The Official Kaveri Review Thread (A10-7850K, etc)

[AtenRa]

I was looking at CPU performance.

I was comparing Desktop SKUs, A8 Richland vs A8 Kaveri at the same 45W TDP to arrive to the Mobile SKUs conclusion.

And I can't see any appreciable difference between A8-7600 and A10-7850k in GPU performance as an effect of shader number.

According to AT review, A10-7850K is 15-20% faster on average than A8-7600 65W in games and 25%+ in OpenCL and Compute. And dont forget this is only coming from shader performance alone, since both have 2133MHz memory and 720MHz iGPU clocks.

 

[Shivansps]

Now, imagine a 35W Mobile Kaveri against 35W Mobile Richland. And that's only with cheap 28nm bulk.

No if the Kaveri 7600 is cheating with the TDP, and you cant really be sure right now whiout investigating intro it.

 

[coercitiv]

No if the Kaveri 7600 is cheating with the TDP, and you cant really be sure right now whiout investigating intro it.

Care to elaborate? Proof would be nice.

 

[Shivansps]

ive already elaborared it on the previus page.

The A8-7600 is a bit strange to me, it uses way too much power to dissipate just 45W, way too close to 7850K numbers, them we have that thing of using more power than a 45W Richland, but the 6800K uses more power than a 7850K?

It its odd, and all that could be easily explained if its using a 2nd tdp like Intel does to get way over the CPU tdp as long the cooler allows it.

As i said before, in desktop does not matter, but it will on mobile, as for proof, unless i can get my hands on one there is just no way, but i think i made my case.

 

[Haserath]

I'd assume AMD has balanced the TDP better with Kaveri. And the reason the 7850 and 7600 perform and use power identically is because the CPU by itself can use around 65W of the TDP.

http://techreport.com/review/25908/amd-a8-7600-kaveri-processor-reviewed/12
45W TDP on the same chip limits it. The Richland chips are also well below their own TDP limit.
And here:

http://www.tomshardware.com/reviews/a10-7850k-a8-7600-kaveri,3725-15.html
7850K and 7600 running roughly equal power throughout, but the entire system would only go around a sustained 100W from an idle 35W.

I don't know of any reviews that measure power for a realistic(not furmark&prime) CPU&GPU load.

 

[AtenRa]

No if the Kaveri 7600 is cheating with the TDP, and you cant really be sure right now whiout investigating intro it.

There is no TDP cheating, how many times should we say that TDP is NOT power Consumption.

Core i5 4430 has the same TDP rating of 84W as Core i7 4770K. They obviously dont have the same power consumption but nobody have come here and say that Intel is Cheating with the TDP.

 

[Abwx]

Theses graphs are for systems power usages and totaly
useless for CPU power comsumption evaluation.

More professionaly Hardware.fr measured 110W at the main
and 60W on the 12V rail wich is the comsumption of the CPU
+ VRMs losses , that with a 7850K and MT Fritzchess bench.

 

[Enigmoid]

There is no TDP cheating, how many times should we say that TDP is NOT power Consumption.

Core i5 4430 has the same TDP rating of 84W as Core i7 4770K. They obviously dont have the same power consumption but nobody have come here and say that Intel is Cheating with the TDP.

Well generally cheating is when something is said to be/have less than stated, not more.

Sure the i5 has a higher tdp rating that it really should but thats much different than it having a lower rating than it should.

A product guaranteed to be up to X should be able to attain X, if it achieves X+ 20% than thats fine. TDP should be representative of power use but in the event that its not, overstating is much less devious than understating.

And before we get into tdp and cheating, a lot of companies 'cheat'. If anything AMD is going what intel is doing on mobile. TDP = max power at default clocks. If thermal headroom is available use more power to reach boost clocks.

Intel does this on mobile where ULV chips cannot run full load on igp + CPU at 15W. I'm guessing amd is doing something similar on the desktop.

 

[Enigmoid]

Theses graphs are for systems power usages and totaly
useless for CPU power comsumption evaluation.

More professionaly Hardware.fr measured 110W at the main
and 60W on the 12V rail wich is the comsumption of the CPU
+ VRMs losses , that with a 7850K and MT Fritzchess bench.

But if it requires a much more power hungry chipset then that has to be taken into account.

Its similar to a GPU who uses the same amount of power on chip for the same amount of work but requires additional CPU power, increasing system power consumption.

 

[frozentundra123456]

By the same token, one cannot say TDP is meaningless or subjective, and at the same time cite the TDP of Kaveri to show how power efficient it is. In fact if you look at Intel, Haswell gives much higher battery life relative to IVB than the TDP would indicate.

The "45 W" TDP A8 does indeed look promising, but the low TDP does not necessarily mean it will be a very efficient laptop chip. In fact, idle states and low power management are much more inportant in mobile that max power consumption.

