaningful statistical sample. ARGA were produced before AXIA but not too early, so it might still be alright.
My goal was to switch between 10*100 at 1.55V and 10*133 at stock voltage (for the occasional gaming or Xvid re-encoding use). I will probably have to lower the multiplier though, because I would be surprised if it would fare as well as an AXIA.

Le_Gritche · Post by **Le_Gritche** » Thu Aug 31, 2006 10:15 am

Le_Gritche wrote:Mine is an ARGA stepping sadly. I couldn't find more than a handful of those in OC database online, not enough to make for a meaningful statistical sample. ARGA were produced before AXIA but not too early, so it might still be alright.
My goal was to switch between 10*100 at 1.55V and 10*133 at stock voltage (for the occasional gaming or Xvid re-encoding use). I will probably have to lower the multiplier though, because I would be surprised if it would fare as well as an AXIA.

Quick follow-up :

I had to reduce my goals seriously, my ARGA stepping Athlon 1GHz would boot up at 1050 MHz, but not at 1070MHz (even with 1.850 V).

At 1050MHz the CPU is prime95 stable at 1.675 V but not 1.650 V.
At the stock 1 GHz, 1.550 V is not enough but 1.575 V is fine. That's a 10% undervolt from the stock voltage of 1.750 V and it's my current setting.

I haven't bridged the L1 gaps yet to allow for the multiplier mod, so I can't test with a 133 MHz FSB, but I would be surprised if I could reach 1066MHz (8*133). I'm not even sure that would be useful then, as 1000 MHz with FSB133 (7.5*133) could be slower than 1050 MHz with FSB100 (10*105).

My Duron 600 proved to be a worthy overclocker, at 1.60 V it reached 900 MHz. At the stock 1.50 V the OS booted at 850 MHz, but not 892 MHz. I did not test in-between frequencies or higher voltages, as it was mainly a training before going for the real deal with the Athlon.

I bought a "non-tested" Athlon 1200 on ebay. Sure enough it didn't boot, but the seller maintain that it can boot at higher voltages than the stock 1.750 V. Maybe I will try that one day, when I'm in a playful mood.

autoboy · Post by **autoboy** » Thu Aug 31, 2006 1:07 pm

I have my athlon 2000+ throughbred running at 1Ghz and 1.2V also. Sounds like we got similiar unvervolts. However, i have no problem running a 133mhz bus. I never tried lowering it. I can cool mine passive with a simple extuded heatsink but i prefer to run it with a simple cheap 80mm fan at 5V. It sits in the garage which can get 100 degrees F. The processor never has problems. I worry a little more about the harddrives with only a 120mm tricool on low

0.13µm AthlonXP 866MHz vs 0.18µm Pentium III 866MHz

0.13µm AthlonXP 866MHz vs 0.18µm Pentium III 866MHz

Re: 0.13µm AthlonXP 866MHz vs 0.18µm Pentium III 866MHz

Re: 0.13µm AthlonXP 866MHz vs 0.18µm Pentium III 866MHz

Re: 0.13µm AthlonXP 866MHz vs 0.18µm Pentium III 866MHz