AMD TDP definition

[Idontcare]

It seems that AMD disagrees with some here. TDP = Power consumption:


http://support.amd.com/us/Processor_TechDocs/43374.pdf

TDP. Thermal Design Power. The thermal design power is the maximum power a processor can draw for a thermally significant period while running commercially useful software. The constraining conditions for TDP are specified in the notes in the thermal and power tables.

TDP is measured under the conditions of all cores operating at CPU COF, Tcase Max, and VDD at the voltage requested by the processor. TDP includes all power dissipated on-die from VDD, VDDNB, VDDIO, VLDT, VTT and VDDA.

The key here is the reference to "under the conditions of all cores operating at...Tcase Max".

I've tried to point this out before. There is a reason AMD coincidentally stopped publishing their max allowed operating temperature for piledriver processors at the same time that people (professionals, not just myself, like the engineers at MSI) started noticing the processors do use more than 125W (or get themselves throttled) when running commercially useful software like Prime95 (finding Mersenne primes) and encoding.

The trick AMD pulled is they spec'ed a TDP value but intentionally refuse to spec the accompanying max allowed operating temperature. That makes the TDP value itself meaningless and ill-defined.

They may as well say "125 Unicorn TDP" at this point in time. Some clever people at AMD know this too, that is why the decision was made to withhold the key operating parameters that must be published alongside the TDP spec value.

 

[USER8000]

MSI seems to have a history of motherboard problems though. Even on another forum,they listed VRM failure incidents:

http://www.overclock.net/a/database-of-motherboard-vrm-failure-incidents

Two people I know had their high end P55 motherboards die,with Core i5 CPUs in them and both were at stock clockspeeds.Both went due to VRM problems,which in one case took out the CPU. As a result I would not touch them with a bargepole. Asus and Gigabyte probably outsell them and still have far less problems,since they actually bother to do better internal testing and make sure their motherboards actually work with the CPUs they list on their websites.

 

[Cerb]

I don't think we'll get rid of the TDP as we know it. No matter how useful SDP could be, how coarse grained your thermal/load management is and and how far from TDP we stray in our daily using, engineers still need a ceiling to target when designing cases/coolers, and this ceiling is the TDP.

For cooling limits, yes. I don't mean getting rid of the need to cool it, but allowing a performance policy to be set so that instead of guessing on battery life, and typical rough throttling (resulting in jerky performance), and the notebook warming up and blasting the fan, etc., it could keep the nominal performance at levels that simply wouldn't get very hot in the first place, and could preemptively, and gracefully, change power states and manipulable speeds so that it averages some reasonable lower power (most useful in a battery-limited situation). It might not be the easiest thing to do, but for future notebooks and tablets, it would be useful, and with our CPUs being SoCs or very near it, high performance still needing fairly high current, and most peripherals having fairly predictable power consumption, I don't think it should have to be a pipe dream.

I think the original topic went that way ^_^ ->

 

[mrmt]

Two people I know had their high end P55 motherboards die,with Core i5 CPUs in them and both were at stock clockspeeds.Both went due to VRM problems,which in one case took out the CPU. As a result I would not touch them with a bargepole. Asus and Gigabyte probably outsell them and still have far less problems,since they actually bother to do better internal testing and make sure their motherboards actually work with the CPUs they list on their websites.

Did you notice that almost every failed VRM accident on that DB was with AMD processors?

 

[Cerb]

Did you notice that almost every failed VRM accident on that DB was with AMD processors?

But, Intel has a showing, including mostly MSIs.

An awful lot of the AMDs are listed as running at stock speeds, though...

 

[USER8000]

But, Intel has a showing, including mostly MSIs.

An awful lot of the AMDs are listed as running at stock speeds, though...

Yet very few of all the other companies though,like Asus,Gigabyte,Biostar,Foxconn,ECS,Shuttle,Zotac or even AsRock.

Many of those companies outsell MSI last time I checked. They have a tendency just to re-use old designs,tart them up and give them a new model number.

They are the XFX of the motherboard world and probably are worse. They are cheap for a reason.

 

[bononos]

Agreed. It seems OTT of a mod to be editing someones answer just because they PMed some concerns about a thread. If it is not a concern,just ignore it.

