90nm done right, winchester 20W cooler than newcastle at same clockspeed!

[saaya]

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

very nice article, sounds like the 90nm a64s are very promising :D
cant wait to see until the 90nm chips from monarch and newegg some people here ordered arrive and they post results :D

[Lithan]

This is nice. I can't wait until we see some 20-25w mobile A64 .09 procs. Maybe not passive coolers, but definately cheap and silent coolers.

Incidentally, what's the power consumption of p4-m's? What about Dothans? Would a .09 mobile A64 possibly be able to compete with those in terms of battery life?

[saaya]

yes they might :D

a dothan 2ghz has 22W, the 533fsb dothan will be around 25W or higher though as far as i know.

it sems amd is only making 512kb L2 chips so far targeting the mainstream and mobile market, wich makes sence as thats where they can save most money in production costs. i think the next move will be 90nm chips with just 256kb L2 cache and only then we will see 1mb 90nm parts.

[saaya]

the rating for a64s is 89W max, right...

an fx55 consumes 104W max, a fx53 still fits in the 89W max power rating, so we can assume the pump in clockspeed of 200mhz means around 15W.

those numbers from amd are for 64bit full load afaik, and a64s run around 10W cooler in 32bit under full load compared to 64bit under full load afaik.

so we get this:

fx55 ~104W(64bit) ~94W(32bit)
fx53 ~89W(64bit) ~79W(32bit)
fx51 ~74W(64bit) ~64W(32bit)

single channel memory controller and only half the cache will make almost no difference in heat dissipation, maybe 2W at most.

so a newcastle 2.2ghz chip produces as much heat as an fx51 (2.2ghz) and the winchester produces 20W less heat than the newcastle at the same clockspeed, then it means it produces only 54W in 64bit and 44W in 32bit :eek:

this sounds incredibly low, but i think its right... this one site that tested winchesters say they run as cool as 1700+ tbred Bs wich produce around 45-50W...

and it also makes sence as the mobile a64s based on 90nm are supposed to produce no more than 25W. the reviews all state that the winchesters would run stock speeds with as low as 1.1-1.2v, that plus a lower clockspeed than 2.2ghz, around 2ghz, should be enough to go down to 25W.

damn i cant wait for the winchesters to arrive in retail here so i can finally upgrade :D

[=[PULSAR]=]

I remember seeing a program out there not to long ago that displayed your cpu's thermal output and when you overclocked it, it would calculate it based on speed, voltage and the type of processor. Anyone remember the name of it and where I can find it?

[saaya]

ocrechner does it :D (rechner=calculator/pc in german)
as you said, its just playing arund with numbers and not THAT accurate... but its a good reference and not very far off.

http://members.lycos.co.uk/wiezaq/download/

[slavik]

CPUs don't produce heat, they waste energy/power and give it off as heat ...

also, the watt rating is power consumption ...

[saaya]

slavik, cpus and all ics dissipate 99% of the energy they conume as heat so its the same thing.

and whats the difference between saying a cpu "produces heat" and "gives heat off"?
i dont see a difference.

[musk]

I downloaded a prog called the overclockulator.

http://www.pimprig.com/downloads/index.php?action=downloadfile&filename=oc1_2_2_beta_setup.exe&directory=Apps and Utilities&PHPSESSID=11891e8a029317679b52ef2aa93e69a1

Its from pimprig site.

musk

[slavik]

purely preferential :)

gives off heat = implies waste (to me)

generates heat = implies wanted (to me)

also, it is wrong IMO to use the measure of power for measuring heat outsput ... we should use Cal/sec instead of watt ...

power consumed = watt
heat generated = Cal/sec

also, we do not know IC efficiency, unless someone can measure it correctly ...


slavik, cpus and all ics dissipate 99% of the energy they conume as heat so its the same thing.

and whats the difference between saying a cpu "produces heat" and "gives heat off"?
i dont see a difference.

[saaya]

purely preferential :)

gives off heat = implies waste (to me)

generates heat = implies wanted (to me)

also, it is wrong IMO to use the measure of power for measuring heat outsput ... we should use Cal/sec instead of watt ...

power consumed = watt
heat generated = Cal/sec



dude, watt IS cal/s :P :lol:

1W=1joules/s

1 joule is ~0.25 colories so 4W=~1cal/s ...




also, we do not know IC efficiency, unless someone can measure it correctly ...

