P4 3.8Ghz (Prescott, 90nm) vs A64 s939 (90nm) power draw

halcyon · Post by **halcyon** » Mon Oct 04, 2004 11:22 am

How does this sound:

Athlon 64 socket 939 3500+ (90nm fabrication process) uses the same amount of power under full stress than Intel Preshott 3.4GHz uses at IDLE! And the Athlon is also faster (in real world use).

http://techreport.com/onearticle.x/7417

Less power draw translates to lower heat, less cooling needed and thus, less noisy system.

Now if the rest of the world would find out this and appreciate it as well, we'd probably save a few nuclear power reactors by switching to Athlon 64 CPUs :)

EDIT: Changed title to be less flammable :)
EDIT2: See 3rd post by Halcyon for link to 3.8Ghz P4 power draw

Bluefront · Post by **Bluefront** » Mon Oct 04, 2004 12:00 pm

Oh crap....another AMD fanboy, saving the world with fewer nuclear reactors, and less fan noise. Your logic stretches over the edge for me. I happen to like my Prescott, with its' slightly hotter running characteristics, and its' powerful performance.

The fact is.....the whole truth about performance doesn't show up in bench-marking. You have to own one for a length of time to appreciate how well it runs.

the_smell · Post by **the_smell** » Mon Oct 04, 2004 12:51 pm

I guess they are both on the powerful side, but how many people have used both a high end althon 64 and a P4 prescott for a significant amount of time doing similar things to be able to notice the difference (appart form the heat).

Heat's not a bad thing - I've dryed clothes on the exhaust of my poor old little barton 2800+, can you imagine how many towels I could dry using a prescott!?!

In fact I wander how many prescotts would be needed to heat a house...

Anyway, at the moment the humble prescott 3400 is tempting purely because they cost 25% less - so the athlon 3500 better be faster and cooler!

Tibors · Post by **Tibors** » Mon Oct 04, 2004 1:00 pm

That techreport article is already being discussed in this thread.

The article is really a comparison between the two AMD procs. The comparison with the Prescott is not "scientifically valid" since it uses another mobo and video card. Even the comparison for the AMD's is questionable since he didn't look at the Vcore.

We already know the Prescott is hot, but it looks as if you can built a silent system with an Intel Prescott, Thermalright XP-120, Nexus Real Silent 120mm fan and an undervoltable motherboard.

The only real world use I have for the maximum output of my processor is F@H. For that a hyperthreaded Pentium still outperforms an AMD 64 of equal price. Now as soon as that changes I'll buy an AMD processor for my next system.

Edit: typo

Beyonder · Post by **Beyonder** » Mon Oct 04, 2004 5:19 pm

Bluefront wrote:Oh crap....another AMD fanboy, saving the world with fewer nuclear reactors, and less fan noise. Your logic stretches over the edge for me. I happen to like my Prescott, with its' slightly hotter running characteristics, and its' powerful performance.

The fact is.....the whole truth about performance doesn't show up in bench-marking. You have to own one for a length of time to appreciate how well it runs.

In defense of Halcyon: I certainly am not any "fanboy" of one chip manufacturer. I'm writing this on a Pentium-M laptop (great machine), I code at work on a [email protected] ghz (great machine), and upstairs I have an old Athlon 1500+ that I really like (great gaming rig, unbelieveable performance for the price at the time of purchase, etc.).

With that said, let me also say this: the prescott isn't that great of a chip, IMO. Does it get the job done? Heck yes. Is it a good deal? Not neccessarily.

It runs smoking hot, the performance isn't anything AMD can't offer (nor vice-versa for that matter), and the price point isn't very compelling to me. For the first reason, Intel is going to be backing away from the P4 architecture, which was almost driven by "mhz rules" marketing hype since its inception. For those three reasons, I really don't feel much lure by Prescott. Again, this isn't to say it's a bad processor--I just think there's better deals out there. In fact, were Intel to offer an affordable Pentium-M motherboard that supported a dothan, I'd already own one, since I think the Dothan is perhaps the most efficient, realistic processors available at the moment. Until then, however, I think the best price/performance ratio, as well as the ease of silencing, is clearly in AMD's ballpark.

