With the current rate of Intel CPU performance increases, could AMD be catching up?

[Homeles]

If i was wrong and they didnt need to redesign the HSF then IvyBridge should have lower operating temperatures than SB.

That's not what you said though.

 

[AtenRa]

That's not what you said though.

Also IvyBridge is not soldered to the Heat spreader like FX8350 and it effects the HSF design and size.

I was talking about the performance of the HSF not the physical design for the socket. I was simple stating that the HSF needed to be redesigned in order to keep the die at lower temps.

For example take the Phenom II TDP table above. The same HSF fitted to a 95W TDP CPU will lower the Tcase max(CPU temperature) to 56,9c. At 75W TDP the Tcase Max will be at 55C.

If Ivy didnt need a new HSF then it would operate at a lower temperature than SB does because the TDP is at 77W when SB at 95W. But due to 22nm, smaller die size, no IHS soldered etc using the same HSF raises the Tcase max.

That means that all the above Affects the HSF design/size/materials

 

[Fjodor2001]

This is not true at all. Actually look at the graph's points. At 125W, the i7-3770K is at 4.6GHz, which is a 31% increase.

You're missing including headroom for Turbo mode. See this post.

 

[Lonbjerg]

Remember...different laws of physics applies to AMD than Intel according to some posters...*chough*

 

[ShintaiDK]

Just no,

Let me explain what TDP is for,

The following is from the AMD Phenom II TDP specs.



So, for the CPU to work as it should the HSF should meet the above specs.

That means for a 125W TDP CPU, the HSF at 38c Ambient should be able to keep the CPU temperature at or bellow 61c.

We all know this is not true for the FX. Its a proven fact and mobo makers says it as well.

You however keep denying it, even tho the edvidence is clear and against you.

The CPUs are not 125W, but ABOVE 125W. Rather 140W area.

AMD is simply scamming people and putting peoples motherboards at risk.

 

[Phynaz]

The FX8350 die size is 315mm2, Ivy is just 160mm2. Not only that but FX module topography(placement) is completely different than Ivy.
Same goes for the 130W TDP 2011 dies, they are 400mm2+ and the core placement is completely different than Ivy 1155 die.

BD/PD power maps are completely different than Ivy and Sandy.

So you have no proof that what you are talking about affects cooling design. You could have just said that, instead of stating it as fact.

Good to know that all those generic one size fits all aftermarket coolers didn't have it wrong by not having a special Intel design.

 

[Ajay]

Uh, getting a bit OT?

 

[2is]

AMD is simply scamming people and putting peoples motherboards at risk.

Pretty much, and I believe this is done so they can at least be competitive on price, particularly on the low end retail machines which is what most people buy. If mobo manufacturers had to beef up their boards to properly support the CPU, costs go up, price goes up, sales go down.

 

[AtenRa]

http://www.intel.com/content/dam/ww...rd-gen-core-family-mobile-vol-1-datasheet.pdf

Page 74.

5.5 Thermal and Power Specifications
The following notes apply to the tables in this section.
Note Definition

1.The TDPs given are not the maximum power the processor can generate. Analysis indicates that
real applications are unlikely to cause the processor to consume the theoretical maximum power
dissipation for sustained periods of time.

2.TDP workload may consist of a combination of a processor-core intensive and a graphics-core
intensive applications.

3.The thermal solution needs to ensure that the processor temperature does not exceed the maximum junction temperature (Tj,max
) limit, as measured by the DTS and the critical temperature bit.

http://www.intel.com/content/dam/ww...-guides/3rd-gen-core-lga1155-socket-guide.pdf

Page 42

5. Thermal Design Power (TDP) should be used for processor thermal solution design targets. TDP is not the maximum
power that the processor can dissipate. TDP is measured at DTS = -1.
TDP is achieved with the Memory configured for DDR3 1333 and 2 DIMMs per channel.

Page 43

Can anyone now stop taking the TDP as the maximum power consumption of AMD and Intel CPUs ???

 

[Exophase]

If someone shows some program at full load not exceeding stated TDP on an IB and grossly exceeding TDP on a Vishera what will your response be? Because I'm pretty sure that's not hard to demonstrate..

 

[AtenRa]

If someone shows some program at full load not exceeding stated TDP on an IB and grossly exceeding TDP on a Vishera what will your response be? Because I'm pretty sure that's not hard to demonstrate..

When you say exceeding the TDP it means that the CPU temperature will exceed Tcase Max not 77W/125W of consumption.

Now does everyone understands what TDP is ???

