Is this plausible? <FX-55 bows to the Pentium M?>

[Super Nade]

Folks,

I happened to chance across an Anandtech article about the ASUS CPU Slot adapter for the Pentium M.

According to their benchmarks, The P-M beats/rides with the FX-55 in games (when overclocked and beats the 6xx Intels at stock). Do you guys think that is plausible? Even the Intel 6xx's with their 2 Mb cache's are blown out of the water!

http://anandtech.com/cpuchipsets/showdoc.aspx?i=2382&p=12

http://anandtech.com/cpuchipsets/showdoc.aspx?i=2382&p=7

I really have my doubts that the P-M is that good of a performer. Why is an overclocked 3000/3200 ranked that low? I have my doubts about this test.

Comments would be appereciated. Can somebody tell me if this is due to any groudbreaking architectural innovation by Intel? (I'm unaware of any).

Edit*
I could have put this under Intel or AMD sections. I chose AMD becasue of the higher traffic

 

[mudpark41]

I was under the impression that the P-M needed a different chipset and could not be used in a normal P4 motherboard. The P-M is more along the lines of the XP-M series in how it works, and yes they have been saying for a while that an overclocked P-M can compete with an FX-55.

 

[Super Nade]

Wow really? I wasn't aware of that.

A quick search threw this up, pretty interesting:
http://www.gamepc.com/labs/view_content.asp?id=dothangaming&page=6

 

[Quailane]

Wow really? I wasn't aware of that.

A quick search threw this up, pretty interesting:
http://www.gamepc.com/labs/view_content.asp?id=dothangaming&page=6

Ok, but you are not looking at the facts. Maybe a pentium m when overclocked can reach an fx-55, but what happens when you overclock the fx-55? What happens when you overclock the pentium 4's? This article is silly because it doesn't tell you much except that an fx-55 will always be better than a pentium m because it takes an overclocked pentium m to hang with the stock '55.

 

[Super Nade]

Mate,
I agree with you.

Look at the clocks on their benches, clock for clock, the P-M is a tremendous leap over the P4's. But even at stock it easily beats the 3400, 3200 and 3000 and probably the 3500 (if they had it in there). Why are they(Anandtech) running CPU benchmarks at 1024 x 786? Shouldn't they be running 800 x 600 tests?

 

[dropadrop]

I have allways been wondering why Intel don't bring the pentium -m to the desktop, and this pretty much awnseres my question. It seems to only do well in very specific types of programs, and even when overclocked by quite a bit, it is still mainly just as fast as an FX at the same clockspeed.

It would be quite a task for Intel to (currently) get the impact they would want by bringing a whole new desktop enviroment to the market.

I guess a new processor version with higher clocks, higher fsb and dualcore could easily change that though.

 

[Gnerma]

Clock for clock the Pentium M Dothan is the most powerful CPU on the market. The power consumption and heat dissipation is also excellent although a Winchester isn't too bad until you clock it up past ~2.2 and it starts leaking. Does that mean it is a viable choice for overclockers? Right now, no not really. The CPUs are expensive and the boards very few and limited. But things might not stay this way. We'll just have to wait and see. I think a dual core Dothan (or its .65nm successor) would be an awesome chip.

BTW this thread probably belongs in general CPU discussion.

 

[RoadWarrior]

Well 1024x768 doesn't stress modern gfx as much as it used to.

I think the Pentium M is a kind of tuned up PIII core, if you look at the last of the P3 line, the Tualatins, they were keeping up handily with Athlon XPs. The P2/3 architecture was always intended to take advantage of a LOT of cache, throughout most of it's life Intel cheaped out in that regard, and it never hit it's maximum potential, even 512K was hurting it. There's an article out on the web somewhere that tells how the original P2s were intended to have as much as 2Mb cache, the architecture was designed for it, but due to the expense of cache RAM it never even made it to production on the slotties with the off die cache.

So basically, what I'm saying is that Pentium Ms are only now fulfilling the potential of their original basic design, dating back to the mid 90s. No particular recent innovation, they just finally did what they should have done years ago.

Road Warrior

 

[yeha]

it's true that the dothan has a better ipc rating than the a64's, however in many cases it's starved for bandwidth. pairing it with a decent chipset (that can push a 200+ fsb, unlike the 855) evens this out quite a bit. if you have an a64 and dothan both at the same clock speed (with 200mhz bus speeds) the dothan will come out on top most every time, however the a64 still wins when latency is the major player.

gnerma: i've been wondering about the winchesters leaking as you say - i've been unable to find any tests where people have used watt-meters to see how much power the winchesters draw at high frequencies, in fact i'm not sure if such a test has ever been done above 2.2 ghz. the power draw is linear with frequency for 2.2 ghz and below (this has been tested in the spcr forums), how can anyone know if the linear relationship breaks down and leakage increases unless it's been tested? i've been curious about that for a while.

