Why have AMD APUs failed on the market?

[Idontcare]

AMD failed because it simply lacked any kind of understanding on what the market wanted. First of all performance/watt.

Ironic given that AMD spent a good 4-5 years beating Intel up over the fact their SOI-enabled process nodes delivered superior performance/watt over Intel's bulk-Si produced Pentium 4 processors.

 

[bononos]

What i reported was Total System Power Consumption, not APU power consumption.
Also, that power consumption gives you %50-100 more iGPU performance.

I'm very sure you were responding to someone who posted that the APU sucked down more power than was stated in the TDP. AMD seems to have exceeded its TDP specs with the A8-7600 which sucks just as much power as the i7 ivy.

http://www.bit-tech.net/hardware/cpus/2014/01/14/amd-a8-7600-kaveri-review/12
http://www.guru3d.com/articles-pages/amd-a8-7600-apu-review,7.html

 

[DrMrLordX]

Mantle/DX12 aside, what about the everyday user of desktop hardware that doesn't game? (As we have seen AMD A8 and A10 Pre-built desktops appear to be at a massive price disadvantage compared to the Intel competition)

AMD may just have to suck it up and accept the fact that Intel can undersell them thanks to the 22nm process advantage and all the economic benefits that come along with it (not to speak of 14nm). My guess is that AMD wants low-end users to buy 1M + GCN chips and rely on HSA/OpenCL for acceleration often enough that the AMD chips can still compete with Pentiums and Celerons effectively. There are probably some opportunities out there for such a solution to work occasionally, but it'll be pretty inconsitant. Really, the machine with an SSD is going to "feel" fastest for that use case anyway.

Or maybe they can convince an OEM to start selling cheap A8-7600 boxes, but the tray price on the 7600 would have to come down for that to happen. And again, process process process . . .

P.S. Regarding the use of igpu to boost dGPU gaming, I don't know much about that now but I do wonder how soon it will pick up? I also wonder how well it will do for a person to spend extra money on an iGPU vs. having a better cpu (but worse iGPU) for dGPU gaming. Something tells me having the better cpu is still going to be the better option.

Bear in mind that, in some circumstances, you won't be spending more for an iGPU. Consider what most people recommend for a budget-to-midrange machine nowadays: a Haswell i5 on LGA1150. The cheapest is the i5-4440 (so far as I can tell) which is around ~170. In contrast, you can grab an A10-7850k for ~$135-$140, depending on where and when you buy it. If DX12 shows up big for iGPUs and the 7850k starts consistently beating i5s in new games while running a 290x or 380x or whatever, then you haven't spent extra money buying the better DX12 gaming CPU.

Or to present another scenario, let's say people are picking between Broadwell Iris Pro and 4790k or a 5820k. I have no idea how much that Iris Pro will cost (probably not too much, it's only a 65W part or something?), but I expect it won't be more than the 4790k or the 5820k. Now in this scenario, sure, Iris Pro might come out costing quite a bit more than a bone-stock i5, but you are (presumably) already dealing with a beefier CPU than some low-end, locked Haswell i5, so you've got to bring more powerful chips into the comparison before you can honestly say you're comparing a "small" CPU + "big" iGPU to a "big" CPU.

In what is an equally-probable (if not more probable) scenario, it is totally possible that DX12 titles will show up big on Broadwell Iris Pro + dGPU as compared to some previous-gen Intel processors that most people would expect to be "better for the serious gamer".

Now as to "how soon", I expect the first example of this phenomenon will be DX12-capable 3DMark, if they finally start allowing physics scores to be affected by a GPU (iGPU or otherwise).

 

[mrmt]

I'm very sure you were responding to someone who posted that the APU sucked down more power than was stated in the TDP. AMD seems to have exceeded its TDP specs with the A8-7600 which sucks just as much power as the i7 ivy.

http://www.bit-tech.net/hardware/cpus/2014/01/14/amd-a8-7600-kaveri-review/12
http://www.guru3d.com/articles-pages/amd-a8-7600-apu-review,7.html

AMD converted its TDP specs in a marketing bullet point years ago and it's common knowledge that their products breach their published TDP specs by a significant margin. This may not be a factor for the consumer, because he buys a finished products, but for AMD partners, who have to bear the costs of designing components able to handle the extra juice AMD processors need, it's an entire different matter. And given the way some AMD partners are dealing with the problem (255C sensor readings), I don't think AMD is being forthcoming with them on the matter.

 

[coercitiv]

AMD converted its TDP specs in a marketing bullet point years ago and it's common knowledge that their products breach their published TDP specs by a significant margin.

Could you please point to this common knowledge you speak of?