 

[Exophase]

Right, so when the companies say the TDP is 77W and 125W respectively that doesn't mean the TDP has anything to do with power consumption, is that what we're supposed to think?

If you think thermal design power has to do with temperature and not heat (equivalent to power) then you have the wrong idea. Intel and AMD both flat out listing TDP as a power figure on their processor listings and they're not being inconsistent.

Yes, there are temperature limits where the TDP only applies if you meet them, because power consumption is temperature dependent. So there are ambient and cooling requirements. But if, while running that test program, IB doesn't exceed 77W while also not exceeding Tjmax or whatever temperature points are specified, and Vishera does exceed 125W while also not exceeding Tcasemax or whatever it specifies then what will your response be?

And if it exceeds 125W while also exceeding the allowable case temperature under an ambient temperature that's acceptable by their datasheet and running a stock cooler then it means AMD did a bad job with the cooling and yes, are screwing motherboard manufacturers, OEMs, and other customers by providing the wrong specs.

 

[AtenRa]

Right, so when the companies say the TDP is 77W and 125W respectively that doesn't mean the TDP has anything to do with power consumption, is that what we're supposed to think?

It means that the power consumption can be more than 77W/125W and the CPU will still have 77W/125W TDP requirements for the Heat-Sink Fan.

If you think thermal design power has to do with temperature and not heat (equivalent to power) then you have the wrong idea. Intel and AMD both flat out listing TDP as a power figure on their processor listings and they're not being inconsistent.

TDP has to do with the ability of the Heat-Sink Fan to dissipate 77W/125W of power and keep the CPUs at or lower than their Tcase Max as illustrated by the tables above.

Yes, there are temperature limits where the TDP only applies if you meet them, because power consumption is temperature dependent. So there are ambient and cooling requirements. But if, while running that test program, IB doesn't exceed 77W while also not exceeding Tjmax or whatever temperature points are specified, and Vishera does exceed 125W while also not exceeding Tcasemax or whatever it specifies then what will your response be?

Two things, that the HSF is indeed doing its job by keeping the CPUs at the nominal operating temperatures and secondly that the CPUs are as specified of being 77W and 125W TDP(if we know that the HSF are made for those TDPs).

And if it exceeds 125W while also exceeding the allowable case temperature under an ambient temperature that's acceptable by their datasheet and running a stock cooler then it means AMD did a bad job with the cooling and yes, are screwing motherboard manufacturers, OEMs, and other customers by providing the wrong specs.

If it would exceed the Tcase Max, it will throttle down meaning it overcome its nominal TDP. Using/consuming more than 125W means that the Tcase Max was not reached thus the CPU was operating with in TDP limits. You have to understand and everyone else that a CPU or any IC can use more "power" than its TDP rating and still be within its nominal temperature. That can happen because 1) not all of the power is dissipated through the HSF, 2) the modern CPUs have Heat-Spreaders to assist the HSF, 3) Core and Logic placement (die topology) plays a significant role in CPU thermals and more.

 

[preslove]

This story is an example of how AMD may never pass Intel long term. AMD has always competed with Intel because the 2 companies set their chip prices to be competitive.

However, with the K7, AMD had the dominant CPU for quite a while. Intel recognized it, and pushed R&D hard, and pretty quickly passed AMD (quick for chip design. I believe it was within 1 generation) and hasn't looked back.

So, yes, I think AMD might be catching up, but Intel's size and resources allow it to get back in front

Intel used collusion to prevent AMD from exploiting their performance dominance, earning them over a billion $ in fines (not enough). If AMD had been able to sell chips in Dell and other big OEMs in the early aughts, they would have had the $ for R&D to keep up with Intel.

 

[Idontcare]

Can anyone now stop taking the TDP as the maximum power consumption of AMD and Intel CPUs ???

I'm with AtenRa on this one. The bottom line is that TDP has always been a poor proxy for estimating power consumption of a CPU, from either manufacturer.

The utility of a TDP rating is limited, and intentionally so. So too is true for TCasemax. It is a spec value which has been defined in such a way as to really only be useful and relevant to engineers who are charged with designing cooling solutions for entire systems.

What we ought to be asking ourselves is "what is relevant?" Rather than getting more and more pedantic about what TDP means or doesn't mean, instead lets just get to the heart of the matter and talk about why we find ourselves compelled to keep talking about TDP.

Because regardless whether or not we all come to an agreement on the purpose and utility of the TDP value, that alone will not negate the reason it continues to crop up in debate after debate in these forums.

