Pentium M 753 vs. Core Solo U1400

All about them.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

NoiseFreeGuy
Posts: 332
Joined: Mon May 15, 2006 7:16 pm
Location: Guelph, Canada

Pentium M 753 vs. Core Solo U1400

Post by NoiseFreeGuy » Fri Dec 15, 2006 3:18 pm

Hi Folks,

I defer this question to all the experts here on CPU's.

I've been doing some research on Intel's lowest power CPU's
and have come up with the Pentium M 753 and the Core Solo U1400 in terms of lowest power and highest processor speed.
The 753 uses a bit less power than the U1400.

Can someone here please tell me what are the main differences between these two CPU's, from a performance standpoint, technical and whatever else comes to mind?

BTW, do they both use a socket 479?

Thanks very much.

Mats
Posts: 3044
Joined: Fri Dec 26, 2003 6:54 am
Location: Sweden

Post by Mats » Fri Dec 15, 2006 5:10 pm

I can't give you all the facts, but the most important fact here is the socket. PM/CM 300 and C1/C2/CM 400 use different sockets so you should be aware of which socket you really want. Both are also called 478 or 479, confusingly.

Low power CPU's like the two you mentioned have a ball grid array that are soldered to the motherboard by the manufacturer, exactly like chipsets are. They're only available in a few laptop models AFAIK. You can't buy these CPU's in stores, you can't mount them, and you can't buy a motherboard without a CPU mounted.

NoiseFreeGuy
Posts: 332
Joined: Mon May 15, 2006 7:16 pm
Location: Guelph, Canada

Post by NoiseFreeGuy » Fri Dec 15, 2006 5:38 pm

Mats wrote:... but the most important fact here is the socket. PM/CM 300 and C1/C2/CM 400 use different sockets so you should be aware of which socket you really want. Both are also called 478 or 479, confusingly.... You can't buy these CPU's in stores, you can't mount them, and you can't buy a motherboard without a CPU mounted.
Thanks for replying Mats.
So if I understand you correctly, I won't be able to buy a motherboard and stick either of these CPU's into it.

So the next logical question I guess would be, do they sell MB's with either of these two CPU's already mounted?

Obviously it would be cost prohibitive to buy a notebook simply for its motherboard and CPU so that you could build a desktop with it.

autoboy
Posts: 1008
Joined: Fri Dec 10, 2004 8:10 pm
Location: San Jose, California

Post by autoboy » Fri Dec 15, 2006 5:51 pm

I have never seen one.

NoiseFreeGuy
Posts: 332
Joined: Mon May 15, 2006 7:16 pm
Location: Guelph, Canada

Post by NoiseFreeGuy » Fri Dec 15, 2006 8:06 pm

autoboy wrote:I have never seen one.
Ouch. So it looks like I'm going to have to spend ~$2500-3000 to get one of these fanless P753 or U1400 notebooks.
Darn industry!

jaganath
Posts: 5085
Joined: Tue Sep 20, 2005 6:55 am
Location: UK

Post by jaganath » Sat Dec 16, 2006 2:13 am

I would be shocked if you couldn't get a normal PM down to ULV/LV levels by underclocking+volting.

Aris
Posts: 2299
Joined: Mon Dec 15, 2003 10:29 am
Location: Bellevue, Nebraska
Contact:

Post by Aris » Sat Dec 16, 2006 2:33 am

Just go with a regular P-M or regular Core Solo/Duo. like jag said, you can undervolt/underclock them if they still produce too much heat for you at stock settings.

Just remember if it specifies "Core" then only the core solo/duo will fit, not the Pentium-M. If it specifies "pentium-M" then the opposite is true. They both use a "socket 478/479". Motherboard manufacturers will pretty much use 478 and 479 interchangably, and dont always specify if its P-M or Core Solo/Duo. It makes it very confusing. Just remember that if it works with one, it wont work with the other. They both have the exact same number of pins, intel just moved one of them so they dont physically fit in the same socket.