 

[NTMBK]

AMD converted its TDP specs in a marketing bullet point years ago and it's common knowledge that their products breach their published TDP specs by a significant margin. This may not be a factor for the consumer, because he buys a finished products, but for AMD partners, who have to bear the costs of designing components able to handle the extra juice AMD processors need, it's an entire different matter. And given the way some AMD partners are dealing with the problem (255C sensor readings), I don't think AMD is being forthcoming with them on the matter.

For the thousandth time, thermal dissipation power is not the same as peak power draw. It is a specification of cooling capability, not power delivery circuitry. Intel does exactly the same thing with its Turbo Boost technology, exceeding TDP within a narrow window before dropping the clock back down before it becomes thermally significant.

 

[ayylamayo]

For the thousandth time, thermal dissipation power is not the same as peak power draw. It is a specification of cooling capability, not power delivery circuitry. Intel does exactly the same thing with its Turbo Boost technology, exceeding TDP within a narrow window before dropping the clock back down before it becomes thermally significant.

What? At 3.9ghz turbo boost on a Haswell i5 the CPU power draw is much less than the 84w TDP. Even a modest overclock fares lower than the tdp target.

 

[coercitiv]

What? At 3.9ghz turbo boost on a Haswell i5 the CPU power draw is much less than the 84w TDP. Even a modest overclock fares lower than the tdp target.

And a 47W TDP mobile Haswell i7 can go to 58W power usage for short periods of time. Your example only enforces what NTMBK said: thermal dissipation power is not the same as peak power draw.

 

[mrmt]

For the thousandth time, thermal dissipation power is not the same as peak power draw. It is a specification of cooling capability, not power delivery circuitry. Intel does exactly the same thing with its Turbo Boost technology, exceeding TDP within a narrow window before dropping the clock back down before it becomes thermally significant.

Oh, come on, you know that I'm talking about the FX throttling and MSI developing the 255 degrees workaround. No need to beat this dead horse anymore. I didn't know you found a Intel chip that throttles down below base clock, as AMD chips do.

 

[bononos]

AMD converted its TDP specs in a marketing bullet point years ago and it's common knowledge that their products breach their published TDP specs by a significant margin. This may not be a factor for the consumer, because he buys a finished products, but for AMD partners, who have to bear the costs of designing components able to handle the extra juice AMD processors need, it's an entire different matter. And given the way some AMD partners are dealing with the problem (255C sensor readings), I don't think AMD is being forthcoming with them on the matter.

That was kind of my point. I was pretty sure Atenra admitted earlier in this thread that the 7600 APU consumed more power than the TDP. Then he put aside this fact and continued to trumpet the 'low' 45W TDP of the A8-7600 as a selling point and even bolded the '45W' in post #494.

After I pointed this out, he said he was referring to "total power consumption" alluding that the test systems were loaded down with peripherals/drives/cards which drove up the 'total power consumption', which doesn't make sense since the i3/i5/i7 test beds were consuming less or similar amount of power.

 

[Idontcare]

Oh, come on, you know that I'm talking about the FX throttling and MSI developing the 255 degrees workaround. No need to beat this dead horse anymore. I didn't know you found a Intel chip that throttles down below base clock, as AMD chips do.

MSI engineers and company are by far the most frank and forthright individuals you'll come across in the OEM industry. It is their corporate culture, to their credit.

 

[AtenRa]

I was pretty sure Atenra admitted earlier in this thread that the 7600 APU consumed more power than the TDP. Then he put aside this fact and continued to trumpet the 'low' 45W TDP of the A8-7600 as a selling point and even bolded the '45W' in post #494.

After I pointed this out, he said he was referring to "total power consumption" alluding that the test systems were loaded down with peripherals/drives/cards which drove up the 'total power consumption', which doesn't make sense since the i3/i5/i7 test beds were consuming less or similar amount of power.

Get your facts straight, I said 88-89W peak total system power consumption not APU power consumption.

Gaming with the A8-7600 at 45W TDP with an ASUS A88XM-Plus and using the ThermalTake TR2 380W 80plus Bronze PSU, the maximum peak power draw from the wall for the entire system was 88-89W (in one or two games) if I remember correctly.

 

[NTMBK]

Oh, come on, you know that I'm talking about the FX throttling and MSI developing the 255 degrees workaround. No need to beat this dead horse anymore. I didn't know you found a Intel chip that throttles down below base clock, as AMD chips do.

Literally any modern Intel chip will throttle to below base clocks if the cooling is insufficient- that's what they are designed to do.

But yeah, some AMD parts lately have very bizarre throttling behaviour. Kaveri was a real mess for it. Given that AMD have been dealing with this stuff since the Cool'n'Quiet days you'd think they could have figured it out by now :\

 

[mrmt]

Literally any modern Intel chip will throttle to below base clocks if the cooling is insufficient- that's what they are designed to do.