IMO power consumption is a big issue for the members of this forum because it negates one major selling point to AMD's FX lineup - price/performance. Once you factor in the cost of paying to operate the FX chips it makes the argument to buy one, albeit at lower price than Intel's, less and less justified.

And I think that is why people are so keen to debate the power consumption topic by proxy through the TDP spec value. Get rid of the TDP spec value argument, it in itself is the red herring here, and get back to discussing the real meat - performance/watt and how that impacts performance/dollar in a TCO-based reality.

 

[Idontcare]

Intel used collusion to prevent AMD from exploiting their performance dominance, earning them over a billion $ in fines (not enough). If AMD had been able to sell chips in Dell and other big OEMs in the early aughts, they would have had the $ for R&D to keep up with Intel.

AMD's market opportunities were supply limited on the AMD side of the equation. Despite what Intel did to the demand side, the reality is that AMD had a limited supply of chips to sell and they sold everything they could get to leave the fab.

The complaints are numerous and many system builders are expressing their frustrations. "It's a fiasco. There's no product in the channel. It's all going to Dell. AMD is divorcing the channel," said Glen Coffield, president of CheapGuys. Coffield went on to say "I'm being treated like the red-headed stepchild. Intel was always a company for the big guys, and AMD was always a company for the small guys like us. The channel made AMD. Now they're at the dance with Dell, and they've thrown us to the wolves."

Another system builder that chose to remain anonymous remarked that "There's nothing out there. They're screwing the channel. All the stuff is going to HP and Dell, and the only inventory available to us is old legacy stuff, single-core and low-power stuff no one is buying. There's little doubt in my mind that it's because of Dell and HP."

http://www.dailytech.com/article.aspx?newsid=4782

Rebates did not prevent AMD from building more fabs. Spending $5.4B on ATI did.

 

[Exophase]

It means that the power consumption can be more than 77W/125W and the CPU will still have 77W/125W TDP requirements for the Heat-Sink Fan.

TDP has to do with the ability of the Heat-Sink Fan to dissipate 77W/125W of power and keep the CPUs at or lower than their Tcase Max as illustrated by the tables above.

If it consumes more than 77W/125W then it will be dissipating more than 77W/125W of heat. Any type of energy emitted by the CPU in other forms like electromagnetic radiation is small enough to be negligible. I don't know if you're arguing against that point but it sounds like you are.

If it would exceed the Tcase Max, it will throttle down meaning it overcome its nominal TDP. Using/consuming more than 125W means that the Tcase Max was not reached thus the CPU was operating with in TDP limits. You have to understand and everyone else that a CPU or any IC can use more "power" than its TDP rating and still be within its nominal temperature. That can happen because 1) not all of the power is dissipated through the HSF, 2) the modern CPUs have Heat-Spreaders to assist the HSF, 3) Core and Logic placement (die topology) plays a significant role in CPU thermals and more.

Sure, a CPU can use more power than its rated TDP and still be within specified temperature. But that greatly brings into question as to what the TDP figure actually means.

I don't see how heat spreaders and die topology changes anything for the amount of power the HSF sees. Temperature, yes, but TDP isn't a temperature rating, even if it's qualified for some temperature range.

You raise a fair point that not all of the heat is dissipated through the HSF, but I talked to an EE at work and he says that only a very small amount of heat will go through the board or the air. That's because the motherboard PCB isn't designed with good thermal pathways to conduct heat, and the air is mostly blocked off between the CPU. You have TIM going from the CPU die to a heatspreader to the HSF, that is going to be the vastly dominating thermal path.

So I think we can agree that the HSF will see the vast majority of the power that the CPU consumes. So what are you saying that 125W number is supposed to mean? The value you get while under Tcasemax? You say that exceeding Tcasemax for any period of time isn't even possible because the CPU will throttle. But exceeding the 125W power consumed with a reasonable program load is clearly possible, even when Tcasemax is maintained. Do you think this is irrelevant to HSF manufacturers? Or what are you trying to say?

 

[ElFenix]

AMD's market opportunities were supply limited on the AMD side of the equation. Despite what Intel did to the demand side, the reality is that AMD had a limited supply of chips to sell and they sold everything they could get to leave the fab.



Rebates did not prevent AMD from building more fabs. Spending $5.4B on ATI did.

Amd could have sold their processors at a higher price. Supply-demand and all.

 

[Pilum]

Amd could have sold their processors at a higher price. Supply-demand and all.

To a certain degree; it's not as if they could have sold each K8 at $300+. Besides, they in all likelihood needed to bin the CPUs anway, I don't think AMD ever had their manufacturing perfected so that all CPUs would have reached top-bin frequency/TDP limits. Combined with the fact that the later P-IV's weren't that bad, the extra profits AMD could have made probably weren't earth-shaking. The $1.25 billion payment from Intel probably covered that pretty well.