Does that mean,everytime someone complains or reports a post or thread,the mods are going to edit the complainers/reporters post,saying they complained??

Edit!!
Wasn't there an older thread about this anyway??

Its a good thing we saw who complained because this is an interesting thread which digs into a real issue and should not be closed because some people are embarrassed to discuss the failings of some company they admire.

There was an older thread but its deep inside some other thread and previous discussions were hard to follow because it was interspersed with other material.

 

[Arachnotronic]

What is the purpose of the thread exactly? We have topics discussing this and similar specs of AMD CPUs. Unless you had in mind instigating a flame war...

Oh and you forgot to underline and bold the most important parts above:
"while running commercially useful software."
"TDP includes all power dissipated on-die from VDD, VDDNB, VDDIO, VLDT, VTT and VDDA. "

Since you reported this thread; we're of the opinion that it's useful, disagree that it's flamebait (though there's an obvious need to keep an eye on it) and will be keeping this thread open
-ViRGE


EDIT by me:

Wow you are so not biased and helpful. Thank you so much.
-inf64

inf64,

Let it go, dude. We get it. As far as you're concerned AMD is great & awesome, and they will rise from the ashes to beat Intel...but the rest of us are trying to determine the truth about AMD's products.

 

[bononos]

I guess the goal of this topic is to warn people from making mistakes, when the next time they want to write something like "Intel TDP = Average Power, AMD TDP = Max. Power." Unfortunately, this wrong idea was inspired by some (former) AMD employees:
http://www.semiaccurate.com/forums/showpost.php?p=7949&postcount=86
....
In fact AMD definition almost identical to the Intel's definition. Personally I didn't know that.

The OP's link to the 10h family spec also uses max power which is defined as "The maximum sustained power dissipated by the processor at nominal voltage and maximum specified case or die temperature."
And it looks like the max power in the spec is always a little higher than the listed tdp for a processor model in the phenom/phenomII family, so it looks like the tdp is sort of an average like you said.

 

[jvroig]

Hi, just dropping a mod note to remind people that should you have moderation-related issues, please do not derail threads by talking about them in the thread itself - it should be done in the Moderator Discussions sub-forum.

1.) We want to keep threads on-topic. When you have an issue with the moderation, the next ten, hundred, or even thousand (for really popular threads) aren't interested in your moderation issues - they came here for the topic.
2.) We also want to avoid grand-standing. You may be well-meaning individuals, but unfortunately, not everyone is. If we allowed this behavior of challenging mod actions directly in threads instead of Moderator Discussions, there would be quite a handful of trolls who would enjoy nothing more than to grand-stand in public and try to recruit a few impressionable individuals into an anti-mod "screw the police" rally. It is a floodgate we cannot afford to open, as it is easily exploited. I hope you understand.


I'm only locking the thread TEMPORARILY until I get another message in, then I will unlock it and you can continue, on-topic, minus all the derails and callouts. Thank you.

Moderator jvroig

 

[jvroig]

I would like to clarify the issue of "Just find it wrong to make PMs a public thing" and similar feelings.

TL;DR: ViRGE did not make any PMs public.

What actually happened is inf64 pushed the "report post" button to send a report into the moderation queue. In the report reason box, he specified that he thinks this thread is flame bait.

ViRGE happened to be the first mod to arrive at this specific report, took a look at it, and decided it's not quite flame bait but recognizes that it could become a nightmare thread. Instead of just shrugging it off and leave inf64 hanging, he replied to inf64 by putting an answer directly into his post (inf64 was issued no infractions or warnings by ViRGE; it was purely a reply).

So please, instead of jumping to unfounded conclusions, please consider getting the facts. It is merely the normal demands of etiquette and decorum that we do so.

And remember, when you do have issues with moderation, please go to Mod Discussions - it doesn't matter if you are wrong, or if you are perfectly justified in your complaint, it still should go to Mod Discs.

Some of you may now find yourself getting a warning or infraction for your participation in mod callouts or, at the very least, thread derailment; it is regrettable, but you guys should not have had to be reminded in the first place. You know the rules. And simply because a moderator - ViRGE or otherwise - disagreed with your own judgment doesn't give you the right to override the guidelines. Please keep this in mind.

