Why have AMD APUs failed on the market?

[Phynaz]

To emphasize the importance of the low TDP.

Low TDP isn't important, low power consumption is.

 

[Cerb]

1156 and 1155 had a LOT more itx boards, the 1155 where cheap too... hell even 775 had more ITX boards than FM2.

Asus and Biostar does not even have FM2 itx, Gigabyte has 1 or 2, Asrock has 2, MSI 1 and ECS has 1 thin itx with very little vrms for some reason.

And thats what about it, there maybe 1 or 2 boards that im missing but thats it, its at least odd, something doesn't add up.

I'm about to buy an Asus one and put a Core i3 in it. The Intel GPU's capabilities are frankly overkill, so I didn't even look at AMD's offerings, after finding that the Intel option was well under budget.

AMD's market share is low, and I suspect that it's just not worth the risk for niche DIY markets. Quickly looking at 1156 and 1155 boards, I see quite a few with what looks like >4 phase voltage regulation, some with coils and VRMs on both sides of the board. Prices were higher than the newer Haswell ones, but not unreasonably so. Even the current 1150 selection for MiniITX is not exactly huge, and those using Realtek or Intel NICs bring it down to almost nothing.

 

[Blue_Max]

1156 and 1155 had a LOT more itx boards, the 1155 where cheap too... hell even 775 had more ITX boards than FM2.

Asus and Biostar does not even have FM2 itx, Gigabyte has 1 or 2, Asrock has 2, MSI 1 and ECS has 1 thin itx with very little vrms for some reason.

And thats what about it, there maybe 1 or 2 boards that im missing but thats it, its at least odd, something doesn't add up.

That is such a huge point! Seriously, the APU just begs to be a tiny SFF - that's the whole damned point! Everything on one chip in order to make things simple and compact... so why on earth didn't it happen?? Where's the tiny, sexy prebuilts with an APU?

 

[Yuriman]

I'm about to buy an Asus one and put a Core i3 in it. The Intel GPU's capabilities are frankly overkill, so I didn't even look at AMD's offerings, after finding that the Intel option was well under budget.

AMD's market share is low, and I suspect that it's just not worth the risk for niche DIY markets. Quickly looking at 1156 and 1155 boards, I see quite a few with what looks like >4 phase voltage regulation, some with coils and VRMs on both sides of the board. Prices were higher than the newer Haswell ones, but not unreasonably so. Even the current 1150 selection for MiniITX is not exactly huge, and those using Realtek or Intel NICs bring it down to almost nothing.

Both Gigabyte and Asus's H97 ITX boards have Intel NICs, while MSI's uses Realtek. It's only ASRock that doesn't, really. My Zotac Z77 ITX board has dual Realtek NICs.

 

[bononos]

To emphasize the importance of the low TDP.

But didn't you admit earlier in this thread that the power consumption of the APUs can exceed the rated TDP by quite abit, around 90+W?


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

 

[AtenRa]

But didn't you admit earlier in this thread that the power consumption of the APUs can exceed the rated TDP by quite abit, around 90+W?


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

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

 

[BSim500]

But didn't you admit earlier in this thread that the power consumption of the APUs can exceed the rated TDP by quite abit, around 90+W?

What I find amusing over the past 2-3 years is how 'irrelevant' +100-200w higher power consumption on FX8xxx/FX9xxx chips vs i5's were regularly declared to be "because it's only $10-$20 per year cost", yet now 45w AMD vs 55w Intel is now the center of the universe (and only then if you ignore the 35w i3-T's) despite "brick" Pico-PSU's in small Mini-ITX cases easily reaching up to 160w... :sneaky:

 

[AtenRa]

What I find amusing over the past 2-3 years is how 'irrelevant' +100-200w higher power consumption on FX8xxx/FX9xxx chips vs i5's were regularly declared to be "because it's only $10-$20 per year cost", yet now 45w AMD vs 55w Intel is now the center of the universe (and only then if you ignore the 35w i3-T's) despite "brick" Pico-PSU's in small Mini-ITX cases easily reaching up to 160w... :sneaky:

What i find amusing is that people still confuse TDP with power consumption.
The emphases of the 45W TDP was made because it directly impacts performance and it also makes the APU suitable for SFF and smaller PSUs.

