Prescott Survival Kit

[ANova]

the althons will also win in code compiling, physics modelling and maths programs.

Office stuff tends to be a split between the 2.


Since this is a 3d forum though we like games so it's not suprising that there will be alot of amd users here.

That is because of the Athlon's higher number of IPCs. I'm interested in seeing how Pentium Ms compare in games since they are available for desktops now.

 

[Fox5]

the althons will also win in code compiling, physics modelling and maths programs.

Office stuff tends to be a split between the 2.


Since this is a 3d forum though we like games so it's not suprising that there will be alot of amd users here.

That is because of the Athlon's higher number of IPCs. I'm interested in seeing how Pentium Ms compare in games since they are available for desktops now.

They are? I thought those were only special cases, where companies where taking labtop parts and putting them into a pc. Anyhow, pentium ms definetely fall behind on a dollars to performance comparision, but I think with a 400mhz fsb they're about 100-200mhz behind the athlon 64 in most cases, and perhaps with a 533mhz fsb slightly ahead.

 

[ANova]

http://www.newegg.com/app/ViewProdu...&propertycode=&propertycodevalue=5170

The 2.0 and 2.13 GHz versions with 533 fsb will be coming out Q4.

 

[Fox5]

http://www.newegg.com/app/ViewProduct.asp?catalog=343&DEPA=0&submit=property&mfrcode=0&propertycode=&propertycodevalue=5170

The 2.0 and 2.13 GHz versions with 533 fsb will be coming out Q4.

Those chips have been available like ever since centrino came out. Just because you can buy them doesn't mean they work in anything, show me a desktop centrino motherboard.(and not a labtop that's been chopped up)

 

[ANova]

True, motherboards are few and far between but there are a couple.

http://www.radisys.com/oem_products/ds-page.cfm?productdatasheetsid=1158

http://www.globalamericaninc.com/products_services/sbc/2807598.php

What I'm confused about is newegg listing the Pentium Ms as socket 478.

 

[Fox5]

True, motherboards are few and far between but there are a couple.

http://www.radisys.com/oem_products/ds-page.cfm?productdatasheetsid=1158

http://www.globalamericaninc.com/products_services/sbc/2807598.php

What I'm confused about is newegg listing the Pentium Ms as socket 478.

Pentium Ms are socket 478, they just don't work in normal pentium 4 motherboards. Or at least I think they are, one of those motherboards says 479, but I thought pentium m's were physically, but not electrically compatible with p4 motherboards.

 

[Althornin]

BS.
The intel temp sensor is waaaaay low.
besides that, runs at 40-50 C with what heatsink? And ambient?

LOL, where do you get your information? My P4 runs exactly where I said it does, between 40-50 C; that's with the stock hsf btw. Temp sensors can be incorrect but saying all Intel sensors are is just rediculous. When was the last time you checked on P4 temperatures?

Your p4 runs at those temps according to what?
I get my information from empirical evidence, where do you get yours, from the notoriously inaccurate (and COOL READING) temperature sensors in the p4's?
You completely miss the point, and then attempt to validate it with "evidence" from said problematically low temperature sensing diode.
Look, dont take it personally - tis just marketing.
do a little thought experiment, ok?
You say 40-50 degrees Celcius, right?
Lets say 50. And lets say ambient is 25 degrees.
Ok, so now we have a delta T of 25 degrees.
and the proc puts out, say, 100 watts of heat dissiapation (typical is 85 watts, max exceeds 100).
Ok, so the C/W for your aircooler is a nice .25
Yeah, right. That is quite simply impossible, considering that the best watercooling solutions barely approach a C/W of .2

I know it must come a quite ablow to you that intel uses a purposefully low reading thermal sensor as a marketing tool, but get over it.

Did you know, for example, that the thermal shutdown sensor is not the same sensor that passes temps onto the motherboard? Ever wonder why?
Its called marketing, and its worked on you.

 

[ANova]

Your p4 runs at those temps according to what?
I get my information from empirical evidence, where do you get yours, from the notoriously inaccurate (and COOL READING) temperature sensors in the p4's?
You completely miss the point, and then attempt to validate it with "evidence" from said problematically low temperature sensing diode.
Look, dont take it personally - tis just marketing.
do a little thought experiment, ok?
You say 40-50 degrees Celcius, right?
Lets say 50. And lets say ambient is 25 degrees.
Ok, so now we have a delta T of 25 degrees.
and the proc puts out, say, 100 watts of heat dissiapation (typical is 85 watts, max exceeds 100).
Ok, so the C/W for your aircooler is a nice .25
Yeah, right. That is quite simply impossible, considering that the best watercooling solutions barely approach a C/W of .2

I know it must come a quite ablow to you that intel uses a purposefully low reading thermal sensor as a marketing tool, but get over it.

