AMD Wants To Stop Being Known As The “Cheaper Solution”

[mrmt]

If AMD wants to move HSA into Android it would probably help to have a good ARM CPU for that process.

Unfortunately some of this will depend on AMD's Linux graphics drivers as well as other projects competing for development dollars.

Interestingly enough AMD consumer CPU business is posed to stay stable or lose revenue with Zen, this after the 70-80% drop caused by the Bulldozer family and the lack of investments on the cat family. That tells me that if HSA is to survive, it is probably in the form of a much more focused initiative, probably targeting servers or the professional market.

On a side not these premises of stability or small drop applies for the entire consumer business at AMD, meaning that their hopes for Zen APU and Zen FX on the consumer market are anything but high. Not only they don't believe in a high sales number, they also forecast almost neutral operating margins. Part of this is caused by the dismantling of the sales and marketing structures in the last few years, making almost impossible to find AMD products in certain markets, but part of it is caused by the products themselves.

 

[R0H1T]

Interestingly enough AMD consumer CPU business is posed to stay stable or lose revenue with Zen, this after the 70-80% drop caused by the Bulldozer family and the lack of investments on the cat family. That tells me that if HSA is to survive, it is probably in the form of a much more focused initiative, probably targeting servers or the professional market.

On a side not these premises of stability or small drop applies for the entire consumer business at AMD, meaning that their hopes for Zen APU and Zen FX on the consumer market are anything but high. Not only they don't believe in a high sales number, they also forecast almost neutral operating margins. Part of this is caused by the dismantling of the sales and marketing structures in the last few years, making almost impossible to find AMD products in certain markets, but part of it is caused by the products themselves.

Is it though? Intel bringing CPU/GPU shared memory pool with Skylake plus Gen9 graphics & Nvidia following suit with Pascal tells me otherwise. Maybe HSA isn't all hot air as you're making it out to be.

 

[ShintaiDK]

Is it though? Intel bringing CPU/GPU shared memory pool with Skylake plus Gen9 graphics & Nvidia following suit with Pascal tells me otherwise. Maybe HSA isn't all hot air as you're making it out to be.

HSA, haha...funny one.

Ever heard about Pixelsync and InstantAccess for example? HSA like the other GPGPU compute things isnt going anywhere anytime soon. And HSA is pretty much gone.

 

[R0H1T]

HSA, haha...funny one.

Ever heard about Pixelsync and InstantAccess for example? HSA like the other GPGPU compute things isnt going anywhere anytime soon. And HSA is pretty much gone.

It doesn't have to be go by the name of HSA does it, ever heard of Mantle & it's offshoots? I know how you dispute every positive thing AMD's ever done but so long as the concept works & lives on it doesn't even have to go by the name of HSA.

 

[ShintaiDK]

One of the things AMD should have stopped if they dont want to be associated with being cheap and rubbish is to stop selling chips like the AMD E1-2100 in PCs that goes straight from factory to the recycling plant.

But it seems AMD says one thing and does another.

 

[R0H1T]

One of the things AMD should have stopped if they dont want to be associated with being cheap and rubbish is to stop selling chips like the AMD E1-2100 in PCs that goes straight from factory to the recycling plant.

But it seems AMD says one thing and does another.

So long as there's market for cheap (& crap) we'll continue to see the likes of E1 being sold to the (uninformed?) masses, the same goes for certain sub 100$ phones & sub 200$ tablets that shouldn't exist even if they were given away for free D:

 

[mrmt]

Is it though? Intel bringing CPU/GPU shared memory pool with Skylake plus Gen9 graphics & Nvidia following suit with Pascal tells me otherwise. Maybe HSA isn't all hot air as you're making it out to be.

HSA isn't just unified memory pool, but also an entire development framework. Intel and Nvidia surely aren't pushing HSA-compliant hardware or development tools.

 

[mrmt]

It doesn't have to be go by the name of HSA does it, ever heard of Mantle & it's offshoots? I know how you dispute every positive thing AMD's ever done but so long as the concept works & lives on it doesn't even have to go by the name of HSA.