100W power consumption is 99W heat dissipation, believe me...

or where do you think does the energy go? to the other parts of the system? they produce the same heat whether they are connected to the cpu or not, so that cant be. where else should all the power go? light? geek radiation that mutates humans? :P

[Lithan]

We use watts for heat as well because our primary use for the data is to perform calculations related to cooling that heat. And watts work best for that.

[blinky]

a joule is the amount of energy required to taise the temperature of one CC of water 1 degree celsius

right?

[saaya]

no, a calorie is :)

[blinky]

no, a calorie is :)
o, damit

[Lithan]

A joule is a W*s and a calorie is equal to a little over 4 of them. 4.185 maybe?

[saaya]

1 calorie = 4.18400 Joule

:D
http://www.google.de/search?hl=de&ie=UTF-8&q=calorie+to+joule&meta=

i love google :p:

[slavik]

hmm, sounds about right ...

still, a watt = joule/second ... not Cal*4.184/second ...

invent a unit which means that and I'll be happy :)

[saaya]

whats the deal? its almost 1=0.25, so its very easy to calculate numbers, dont even need a calculator :)

this inch gallon pound crap is 10X worse! :D

[Torin]

Talk about OT....

[HKPolice]

Don't be so happy yet....

http://www.sudhian.com/showdocs.cfm?aid=610&pid=2297

[slavik]

that's why US should move to metric system :)


whats the deal? its almost 1=0.25, so its very easy to calculate numbers, dont even need a calculator :)

this inch gallon pound crap is 10X worse! :D

[Hallowed]

Don't be so happy yet....

http://www.sudhian.com/showdocs.cfm?aid=610&pid=2297

Been talking to joel about this.

To hazard a guess I'd either say earlier review samples run warmer, or simply some boards BIOS's incorrectly report temps.

[***Deimos***]

Hallowed:
Or perhaps you just cant assume the better of two view is correct, and AMD has same problems with 90nm that everybody else in the industry had.. they're not infalible. Hopefully you can run stock Ghz at 1.3V. Maybe even 1.2V.

Aside: I'm sure when 'Duel-Core' CPUs come out, some modders will create 'Dual'-water cooling system with a block on top, and one below (up against the PCB on the back), to 'combat' the heat problem.

[Hallowed]

Its not an assumption when a majority of the reviewers and owners confirm their chips run cooler.

I'm trying to discover WHY there is a variance, since there is no odd man out.

Svethardware.cz
Techreport.com
Madshrimps.be
And a few at Overclockers claim their chips run cooler, OC about the same.

Sudhian
AMDZone
GZeasy all show results of higher temps. I'm tempted to discount AMDZone for a previous record of poor reviews, but Sudhian & Joel Hruska know what theyre doing.

[Lithan]

Hallowed:
Or perhaps you just cant assume the better of two view is correct, and AMD has same problems with 90nm that everybody else in the industry had.. they're not infalible. Hopefully you can run stock Ghz at 1.3V. Maybe even 1.2V.

Aside: I'm sure when 'Duel-Core' CPUs come out, some modders will create 'Dual'-water cooling system with a block on top, and one below (up against the PCB on the back), to 'combat' the heat problem.


Seems unlikely since there's nothing on the back of the motherboard too cool really... except a lot of stuff you don't want to be pressing copper against.

[drunkenmaster]

again, dual cores aren't the same as dual cpu's, its not twice the power at all. Its spreading out the logic, adding more cache but lowering speed of cores individually. But as it can effectively work as two logical cpu's there are many advantages. AMD's stuff shows their dual core cpu's on 0.09 will be 89W max, you just can't dissapate more heat than that on stock cooling, not easily, and these cpu's will be in average desktop systems too.

[saaya]

Hallowed:
Or perhaps you just cant assume the better of two view is correct, and AMD has same problems with 90nm that everybody else in the industry had.. they're not infalible. Hopefully you can run stock Ghz at 1.3V. Maybe even 1.2V.

Aside: I'm sure when 'Duel-Core' CPUs come out, some modders will create 'Dual'-water cooling system with a block on top, and one below (up against the PCB on the back), to 'combat' the heat problem.