Did you know, for example, that the thermal shutdown sensor is not the same sensor that passes temps onto the motherboard? Ever wonder why?
Its called marketing, and its worked on you.

I know your refering to the fact that the thermal sensors are located away from the center of the die where most of the heat is generated; thus the temperatures the motherboard reads aren't exactly true to actual temperatures. However, it's rarely more then a 2-8 degree difference and as long as the cpu works with no stability issues I fail to see the importance. My original point still stands, Prescotts run around 10 degrees hotter then Northwoods.

 

[Vortigern_red]

My original point still stands, Prescotts run around 10 degrees hotter then Northwoods.

Here are some figures from Prescott reviews.

Hocp

16 degrees


TR

14 degrees

Under load, the Northwood P4 3.4GHz hit temperatures of 64 degrees C (148 degrees Fahreneheit). In the same conditions, the Prescott raced past Northwood's peak temperature on its way to a steady peak of 78 degrees C (or 173 degrees Fahrenheit).

Aces

approx 19 degrees

The temperature difference was big, though. After running a 3DSMax rendering and restarting the PC, the BIOS reported that the 3.2 GHz Northwood was at about 45-47°C, while Prescott was flirting with 64-66°C. Mind you, this is measured on a motherboard completely exposed to the cool air (18°C) of our lab.

TR also supply figures for the Athlon64 3400+ on the same page

After running under load for quite some time, the Athon 64 3400+ reached a steady peak of 52 degrees C. I saw the indicator jump up to 53C once

The Prescott according to TR numbers is nearly 50% hotter than a similar performing (faster on some things slower on others ) Athlon64. If I was AMDs marketing Dept I think I would be making some noise about this as well!

 

[ANova]

Those numbers are very inconsistant. And of course the Athlon 64 has better temperatures, it uses SOI and doesn't run at 3.0 GHz+.

 

[Fox5]

Those numbers are very inconsistant. And of course the Athlon 64 has better temperatures, it uses SOI and doesn't run at 3.0 GHz+.

Isn't northwood cooler than athlon 64 though, and it runs at 3.0+ ghz?

 

[ANova]

Northwoods are about the same, maybe slightly less. Thing is, the Northwood operates with a 20 stage pipeline, the Prescott's is 31, the Athlon 64's is somewhere around 12-15 I believe.

 

[Vortigern_red]

From the tech report article:

Athlon64=53
Northwood=64
Prescott=78

Granted the figures I posted in my previous post are not as consistant as your "about 10 degrees" so you must be right

I don't see the relevence of the number of pipeline stages, nor for that matter clock speed, we should be comparing performance to temp surely!

IE the AthlonXP has fewer stages than the Athlon64 (10 V 12) however at a given clock speed it is much HOTTER for LESS performance.

The Northwood has fewer stages than the Prescott (20 v 31) however at a given clock speed it is much COOLER for MORE performance.

It makes more sense to compare performance to temp IMHO.

I can't belive that you are trying to argue that Prescott does not run hot, irrespective of how cool the Athlon64 runs the Prescott is about 15 degrees (In your, consistant, opinion 10 degrees) hotter than the previously hottest chips on the market, the Northwood 3.2/3.4 and the AthlonXP 3200+. Thats pretty warm.

The temp increase might not make much difference to you but it will to a lot of people.

Anyway Gamepc has some,very limited, Dothan benchmarks but have a look at them. That chip is a much better competitor for Athlon64 than Prescott IMHO and the sooner its *really* available for desktops the better for everyone (except perhaps AMD )

 

[ANova]

IE the AthlonXP has fewer stages than the Athlon64 (10 V 12) however at a given clock speed it is much HOTTER for LESS performance.

The Northwood has fewer stages than the Prescott (20 v 31) however at a given clock speed it is much COOLER for MORE performance.

It makes more sense to compare performance to temp IMHO.

The longer the pipeline the more heat generated. The A64's increase in it's pipeline is a mere 2 over the XP which is offset by the fact that it uses SOI. That is the only reason why it runs less hot then the XP. The Prescott, on the other hand, has an increase of 11 in it's pipeline over the Northwood and does not use SOI or low-k. I just think Intel took the wrong path in designing the Prescott, it's inefficient.

