Dothan Review: Powerful and low-power

Mephistopheles · Jul 21, 2004

Anandtech has just <A HREF="http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2129&p=1" target="_new">posted</A> this very interesting review of the new Pentium M. It's really very sad to see this architecture being passed away as a possible desktop dual-core solution; if it were given all the bells and whistles (current P4 FSB, dual-channel memory controller, x86-64 technology) it would be a worthy contender even to the 2Ghz A64!

Why try to put two prescotts on one die, if you can put two 2Ghz Dothans on a die (they're even slightly smaller) and end up with a 4MB cache, 2x2Ghz cpu with 42W of heat dissipation? One dual-core dothan would consume far less than a single-core A64 or P4, easily.

Why don't you call in your arbiter chip in dothan and release it instead of a dual-core prescott, Intel?....

You never change the existing reality by fighting it. Instead, create a new model that makes the old one obsolete - Buckminster Fuller

TheRod · Jul 21, 2004

I fully understand your point! I personnaly think that Dothan have the potential to be the Athlon 64 "killer". But, Intel must make money with the Prescott, they need ROI on this technology and most home users don't care about heat issue and power requirements.

Dothan was designed with mobility in mind and Intel succeeded with this design and it's so bad for users like us who would consider this chip for desktop use if they were available.

--
Asus A7N8X / AMD Sempron 2800+ (tbred @ 167x12)
Kingston DDR333 2x256Megs
Built by ATI Radeon 8500LE 128Megs @ C:275/M:290

aaron2504 · Jul 21, 2004

yeh it is a good idea for a dual core dothan. but intel are being themselfs. (being stupid)

trooper11 · Jul 21, 2004

once again, dont forget that amd could easily release dual core opterons with a total of 66W if using the current low voltage 33W model and thats on 130nm. if they can pull of 90nm without a hitch, you could see it at the same low voltage as dothan. as far as performance, they are oh so clsoe in many areas, but the problem for intel is, if they switch over to dothan, they will sacrifice the elads they have in a few areas over the athlon 64 in the short term. for instance in video editing and rendering, which have been intel strongholds. i dont think they want to ever hear anyoen say athlon is better across the board performance wise, no matter what the cost lol.

Mephistopheles · Jul 21, 2004

"If" can sometimes be a rather pointless word, isn't it?

I mean, if Intel has a 1Ghz ULV Dothan with 5W TDP and 2MB cache, they could put 16 cores and end up with a 16-core, 80W-thermal output processor which would, when executing massively parallel code, kick ass of just about anything.

Mind you, Intel could also put 4 dothan cores and end up with a 84W 4-cored processor*.

What is important in this whole story is who gets a dual-core processor out the door first, and how and with what configuration. Dothan has the added advantage that you don't even need a clock decrease to keep power down, that's why it's a good candidate.

In any case, because I don't understand jack about chip engineering and manufacturing (all I know is using them), I don't think I really know what will happen.... Interesting things are bound to happen anyway because of performance races! So why worry and speculate? I say let's enjoy it step by step, one step at a time...

You're right to say, AMD could end up releasing a better dual-core eventually... if they do it faster, then intel has real trouble (here's that "if" word again)...

* - BTW, a 4-core dothan-based Xeon has been reported to be in development in Intel's India Labs, but is still a long road ahead.

You never change the existing reality by fighting it. Instead, create a new model that makes the old one obsolete - Buckminster Fuller

trooper11 · Jul 22, 2004

yes dothan wouldnt have to lower clock speed to keep power down, but that would be the same for the low voltage opterons. i consider both of those chips very good candidates for dual core. prescott is not. opterons have the added advantage of having be designed to include the infrustructure for dual core to begin with lol.

besides, i doubt we will see dual core form either of them till next year.

Mephistopheles · Jul 22, 2004

Absolutely, dual core from either one of them will only come in 2005.

Besides, AMD still has the 90nm transition to do.

And Intel still has to debug the 90nm chips (the "J" edition) and implement 64-bit functionality.

There are a few things on the to-do list before going dual-core on both companies.