Mats
Posts: 3044
Joined: Fri Dec 26, 2003 6:54 am
Location: Sweden

Post by Mats » Sat Dec 16, 2006 4:04 am

Remember that it's much easier to get a fanless desktop than a fanless notebook. Just because manufacturers use these for fanless notebooks it doesn't mean that it's your only choice for a desktop. A notebook heatsink is quite small compared to a Scythe Ninja for instance.

Edit: If you want ATX for your Core Duo then you have two alternatives:

- AOpen i975Xa-YDG, reviewed here at SPCR

- MSI 945GT Speedster-A4R

There are two µATX mobos that I think looks better than most others since they have better mounting holes for the CPU heatsink, regular RAM and PCIE x16:

- MSI 945GT Speedster-A4R

- Gigabyte GA-8I945GMMFY-RH

As you can see, the board from MSI can be used as either ATX or µATX, you can get a total of 6 slots with an expansion board.

NoiseFreeGuy
Posts: 332
Joined: Mon May 15, 2006 7:16 pm
Location: Guelph, Canada

Post by NoiseFreeGuy » Sat Dec 16, 2006 6:21 pm

Aris wrote:Just go with a regular P-M or regular Core Solo/Duo. like jag said, you can undervolt/underclock them if they still produce too much heat for you at stock settings.

Just remember if it specifies "Core" then only the core solo/duo will fit, not the Pentium-M. If it specifies "pentium-M" then the opposite is true. .
Thanks for the tip on the socket and what will work with it.

As much as I'd like to go with a regular P-M or Core Solo chip I don't think it would work for me.

Can you see undervolting/clocking these chips to get these results:
P-M 753 4-7 Amps, 3-5 Watts.
U1400 6-8 Amps, 5-5.5 Watts

You are reading that correctly!
Judging by my sensitivity so far to Athlon X2 chips (3800, 4200)
Turions and many others, I don't think undervolting the regular chips would work.

Heck I simply went from my P-233 to an AMD K-2 350 Mhz (that's right!)
and I could feel the chip's EMFs from an adjoining room (cables through the wall).

Unless I could achieve the above power ratings with a regular chip I don't think it would for me.

NoiseFreeGuy
Posts: 332
Joined: Mon May 15, 2006 7:16 pm
Location: Guelph, Canada

Post by NoiseFreeGuy » Sat Dec 16, 2006 6:27 pm

Mats wrote: If you want ATX for your Core Duo then you have two alternatives:

...As you can see, the board from MSI can be used as either ATX or µATX, you can get a total of 6 slots with an expansion board.
Thanks Mats. As much as I'd like to take advantage of the core duo speed/power, it just won't work for me. Only an ULV core solo will work or an ULV P-M chip. If you can tell me how to undervolt a Core Duo to achieve 4-7 Amps, 3-5 Watts power like can be achieved with an ULV P-M, then I'd go for it.

jaganath
Posts: 5085
Joined: Tue Sep 20, 2005 6:55 am
Location: UK

Post by jaganath » Sun Dec 17, 2006 3:01 am

Can you see undervolting/clocking these chips to get these results:
P-M 753 4-7 Amps, 3-5 Watts.
U1400 6-8 Amps, 5-5.5 Watts
Yes. Remember power consumption scales with the square of the voltage; the ULV chips have a supply voltage that is roughly 40% less than the standard PM; downclocking is just a way to get the chip to run stably at lower voltages.
Judging by my sensitivity so far to Athlon X2 chips (3800, 4200)
Turions and many others, I don't think undervolting the regular chips would work.

Heck I simply went from my P-233 to an AMD K-2 350 Mhz (that's right!)
and I could feel the chip's EMFs from an adjoining room (cables through the wall).
This suggests the problem is not the chips themselves but inadequate shielding; if you surround the PC with steel (ie PC case with no openings, so fanless PC like HushPC for example) and put ferrite cores on all the cables leaving the PC, EMF cannot escape.

hmsrolst
*Lifetime Patron*
Posts: 427
Joined: Sun Aug 11, 2002 3:26 pm
Location: Arlington, VA USA

Post by hmsrolst » Sun Dec 17, 2006 3:30 am

I don't know what power draw this translates into, since I've never attempted to measure it, but in my pretty extensive experience with P-M's in desktops and some now with Core Duo's, much lower undervolting with the former is possible.

