Why have AMD APUs failed on the market?

[mrmt]

Given the amount of posters defending AMD APUs in the AMD Q414 results thread, I'd like to open a new thread to explore why the APUs failed on the market. Regarding the consumer market, AMD lost share in every single market bracket where it fielded APUs and the bleeding is yet to stop.

I'm rather curious to see the opinions of the people here on what are the APU value proposition strong points and why it failed on the market despite these strong points.

 

[Fjodor2001]

The AMD APUs are doing very well in the PS4/XBONE.

On desktop/server I think the Bulldozer uArch and old process tech is the explanation of the lackluster sales. But we're at the end of that era now.

Hopefully the new Zen uArch and 14 nm will change that. It will be competing with Intel's Core uArch from 2006, and Intel will be on 14 nm too at the same time. So AMD's chances should be better. AMD will also have a vastly superior iGPU arch compared to Intel, as AMD has had for a long time.

 

[Shehriazad]

The CPU side was simply too weak.

Richland/Kaveri would both have needed a 25-30% stronger IPC to have a chance at competing with Intels stuff.

But even if AMD WAS able to do that...they would have had to sacrifice their FX series to do that...and they have no means to deliver a meaningful replacement for that, yet.


Also Kaveris GPU performance came like a generation too late and HSA is a few years too early. AMD always manages to bring in things at the worst timings (that's just my opinion)

 

[mrmt]

On desktop/server I think the Bulldozer uArch and old process tech is the explanation of the lackluster sales. But we're at the end of that era now.

They are also bleeding share on the mobile market with Beema, and Beema is no Bulldozer.

 

[sandorski]

Process disadvantage and Power Consumption, both related to each other.

 

[mrmt]

Process disadvantage and Power Consumption, both related to each other.

So in your opinion the concept and the design are fine, they just don't have a competitive foundry partner?

 

[sandorski]

So in your opinion the concept and the design are fine, they just don't have a competitive foundry partner?

Yes.

 

[CHADBOGA]

The single core performance is simply too weak compared to what Intel is offering and the GPU part isn't better by enough, to matter.

 

[senseamp]

AMD thought that having a good GPU is enough to offset having an inferior CPU.
But that's not how reviewers and users think. They don't do weighted sum comparisons, but they compare on one metric first, then followed by another metric. The ranking of the metrics is crucial.
In PC space, most users don't care about good gaming performance (and those who do often go for a dGPU), so an APU is a CPU that happens to have a GPU on it.
As such, it's judged as a CPU first, and a graphics chip second as a tie breaker.
AMD is not big enough to change that mindset in the PC space, where that's the way computers have been compared for many years.
In consoles, most users do care about good gaming performance, so it's a graphics chip that happens to have a CPU on it, so it's judged as a graphics chip first, and a CPU second, and comes out looking more favorable.

 

[mrmt]

In PC space, most users don't care about good gaming performance (and those who do often go for a dGPU), so an APU is a CPU that happens to have a GPU on it.

Isn't that because you can supplement GPU performance with a dGPU while you can't supplement CPU performance, as there are no substitutes for the CPU in a PC?

 

[Sweepr]

They basically overpromised and underdelivered on the promise of having a reasonably fast CPU next to mainstream class dGPU performance. There is a massive amount of gamers using iGPUs, but most of them are not exactly playing the latest X1/PS4 ports. The amount of gaming (mostly older titles and casual games) you can do with a lowly Bay Trail IGP is amazing and Haswell/Broadwell graphics (being much faster than BT's IGP) serves most people just fine. They don't need/care about faster graphics. Then there's the hardcore gamers who build/buy systems to play the latest X1/PS4/exclusive titles, and taking a look at CPU sales from Amazon even the ancient Vishera FX chips are more popular than APUs, which kinda indicates that hardcore gamers still look for a faster CPU + discrete graphics card solution. All that's left is a niche of users who want faster gaming performance but for some reason are not willing to add a dGPU anytime soon.

 

[coercitiv]

I'm rather curious to see the opinions of the people here on what are the APU value proposition strong points and why it failed on the market despite these strong points.

Memory bandwidth. Thy need to fix that to make it truly competitive.

 

[MisterLilBig]

Why have AMD APU's failed?

Horrible marketing towards their features to consumers. And nowhere near as good ($$$)partnerships with OEM's like Intel has. Last point is the most important one.

Did the high end mobile Kaveri even reach the market? I was totally waiting for that! But month after month of nothing in the market, I quit looking.


"Lack of CPU performance" is only an excuse used by fanatics. The same ones that will keep pushing single thread performance above all else.

I have a Nexus 5, its fast. It has a Snapdragon 800 SoC. My desktop has an A10-5800k, that's like 4-6 times faster, or more, than the Snapdragon 800, and it does everything I want, fast.

What am I trying to say? What Intel is great at, and it is great at that, does not matter to the average Joe and Jane buying the millions of PC's each year. But they do, because that's what they are fed, constantly.



(Which is why I think AMD should do a semi-custom type pre-order and pre-purchase program for the "hardcore" customers. Tho, that's what might be happening with the larger companies to their semicustom business.)

 

[jpiniero]

You have to realize that consumers aren't buying PCs. They are buying smartphones/phablets. Now, Corporate users are buying PCs, but they don't buy AMD and wouldn't be interested in games obviously.

Even if the GPU performance was better, it wouldn't make much of a difference.

 

[Vesku]

With laptops it's lack of variety and that they trickle onto the market slowly.

 

[monstercameron]

Marketing, bad products and stiff competition.

I have never seen an ad on tv for AMD
I have yet to see an Amd in a decent product
Also as much as I hate to admit it but Intel makes some dam good products.