Mantle certainly isn't HSA, and it certainly didn't bring much benefits for AMD, so I don't see it as something positive at all.

So long as there's market for cheap (& crap) we'll continue to see the likes of E1 being sold to the (uninformed?) masses, the same goes for certain sub 100$ phones & sub 200$ tablets that shouldn't exist even if they were given away for free D:

There is no problem in selling things like E1, but when E1 become one of your most visible products, then you have a problem.

 

[R0H1T]

HSA isn't just unified memory pool, but also an entire development framework. Intel and Nvidia surely aren't pushing HSA-compliant hardware or development tools.

So you say but aren't they're all (the latter two) doing their own version of it? Intel with Skylake & Gen9 graphics accelerating apps through OpenCL, Nvidia with Pascal & Power9 working inside an HPC environment. If they're making provisions for it then I see no reason why a given software utilizing OpenCL (or any other suitable programming language) can't make use of preexisting features.

 

[R0H1T]

Mantle certainly isn't HSA, and it certainly didn't bring much benefits for AMD, so I don't see it as something positive at all.

Of course it isn't but it'll have the same effect that Mantle did, I doubt even you can deny that the latter had a profound impact on DX12, Metal, Vulcan et al.

There is no problem in selling things like E1, but when E1 become one of your most visible products, then you have a problem.

I wasn't advocating the sale of E1, maybe I should've clarified that OEM's are at fault for selling such cheap stuff & duping uninformed customers. AMD is also at fault for (still) making such low end stuff.

 

[mrmt]

So you say but aren't they're all (the latter two) doing their own version of it? Intel with Skylake & Gen9 graphics accelerating apps through OpenCL, Nvidia with Pascal & Power9 working inside an HPC environment. If they're making provisions for it then I see no reason why a given software utilizing OpenCL (or any other suitable programming language) can't make use of preexisting features.

Intel and Nvidia doing something means real business, AMD doing something is usually hot air. There's a massive difference in resource levels from AMD to Intel and a significant one from AMD to Nvidia. So just because these two companies are doing something doesn't mean that AMD should do it, let alone be able to deliver it. In fact, given AMD minute market share on the consumer and server market, it's better for their sake to go along Intel or Nvidia than to develop their own thing.

Of course it isn't but it'll have the same effect that Mantle did, I doubt even you can deny that the latter had a profound impact on DX12, Metal, Vulcan et al.

On DX12? Probably. Microsoft was sitting on DX12 for years because their gaming strategy is totally built around the XBOX, and the Khronos group is too incompetent to move things forward alone (I expect them to adopt Vulcan but fail miserabily on the market, like their always did), but Metal? Apple needs something to replace OpenGL on the iOS, that's not dependent on AMD at all.

And no, HSA won't have the same impact as of DX12 and OpenGL. In this case we have Microsoft actively sabotaging the PC platform as far as gaming is regarded, and the Khronos group is too incompetent to move things forward, so there was actually some interesting low hanging fruits to be taken. HSA is a much tougher nut to crack, as the software companies are too many and their needs too diversed, also there are much heavier hardware IHVs/IDMs to compete against.

Think about it, Mantle, a gaming API, died without a public SDK, probably because the thing was too broken for prime, do you think AMD can muster enough resources to develop something on par with what Microsoft, Intel, Oracle, Nvidia have been trying to push?

AMD has been chasing a lot of rainbows in the last few years, Semi-custom, HSA, Seamicro, Skybridge, ARM, HSA, Mantle, and from these only semi-custom generated fruits in the form of the console chips, and semi-custom is below-par of what they were saying in 2013. Lisa has been slowly bringing AMD back to its sense and killing/leaving to die these distractions and focusing on real, proven markets. I think HSA will eventually find this fate, or at least be severely reduced in terms of scope in order to fit on AMD R&D budget.

AMD financial projections for the next 9 months aren't exactly inspiring, I wouldn't be surprised if we see another round of cuts along with more projects being canned if things don't go that well, and this is a real possibility, given that every single node introduction wreaked havoc on AMD balance sheet in the last 8 years. HSA should be in the forefront in this case.