You never change the existing reality by fighting it. Instead, create a new model that makes the old one obsolete - Buckminster Fuller

P4Man · Jul 22, 2004

I'm not so impressed. A 90nm chip with 2 MB cache performs on par with a 130nm, 512kb part from AMD. Wow..I'm speechless

A few things to consider:
1) Anand crippled all the cpu's equally (single channel, slow memory,..), which isnt unfair, but does give a bigger advantage to the huge cache of Dothan. Had he added a P4EE, I think the results would be far less promising.
1.5) upping the FSB and adding dual channel will do much less for Dothan than it would for the P4C, and to a lesser extend, the A64.
2) it remains to be seen how high Dothan can be clocked. Clearly, thermals aren't the problem, but other factors are much more likely to limit Dothan to clock speeds well below what is achieveable for A64 on a similar process. Loosen the thermals, up the Vcore, and ~2.5 should be within reach, but that will be less than even the initial 90nm A64's. 3+ GHz is out of the question IMHO.
3) dual core is nice and all, but not very different from SMP performance wise. How many people are running 2 way systems here ? Why do you think there arent any more ? A 2 way Athlon MP 2400+ is cheaper than a 3800+. Simple anwser: the benefit isnt there for >90% of the applications. Single threaded performance still matters much more on the desktop. Besides, by the time you'll see dual core "dothans", you'll see dual core K8's. You may well see the K8's earlier, either as opteron or Athlon FX as well.

In short: what is impressive about Dothan is its extreme small size and low power consumption. Great for mobile, not that relevant for the typical desktop. You can't just trade them for more performance.

= The views stated herein are my personal views, and not necessarily the views of my wife. =

Xeon · Jul 22, 2004

In short: what is impressive about Dothan is its extreme small size and low power consumption. Great for mobile, not that relevant for the typical desktop. You can't just trade them for more performance.

Your vision messed?
1) Why would slower FSB benefit the larger cache? It theory it should hinder it as it did the P4 and A64.
2)Why is that, how could upping the FSB and adding dual channel do less than upping it on the P4 or A64. Every single piece of silicon that moved to dual channel and a faster FSB always benefited even if it was only marginal.
3) The CPU is designed around power-saving so upping the V-core would probably just aggravate the silicon. But hey you always seem to sound like you are an IC engineer so you must know :tongue: .
4) Dual core latencies will be far lower than that of a SMP system.

In short: what is impressive about Dothan is its extreme small size and low power consumption. Great for mobile, not that relevant for the typical desktop. You can't just trade them for more performance.

Well considering it puts out near 1/2 of the thermal output an A64 does, doesn’t have an on die memory controller, or a screaming fast FSB and a shaky micron process. I think it holds out extremely well.

With considerations it has no SSE2 and will probably never have it means it will never make it “the” desktop processor for Intel. But nothing is out of the realm of possibly.

AMD should fear this silicon since clock for clock it does more work while putting out far less heat with less FSB and other goodies that give the A64 and the P4 the edge on it.

Xeon

Post created with being a dickhead in mind.
For all emotional and slanderous statements contact THG for all law suits.

Mephistopheles · Jul 22, 2004

I'm not so impressed. A 90nm chip with 2 MB cache performs on par with a 130nm, 512kb part from AMD. Wow..I'm speechless

You're completely missing the picture here, I'm afraid... The thing is that Dothan is currently completely geared towards lower heat output, and A64 as an architecture couldn't care less about it - it's geared toward speed. What happens to be the case is that Dothan, while still NOT having as a main objective performance, is on par with a highly performance-driven architecture while still mantaining its own merit in what it was designed to do: be a mobile solution.

While I understand your things to consider, I also don't think you're highly qualified (sorry 'bout that) to accurately predict to what extent Dothan can benefit from dual-channel memory and 1066FSB. Neither are we, for that matter, but I'm sure it should have an impact. Why else would Intel have rigged Banias and Dothan with a quad-pumped FSB? Because it couldn't benefit from it?....

Also bear in mind that most of Dothan's architecture is secret. So unless any of us actually works for Intel's R&D department, all we can do is speculate... and try to stay unbiased and with good will.