I have a P-M 750 running in a Shuttle SD11G5 using Notebook Hardware Control. At the 6X multiplier it undervolts to .70 and at the 14X it's at 1.05. Most Dothan's seem to be able to do similarly.

Core Duo's unfortunately seem to be locked to a lowest voltage of .9375 no matter what the multi. I have a T2500 in an AOpen i945GTm-VHL which runs at 1.00 at 6X. I haven't found a good utility for further undervolting it yet.

I also have a Dell D420 with a U2500 Core Duo. Even with this ULV chip, NHC only goes down to .9375, although at 9X I have it a .975 (and haven't tried it lower).

So if you're spending a lot of time at idle or low usage (as most of us are), P-M's might be better than you think.

NoiseFreeGuy
Posts: 332
Joined: Mon May 15, 2006 7:16 pm
Location: Guelph, Canada

Post by NoiseFreeGuy » Sun Dec 17, 2006 7:52 pm

jaganath wrote:
Judging by my sensitivity so far to Athlon X2 chips (3800, 4200)
Turions and many others, I don't think undervolting the regular chips would work.

Heck I simply went from my P-233 to an AMD K-2 350 Mhz (that's right!)
and I could feel the chip's EMFs from an adjoining room (cables through the wall).
This suggests the problem is not the chips themselves but inadequate shielding; if you surround the PC with steel (ie PC case with no openings, so fanless PC like HushPC for example) and put ferrite cores on all the cables leaving the PC, EMF cannot escape.
I wish it were that simple Jag.
In a discussion a few months ago about EMF's several others
said that they are extremely difficult to block.
That a Faraday chamber would be in order to stop them!

I have a very old case right now (all steel) with a couple of fan openings.

Ric

pxa270
Posts: 10
Joined: Fri Aug 18, 2006 2:38 pm

Post by pxa270 » Mon Dec 18, 2006 5:53 am

Could you tell us more about the intended application of your setup and your budget constraints? That might get you some more useful advice.

NoiseFreeGuy
Posts: 332
Joined: Mon May 15, 2006 7:16 pm
Location: Guelph, Canada

Post by NoiseFreeGuy » Mon Dec 18, 2006 7:32 pm

pxa270 wrote:Could you tell us more about the intended application of your setup and your budget constraints? That might get you some more useful advice.
PXA, you're quite right, I will fill in more detail now.

This rig is intended to be my main computer which will be used for e-mail, browsing and some music apps; no games at all though.

Budget is not an issue at all right now. I am quite prepared to spend whatever it takes to get not only a quiet system, but one where I can't feel the EMF's.

For example I have tried several notebooks and desktops in the past few months where the EMF's from the CPU's/hard drives have done a number on my head.

7200 rpm hd's are definitely out of the picture. I've tried them (the new ones too) and they don't get along with my head, even in the next room.

I am looking at the lowest power CPU/motherboards now because of a startling discovery a few months ago when I simply exchanged my P-233 with an AMD K6-2 350 Mhz. No other changes to the system and I could feel the extra vibes in my head (again, from the next room).

This got me looking into ULV CPU's. I tried the U1400 1.2G and it was OK with my head from a few feet away.

So I don't mind taking quite a performance hit as well, so long as the system is pretty much silent (as my P-233 now is) and doesn't emit EMF's that I can feel.

To give you an even better idea of my sensitivity. I need to unplug a printer's power adapter if it is in the same room as me.

pxa270
Posts: 10
Joined: Fri Aug 18, 2006 2:38 pm

Post by pxa270 » Wed Dec 20, 2006 2:00 am

Unless you need Windows specific apps, perhaps you could try a used iBook G4. From the German http://www.macgadget.de:
The German test-magazin 'Oeko-test' had tested eight notebooks for electro-smog (radiation of high- & low frequency electric fields & electrostatic). And THE ONLY ONE which passed the tests and being totally okay with the very severe Swedish TCO-Norm has been the Apple iBook G4. All the other notebooks (for instance a Samsung X30 WVC 1700 and a Sony Vaio PCG-GRT916Z) did radiate electro-smog 10 to 40 times higher than allowed by TCO. The worst results of all notebooks had a Inspiron 510m by Dell.