 

[BSim500]

What i find amusing is that people still confuse TDP with power consumption.
The emphases of the 45W TDP was made because it directly impacts performance and it also makes the APU suitable for SFF and smaller PSUs.

It's also wildly exaggerated and as soon as you introduce low-end dGPU's, direct TDP comparisons go straight out the window since 2x heatsinks & fans have twice the cooling capacity. Even though they may draw a little more power (but provide far higher perf-per-watt), it's not as simple as "45w TDP is better than 50w + 50w TDP") if the CPU cooler ends up the same 50w class size on both, and in Mini-ITX "toaster" cases the dGPU is sucking outside cold air in directly through the side vent), and both can easily be powered by a 120-160w brick PSU but one has triple the fps of the other at less than double the power consumption...

 

[AtenRa]

It's also wildly exaggerated and as soon as you introduce low-end dGPU's, direct TDP comparisons go straight out the window since 2x heatsinks & fans have twice the cooling capacity. Even though they may draw a little more power (but provide far higher perf-per-watt), it's not as simple as "45w TDP is better than 50w + 50w TDP") if the CPU cooler ends up the same 50w class size on both, and in Mini-ITX "toaster" cases the dGPU is sucking outside cold air in directly through the side vent), and both can easily be powered by a 120-160w brick PSU but one has triple the fps of the other at less than double the power consumption...

We are talking about APUs, dGPUs have nothing to do here. A8-7600 as a product(APU) against what the competition have, which is the Intel Pentium and Core i3.

 

[Qwertilot]

Well it does also depend on having APU focused cases/motherboards. Very rare nowadays but it should be possible to build something properly small that'll cool a 100+w APU very nicely.

Part of AMD's problems that of course, that they don't get that much stuff built around their chips.

 

[BSim500]

We are talking about APUs, dGPUs have nothing to do here.

Who is this "we" collective you keep trying to own? Only you and one other person here seem obsessed with 45w restrictions (and the A8-7600 in general) when even consoles "APU's" run double to triple that with better & more detailed gameplay at higher resolutions with most PS4/XB1 owners caring not one jot about +130w TDP or higher power consumption in a small box that sits next to the TV. And they're the biggest actual real-world competitors to low-budget low-performance AMD APU's in the sub $400 price bracket...

 

[AtenRa]

Who is this "we" collective you keep trying to own?

OP topic

"Why have AMD APUs failed on the market?"

It is only you that keeps bringing dGPUs in to the discussion when AMD APUs have the performance lead over Intel, especially at low TDPs.
Not only that, you CPU+dGPU price bracket is way higher than the APU (A8-7600 for this matter) making the comparison apples to oranges.

 

[BSim500]

Not only that, you CPU+dGPU price bracket is way higher than the APU (A8-7600 for this matter) making the comparison apples to oranges.

And in reality many people building a rig for gaming are willing to pay a $70-100 premium for a disproportionately huge +150-200% jump up in performance which is why AMD APU sales have been low. One of the key reasons for the huge price cuts to APU's is hardly "off topic" given the thread title. LOL.

 

[Yuriman]

OP topic

"Why have AMD APUs failed on the market?"

It is only you that keeps bringing dGPUs in to the discussion when AMD APUs have the performance lead over Intel, especially at low TDPs.
Not only that, you CPU+dGPU price bracket is way higher than the APU (A8-7600 for this matter) making the comparison apples to oranges.

APUs have only failed on the market because there are dGPUs. I'm sure gamers would buy AMD's APUs if dGPUs weren't available. dGPUs do everything that an APU can do, but better, and you can select one that fits within your power budget.

/thread

 

[mrmt]

And in reality many people building a rig for gaming are willing to pay a $70-100 premium for a disproportionately huge +150-200% jump up in performance which is why AMD APU sales have been low. One of the key reasons for the huge price cuts to APU's is hardly "off topic" given the thread title. LOL.