You never change the existing reality by fighting it. Instead, create a new model that makes the old one obsolete - Buckminster Fuller

trooper11 · Jul 22, 2004

well first of all, just because intel made a choice in design doesnt mean its gold or that it was a good thing, i never just trust either company that everything they impliment must be good cause they 'know better' we have seen thats not the case.

lets look at this antoher way, dont forget that the price of a pentium m on its own right now is about on par with the low voltage opterons, so if you compare it based on that, they are really close in wattage. if you want to compare dothan to desktop cpus, then you have to consider what ti would cost to build a p-m system if you wanted one, otherwise this whole review was useless. i agree dothan is an achievement, but only in the mobile arena. until we see desktop chips at reasonable prices, you cant compare it to the desktop chips, which can offer better support parts like ram and bus speed.

it would have been nice to see it compared to the 35w athlon 64 low voltage mobiles too.

is the dothan a success? yes definitely
does it have no competition in the low voltage area? yes it does

this review just brings up more questions and not enough answers for me lol. ill wait till i can see it on adesktop motherbaord.

P4Man · Jul 22, 2004

>Your vision messed?

Not really.

>1) Why would slower FSB benefit the larger cache? It theory
>it should hinder it as it did the P4 and A64.

A bigger cache goes a long way hiding slow I/O and high memory latencies. Thats what it is therefore. The faster the memory subsystem, the less a big cache will pay off. So 2 MB does wonders with single channel slow memory, but it will provide less of a benefit (relative to small cache P4s and A64s) with faster fsb's and memory subsystems. IOW, the P4 and (to a lesser extend) the A64 will benefit more from faster I/O than Dothan. If you don't believe that, just compare P4B scores with P4C and/or single channel versus dual channel. Pretty damn big difference, and bigger than Dothan will enjoy.

> Every single piece of silicon that moved to dual channel
>and a faster FSB always benefited even if it was only
>marginal.

Obviously they will all benefit; question is how it would change there performance compared to each other. As I pointed out, Dothan will most likely gain less. IOW, P4 and A64 suffered more from the I/O castration Anand performed than Dothan. Give all cpu's a 66 MHz EDO RAM interface, and Dothan may well be 2x faster than the 512Kb P4.

>The CPU is designed around power-saving so upping the V-
>core would probably just aggravate the silicon

It is designed to operate at low voltages, which typically hinders clockscaling. All cpu's I'm aware off, with maybe the exception of K8 (built on SOI) benefit substantially from higher voltages to achieve higher clocks. Do'nt tell me you don't know that..

> Dual core latencies will be far lower than that of a SMP
>system.

And why is that ? I don't see the difference.. in both cases both cores will share the same FSB and memory controller. Only intercpu (intercore) traffic will be faster for things like cache snoops. Don't hold your breath..

>Well considering it puts out near 1/2 of the thermal output
>an A64 does, doesn’t have an on die memory controller, or a
>screaming fast FSB and a shaky micron process.

TDPmax isnt anywhere near 2x that of A64 on the same process, they are in fact pretty close at the same clock, same process, especially if you take into account the power dissapated by the northbridge and FSB which is not included in the Pentium M numbers, but which is included in A64. And intels 90nm is just fine; compare Dothan to Banias and you'll see they did pretty well.

>With considerations it has no SSE2

Sure it has.

>AMD should fear this silicon since clock for clock it does
>more work while putting out far less heat with less FSB and
>other goodies that give the A64 and the P4 the edge on it.

As I said, its a very nice mobile cpu, but I wouldn't fear it as a desktop/workstation chip unless I grossly underestimate its clocking potential (which I think falls far short of A64s).

= The views stated herein are my personal views, and not necessarily the views of my wife. =

P4Man · Jul 22, 2004

>You're completely missing the picture here, I'm afraid...

I doubt it..

>The thing is that Dothan is currently completely geared
>towards lower heat output, and A64 as an architecture
>couldn't care less about it

Well, the A64 doesnt do too bad then in the power consumption department if it "couldn't care less about it". Its pretty damn close to Pentium M in terms of TDP at similar clockspeeds, same process nodes and roughly equivalant performance.