NoiseFreeGuy
Posts: 332
Joined: Mon May 15, 2006 7:16 pm
Location: Guelph, Canada

Post by NoiseFreeGuy » Wed Dec 20, 2006 9:57 pm

pxa270 wrote:Unless you need Windows specific apps, perhaps you could try a used iBook G4. From the German http://www.macgadget.de:
Well basically I do need Windows. All my apps are Windows-based.
If I were to go iBook I'd need to start from ground zero with all my software and hardware.
But it's something to consider.

Thanks for the link! They are much more aware of the environment and its effects in Europe than we are over here in N.A.

Aris
Posts: 2299
Joined: Mon Dec 15, 2003 10:29 am
Location: Bellevue, Nebraska
Contact:

Post by Aris » Wed Dec 20, 2006 10:49 pm

you can get a VIA CPU/Motherboard where the CPU consumes around 5 watts.

Like this one: 800mhz

http://www.logicsupply.com/product_info ... cts_id/306

If your willing to go up to around 10 watts for the CPU, you can jump up to the C7's.

Like this one: 1.2ghz

http://www.logicsupply.com/product_info ... cts_id/688

Both come stock passive in a mini-itx form factor. Ive never seen ULV's for sale in retail channels. Other than them, these VIA cpu's are the fastest ones i know with cpu energy utilization down around that same level.

pxa270
Posts: 10
Joined: Fri Aug 18, 2006 2:38 pm

Post by pxa270 » Thu Dec 21, 2006 2:39 am

The VIA's are awful performers. You can get a complete IBM Thinkpad X40 ultraportable laptop with a 1.0Ghz ULV Pentium-M for $500 and it will run circles around a 1.2Ghz VIA C3.

NoiseFreeGuy: I would really look into shielding solutions. The iBook's PowerPC G4 CPU has an output comparable to regular (not ULV) Pentium-M's. I suppose its low emmisions result from better RF shields inside the case.

NoiseFreeGuy
Posts: 332
Joined: Mon May 15, 2006 7:16 pm
Location: Guelph, Canada

Post by NoiseFreeGuy » Thu Dec 21, 2006 2:15 pm

pxa270 wrote:The VIA's are awful performers. You can get a complete IBM Thinkpad X40 ultraportable laptop with a 1.0Ghz ULV Pentium-M for $500 and it will run circles around a 1.2Ghz VIA C3.

NoiseFreeGuy: I would really look into shielding solutions. The iBook's PowerPC G4 CPU has an output comparable to regular (not ULV) Pentium-M's. I suppose its low emmisions result from better RF shields inside the case.
Thanks for these suggestions guys, I never would have thought of them.
I'll definitely look into shielding ideas.

I'm looking at an industrial MB that will accept P-M chips (not ULV ones though). Do you think that I could undervolt a P-M to achieve a 5-10W CPU output? If I could, then the MB plus good shielding may be an option.

What is the criteria for being able to undervolt a P-M chip?
Is it the BIOS, the chip or something else?
I know there's software out there, but I've read here that it doesn't always work with a CPU.

Thanks again!

hmsrolst
*Lifetime Patron*
Posts: 427
Joined: Sun Aug 11, 2002 3:26 pm
Location: Arlington, VA USA

Post by hmsrolst » Thu Dec 21, 2006 4:36 pm

NoiseFreeGuy wrote: I'm looking at an industrial MB that will accept P-M chips (not ULV ones though). Do you think that I could undervolt a P-M to achieve a 5-10W CPU output? If I could, then the MB plus good shielding may be an option.

What is the criteria for being able to undervolt a P-M chip?
Is it the BIOS, the chip or something else?
I know there's software out there, but I've read here that it doesn't always work with a CPU.