Moderator jvroig

 

[CHADBOGA]

Wouldn't it just be easier for everyone concerned if you guys handed over the running of the forum to inf64? D:


You just added a thread derailment and a member callout right after a mod post.

Suffice it to say, this is not allowed and I hope you do not repeat this in the future.

Moderator jvroig

 

[SlowSpyder]

MSI seems to have a history of motherboard problems though. Even on another forum,they listed VRM failure incidents:

http://www.overclock.net/a/database-of-motherboard-vrm-failure-incidents

Two people I know had their high end P55 motherboards die,with Core i5 CPUs in them and both were at stock clockspeeds.Both went due to VRM problems,which in one case took out the CPU. As a result I would not touch them with a bargepole. Asus and Gigabyte probably outsell them and still have far less problems,since they actually bother to do better internal testing and make sure their motherboards actually work with the CPUs they list on their websites.

On the other side of the spectrum, I've had my MSI board for over five years now (purchased March 2008), always running power hungry and overclocked/overvolted Phenom processors. My first was a PhI 9850 @ 2.7/2.8GHz and voltage. Then a PhII 940BE @ 3.6GHz/~1.46v. Finally my Thuban 1090T @ 4.03GHz and 1.48v.

I am not saying is MSI motherboards are good or bad. But when your sample size is only a board or two, its hard to draw any real conclusions.

*edit - I pretty much always end up editing my posts... haha. Anyway, I should mention this disclaimer, that I am unable to see anything in your link. Depending on what information is in that link, maybe there is a clear trend with a large number of users having issues, like I said, I don't know as I can't see it. I'm at work with my crappy work computer, I'll check the link when I get home.

 

[Magic Carpet]

So do AMD still use ACP in their advertising?

AMD has said that the ACP and TDP values of the processors will co-exist, and do not replace one another. All server products will see two power figures starting from the codenamed Barcelona server processor onwards. More.

Personally, I couldn't care less. I always do my own testing. TDP is just a guideline, nothing more. Same, when you speed on a highway

 

[mrmt]

Just for the record, AMD dropped ACP from their sites. They only specify "Wattage", which is the same as the TDP in the processors I compared.

 

[Exophase]

I don't think there's actually that much wiggle room in "thermally significant" or "commercially useful software." I don't think anyone can argue that sustaining an average power consumption over multiple minutes constitutes as thermally significant. I don't think anyone can argue that using software that could survive commercially for years constitutes as commercially useful. These are the conditions I've seen people use when they say that their FX-8350s are measured to be using a lot more than 125W.

If the short power spikes don't impact battery life in any meaningful way and they don't affect the cooling requirements either then the only thing that they impact is the power regulation/power circuitry and as users that's not that interesting to us. So while people may need to be reminded now and again that we're supposed to be talking about average power consumption over a fairly long period of time while performing something intensive this isn't of much more than academic interest.

using amds metrics would that mean that...bit of a jump here...arm soc tdps are closer to actual power draw?

What are these ARM SoC TDPs you refer to? I've seen maximum power consumption in some SoC datasheets. I don't think I've yet to come across TDP.

 

[Exophase]

Most power a system can draw:
Intel = Max power
AMD = TDP

Typical power draw for standard applications:
Intel = TDP
AMD = ACP

In fact AMD definition almost identical to the Intel's definition. Personally I didn't know that.

I remember in the S|A thread where JF-AMD was posting that stuff Ungo (a clearly quite experienced EE) showed how much error there was in everything he was saying.

Frankly JF-AMD struck me as being wrong a lot. Between this, his claims that Bulldozer can do 2x AGU + 2x ALU where K10 could only do 3x AGU or ALU (contradicting the long standing knowledge of 9 ops dispatched per cycle) and the infamous IPC blunder I'm surprised that AMD sat by and let him speak for their engineers on forums.

 

[TuxDave]

Maybe it's the post-lunch coma but what are we trying to argue? TDP specs serve two purposes. Personally I think it tells the OEMs what type of cooling they need to provide as well as the power delivery it needs to package.

Or are we debating what "maximum power on a commercial software" is supposed to mean?