 

[Soulkeeper]

The reason I havn't purchased trinity/richland or kaveri is because the CPU performance is underwhelming.
I liked the socket and motherboards, but the fact that they offer no 3 or 4 module options without a GPU is the killer for my use case.
Owning this llano overclocked 10% I usually wish it was an X6 without the GPU, and know another APU won't close that gap. AM3 just seems old and even seemed old when I got this llano a few years ago.

 

[monstercameron]

Yeah I have a review of high end mobile Kaveri, I'll post about it next week.

 

[coercitiv]

Did the high end mobile Kaveri even reach the market? I was totally waiting for that! But month after month of nothing in the market, I quit looking.

It did, shops in my country sell laptops from Dell, HP, Lenovo and Asus equipped with 35W or 19W Kaveri SKUs. I'll have to keep an eye on stocks to see if they sell or not.

 

[Blitzvogel]

If we're talking big die APUs, then a number of things really screwed AMD :

1. Process disadvantage - the root of AMD's problems
2. TDW disadvantage, especially in mobile where it matters
3. Performance disadvantage
4. Software ecosystem to really take advantage of the APU just doesn't exist. All that HSA hooplah hasn't meant anything really.
5. IMHO the iGPU is too big. Consumers who don't give a shit about gaming or graphics intense programs have to absorb the cost of the larger die. And current APUs don't really challenge dGPUs well enough to take a bite out of their market.
6. Not enough memory bandwidth to make the most of the iGPU
7. Poor AMD marketing
8. Lack of design wins.

I think Llano was a good product (the full 4 core, 320/400 SP models) simply because it was better positioned GPU and iGP wise to compete with what Intel had at the time. Intel just did not have an iGP that could come near it's level of performance. Despite power and efficiency issues, it could outperform the i3's in terms of overall performance on the CPU side of things but of course had the big iGP and was priced reasonably. It was a decent budget gamer chip as the 360 and PS3 had grown long in the tooth, and it's 400 SPs were enough to play multiplatform games reasonably well at 720p.

You could say that AMD set the bar too high with Llano, and the subsequent products have failed to truly succeed it. AMD could've cut the iGP size but that would've looked weak in the eyes of OEMs and consumers.

As for the Brazos platform and the Bobcat CPUs, it was a runaway success. It really did appear at the right time and at the right price for OEMs still hot on the boom of netbooks and cheap laptops. As a total product it collectively crapped all over Intel Atoms. AMD was also not at as bad a process disadvantage because Atom was IIRC still on 45 nm. AMD's biggest failure here was not capitalizing on it's success by releasing the quad core Krishna variant of Bobcat, even if it would've competed against low end variants of Llano. Jaguar also took took too long to succeed Bobcat which gave Intel time to get some more roots into the low end market which ate up much of AMD's potential OEM base.

 

[meloz]

1. AMD CPUs are still weak. Bulldozer is a gift that keeps giving.
   
   
2. AMD iGPU is not strong enough to matter in real world. The competetion is not Intel's iGPU, it is their and nvidia's entry level DGPU cards.
   
   
3. People who depend on iGPU find modern (IVB & later) Intel GPUs 'good enough'. Those who truly need a powerful GPU buy a dedicated card.

If AMD want APUs to be a successful they will have to pack in a much powerful iGPU. Powerful enough that it plays most common games at 1080p in low-to-medium settings.

But they won't do that, because doing so will hurt their dedicated GPU business.

So, essentially, AMD cripple themselves. Just like Intel cripples Atom line of products in fear it might become 'too powerful' and ruin their precious Celeron and Pentium profits.

 

[mrmt]

You have to realize that consumers aren't buying PCs. They are buying smartphones/phablets. Now, Corporate users are buying PCs, but they don't buy AMD and wouldn't be interested in games obviously.

Even if the GPU performance was better, it wouldn't make much of a difference.

Nvidia is breaking revenue record after revenue record, partially because of growth in the gamer market.

 

[frozentundra123456]

As much as I criticize the gaming performance (and I still feel the same way), the real problem is that there are not enough applications that use HSA. There are a lot of other problems also, such as low bandwidth, GF inferior process nodes, Bulldozer's problematical at best cpu architecture, etc. but it all could have been overcome if there were sufficient apps that could leverage HSA. The problem is that it is very difficult to drive software innovation when you have a small share of the market, and I am not a programmer, but it seems a lot of apps do not adapt well to gpu compute anyway.

 

[Abwx]

The amount of gaming (mostly older titles and casual games) you can do with a lowly Bay Trail IGP is amazing and Haswell/Broadwell graphics (being much faster than BT's IGP) serves a massive amount of people just fine. They don't need/care about faster graphics.

Apparently they dont care for CPU perfs either, assuming it s an Intel Baytrail, the lower performing part of the market, suddenly the level at wich CPU perf is deemed acceptable is much reduced to fit your double standard, well below AMD CPUs but thoses ones have no right to be good enough, isnt it...

 

[ShintaiDK]

Its quite clear that AMD havent learned a single thing since their K8 success. And its shining through all their segments.

The complete lack of focus on performance/watt is a huge problem for APUs, CPUs, GPUs and SoCs.

At the same time they waste precious resources on useless features like HSA and such that only looks good on paper. They was late to the integrated IGP as well: Completely wasting their ATI purchase in that regard instead of capitalising on it.

The uarch failure for the big core is again based on their feverish wishes of coming back into the server space. And the current cores in development shows they still havent learned anything and will just be even more painful in the future.

The small cores simply drops straight between 2 chairs. Its unfit for low power, and its unfit for performance.

And all them suffers from the too much IGP focus for something that isnt there to begin with. Its too slow to be useful. And its too fast (if you can say that) for the basic tasks its supposed to handle.

It just seems its a company completely out of touch with the consumers/OEMs.