Since AMD fielded APUs as they main bet on the market in 2011 their CPU business shrank 70-75%. Maybe the pattern is too subtle, you know, they had more market share in every single market bracket they were present before the introduction of this botched APU concept and now they have less. It must be a real analytical leap to realize the cause and effect relationship between these two events and ergo understand that people must be buying others types of products than the combination of poor CPU + bandwidth choked iGPU AMD offers us.

 

[Blitzvogel]

Since AMD fielded APUs as they main bet on the market in 2011 their CPU business shrank 70-75%. Maybe the pattern is too subtle, you know, they had more market share in every single market bracket they were present before the introduction of this botched APU concept. It must be a real analytical leap to realize the cause and effect relationship between these two events and ergo understand that people must be buying others types of products than the combination of poor CPU + bandwidth choked iGPU AMD offers us.

AMD bet they could get away with mediocre CPU performance by having alot of graphics processing power to "augment" their lackluster CPU performance. Even with Kaveri, the integration in both software and hardware has not far along enough to justify this path.

With a smaller iGP, AMD APUs would've cost much less for the mobile market, where there is more money for AMD to make too. After Carrizo, AMD should cut the IGP die size (256 SP?). Even a 128 SP size IGP is enough for any kind of media playback and very light gaming with enough clock. They should move as fast as possible to stackable APUs where the OEM customer can choose to stick with a generic 128 SP IGP, or add a graphics die w/ HBM on top if they feel it suits a specific market. Honestly, the graphics die sounds like a waste of time and money though, until AMD's CPU performance actually catches up to Intel to be worth pairing with discrete level graphics.

In the APU fervor, AMD pretty much forgot about the higher end desktop market where there was and still is some hope, even if it IS a declining market, but one that is starting to level off at some natural level of long term equilibrium because of it's efficient value for businesses, cheap consumer desktops, and system builders. It's not going away anytime soon.

 

[ShintaiDK]

In the APU fervor, AMD pretty much forgot about the higher end desktop market where there was and still is some hope, even if it IS a declining market, but one that is starting to level off at some natural level of long term equilibrium because of it's efficient value for businesses, cheap consumer desktops, and system builders. It's not going away anytime soon.

But AMD didnt have a CPU core that could compete in that market. Not that it really mattered.

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

 

[Blitzvogel]

But AMD didnt have a CPU core that could compete in that market. Not that it really mattered.

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

I think a half shrink to 28 nm and a 3/4 module Steamroller + 128 SP FX chip on FM2+ (if possible) might have been a good product. The FX chips currently lack any IGP which of course requires a mobo IGP or graphics card, raising their overall product cost in prebuilt desktops.

Even if it is barely a step up above the last round of FX chips that were released (Vishera?), a new product is sometimes just enough to keep the market moving for AMD. Being idle can be foreboding to OEMs who lose confidence in their supplier.

 

[cbn]

A8-7600 as a product(APU) against what the competition have, which is the Intel Pentium and Core i3.

On the DIY market A8-7600 is slotted between Pentium and Core i3.

But in the OEM desktop market it appears the A8 APU (with 6GB to 8GB dual channel RAM) is positioned anywhere between $70 and $100 higher than Core i3 (with 4GB single channel RAM) going by those price comparisons I pulled up from one year ago and few days ago.

Not sure why A8 APUs are priced so high in OEM desktop, one reason I suspect is because AMD just doesn't put enough volume and/or OEM does not see much value in paying for extra iGPU.

So this begs the question what is wrong with AMD releasing smaller iGPU APUs based on a quad core or hex core die? Like I mentioned earlier, a quad core with reduced DDR3 PHY and small iGPU would be half the size of Kaveri, but still provide the steamroller quad core. The small iGPU APU would also work well with single channel RAM lowering price to the OEM further.

As I see it such a big core small iGPU APU (working well with single channel) would be a benefit to both the DIY and OEM desktop market.

P.S. If AMD were to get two steamroller quad core/small iGPU dies for the same amount of silcon as one Kaveri die they could sell each reduced iGPU quad core for half the price of Kaveri. So $80 for each reduced iGPU processor would be like getting $160 for each Kaveri or $70 for each reduced iGPU processor would be like getting $140 for each Kaveri. Even $60 for each reduced iGPU quad core processor works out to getting $120 for one Kaveri.