 

[R0H1T]

Intel and Nvidia doing something means real business, AMD doing something is usually hot air. There's a massive difference in resource levels from AMD to Intel and a significant one from AMD to Nvidia. So just because these two companies are doing something doesn't mean that AMD should do it, let alone be able to deliver it. In fact, given AMD minute market share on the consumer and server market, it's better for their sake to go along Intel or Nvidia than to develop their own thing.

That doesn't make any sense, Intel or Nvidia doing HSA is business while AMD doing it is just for kicks?

It's like saying heads I win tails you lose, no point in drawing this any further when you put it this way.

 

[mrmt]

That doesn't make any sense, Intel or Nvidia doing HSA is business while AMD doing it is just for kicks?

AMD cannot develop a relevant compiller for the x86 market, do you think they can develop an entire development framework?

 

[R0H1T]

AMD cannot develop a compiller for the x86 market, do you think they can develop an entire development framework?

Do they have to, what are the other partners for? If an initiative, like this, depends on the resources & time of a single company then it's bound to fail.
If AMD & their software/hardware partners aren't going to contribute in any meaningful way then it may well fail but that's also true for every other thing in life & business as well.

 

[ShintaiDK]

Always someone else needs to carry the burden and you wonder why it fails flat?

 

[R0H1T]

Always someone else needs to carry the burden and you wonder why it fails flat?

Not carry but share it, you think MS & Windows would be where they are without AMD/Intel or Android without Samsung, HTC, LG, Sony et al ?

 

[mrmt]

Do they have to, what are the other partners for?

Apparently AMD thinks they do, otherwise they wouldn't be crying to mommy FTC because Intel compiller does not run the optimum code path on their processors, and apparently AMD doesn't have the resources to develop this compiller, the results of their compiller are real bad.

If an initiative, like this, depends on the resources & time of a single company then it's bound to fail.
If AMD & their software/hardware partners aren't going to contribute in any meaningful way then it may well fail but that's also true for every other thing in life & business as well.

Not really. You need software partners to adopt and provide feedback to what you develop, but by no means you need your partners to do your job. Adobe uses CUDA, but by no means Adobe is developing CUDA for Nvidia, that's Nvidia job. Same with AMD, if AMD is unable to develop HSA then it cannot expect their partners to do their job.

 

[ph2000]

i dont see how following Intel or Nvidia way into compute will provide any benefit at all
intel heteregenous model is pretty much same like HSA, only AMD take it further by adding HSAIL while intel stick to openCL
Nvidia compute means CUDA, which is off limit to anyone

 

[Ketchup]

The A8-7650 Review just hit AT for those who haven't seen it:
http://www.anandtech.com/show/9217/the-amd-a8-7650k-apu-review-also-new-testing-methodology

I find it interesting that it is priced a hair below the i3 4130 ($15), but it performs SLOWER or very close in all of the tests they ran until they compared integrated graphics.

I guess the changes haven't started yet, or they think if they price it higher, people will just think it's better.

Here's what I would tell AMD : Remember when your chips were faster and cheaper than comparable Intel chips? Doing the opposite on both counts is not the way to make things better.

 

[R0H1T]

Apparently AMD thinks they do, otherwise they wouldn't be crying to mommy FTC because Intel compiller does not run the optimum code path on their processors, and apparently AMD doesn't have the resources to develop this compiller, the results of their compiller are real bad.

Does Intel compiler also makes OpenCL coded programs run slow(er) on Intel CPU's & if not that's your answer.

Not really. You need software partners to adopt and provide feedback to what you develop, but by no means you need your partners to do your job. Adobe uses CUDA, but by no means Adobe is developing CUDA for Nvidia, that's Nvidia job. Same with AMD, if AMD is unable to develop HSA then it cannot expect their partners to do their job.

Apparently Adobe is also betting on OpenCL a lot, maybe because of Apple, so they're certainly up for the job as & when the opportunity presents itself.