 

[ElFenix]

ive already elaborared it on the previus page.

The A8-7600 is a bit strange to me, it uses way too much power to dissipate just 45W, way too close to 7850K numbers, them we have that thing of using more power than a 45W Richland, but the 6800K uses more power than a 7850K?

It its odd, and all that could be easily explained if its using a 2nd tdp like Intel does to get way over the CPU tdp as long the cooler allows it.

As i said before, in desktop does not matter, but it will on mobile, as for proof, unless i can get my hands on one there is just no way, but i think i made my case.

i'm still trying to figure out how you managed to decide it uses too much power to just be drawing 45 watts

 

[Enigmoid]

i'm still trying to figure out how you managed to decide it uses too much power to just be drawing 45 watts

Comparison to i3 and i5/i7, The 65W chip is drawing nearly as much power. Furthermore drawing that much power on a CPU only load is not good, what happens when the GPU is loaded?

What I think Shivapsps is saying is that it looks like AMD is labelling tdp, similarily to how intel does for ULV, except without a hard cap. The CPU is designed to use 45/65W but can go over that limit to maintain turbo (similar to amd/intel on mobile where you will most definitely not be able to get full clocks under full load only instead of throttling down to stay in the power envelope, the cpu is sucking more power to maintain turbo).

 

[coercitiv]

ive already elaborared it on the previus page.

The A8-7600 is a bit strange to me, it uses way too much power to dissipate just 45W, way too close to 7850K numbers, them we have that thing of using more power than a 45W Richland, but the 6800K uses more power than a 7850K? It its odd, and all that could be easily explained if its using a 2nd tdp like Intel does to get way over the CPU tdp as long the cooler allows it.

As i said before, in desktop does not matter, but it will on mobile, as for proof, unless i can get my hands on one there is just no way, but i think i made my case.

There seems to be some confusion here as to how (configurable) TDP limits work even in Intel processors. TDP and maximum power consumption are not one and the same, even if they may be equal in certain scenarios. For example, the mobile Core i7 4700HQ has a 47W TDP and power limits described as follows: a 47W long term Turbo Boost ceiling and - attention - a 58.75W short term Turbo Boost Max ceiling. So if we were to measure peak power drawn by this processor under a high load scenario, our measurements would indicate 59W!

However, even if TDP and peak power were to be the same, there is still enough data to show you suspicions are unfounded.

As I have already mentioned in a previous post, if we consider a 45W power budget for the A8 7600 and a 20W power consumption for MB + RAM + SSD + Fans, take into account a 90% efficiency in the cpu power regulation circuitry and consider an unlikely high 85% PSU efficiency (for such a low load), we still end up with a total power consumption of 83W, which is quite in line with what reviews have shown.

Furthermore drawing that much power on a CPU only load is not good, what happens when the GPU is loaded?

Tech spot tested system power draw with Prime95&Furmark and got 86W.

 

[Shivansps]

Re-read what i say, i never linked power to TDP, only for that part of "kavery uses way too much power to just dissipate 45W".

Comparison to i3 and i5/i7, The 65W chip is drawing nearly as much power. Furthermore drawing that much power on a CPU only load is not good, what happens when the GPU is loaded?

What I think Shivapsps is saying is that it looks like AMD is labelling tdp, similarily to how intel does for ULV, except without a hard cap. The CPU is designed to use 45/65W but can go over that limit to maintain turbo (similar to amd/intel on mobile where you will most definitely not be able to get full clocks under full load only instead of throttling down to stay in the power envelope, the cpu is sucking more power to maintain turbo).

Exactly, but Intel allows that for every CPU, it happens on standart mobile too, its up to OEMs to use it our not, ive encountered a few Lenovos that, for example, as OEM can customise the TDPs, they set a 17W part to run at 10W tdp, and 17W as 2nd tdp, meaning they go for a cheap cooling and the big impact will be the turbos will kick in for a short time.

OpenHardwareMonitor reports the CPU TDP on real time, at least for Intel.

 

[ElFenix]

Re-read what i say, i never linked power to TDP, only for that part of "kavery uses way too much power to just dissipate 45W".

again, how did you come to that conclusion?

Comparison to i3 and i5/i7, The 65W chip is drawing nearly as much power.

so? the intel chips may not be running at full TDP.

what happens when the GPU is loaded?

cpu clocks drop.

 

[Shivansps]

again, how did you come to that conclusion?

By looking at other cpus, even considering the errors and components it seems a bit too much because its almost doubling the tdp whiout a dgpu on it, the only other cpus/apu ive seem doing that re the very low TDP ones, like the E-350, Atoms, etc, but that just the last of the reasons, im not sure about it.

 

[Abwx]

By looking at other cpus, even considering the errors and components it seems a bit too much because its almost doubling the tdp whiout a dgpu on it, the only other cpus/apu ive seem doing that re the very low TDP ones, like the E-350, Atoms, etc, but that just the last of the reasons, im not sure about it.