 

[Gnerma]

gnerma: i've been wondering about the winchesters leaking as you say - i've been unable to find any tests where people have used watt-meters to see how much power the winchesters draw at high frequencies, in fact i'm not sure if such a test has ever been done above 2.2 ghz. the power draw is linear with frequency for 2.2 ghz and below (this has been tested in the spcr forums), how can anyone know if the linear relationship breaks down and leakage increases unless it's been tested? i've been curious about that for a while.

I have not seen such tests either, that comment was mostly based on my own experiences with temps when running at various speeds and what I have read. I do have the means to do a power draw test here with one problem, I have to increase voltage when going over 2.2. So I suppose it is just educated speculation on my (and others') part. If we had a member who owned a Winchester that could run ~2.6 at stock voltage and a kill-a-watt we could get some hard numbers but I doubt we do

 

[RoadWarrior]

Here's an interesting read....
http://www.cpuid.com/PentiumM/index.php

Still looking for that article about intels "mistakes" with the PII line and how the architecture sucked with the amount of cache they gave it.

 

[yeha]

gnerma: i know this is getting o/t, but you could keep the voltage constant at say 1.5/1.6 throughout the range of frequencies, as that will still show a change in the linear relationship if leakage is increasing. be sure to keep ram speed constant throughout though, and htt speed should be documented so chipset power usage at different htt's can be accounted for.

man i wish i had an a64 to play with.

 

[Gnerma]

gnerma: i know this is getting o/t, but you could keep the voltage constant at say 1.5/1.6 throughout the range of frequencies...

I considered that but I suspect that would skew how the power consumption behaves. The numbers may be viable but only at that Vcore if you know what I'm saying. Not to mention the poor behavior of my Vcore. If you would like me to run a few tests I will, although we should probably move over to SPCR. I post there sometimes, just point me to a relevant thread.

 

[Aslan]

Clock for clock the Pentium M Dothan is the most powerful CPU on the market. The power consumption and heat dissipation is also excellent although a Winchester isn't too bad until you clock it up past ~2.2 and it starts leaking. Does that mean it is a viable choice for overclockers? Right now, no not really. The CPUs are expensive and the boards very few and limited. But things might not stay this way. We'll just have to wait and see. I think a dual core Dothan (or its .65nm successor) would be an awesome chip.

Agreed. The Dothan has a great deal of potential in terms of OC'ing, but the drawbacks outweigh the benefits. Even Asus' slot adaptor has many negatives that hurt the casual overclocker, and IMO only the people who are inclined to modding will see any major benefits. However, DFI and some other companies are planning i915 mobile desktop boards even for the Dothan, so there may be some life left still.

I'm hoping that when Yonah comes out, there will be some more OC friendly boards out there. I'm strongly interested in making my next system Yonah based, but it only depends on the cost and difficulty level.

 

[Super Nade]

Guys,

What is SPCR? Are you guys saying P3's were way more efficient than the current P4's?
Are these Donathon's 90nm? How does the Donathon stack up agaist the Turion?

RW thats an excellent link

 

[Gnerma]

What is SPCR? Are you guys saying P3's were way more efficient than the current P4's? Are these Donathon's 90nm? How does the Donathon stack up agaist the Turion?

http://www.SilentPCReview.com. Clock for clock yeah the P3 is much more efficient than a P4. Dothans are 0.09nm. How do they stack up again Turinos? Not sure, are Turinos even available yet?

 

[Super Nade]

Yes the Turions <spelling error on my part> are available in the market.
http://techbits.ca/modules.php?name=News&file=article&sid=972

Tested at xtremesys:
http://www.xtremesystems.org/forums/showthread.php?t=55174

Counter-claims:
http://www.theregister.co.uk/2005/03/15/amd_turion_bench/

 

[yeha]

the turions are an interesting bunch, i'm yet to see any tests with oc results above 2.0 ghz. sure, the one that got to 2 ghz on a bit over 1 volt is impressive, but these chips are designed especially for low voltage and low frequency, and may not scale anywhere near as high as their desktop brethren.

previous 'mobile' chips from amd were just cherrypicked out of normal batches iirc, which meant they scaled up in frequencies just as well as the normal chips. i doubt that will be the case with turions, they'll hit a clock speed wall and voltage won't help. i'd love to be proved wrong though

 

[man_utd]

Note: P-M and P3 are similar, but by no means remotely the same. P-M was a completely new design.

 

[Super Nade]

the turions are an interesting bunch, i'm yet to see any tests with oc results above 2.0 ghz. sure, the one that got to 2 ghz on a bit over 1 volt is impressive, but these chips are designed especially for low voltage and low frequency, and may not scale anywhere near as high as their desktop brethren.

previous 'mobile' chips from amd were just cherrypicked out of normal batches iirc, which meant they scaled up in frequencies just as well as the normal chips. i doubt that will be the case with turions, they'll hit a clock speed wall and voltage won't help. i'd love to be proved wrong though

So you are saying, these Turions are an entirely different architecture? Not much similarity with the Venice core?