Thanks again!
From my experience, it's not so much the CPU (although there is some variation) as it is a matter of the motherboard. Almost all the P-M CPU's will undervolt pretty easily to around 1.10-1.15v either at full speed or overclocked by about 30%. However, if you're willing to take the performance hit (which isn't probably noticeable for the things you're planning to do), getting a board which will work with a utility like Notebook Hardware Control (the only one I have experience with) will allow you to lock the multiplier at 6x and run at around .70v.

I've had success with an AOpen i855GMEm-LFS, a P-M 755 and NHC (although my wife's machine has the same MB, but an earlier one, and I haven't been able to do it with hers). At 6x (600 MHZ) it will run stably at .70v. This MB also allows undervolting in the BIOS, but doesn't seem to work very stably playing around with the multi in the BIOS.

As I mentioned in an earlier post, I've also had success with a P-M 750 in a Shuttle SD11G5 running at 6x (800 MHz) and .70v.

I'm not competent to say whether these would result in the 5-10W levels you're looking for.

NoiseFreeGuy
Posts: 332
Joined: Mon May 15, 2006 7:16 pm
Location: Guelph, Canada

Post by NoiseFreeGuy » Thu Dec 21, 2006 5:22 pm

hmsrolst wrote: From my experience, it's not so much the CPU (although there is some variation) as it is a matter of the motherboard.... getting a board which will work with a utility like Notebook Hardware Control (the only one I have experience with) will allow you to lock the multiplier at 6x and run at around .70v.

I've had success with an AOpen i855GMEm-LFS, a P-M 755 and NHC ... At 6x (600 MHZ) it will run stably at .70v. ...
As I mentioned in an earlier post, I've also had success with a P-M 750 in a Shuttle SD11G5 running at 6x (800 MHz) and .70v.

I'm not competent to say whether these would result in the 5-10W levels you're looking for.

Thanks very much hmsrolst! Your posts give me hope, as you seem to have the type of systems that I could live with. Interesting about the AOpens, one works with NHC and the other doesn't.


Jaganath:
"I would be shocked if you couldn't get a normal PM down to ULV/LV levels by underclocking+volting."

Hey Jag, I need your help here! Would undervolting a P-M with NHC (say 750 or 755) get my CPU power output down in the 5-10W range?

hmsrolst
*Lifetime Patron*
Posts: 427
Joined: Sun Aug 11, 2002 3:26 pm
Location: Arlington, VA USA

Post by hmsrolst » Thu Dec 21, 2006 5:57 pm

I'm not sure that the problem with my wife's is not a Windows issue, since I get an error message when I try to start NHC. I got the other one working on a clean install.

pxa270
Posts: 10
Joined: Fri Aug 18, 2006 2:38 pm

Post by pxa270 » Fri Dec 22, 2006 3:17 am

If you can get them working stably at low voltage (<1.0V) you're set. At 600Mhz 0.70V it will almost certainly dissipate less than 3W. About the only difference (besides the socket) between the regular and the LV/ULV chips are that the latter have been selected and certified by Intel as working stably at low voltage.

For undervolting regular chips, it's just the luck of the draw (and it depends on how much you're willing to cut the clock speed). From http://www.sandpile.org/impl/pm.htm you get 5-10W if you can get them running at 0.940V. A good chip will go over 1Ghz at that voltage.

NoiseFreeGuy
Posts: 332
Joined: Mon May 15, 2006 7:16 pm
Location: Guelph, Canada

Post by NoiseFreeGuy » Fri Dec 22, 2006 1:44 pm

pxa270 wrote:If you can get them working stably at low voltage (<1.0V) you're set. At 600Mhz 0.70V it will almost certainly dissipate less than 3W. About the only difference (besides the socket) between the regular and the LV/ULV chips are that the latter have been selected and certified by Intel as working stably at low voltage.

For undervolting regular chips, it's just the luck of the draw (and it depends on how much you're willing to cut the clock speed). From http://www.sandpile.org/impl/pm.htm you get 5-10W if you can get them running at 0.940V. A good chip will go over 1Ghz at that voltage.

Thanks for this info PXA, it is extremely helpful to me!

Just so that I understand Intel's manufacturing process a bit better, is this the way they do it?