> What happens to be the case is that Dothan, while still
>NOT having as a main objective performance, is on par with
>a highly performance-driven architecture

Is it really ? The fastest $600 Dothan can keep up with a crippled bread and butter 2 GHz A64 that only has a quarter of the cache, and that is produced on an older process. On 130nm, Banias scaled to 1.7 GHz, A64 to 2.4. Want to compare them, with full fledged I/O and claim they are on par ? Or even close ?? Want to compare a 2.13 GHz Dothan with a 90nm Athlon 4000+ or FX55 ?

I thought not, so what you *are* saying is that this mobile chip could *perhaps* go toe to toe with the A64 *if* it is given a much higher clock, faster fsb, better FPU, dual channel, etc, etc. Which could all be true, but remains to be seen. Its your speculation, but I'm more sceptical than you.

> I also don't think you're highly qualified (sorry 'bout
>that) to accurately predict to what extent Dothan can
>benefit from dual-channel memory and 1066FSB. Neither are
>we, for that matter, but I'm sure it should have an impact

OF course it will have an impact; however, common sense tells me the raison d'etre of a L2 (or L3) cache is to hide memory latency. The bigger the cache, the less of an impact a slow FSB or slow I/O has. Vice versa, the faster the I/O, the less a big cache will help (think K8). You don't need any qualifications to understand this principle. Dothan was outfitted with a big cache for a reason: keep performance good even on slow (less power hungry) I/O subsystems in notebooks. A64 cache reduced because it has such fast I/O that the bigger caches barely made a difference. Now if you cripple the A64 (and much more even, the bandwith hungry P4C!) to make an "even" comparison with Dothan, you can't extrapolate these results to predict what will happen if you outfit all cores with an equally fast I/O. Dothan will benefit less than P4C, there is no way around that. How much less remains to be seen, but it will be less period.

>and try to stay unbiased and with good will

I tend not to lack either much if I dare say so myself, but feel free to disagree. But claiming two cores are "on par" when comparing the newest, fastest chip of one with a crippled, low clocked part produced on an older process is not my idea of seeing the picture or showing a lack of bias.

= The views stated herein are my personal views, and not necessarily the views of my wife. =

Xeon · Jul 22, 2004

All cpu's I'm aware off, with maybe the exception of K8 (built on SOI) benefit substantially from higher voltages to achieve higher clocks

Isn't the P4E's build with Intel’s SOI technology as well? Also this core is built around low voltages which means is logic units will be running slim too. I really don’t see that particular silicon taking a likening the increased voltages and clock speeds since the core is designed around clock efficiency and increasing voltages changes the behaviors of the silicon logic.

And why is that ? I don't see the difference.. in both cases both cores will share the same FSB and memory controller. Only intercpu (intercore) traffic will be faster for things like cache snoops.

On SMP systems there is quite some space between the cores, on a dual core machine it would be a matter of microns. I see the latencies being lower on the dual core systems in comparison to the SMP machines of today.

TDPmax isnt anywhere near 2x that of A64 on the same process, they are in fact pretty close at the same clock, same process, especially if you take into account the power dissapated by the northbridge and FSB which is not included in the Pentium M numbers, but which is included in A64.

<A HREF="http://www6.tomshardware.com/cpu/20040201/images/cpu_history_big.gif" target="_new">Here</A>
It states the power dissipation is near 84-89watts for the A64's, that is no where near the maximum 21watts from the current Pentium-M.

And intels 90nm is just fine

It's no where near from perfected as Intel had its 0.13u process down.

Sure it has.

Ya your right nice catch on that one, but it goes to say its performance is lackluster, and the "digital world" can't have a lackluster digital media processor.

As I said, its a very nice mobile cpu, but I wouldn't fear it as a desktop/workstation chip unless I grossly underestimate its clocking potential (which I think falls far short of A64s).

I don’t see clock speed being an issue the chip is going to be over 2 GHz in Q4. I think it’s a manufacturing process issue since the cores thermal specs have to stay in that 30watt area.

Xeon

Post created with being a dickhead in mind.
For all emotional and slanderous statements contact THG for all law suits.