 

[Ajay]

To a certain degree; it's not as if they could have sold each K8 at $300+. Besides, they in all likelihood needed to bin the CPUs anway, I don't think AMD ever had their manufacturing perfected so that all CPUs would have reached top-bin frequency/TDP limits. Combined with the fact that the later P-IV's weren't that bad, the extra profits AMD could have made probably weren't earth-shaking. The $1.25 billion payment from Intel probably covered that pretty well.

Yes, but most thought they could get more. The didn't, it seems, because they really, really wanted choice in the foundry partner. Heck of allot of good that did them (since they are still wedded to GloFo, the WSA and contribute to GF's development costs).

 

[Cookie Monster]

Outside the "TDP" marketing nonsense, isn't TDP one of many specifications for heatsink design engineers to follow? For instance, doesn't TDP of 125W just means that the accompanying heatsink must have the capability to dissipate at minimum 125Ws without it failing?

Atenra also has a point about IB and why its hotter. Normally you don't want anything in between the heatsink and the processor die because the heatspreader/TIM all degrade the temperature transfer (hence why you want a thin layer of TIM as possible) assuming 100% contact between the two. Given the worse TIM between the heat spreader and the die while also being a small die i.e. area for heat transfer, heat transfer is actually worse than SB (according to a quick look at Fourier's law).

Since there are no moving parts to a processor, all the energy consumed must be dissipated as heat i.e. TDP could roughly approximate the general power consumption of a CPU at load.. but its still a vague number for determining the actual power consumption of the processor. For instance, a heatsink design for 150W TDP may well be sufficient for a processor that normally dissipates ~120W with 170W peaks (not sustained).

And from my knowledge, most coolers have a margin (from the TDP rating of the processor) in their capability to dissipate the heat especially aftermarket coolers which are probably designed for +200W TDPs i.e. overkill for stock operation.

 

[Exophase]

Outside the "TDP" marketing nonsense, isn't TDP one of many specifications for heatsink design engineers to follow? For instance, doesn't TDP of 125W just means that the accompanying heatsink must have the capability to dissipate at minimum 125Ws without it failing?

The minimum that the HSF is supposed to tolerate should also be the maximum that the CPU actually produces under a sustained reasonable load. If it can be higher then what's the point of specifying it? What's the number supposed to actually mean?

The HSF is supposed to keep the temperature within spec if the TDP doesn't exceed the rating (for very long) and the ambient temperature is also within spec (that spec is usually higher than what most people will test in so there has to be some slack to account for that). If the temperature goes out of spec the CPU will throttle to keep it down, but if the CPU has to throttle under a "reasonable" load with a properly designed HSF then something isn't doing its job. You could argue, and I think this is what AtenRa's arguing, that if the HSF is keeping the temperature in spec then it doesn't matter if the TDP is exceeding the limit. But that means that a HSF designed only to accommodate the rated TDP and other design information provided by the CPU isn't going to do its job. And TDP really is going to be close to sustained power consumption under a reasonable load. I don't see the flaw here and what it is I'm not understanding. I don't see how a CPU with a TDP rated at 125W is free to consume far more than that while its temperature requirements are being met by the HSF, ie its power consumption isn't out of control due to thermal runaway (in which case it'd throttle anyway).

The temperature rising on IB isn't a problem if it's still in spec, it doesn't matter if it's higher. They could have raised the spec; if the processor can tolerate it it doesn't make a difference. Since we're not seeing mass recalls of IB's it's obviously not a problem except for people trying to overclock it as much as possible.

 

[Cookie Monster]

The temperature rising on IB isn't a problem if it's still in spec, it doesn't matter if it's higher. They could have raised the spec; if the processor can tolerate it it doesn't make a difference. Since we're not seeing mass recalls of IB's it's obviously not a problem except for people trying to overclock it as much as possible.

I dont think anyone was saying that it was a problem. But a minor annoyance for enthusiasts due to the poor thermal characteristics compared to Sandy.

 

[Exophase]

It sucks for enthusiasts, no disagreement there. But it doesn't mean Intel is lying or inconsistent or that their stock cooler is insufficient. I think that's AtenRa's counterargument here. That there's nothing wrong with FX-8350 having a 125W TDP rating but that IB is violating its specs with the temperature increase.

 

[ShintaiDK]

AtenRa just wants the light away from the fact the FX CPUs are 140W CPUs instead of 125W.