Mobile Celeron 1.5/256 -v- Pentium 4M 1.8/512 (both 478pin 400fsb Northwood)

[MrCreosote]

First, apologies for such an obsolete technology question (!) but I need some advice.

I have a trusty Toshiba Satellite 1115-s103 with a Mobile Celeron 1.5/256 and am thinking of putting the best cpu in it since I have it all tore apart for a HD replacement. (160 Scorpio WDC)

I found some old tests http://www.xbitlabs.com/articles/cpu/display/celeron-1800.html where a Celeron 1.3 outperformed a Celeron 1.8 because the 1.3 had more L2 cache.

Also found awesome detailed CPU list at PassMark Software: http://www.cpubenchmark.net/low_end_cpus.html where the ambiguous "Mobile Intel Pentium 4 M", "Mobile Intel Pentium 4-M" and "Intel Pentium 4 Mobile" 1.8's scored about 15% higher than any of the ambiguous Celerons listed. But still very confusing.

But in other articles, the Celeron -v- P4 was just marginally slower. So a 2.0 Celeron -v- 1.8 P4 might actually favor the Celeron.

I'm leaning towards the P-4 however I like the idea that the Celeron 2.0 is a 32w power class (all other upgrades are 30w). I was told that for scientific computing, in many cases, gHz is the primary performance objective. Since my ap is scientific, the Celeron is probably OK.

Open to all opinions,

Thanks in advance,

Tom

 

[e6600]

i dont think you would notice a huge difference between the two
only difference thats probably worth the upgrade is a P4M - Pentium mobile upgrade (dothan) but thats socket 479 and wouldnt work with your setup

the cpu might be a bit expensive too

 

[MrCreosote]

From what I have found, both candidates are P4-M Northwoods where the "P4" has twice the cache and the "Celeron" has higher clock speed.

It is clear that the trend in mobile processors in particular is to pile on the cache. The improvement from 256 to 512 should be greater than 512 to 1024 I would think.

I think it is clear that restriction on the upgrades is 478 pin and 30w power dissipation. (The fact that the 2.0 Celeron is actually the only 32w upgrade so maybe using thermal dissipation as the metric, that would be the most "bang" for the buck.)

BTW, both processors are the same price on eBay.

 

[Evilsizer]

netburst arch needs L2, no it loves it, the more the better. get the P4-M over the celeron, even if the celeron has a higher clock speed.

 

[MrCreosote]

argument for more L2 is probably wise. If you need it the Celeron will provide no improvement. But when L2 is critical, like I saw the benchmark where the 1.3 beat out the 1.8 Celeron.

another way of looking at it: L2 and Hz: I have the choice of 100% L2 increase -v- a 33% Hz increase where the starting point 1.5/256 is probably wanting more of L2
________________

Oh, a side question: I need a BIOS upgrade to do this CPU upgrade. I suspect I will have to do the BIOS upgrade with the old cpu, THEN replace the CPU. If I do the CPU upgrade before the BIOS upgrade, that could get ugly I bet. Thoughts?

 

[Evilsizer]

ok, but this isnt a argument, it is a fact that netburst AKA P4's respond to more L2. in the case of netburst it makes it more efficent per clock with more L2. 512Kb is the min L2 that P4 needs to be good, with 1mb L2 best for netburst architecture.

i dought you do need to upgrade the bios the laptops are not like desktops. a line of cpus is released when that chipset is used. Any cpu from the support list for that chipset would work without a bios update. Unless someone just messed up the coding for a cpu or something like that.

 

[MrCreosote]

From what I have found, both candidates are P4-M Northwoods where the "P4" has twice the cache and the "Celeron" has higher clock speed.

It is clear that the trend in mobile processors in particular is to pile on the cache. The improvement from 256 to 512 should be greater than 512 to 1024 I would think.

I think it is clear that restriction on the upgrades is 478 pin and 30w power dissipation.

BTW, both processors are the same price on eBay.

 

[MrCreosote]

Evilsizer,

I just did some learning on netburst and since both the candidates employ it, the 256 is just not enough. So 512 L2 is basically correcting a "problem"

As far as the BIOS goes, from Toshiba site, the latest BIOS adds the following CPU recognition:

Stepping C1: Pentium 4-M 1.8/512
Stepping D1: Mobile Celeron 1.5, 1.6, 1.7, 1.8, 1.9, 2.0 & Pentium 4-M 1.8/512

NOTE: Stock is C1: Mobile Celeron 1.5

Would there be any advantage of the D1 -v- C1 P4M stepping?

NOTE: It would be nice if the P4M 2.0/512 would work. It is only 32w like the 2.0 Celeron. I wonder how "strict" that compatible list of CPU's is?

 

[Evilsizer]

it would be pretty strict since the cpu support for mobiles is more based by the chipset. if the 2ghz P4-M is using the same fsb and same socket as the P4-M 1.8gh then it should work. that is when though you want to check with the manufacture of what cpus are supported first...

stepping has very little baring when talking about older Netburst cpus. D1 might run cooler then C1 but in the mobile segment intel already made sure to use low heat P4's.

 

[MrCreosote]

ya, the socket, the fsb, L2, etc all the same just 2.0 -v- 1.8 for the P4-M, however Toshiba in their BIOS download description says only C1 and D1 steppings for the 1.8 P4-M.

I'm thinking that maybe they limit the cpu list in the BIOS so you don't over heat or over draw on the power supply. One would have hoped that if they passed the 32w Celeron, they would have passed the 32w P4.

The chipset is an i845 - there are different suffixes depending on where you get specs from. I'll try to edit this when I get home with what I dug up on it.

Toshiba.com spec sheet says 845MZ chipset.

NOTE: While searching about the 845MZ chipset, I read an article saying that the 1.7gHz Pentium 4-M was the most advanced mobile processor available. At that time, I believe the 1.7 sold for about $500. So I guess the 1.8 might be the limit for this laptop. of course, what is detected in the BIOS? Is it the ASCII for the Stepping that does a table lookup? Or does the BIOS merely detect the multiplier in which case the 2gHz Celeron multiplier is handled. What to do.... what to do....