On Hardware.fr test of the 7850K the APU comsumption
including VRMs losses was 55% of the plateform total
comsumption when benching with Fritzchess with 4 threads,
ratio must be much lower with a 45W TDP APU given that it
operate at a point where the PSU has even lower efficency,
likely slightly below 50% of the total comsumption will be
drained by such an APU.

On the TReport test the PSU has about 75% efficency at
80W input so only 60W goes to the plateform wich correlate
the 45W TDP of the APU.

Edit : Tomshardware test of the 7600 is made with
TDP configured at 65W............

 

[Vesku]

But if it requires a much more power hungry chipset then that has to be taken into account.

Its similar to a GPU who uses the same amount of power on chip for the same amount of work but requires additional CPU power, increasing system power consumption.

The TDP rating is just to give OEMs and to a lesser extent consumers an idea of what kind of cooling, air flow + heatsink, will be required for the CPU. Chipsets have their own thermal information which is not even bothered to be passed on to consumers (I've never seen it listed by retail motherboard companies).

Moving on from the whole TDP thing: the A8-7600 is an OK desktop chip in terms of performance and power draw. Just OK, but with the added benefit of having AMD's GCN based graphics. Definitely worthy of consideration in the low end desktop space.

As for Kaveri mobile, AMD will need to squeeze a bit more out of GF 28nm to provide a similarly OK notebook offering.

 

[el etro]

No if the Kaveri 7600 is cheating with the TDP, and you cant really be sure right now whiout investigating intro it.

TDP is a estimation, don't reflect power consumption otherwise with same architecture chips(not compared to other architectures(intel's or amd's) chips).

 

[Enigmoid]

so? the intel chips may not be running at full TDP.

cpu clocks drop.

Intel is quite conservative with TDP on the desktop. I would also say that, in my opinion, CPU or GPU clocks should never drop on the desktop because of TDP requirements. IMO boost should be fully and completely available.

CPU clocks dropping is not a good sign for mobile.

 

[tweakboy]

Well see what happens!!!

 

[Dresdenboy]

Intel is quite conservative with TDP on the desktop. I would also say that, in my opinion, CPU or GPU clocks should never drop on the desktop because of TDP requirements. IMO boost should be fully and completely available.

CPU clocks dropping is not a good sign for mobile.

There was throttling with some P4 models. Did your hear about that?

 

[coercitiv]

I would also say that, in my opinion, CPU or GPU clocks should never drop on the desktop because of TDP requirements. IMO boost should be fully and completely available.

1. The A8 7600 has a configurable TDP - meaning user can choose between 45W and 65W. Is it your opinion that the CPU and GPU clocks should never drop bellow maximum values no matter what TDP you choose?

2. If "boost" is to be "fully and completely available", what is the point of calling it boost? Wouldn't it be nominal speed?

[Later edit] I'll provide a short example to compare with your expectations:

The mobile i7 4700HQ CPU has a 47W TDP, a nominal speed of 2.4Ghz for CPU and 3.2Ghz turbo boost (on all 4 cores).
- When stressed with Prime95 it will boost up to 3.2Ghz and consume 43W.
- When stressed with both Prime 95 & Furmark it will jump to 58W power consumption for a short while, then go back to 47W by lowering clocks to 2.4Ghz (GPU starts and stays at 1200Mhz)

I honestly don't care if this is a mobile CPU or not. It has a given TDP and does the best it can within the power envelope it is given.

 

[AtenRa]

Well generally cheating is when something is said to be/have less than stated, not more.

Cheat
verb \ˈchēt\ : to break a rule or law usually to gain an advantage at something

Sure the i5 has a higher tdp rating that it really should but thats much different than it having a lower rating than it should.

Core i7 4770K has the same TDP value of 84W as Core i5 4430. Core i7 4770K has higher Power Consumption than Core i5 4430. With the same thinking, Intel is cheating in the TDP value of Core i7. Do you get it now ???
Shivansps compared the TOTAL Platform power consumption of different 45W TDP SKUs in Reviews to come to the conclusion that AMD is cheating with Kaveri 45W TDP value simple because another 45W TDP SKU has lower power consumption than Kaveri.
As we have said so many times here before, TDP is not Power Consumption. Two different 45W TDP SKUs can have different power consumptions. Same applies to Intel SKUs both in Desktop, Mobile and Server.

 

[9enesis]

TDP is not Power Consumption

emmm....yes and no. TDP(heat) = max* (electrical)power consumption for real world loads. IOW if one sees APU/CPU electrical power consumption exceed TDP value , then one might start screaming for fraud.

* As to (some)intel's CPUs - those have some transient peak power "feature" thus they might exceed the TDP value for short (state changing etc) bursts, if there is enough temperature headroom.


edited