AMD's Ryzen is here!


#21

So far, it looks like the Ryzen 7’s are for Desktops, but the potential for mobile is HUGE!!! Really excited… One core in particular would be awesome in a V: The Ryzen 7 1700. It’s an awesome CPU that runs at only 65 Watts, yet with 8 cores and 16 threads running at 3 GHz each, and boosting at 3.7 GHz, this has some SERIOUS potential for laptops like the V running for a ridiculous amount of time on the larger battery!


#22

Ehm… 65W - are you seriously comparing that to our 4.5W? :smile:
The most powerful laptop processors are rated at 45W, and those rarely get more than 3h of real world battery life, even without heavy loads.


#23

To be perfectly honest, I didn’t know what watt draw the intel processor that was in the V had, but I was just talking about the potential that this type of technology has in the future for the Mobile world! At this point, we don’t even know when ryzen will come out for laptops, and if they have this kind of tech for Desktops, I’m hoping that they’ll have similar tech for laptops!


#24

Shouldn’t expect much on the mobile processors from AMD until 2019 which is rumoured to be the launch of the 7nm (this ones are 14nm). If they can get good performance/value under 35w they could get something in the netbook/laptop market.


#25

Not the power, it’s the heat.

A 7GHz i7-6700k running at 1.8+ volts vcore cooled with liquid nitrogen consumes waaaaaaaaay more power than 95 Watts, for example.

And thermal throttling normally isn’t done by the CPU itself, but by the motherboard chipset.


#26

At least in my PC, I think it’s the CPU itself that does the throttling. The BIOS has no settings related to that, and the CPU only throttles when it reaches its absolute maximum operating temperature (102C), so I assume it’s a safeguard implemented in the CPU itself. Otherwise, I don’t think the morherboard would let it reach such dangerious temps…


#27

I wasn’t being clear enough or didn’t put this correctly :joy::

And thermal throttling normally isn’t done by the CPU itself, but by the motherboard chipset.

I was trying to say hard limits are generally imposed by the motherboard. (It can’t be “done” without the collaboration with the CPU.)

Throttling is also known as “dynamic frequency scaling”. From my understanding, there are conceptually two (or three) kinds of throttling. All of which involves both the CPU and motherboard and revolves around two concepts: adjusting voltages and frequency. Both of which requires “collaboration” between the motherboard and the CPU.

Since they employ similar techniques, they’re actually the same thing but have different scenarios and triggers.

The first kind is the absolute temperature protection, the kind you described. This is not controllable anywhere - an absolute maximum. Normally this limit is set at around 85 - 100 degrees Celsius. It’s to protect the CPU from frying/melting itself. In extreme cases (e.g. heatsink falls off and the temperature instantly increases pass 100C), the system shuts down.

The second kind is “Thermal Throttling”. This is the kind we’d notice more on notebooks and laptops. It may or may not be configurable from the BIOS/UEFI (or possibly from the OS as well). These are throttle more aggressively than the first kind (i.e. throttles at lower temperatures) and is generally related to the specified TDP of the CPU. (I was talking about this kind)

The third kind is “Dynamical Frequency Scaling”. Largely the same as the second kind, the aim is not to prevent overheat as in the second kind, but to conserve power. If you heard of Intel’s “SpeedStep” and AMD’s “Cool ‘n’ Quiet”, it’s this kind.

On most if not all systems, the first and third kind is always there. On certain systems, the second kind may not exist at all (or indistinguishable from the first kind).


#28

I think AMD is pretty comfortable to destroy a laptop’s Core i7 with Ryzen mobile at 45watts. I mean, Intel hasnt done anything with laptops since 2012…


#29

My case was neither of what you described :slight_smile: When the CPU reaches 101C, it starts slowly reducing the frequency. But it doesn’t shut down. And nope, it’s not configurable anywhere from BIOS… So I’m asuming the CPU does this itself…

@Xify LOL!!! Yeah except making a GIANT breakthrough in thin&light laptops with their Core M and Atom series?


#30

Which is why it’s the first case I described.

It only shuts down in extreme cases, for example starting without a heat sink or removing the heat sink when the CPU is running.

And right, the first case is not configurable.


#31

No, it doesn’t shut down :smile: that’s what I’m trying to say, it doesn’t. It just reduces the frequency!