P4Man · Jul 23, 2004

>Isn't the P4E's build with Intel’s SOI technology as well?

Strained silicon, not SOI.

> Also this core is built around low voltages which means is
>logic units will be running slim too. I really don’t see
>that particular silicon taking a likening the increased
>voltages and clock speeds since the core is designed around
>clock efficiency and increasing voltages changes the
>behaviors of the silicon logic.

Pretty funny paragraph.. If you had a point I missed it, it sort of got lost in all the nonsense you're brabbeling.

>On SMP systems there is quite some space between the cores,

LOL.. yeah, true, I forgot about the "space"

>It states the power dissipation is near 84-89watts for the
>A64's, that is no where near the maximum 21watts from the
>current Pentium-M.

*yawn*.. not again

I'm sorry, I'm getting tired of having to correct each and every statement you make. IF you want to believe A64 consumes 4x as much as Banias or even Dothan, be my guest. but you're off by roughly a factor 3x-4x.

Mobile A64s have a TDP of 35W @1.8 GHz, using a "TDP" that is much more conservative than Intels (Dothan has a TDPmax of ~37W using AMD's way of calculating TDP), that number even includes FSB and (most of the) northbridge power consumption, its achieved without fancy mobile technologies like cache gating and fsb gating, oh.. and on 130nm, not 90nm. Dothan beats K8 when it comes to average powerconsumption and low power states, (which matters a lot in mobile, but not on the desktop), but in TDP (which matters on the desktop and to achieve dual core) they are pretty damn close. Don't be surprised if K8 does even better on TDP than Dothan once it moves to 90nm.

>I don’t see clock speed being an issue the chip is going to
>be over 2 GHz in Q4.

Great, before that time A64 will be at 2.6.. even using 130 nm instead of 90nm. Dothan (90nm) may or may not reach 2.6 GHz as a desktop chip, but I don't believe for a second it could keep up with A64 on a similar process.

= The views stated herein are my personal views, and not necessarily the views of my wife. =

trooper11 · Jul 23, 2004

It states the power dissipation is near 84-89watts for the A64's, that is no where near the maximum 21watts from the current Pentium-M.

yet you cant deny that the mobile athlon 64's at 35W and the opteron EE line at 30W are that high, they can both compete agressively against dothan for low voltage... you cant just exclude those two form this comparision.

Xeon · Jul 24, 2004

Pretty funny paragraph.. If you had a point I missed it, it sort of got lost in all the nonsense you're brabbeling.

The start reading what I type then, when you increase voltages you increase them on a "global" scale on the processor. Logic is very sensitive about voltages, with considerations that logic usually in very dense transistor counts and high draws in comparison to cache.

Upping the voltages on a processor that is designed to run slim will undoubtedly cause problems with the logic, plain and simple. When you OC what do you think goes first, the cache or maybe the prefect engine perhaps?

*yawn*.. not again I'm sorry, I'm getting tired of having to correct each and every statement you make. IF you want to believe A64 consumes 4x as much as Banias or even Dothan, be my guest. but you're off by roughly a factor 3x-4x.

You should go to sleep if you’re yawning. If you want me to believe otherwise find some solid thermal specifications on the A64. With considerations that the block on that chip is just as large as the P4's and RPM's are near the same, I just don’t see people running 40-50 on load on those coolers, running CPU's that aren’t putting out 60-90 watts of heat.

yet you cant deny that the mobile athlon 64's at 35W and the opteron EE line at 30W are that high, they can both compete agressively against dothan for low voltage... you cant just exclude those two form this comparision.

Now we are talking what’s the performance of a mobile A64 vs. a Pentium M clock for clock watt for watt?

Xeon

Post created with being a dickhead in mind.
For all emotional and slanderous statements contact THG for all law suits.

trooper11 · Jul 24, 2004

i frankly dont knwo what your trying to compare....

what is mroe important? wattage? performance? best ratio?

if its performance, then if you compare the mobile ahtlon 64 and the low voltage opterons versus the dothan cpu, then that would answer the performance question. then if you get the wattage numbers, then you add that to the ratio. im telling you though, the opterons would probably have the ebst performance to wattage ratio.

frankly i dont knwo what the squabble is about, if its just low wattage, thats misleading. doth's are good, but they wont see any true potential til they make it to a desktop with all the bus and ram that desktops have.