LOL how? how do you want to cool a 939 cpu from below? :D

and about the temps, id like to know whether they tested with the same vcore for all speeds and cool n quiet was REALLY disabled.

and the temperatures they used were the on-die temp sensor results, right? in that case we can still hope that its meassuring differently than the 130nm on-die temp probe or the bios is reading it out differently....

nobody really mentions how they meassured the temps from what i can see... :/

[saaya]

again, dual cores aren't the same as dual cpu's, its not twice the power at all. Its spreading out the logic, adding more cache but lowering speed of cores individually. But as it can effectively work as two logical cpu's there are many advantages. AMD's stuff shows their dual core cpu's on 0.09 will be 89W max, you just can't dissapate more heat than that on stock cooling, not easily, and these cpu's will be in average desktop systems too.

intel can, but look at the elephant heatsinks it takes :D

[saaya]

ok, that site i linked to meassured the systems power conumption, so tbh i rather believe them than the people who meassured the temperature of the heatsink and dont even say how and where they meassured it... i cant wait until anadtech or another mayor site releases their 90nm article :D

[reject]

i wouldnt trust anandtech with their "magic samples" motherboards and ram, next thing it will be cpus

[saaya]

hmmm yeah, imo all review sites should only review retail boards they get from shops...

[texuspete00]

i wouldnt trust anandtech with their "magic samples" motherboards and ram, next thing it will be cpus

I notice this too and bring it up when someone says X ram is coming that can do XXX (then provides anand link) and no one ever commented. Good to know I'm not alone here. There ram results are just straight from another planet. It comes up often because so often anandtech has the next holy grail of memory and people get all excited and post about it. Kingmax doing 2-2-2!?! It's gotten so bad with their ram I have to doubt them now. LL has never been a Kingmax strength and every other review of there 466 or 500 is just as I would expect. First I went with the "golden sample" theory but now I don't even want to look at their ram reviews ever. There consistent ability to get the greatest results started my questioning, but it only takes a couple Kingmax type instances on top of this fact, where chips seem to have an identity crisis and be all things to all people even when some of the strengths don't really play into the chips abilities... time to call shens. Like in this instance I'm pretty sure you could buy piles of these chips and not get that. Makes no sense. Be like BH5 doing 250 at 2.6v... or Hynix doing 2-2-2.

[saaya]

hynix does 2-2-2 ;)

at 170fsb... :D

yeah i get your point. imo it should go like this:

abit wants their latest board reviewed so they ask some hardware sites, or hardware sites ask them about a review.
guy from the site orders the board or buys it in a local shop, sends abit the bill and they transfair the money back to his account.

THATS how hardware should be reviewed! all those reviews of boards before they are released are nice, but those are PREVIEWS not reviews, its not the same silicon that other people can buy.

some manufacturer picsk the best of its 50.000.000 memory chips and puts them on one stick, sends it to a hardware site and they get amazing scores with it...

nvidia has already done this with their 6800 ultra extreme!!!! and its looking more and more like its a general trend with every new hardware release :mad:

[Petr]

Latest October 2004 OPN from AMD shows 90nm 3000+, 3200+ and 3500+ being 67W TDP chips, much lower than original 89W rating.

http://www.svethardware.cz/sh/media.nsf/v/49C34E0738998A2EC1256F24005DA2B1/$file/OPN_large.gif

[saaya]

cool! then my guess of 74W was not very far off indeed :D

[MUCHO]

abit wants their latest board reviewed so they ask some hardware sites, or hardware sites ask them about a review.
guy from the site orders the board or buys it in a local shop, sends abit the bill and they transfair the money back to his account.

THATS how hardware should be reviewed! all those reviews of boards before they are released are nice, but those are PREVIEWS not reviews, its not the same silicon that other people can buy.

some manufacturer picsk the best of its 50.000.000 memory chips and puts them on one stick, sends it to a hardware site and they get amazing scores with it...

nvidia has already done this with their 6800 ultra extreme!!!! and its looking more and more like its a general trend with every new hardware release :mad:

Agree 100% - Thats one thing I like about Anandtech's recent reviews. They mention several retail results as well.

Still - nothing like this message board.

:toast:

[blinky]

Latest October 2004 OPN from AMD shows 90nm 3000+, 3200+ and 3500+ being 67W TDP chips, much lower than original 89W rating.

http://www.svethardware.cz/sh/media.nsf/v/49C34E0738998A2EC1256F24005DA2B1/$file/OPN_large.gifso is winchester considered D0 stepping?

[CTKP]

yes

[spaceman]

Wait a minute, those 90nm chips are all 939s. :mad: ;) I hope they'll be giving us 754 chips in 90nm soon, too. :confused:

[halcyon]

No they won't.

90nm in socket 754 is ONLY for Sempron's and true mobile socket 754 Athlon 64 (35W). The latter is expected during Q2-Q3/2005 at max speed of 2.2Ghz.

[saaya]

there already are 754 90nm chips, actually the first 90nm chips were sold as mobile 754 chips :D

but they were all sold directly to the big notebook makers afaik, so its almost impossible to get one :/