They set their design and limits for their P-M chips and build a run.
They then take a chunk of these chips that are to be used as their UL/V chips.
They test the chips that they selected. Those that pass the test are kept as UL/V chips, those that don't are thrown back in with the original batch.

If this is the way they do it, I can see how many chips (those that *weren't* tested) could be good UL/V chips.

If they do it another way I'd be interested to learn the method.

Thanks!

pxa270
Posts: 10
Joined: Fri Aug 18, 2006 2:38 pm

Post by pxa270 » Fri Dec 22, 2006 5:18 pm

I'm no expert on the subject, but from what I understand, that's pretty much how they do it. LV/ULV chips don't use different manufacturing techniques, but are results from "binning" the regular P-M/Core chips. They select and test chips likely of high quality (usually from the middle of the wafer) and if they pass, they're sold at higher stock clockspeed or lower voltage. So a regular P-M might have never been tested on a lower voltage, or it might have been rejected (but then it still could run at slightly higher voltage or lower clockspeed than in the test). It depends on lots of things, like market demand (for LV/ULV), maturing production processes (probably a higher percentage of the later P-Ms will run at low voltage). From various forum reports, it seems many regular P-Ms tend to respond very well to undervolting.

You can get some more info from this thread.

NoiseFreeGuy
Posts: 332
Joined: Mon May 15, 2006 7:16 pm
Location: Guelph, Canada

Post by NoiseFreeGuy » Sat Dec 30, 2006 9:03 pm

pxa270 wrote:I'm no expert on the subject, but from what I understand, that's pretty much how they do it. LV/ULV chips don't use different manufacturing techniques, but are results from "binning" the regular P-M/Core chips. They select and test chips likely of high quality (usually from the middle of the wafer) and if they pass, they're sold at higher stock clockspeed or lower voltage. .
Thanks pxa, that makes sense to me.

Just one thing though. I was just reading the specs for some Mobile P's at this site:
http://www.techarp.com/showarticle.aspx ... 347&pgno=1

and let's say we choose the P-M 750 and the P-M 753 (ULV).
We get:
750 12-26 A 10-27 W
753 4-7 A 3-5 W

so it's not even close. No overlap or anything.

Are you saying that the 753's are taken from the same bin as the 750's
and tested. If they pass they are rated 4-7 A and 3-5 W?
and all the other 750's that were either not tested, or failed, are rated
12-26 A and 10-27 W?

Is that really how it works? It's just that there seems to be such a big gap between them even though they come from the same stock.

jaganath
Posts: 5085
Joined: Tue Sep 20, 2005 6:55 am
Location: UK

Post by jaganath » Sun Dec 31, 2006 7:58 am

Ric, it's important to note that all P-M's have to comply with Speedstep, regardless of which particular flavour of P-M they are. So for the 750 the lowest setting is 0.988V @ 600MHz, whereas for the 753 it is 0.812V @ 600MHz, so the only difference is the voltage they are capable of running at. In those states the 753 is probably using <1W and the 750 less than 5W, so the difference is not as big as it looks.

753 datasheet

750 datasheet

See the sections titled "Processor DC Specifications".

http://en.wikipedia.org/wiki/Pentium_M
The power requirements of the Pentium M varies from 5 watts when idle to 27 watts at full load.
The idle state corresponds to the lowest Speedstep state.

hmsrolst
*Lifetime Patron*
Posts: 427
Joined: Sun Aug 11, 2002 3:26 pm
Location: Arlington, VA USA

Post by hmsrolst » Sun Dec 31, 2006 9:20 am

Intel leaves in a substantial margin of error too. I've yet to have a P-M that wouldn't run stably at .7v when operating at 6x.

simeli
Posts: 61
Joined: Wed Jun 28, 2006 5:31 am
Location: Zurich Switzerland
Contact:

Post by simeli » Sun Feb 25, 2007 6:47 am

I'd go and check the mini pc's from digitallogic. they have a host of different fanless machines for industrial applications and even a media center. some run on lv pentiums. if you are prepared to spend some money, they even make you custom boards to exactly fit your needs. http://www.digitallogic.com/english/index.asp

Post Reply