P4Man · Jul 24, 2004

>pping the voltages on a processor that is designed to run
>slim will undoubtedly cause problems with the logic, plain
>and simple.

Plain and simple nonsense. An (U)LV chip is typicaly a cherry picked part that happens to run on a lower voltage. Think mobile 2500+. These chips run just as happily with the higher voltage of their "desktop counterparts" (same chips, just binned differently), and almost always, at a higher frequency. Its just that there is no "desktop" PM yet, but the logic holds just as well.

Dothan is built on the same process as Prescott, there is no reason to assume its transistors could not withstand Prescotts Vcore. In fact, afaik electromigration is more dependant on temperature than voltage, so I wouldn't be surprised if Dothan could handle an even higher Vcore as Prescott, since it will probably run cooler as Prescott at the same voltage. How much extra clockspeed that would bring is another matter entirely though..

>If you want me to believe otherwise find some solid thermal
>specifications on the A64

<A HREF="http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/30430.pdf" target="_new">http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/30430.pdf</A>

Page 24, 35W TDP @1.8 GHz. Including memory controller and hypertransport. Using a way to calculate TDP that is not directly comparable to intel (AMD uses a TDPmax, intel a TDPtypical).

Up Dothans FSB to 800 MHz to match A64's performance, include dual channel memory and add the power the memory controller requires, use the same way to calculate TDP and you're looking at two similar figures. Oh, did I mention the A64 numbers are for 130nm ? Dothan shines for its low average powerconsumption and extremely low power consumption in sleep and deep sleep modes which makes it a terrific mobile chip allowing great battery life, but for maximum power consumption, it is not in a different league as A64 using the same premises.

>I just don’t see people running 40-50 on load on those
>coolers

Indeed not. The last 3000+ I built would not exceed 39°C under load.

= The views stated herein are my personal views, and not necessarily the views of my wife. =

Xeon · Jul 24, 2004

Plain and simple nonsense. An (U)LV chip is typicaly a cherry picked part that happens to run on a lower voltage. Think mobile 2500+. These chips run just as happily with the higher voltage of their "desktop counterparts" (same chips, just binned differently), and almost always, at a higher frequency. Its just that there is no "desktop" PM yet, but the logic holds just as well.

Dothan is built on the same process as Prescott, there is no reason to assume its transistors could not withstand Prescotts Vcore. In fact, afaik electromigration is more dependant on temperature than voltage, so I wouldn't be surprised if Dothan could handle an even higher Vcore as Prescott, since it will probably run cooler as Prescott at the same voltage. How much extra clockspeed that would bring is another matter entirely though..

The same process that’s your argument the same process means they should be able to handle the same voltages?

You are aware that different parts of the processor are made with different types of transistor gates. Different transistor gates handle different voltages, different speeds, different amounts of heat tolerances different densities the whole nine yards man.

So let me again state that the Pentium M's are designed around thermal output limits. Those limits will include core voltage and core draw. Those two variables alone rely heavily on transistor design implementation and electrical characteristics.

I am not saying it is impossible which I realized was what it appeared I was saying. The reality of the situation is if that core comes to desktop one day it will be severely redesigned to accommodate the variables that make a desktop processor different from a mobile one.

Page 24, 35W TDP @1.8 GHz. Including memory controller and hypertransport. Using a way to calculate TDP that is not directly comparable to intel (AMD uses a TDPmax, intel a TDPtypical).

Here's a little formula I use to determine the max thermal output of processors.
DCD*V=TOmax
DCD =Direct current draw, in amps.
V =Max voltage of the silicon at load.
TOmax =Maximum thermal output.