(sorry, I just don't like "fanboy" accusations--pardon my high horse and feel free to kick me a few times while I'm down =)

josephclemente · Post by **josephclemente** » Mon Oct 04, 2004 5:51 pm

Wow, those are some interesting numbers in that comparison.

I plan on skipping Prescott completely.

Project · Post by **Project** » Mon Oct 04, 2004 6:09 pm

i second that!

halcyon · Post by **halcyon** » Tue Oct 05, 2004 12:26 am

Hey relax people : )

I didn't mean to make mockery of your Prescotts.

I was just trying to beat the dead horse about Prescotts power consumption, because things like these tend to get forgotten quickly.

As for the comparison, it was very valid.

Let's look at the details:

Prescott was using DDR2 memory at 1.8 volts. This is consumes less wattage than the Athlon64 setup with PC3200 Corsair XMS XL at 2.75 volts.

Also, the PCI-E version of 6800GT (used on the Prescott platform) is clocked lower for both memory and core than the AGP version of the same card. This means that even the PCI-E card draws less power.

All the other components (sans motherboads, which are part of the platform under evaluation) were the same.

So, with lower power draw memory and video card Prescott was drawing the same amount of power at idle than A64 (90nm) was at full power.

The end results would have been even more in A64 (90nm) favour, had they actually both (P4 and a64) been using equally power hungry memory and ram.

And as for the other comments: those who know me, know I'm a platform agnostic: choose the right tool for the right job.

In this forum the job is silence, hence currently the choice would be 90nm A64 for power crunching jobs as long as Intel keeps Pentium M off the desktop.

Don't get me wrong, I'd really like to use Pentium M on my desktop platforms, but it's not really economically viable with all the hard-to-get mini-itx industry mobos and very expensive cpu prices.

Anyway, what ever platform you use, make the most out of it.

However, if you are now contemplating a silent platform, it'd be foolish not to take the power draw / heat output into consideration.

cheers,
halcyon

PS My thread was posted first

The article was discussed here first, not in the other thread, tsk tsk

Refs:
http://www.ocztechnology.com/products/m ... _Bandwidth
http://www.corsairmicro.com/corsair/pro ... 3200xl.pdf
http://www.bit-tech.net/review/352/

GlassMan · Post by **GlassMan** » Thu Oct 07, 2004 2:15 pm

Is any one familliar with the programs they were using? And why didn't they use prime95 test one as once it starts it's almost cpu only? I thought the difference in cpu power draw (idle to load) would be important.

halcyon · Post by **halcyon** » Sun Nov 14, 2004 11:11 pm

Anandtech did a review of 3.8GHz P4 (LGA775, 90nm) against A64 3500+ (s939, 90nm) and A64 FX-55 (s939, 130nm).

http://www.anandtech.com/cpuchipsets/sh ... =2275&p=13

At Full power 3.8Ghz P4 CPU consumes more than twice the amount of power* than A64 3500+ CPU (* considering that P4 is using lower voltage DDR2).

So, to me it's a no brainer. Unless you are into exotic supercooling, A64 is the easier choice for silent cooling in the high powered CPU front.

The gap is much smaller compared to FX-55, but it's still more than 60 watts in favour of FX-55.

regards,
halcyon

PS I think CPU Burn k7 is the standard for thermal loading of CPUs (consistently produces the highest temps in cpus). However as it does not constitute a 'real world load', I guess that's why it's not used as much as real applications are.

Ref: http://www.systemcooling.com/load_proc-02.html

JohnMK · Post by **JohnMK** » Mon Nov 15, 2004 7:42 pm

Actually, the Geforce 6600 GT AGP has memory clocked 100MHz lower than the PCI-E version of the same card. You had it slightly reversed my friend.

But this is very tentative . . .

halcyon · Post by **halcyon** » Wed Nov 17, 2004 1:25 pm

Actually we were both wrong

The core/mem clocks for 6800GT (not 6600GT) are the same for AGP/PCI-E versions (nvidia Reference boards): 350Mhz/500Mhz.

regards,
halcyon