That chip is a much better competitor for Athlon64 than Prescott IMHO and the sooner its *really* available for desktops the better for everyone (except perhaps AMD )

I agree. Sadly they won't be supporting hyperthreading though. My guess is Intel wants it as an exclusive feature to advertise for it's Prescotts.

 

[Vortigern_red]

The longer the pipeline the more heat generated.

Is this off the top of your head?

The Winamette was 20 stages and ran much cooler than AthlonXP (10 stages) at the same clock speed The AthlonXP Tbred/Barton also ran hotter than the Northwood at the same clockspeed.. The number of piplines has little or nothing to do with heat AFAICS. Also Thunderbird and Palomino had the same number of stages and Palomino ran cooler due to some process tweaks. There are far greater differences between Athlon64 and Prescott than pipeline stages to account for the heat difference.

Can you supply an example that proves your point? To be honest my assumption has always been that, all other things equal, a longer pipeline should equal less heat ie lower IPC = less work done = less heat. But I admit I've never really looked into it.

mere 2

That will be a *mere 20%" then

hat is the only reason why it runs less hot then the XP. The Prescott, on the other hand, has an increase of 11 in it's pipeline over the Northwood and does not use SOI or low-k.

This is what I am saying the chip is the sum of its parts the fact that Intel chose higher clocks, longer pipelines and did not use SOI was their choice. You keep pulling parts of the chip out and using them as an excuse for the Prescotts heat problems. But those things ARE Prescott, without a longer pipeline, high clocks and a low IPC Prescott would be......Dothan, perhaps!

I just think Intel took the wrong path in designing the Prescott, it's inefficient.

Time will tell I suppose, but I'm in real danger of agreeing with you here

I think I'll call it a day here, I know what I'm like I'll be posting for weeks otherwise!

 

[ANova]

Can you supply an example that proves your point? To be honest my assumption has always been that, all other things equal, a longer pipeline should equal less heat ie lower IPC = less work done = less heat. But I admit I've never really looked into it.

The longer the pipeline the more work is required per cycle which means more heat is generated but which also allows for higher frequencies. The P4 takes a brute force approach in that it strives to achieve performance through high operating frequencies. The Athlons on the other hand go the other route, through more IPC. You are correct in that more IPC also = more heat and this is why Athlons have always been no slouch in heat dissipation up until the Athlon 64 either. The Prescott is built around the same architecture as the Northwood so it inherits it's flaws; when you add new instructions and scale operating frequencies even higher heat is bound to increase as a result, especially if you do nothing in an attempt to decrease that heat. Prescotts may run hot but they are able to scale to ~4.0 GHz without much (if any) increase in thermal output. I think if Intel had decided to implement SOI or low-k the numbers would have been more akin to the Northwood and A64 but they decided to opt out for whatever reason.

 

[Vortigern_red]

The longer the pipeline the more work is required per cycle

The longer the pipeline the LESS work is done per clock cycle not more!

 

[ANova]

The longer the pipeline the LESS work is done per clock cycle not more!

Your misunderstanding me. The longer the pipeline the higher the frequencies that are allowed, heat increases with frequency increases. However, since the Prescott has a 31 stage pipe it's rather inefficient since the processor has to throw out alot of data when it misses predictions and thus uses up more clock cycles. This means more work for the cpu and more heat generated as a result.

 

[Vortigern_red]

heat increases with frequency increases.

I agree with this but your "branch misprediction" theory is mad!

The Dothan suffers from branch misprediction as well, if it gets it wrong 80% of the way down the pipe then it has done 80% of the work, if the P4 flushes it 80% of the way down the pipe then it to has done 80% of the work. The P4 has used more clock cycles but has done no more work!! The P4 does not have more branch mispredictions then most other chips it just loses more clock cycles because of them. But it loses the same amount of work done.

I'm really giving up this time. I can see no evidence for your "longer pipeline= higher temps theory whatsoever. You have not posted a single example.

EDIT: baby daughter decided to post this for me before I was finished.

 

[ANova]

heat increases with frequency increases.

I agree with this but your "branch misprediction" theory is mad!

Mad? Or just so crazy that there might actually be some truth to it?!

The Dothan has a much smaller pipeline and thus does more IPC like that of the P3. Let me break it down for you. The Prescott produces 103 watts of heat because it has a longer pipeline, is designed to run at higher frequencies, packs more transitors in a smaller area then either the Northwood or the Athlon 64, does not use SOI or low-k and supports more instructions sets. Need I say more?