 

[DrMrLordX]

So this begs the question

One hopes you are not begging the question . . . but hey, people mis-use the term frequently. No big whoop.

what is wrong with AMD releasing smaller iGPU APUs based on a quad core or hex core die?

Like so many other plans that have fallen by the wayside, the potential for such products has been lost due to a shrinking R&D budget. AMD has already dialed in the product lineup between now and 2016. They don't have the cash, income, or credit line to pivot. Agile is expensive when it means taping out new designs, putting them through testing, and realigning all your SKUs to accomodate the new product.

Anything they started now would come out around the same time as Zen, which hopefully will render most (if not all) discussion of Construction cores moot.

Bottom line is: AMD is hoping for/pushing for a software ecosystem in which low-end users get a good-to-great experience from their 2M + GCN core APUs thanks to widespread adoption of OpenCL/HSA. In games, low-end users get Mantle/DX12 (or similar) which makes equisite use of what is still the best iGPU on the market, at least for now. Broadwell Iris Pro may shake that up a bit . . . we'll see.

For the high-end user, the same software ecosystem ensures that the APU user get good-to-great performance from an overclocked APU + GCN video card, and gets exquisite performance in games thanks to Mantle/DX12 offloading a bunch of work to the iGPU that normally slows down a chip like Kaveri.

AMD doesn't have that software ecosystem yet. That's why APUs are seen as failures. That's why people are asking "why can't they release chips with smaller iGPUs?".

DX12 may finally prove the "big iGPU" way to be the superior one. Looking at Intel's future chips, it may well be their APUs that win the battle, assuming their drivers get better (or at least more consistent).

But it's hard to see things that way when practically nothing that actually exists hints at that future, aside from OpenCL apps where people will cry foul if they point to a Kaveri chip beating a Haswell until a dGPU is applied (it's hard to tell which ones are capable of using both iGPU + dGPU as well. AIDA64 GPGPU benchmark does this . . . does anything else?).

 

[cbn]

Bottom line is: AMD is hoping for/pushing for a software ecosystem in which low-end users get a good-to-great experience from their 2M + GCN core APUs thanks to widespread adoption of OpenCL/HSA. In games, low-end users get Mantle/DX12 (or similar) which makes equisite use of what is still the best iGPU on the market, at least for now. Broadwell Iris Pro may shake that up a bit . . . we'll see.

For the high-end user, the same software ecosystem ensures that the APU user get good-to-great performance from an overclocked APU + GCN video card, and gets exquisite performance in games thanks to Mantle/DX12 offloading a bunch of work to the iGPU that normally slows down a chip like Kaveri.

AMD doesn't have that software ecosystem yet. That's why APUs are seen as failures. That's why people are asking "why can't they release chips with smaller iGPUs?".

DX12 may finally prove the "big iGPU" way to be the superior one. Looking at Intel's future chips, it may well be their APUs that win the battle, assuming their drivers get better (or at least more consistent).

But it's hard to see things that way when practically nothing that actually exists hints at that future, aside from OpenCL apps where people will cry foul if they point to a Kaveri chip beating a Haswell until a dGPU is applied (it's hard to tell which ones are capable of using both iGPU + dGPU as well. AIDA64 GPGPU benchmark does this . . . does anything else?).

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)

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.

 

[DownTheSky]

But AMD didnt have a CPU core that could compete in that market. Not that it really mattered.

But couldn't they make one? FX 8350 stands at 320mm2 and 1/3 of that is L3 cache. I don't think +30% die size @ 32nm would have been a problem, if they could have engineered a CPU that could compete with intel. A 400-450mm2 CPU with 50% more IPC, lower cache and lower clocks would had been much better. From what I've seen the biggest problem with AMD vs intel is the memory access time. Latency is 3-4x higher for AMD.

 

[coercitiv]

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)

I don't know about desktop prices, but Kaveri notebook prices are really competitive vs i3 in my country.

 

[Cerb]

Both Gigabyte and Asus's H97 ITX boards have Intel NICs, while MSI's uses Realtek. It's only ASRock that doesn't, really. My Zotac Z77 ITX board has dual Realtek NICs.

GB, at least, uses Atheros, as well. The dual NICs would be great for the cost, were it not for that.