 

[mrmt]

Frankly JF-AMD struck me as being wrong a lot. Between this, his claims that Bulldozer can do 2x AGU + 2x ALU where K10 could only do 3x AGU or ALU (contradicting the long standing knowledge of 9 ops dispatched per cycle) and the infamous IPC blunder I'm surprised that AMD sat by and let him speak for their engineers on forums.

If you dig deeper you'll trace a lot of rumors to him. Do you remember the "unconnected MCM Clovertrail?", yes, that was him too.

What caught my attention is that AMD so far hasn't release the thermal datasheets even for their opteron line up, something they do since the first K8 opteron chips.

 

[galego]

And the trick is to not release any specs, so you can run the CPU out of spec like the FX series.

It is rather hard to go out of spec without releasing any spec. Moreover, the FX series TDP is correct. MSI problems are just MSI problems.

 

[galego]

It seems that AMD disagrees with some here. TDP = Power consumption:


http://support.amd.com/us/Processor_TechDocs/43374.pdf

TDP. Thermal Design Power. The thermal design power is the maximum power a processor can draw for a thermally significant period while running commercially useful software.

Now compare with Intel definition:

Thermal Design Power (TDP) represents the near maximum power a product can draw for a thermally significant period while running commercially available software.

TDP = Power consumption? No

 

[GreenChile]

It is rather hard to go out of spec without releasing any spec.

Actually every device has a spec that it operates in whether it is published or not.

Moreover, the FX series TDP is correct. MSI problems are just MSI problems.

Have you got anything to back up that statement? It would not be unexpected for AMD to sell a few marginal chips to try and squeeze every last bit of profit from each wafer.

 

[MisterMac]

Let me get this straight.

Naming a moderator in any kind of way is a mod call out?

Even if - the context clearly suggest i'm giving my oppinion to someone(who happens to be a mod?)
That's kinda fucked imho - but ill take whatever is coming then as per rules.

No cussing in the technical forums.
Markfw900
Anandtech moderator


Ontopic:

I think everyone knows AMD started being shady because BD was and is - a powerhungry uARCH that was ...well half finished when released.

It's not really a secret is it?

@Inf64:

It's just that it's indirect flamebaiting on the fact that he's trying to sensationalize something most of us know.

I'm more in the Intel Camp than AMD myself - but i don't go around trying to indirectly downplay AMD on stuff most of us know.

 

[R0H1T]

What a surprise ~ AMD lies to us about TDP D:
AT advertising SDP for Intel, sorta like TDP, yeah not much to see here

 

[galego]

The key here is the reference to "under the conditions of all cores operating at...Tcase Max".

I've tried to point this out before. There is a reason AMD coincidentally stopped publishing their max allowed operating temperature for piledriver processors at the same time that people (professionals, not just myself, like the engineers at MSI) started noticing the processors do use more than 125W (or get themselves throttled) when running commercially useful software like Prime95 (finding Mersenne primes) and encoding.

The trick AMD pulled is they spec'ed a TDP value but intentionally refuse to spec the accompanying max allowed operating temperature. That makes the TDP value itself meaningless and ill-defined.

They may as well say "125 Unicorn TDP" at this point in time. Some clever people at AMD know this too, that is why the decision was made to withhold the key operating parameters that must be published alongside the TDP spec value.

I think all of this was debunked before.

Therefore I will move to Intel and how they work:

http://www.theinquirer.net/inquirer...s-chip-vendors-tdp-figures-are-not-comparable

At this time we are not prepared to talk specifically [about] what workloads we are using. It is something we haven't disclosed publicly to date and so that would be a pretty big decision for us to do that and we are not prepared to do that today, mainly for competitive reasons. One of the key things about TDP is that there is not an industry standard definition for it. We're not actually quite sure what our competition uses. We know a lot of the phone silicon guys don't appear to specify it at all. So it tends to be a little bit of a trade secret from company to company.

You should note that the Atom processors use a different definition of TDP [than Core processors], one that is more oriented towards mainstream or consumption oriented workloads.

And not to forget the weird case of 7W ivy chips

http://www.fudzilla.com/home/item/3...s-actually-13w-tdp-cpu?tmpl=component&print=1

Actually every device has a spec that it operates in whether it is published or not.

Have you got anything to back up that statement?

If the spec is not released you cannot say if the device is out of it or not. And it is the guy who claim that the FX chips are out of spec who has to back his statements.