So when operated within the intended specifications Intel chips don't throttle, right?

But yeah, some AMD parts lately have very bizarre throttling behaviour. Kaveri was a real mess for it. Given that AMD have been dealing with this stuff since the Cool'n'Quiet days you'd think they could have figured it out by now :\

So you are saying that AMD chips are throttling when they shouldn't, right?

 

[monstercameron]

That moment when this moved from an interesting thread to the usual AMD bash thread, go certain poster, you never fail to disappoint.

So can someone explain to us how amd chips throttle within spec?

 

[coercitiv]

That was kind of my point. I was pretty sure Atenra admitted earlier in this thread that the 7600 APU consumed more power than the TDP. Then he put aside this fact and continued to trumpet the 'low' 45W TDP of the A8-7600 as a selling point and even bolded the '45W' in post #494.

After I pointed this out, he said he was referring to "total power consumption" alluding that the test systems were loaded down with peripherals/drives/cards which drove up the 'total power consumption', which doesn't make sense since the i3/i5/i7 test beds were consuming less or similar amount of power.

Based on the data you pointed out in previous posts, what is the average power usage of an AMD A8 APU under load?

 

[bononos]

MSI engineers and company are by far the most frank and forthright individuals you'll come across in the OEM industry. It is their corporate culture, to their credit.

Sounds like they are former military.

 

[bononos]

Get your facts straight, I said 88-89W peak total system power consumption not APU power consumption.

Based on the data you pointed out in previous posts, what is the average power usage of an AMD A8 APU under load?

I don't know but going by guru3d/bittech figures, which shows the 7600 idling at 40+W which is higher than the i3 and very slightly lower than i7, I don't think the A8 is stay true to the 45W limit at load and only momentarily overshooting the 45W limit. The power consumption at load is probably staying above 45W for the majority of the time.

 

[NTMBK]

So when operated within the intended specifications Intel chips don't throttle, right?

No, I am not saying that at all. I am saying that if they get too hot they will downclock, which happens in a fair few shipping products. Check out Anandtech's review of the Gigabyte Brix Pro with i7-4770R: http://www.anandtech.com/show/8175/...-at-the-intel-i74770r-with-iris-pro-hd-5200/7 When the CPU and GPU are fully loaded, the GPU is throttled right down to its 200MHz base clock and the CPU is throttled below it's 3.2GHz base clock.

Throttling happens when you are in a thermally constrained environment; it's meant to happen. It's whether you handle this well or not that makes a real difference, and whether you can make the most of the limited thermal budget. Intel tends to do a pretty good job of it, better than AMD do (see: Kabini launching without a working Turbo).

 

[positivedoppler]

Holy Sh1T, China is making up to a 19 Billion investment with AMD
http://www.vrworld.com/2015/02/09/amds-real-china-play-strategic-investment/

 

[NTMBK]

So AMD now has strategic links with the oppressive Dubai government, the oppressive Chinese government, and a morally bankrupt Korean chaebol... Great :\

 

[positivedoppler]

Not much different than IBM selling their servers to the Chinese. Intel also has all sort of partnership going on in China.

 

[mrmt]

No, I am not saying that at all. I am saying that if they get too hot they will downclock, which happens in a fair few shipping products. Check out Anandtech's review of the Gigabyte Brix Pro with i7-4770R: http://www.anandtech.com/show/8175/...-at-the-intel-i74770r-with-iris-pro-hd-5200/7 When the CPU and GPU are fully loaded, the GPU is throttled right down to its 200MHz base clock and the CPU is throttled below it's 3.2GHz base clock.

Throttling happens when you are in a thermally constrained environment; it's meant to happen. It's whether you handle this well or not that makes a real difference, and whether you can make the most of the limited thermal budget. Intel tends to do a pretty good job of it, better than AMD do (see: Kabini launching without a working Turbo).

From the link you posted:

" After 30 minutes of full CPU loading, we found that the core temperatures were held below 100 C, the clock speeds remained at 3.2 GHz for the CPU (the GPU was 'idling' at 200 MHz) and there was no thermal throttling to be seen. At this point, we introduced Furmark loading into the picture. It becomes clear that the system gives preference to CPU performance. >>The GPU remains throttled at 200 MHz, while the CPU cores don't thermally throttle<<."

See? Base clock is base clock for a reason, at least in Intel processors. That is more than we can say of AMD processors.

 

[coercitiv]

Now we know why they call it Zen.

 

[mrmt]

Holy Sh1T, China is making up to a 19 Billion investment with AMD
http://www.vrworld.com/2015/02/09/amds-real-china-play-strategic-investment/

Gotta wait for the strings attached. AMD isn't exactly in a strong position for a negotiation and we don't know what kind of investment they are going to make. Their 2003 R&D centre didn't really rock the world, did it?