64.8A * 1.42688V = 92.461824W 3.0 GHz P4C
71.5A * 1.41415V = 101.111725W 3.2 GHz P4EE
91.0A * 1.26805V = 115.39255W 3.4 GHz P4E

21A * 1.277V = 26.817W 2.00 GHz PM
21A * 1.277V = 26.817W 1.80 GHz PM
21A * 1.277V = 26.817W 1.70 GHz PM

Now since AMD is unable to break the laws of physics, it's a pretty safe bet that the formula I used for the P4's and PM's will work.

57.8A * 1.5V = 86.7W 2.0 GHz A64 (3000+)
25.2A * 1.3V = 32.76W 800 MHz A64 (3000+) Cool N' Quiet Mode.
57.4A * 1.5V = 86.1W 2.4 GHz FX53

As you can see your 35watts comes from an 800 MHz clock speed attained from Cool N' Quiet, not the full clock speed and our European friend so claims.

Xeon

Post created with being a dickhead in mind.
For all emotional and slanderous statements contact THG for all law suits.

Schmide · Jul 24, 2004

Check out the AMD2800BQX4AX on page 24

<A HREF="http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/30430.pdf" target="_new">http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/30430.pdf</A>

You forget that the current draw and heat production is dependant on the resistance of the chip. (I=V/R and P=I^2*R or (P=V^2/R)). Your equations basically mean nothing. The thermal characteristics of a chip have more to do with how efficiently current is routed through the chip than a simple max amps * voltage.

Dichromatic for your viewing plesure...

trooper11 · Jul 25, 2004

... ok um the 35w mobile ahtlon 64's are low voltage, they dont exceed 35W, thats why they are rated that way. @ 1.8ghz, they dissapte 35W. i dont knwo why you think this would be some kind of lie on amd's part. the rating is at its full 1.8ghz speed. i can show you a prodcut link if you need that to believe me..

Xeon · Jul 25, 2004

You forget that the current draw and heat production is dependant on the resistance of the chip. (I=V/R and P=I^2*R or (P=V^2/R)). Your equations basically mean nothing. The thermal characteristics of a chip have more to do with how efficiently current is routed through the chip than a simple max amps * voltage.

For the limits of the discussion I think the formula works fine. Adding ambient temp, leakage, pipeline depth, transistor count, heat conduciveness of the silicon, the silicon recipe, electrical noise, DC energy loss just isn’t needed when the numbers are within 15% of their real numbers.

But good point but like I said for the current discussion it works.

i dont knwo why you think this would be some kind of lie on amd's part.

They aren’t its clearly labeled move down the chart and you’ll notice additional clock speeds with corresponding draw, and voltages. I also don’t know what’s so hard to understand with this the Athlon XP's put out their fair share of heat on 0.13u and then all of a sudden they have a chip that does near 2x the work per clock and are just as equally clocked yet they put out 80% less heat? Come on common sense guy’s new socket, new heat sinks, new power saving technologies, a better silicon recipe and some damned fine engineering IMO.

Xeon

Post created with being a dickhead in mind.
For all emotional and slanderous statements contact THG for all law suits.

Schmide · Jul 25, 2004

Your equations are misleading as you’re applying power usage as if it’s power conversion to heat. By the conservation of energy you can’t put out more energy than you put in. Heat output will always be less than the consumption of energy. Even the most efficient home space heaters are about 85% efficient. The true travesty of justice is the seems to be that you took the 1600mhz numbers and labeled them 800mhz numbers.

For A64 3200+

2000mhz 1.4v * 42.7a = 59.78w
1800mhz 1.3v * 33.7a = 43.81w
1600mhz 1.2v * 26.5a = 31.8w
800mhz 0.95v * 11.4a = 10.83w

And those are for total energy. Thermal output should be at least 15% less.

Dichromatic for your viewing plesure...

trooper11 · Jul 25, 2004

ok are the 1.8ghz ahtlon 64 parts 35W or not? i say they are, look for instance at this link:

<A HREF="http://www.amd.com/us-en/Processors/ProductInformation/0,,30_118_10220_10221^11030,00.html" target="_new">http://www.amd.com/us-en/Processors/ProductInformation/0,,30_118_10220_10221^11030,00.html</A>
on amd's own page they list the thermal output at 35W... case closed...

Dothan Review: Powerful and low-power

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Distinguished

Share this page