#32

Sorry for confusing you… :dizzy_face:

I mean very extreme cases where the temperature rises from room temperature to 100+ degree Celsius in half a second. Only in these very extreme cases they shutdown.

Increasing it to 100 degree Celsius in 10 seconds is not extreme and would not cause an immediate shut down, but a throttle instead.

(Note: half-second vs 10 seconds is for illustrative purpose only, I do not know the actual threshold for abrupt changes that would result a shutdown.)


#33

FYI

Performance and power consumption are a mystery so far, but there is promise of nice mobile chips.


#34

From what I see in the industry, Intel is not limiting core numbers because of cost nor engineering limitation, they do it simply because they can. If they are capable of putting 22 cores in a CPU, I don’t think putting 6 or 8 is that hard.

Whats interesting is the trend of the industry. The ‘cheaper alternative’ companies like MediaTek (compared to Apple or Qualcomm) and AMD (compared to Intel) tend to give you more cores, while the actual experience is arguably worse. Unlike Intel however, they increase the core numbers because its the only thing they can do to have an advantage over the competition.

You are wrong. The i7/2600K can hit 95W quite easily, and when overclocked, it can also easily consume more than 100W, which is why you need a beefier cooler with it.

Same is true with graphics card. Increasing the “Power Limit” increases, well, the power limit past the TDP that the card is by default designed to run at.


#35

But TDP != power consumption… It’s supposed to show how much heat is emitted. Not sure how they calculate the thermal power, though… But I bet it’s not a simple “all power consumed = heat” equation.


#36

There is a lot of discussion around TDP that misunderstands what the number means:

Thermal Design Power (TDP) is the expected amount of heat the system is expected to dissipate under a typical load.

  • Is it a measure of the maximum amount of heat a processor can generate? No.

  • Is it a reflection of processor power consumption? No, only an indication.

TDP is a manufacturing specification that tells original equipment manufacturer (OEM like Eve) what the level of cooling demand they need to engineer for

  • the central processing unit (CPU)/
  • system on chip (SoC)/
  • accelerated processing unit (APU)/
  • graphical/graphics processing unit (GPU)
    in order for it to perform to specification under typical operating conditions.

Processor power draw is a very contentious issue as there are many ways to measure power draw. It is incredibly difficult to determine what the power draw is as even minute differences in the processor itself and other components (e.g. motherboard, memory etc.) can make a difference.

There is NO ONE truth about the processor power draw. Probably just alternate facts, ha ha…

Here’s an example: Tom’s Hardware tested Intel Core 7700K processor in Germany and the USA with the same methodology and the power draw difference was 18 watts.

Tom’s Hardware had the following power draw figures: Germany 137 watts and USA 119 watts. And in Anandtech’s review the processor consumed 90 watts. And Intel Core 7700K processor’s TDP is 91 watts…

I’ve read about an Intel Xeon processor with a TDP of 145 watts that actually consumed some 400 watts with an advanced vector extension (AVX) test.

So, TDP is not the same as power draw, period.

What the actual power draw is, depends on so many factors that there are as many answers as there are processors and reviewers.

In certain conditions, the power draw is ALWAYS larger than the TDP.


UPDATE: Eve V status
#37

Is TDP then the same as the power a CPU loses as heat (under the expected/typical load) or did I misunderstand you?


#38

According to Core M, yes it is :wink:


#39

First of all, I understand that the current TDP definition is about the cooling system design and requirements more than about the actual processor heat generation even though they are related.

One definition is:

TDP = the expected amount of heat the cooling system is expected to remove from the processor under a typical load.

Another definition from Wikipedia:

“Both Intel and AMD have defined TDP as the maximum heat generation for thermally significant periods, while running worst-case non-synthetic workloads; thus, TDP is not reflecting the actual maximum power of the processor.”

The latter also gives a definition for a typical workload.

In principle these definitions mean the same, but they approach the definition from different directions. First one is about removing heat and the latter about heat generation.

Question for SyrtakiVampir:
Please define what do you understand with your expression: “the power a CPU loses as heat”.

This is probably semantics, but I prefer to talk about generated heat and not heat losses.

And the purpose of a processor is not to generate heat but to execute instructions.


UPDATE: Eve V status
#40

Sorry Paul, your reply was lost on me. Please elaborate.