I have never said that AMD shouldn't do their part of the deal, in fact as the (HSA) lead founder they should push more resources into developing it further. However you know their financials & all such niche (as you & many others put it) initiatives will be put on the back burner, at least for the time being.

 

[OBLAMA2009]

man amd is such a joke, theyre the "cheaper solution" cuz thats how they perform. performance isnt important now anyway, its more about low power and stability

 

[ShintaiDK]

The A8-7650 Review just hit AT for those who haven't seen it:
http://www.anandtech.com/show/9217/the-amd-a8-7650k-apu-review-also-new-testing-methodology

I find it interesting that it is priced a hair below the i3 4130 ($15), but it performs SLOWER or very close in all of the tests they ran until they compared integrated graphics.

I guess the changes haven't started yet, or they think if they price it higher, people will just think it's better.

Here's what I would tell AMD : Remember when your chips were faster and cheaper than comparable Intel chips? Doing the opposite on both counts is not the way to make things better.

Not to mention 95W. No wonder nobody buys them.

 

[Ajay]

AMD has been chasing a lot of rainbows in the last few years, Semi-custom, HSA, Seamicro, Skybridge, ARM, HSA, Mantle, and from these only semi-custom generated fruits in the form of the console chips, and semi-custom is below-par of what they were saying in 2013. Lisa has been slowly bringing AMD back to its sense and killing/leaving to die these distractions and focusing on real, proven markets. I think HSA will eventually find this fate, or at least be severely reduced in terms of scope in order to fit on AMD R&D budget.

Speaking of Lisa Su, I finally watched her presentation on Analyst Day. When she put was talking about the 2015 roadmap and called it 'strong', my draw dropped. Honestly, to suck it up and say that to a bunch of analysts for the sake of shareholders takes some guts.

The 2014/2015 product line-up was gutted because 20nm was late and such a poor performer that most of the big players are just skipping it. Between having to rely on a third tier fab partner (in terms of execution) and the many unforced errors on AMD's part as you noted above (such as trying to expand their product scope even while R&D dollars were shrinking) - it's just amazing that they are still managing to scrape along.

I know I've said it before, but if Zen isn't a home run, AMD will have burnt their last match in the CPU business. They will become a different company after that. The stakes are real high and the pressure is on AMD's CPU team and on GloFo (which really needs to get 14FF right after what happened with 20nm). While I think it would be awesome if, after all that's happened, AMD succeeds with Zen; I won't be putting any money on it.

 

[Ajay]

Not to mention 95W. No wonder nobody buys them.

Had they performed as well as Haswell, nobody would have quibbled over a few watts (at least not in the consumer market). The killer is that AMD's reputation is so bad right now that people who aren't involved in forum's like this probably think AMD is ten years behind Intel. Reversing that image will not only require better products, but bring on a top Ad agency to improve their image (seriously, they can't afford not to). Prior to Bulldozer, AMD did some reasonable effective advertising (IMHO). Now, their brand image is in shambles.

 

[mrmt]

I have never said that AMD shouldn't do their part of the deal, in fact as the (HSA) lead founder they should push more resources into developing it further. However you know their financials & all such niche (as you & many others put it) initiatives will be put on the back burner, at least for the time being.

Yes, they again promised something and failed to deliver, and now will quietly move the initiative to a dark corner.

Speaking of Lisa Su, I finally watched her presentation on Analyst Day. When she put was talking about the 2015 roadmap and called it 'strong', my draw dropped. Honestly, to suck it up and say that to a bunch of analysts for the sake of shareholders takes some guts.

She has to say that. If she doesn't show confidence in the line up, who will? But OTOH they are planning a stable revenue even counting with the low bar set in their Q215 forecast, and they are expecting further OPEX cuts, so she isn't completely out of touch with the reality, and her actions aren't backing up this "strong line up statement". OTOH even this low financial forecast got a lot of flack in the Q&A.

I know I've said it before, but if Zen isn't a home run, AMD will have burnt their last match in the CPU business. They will become a different company after that. The stakes are real high and the pressure is on AMD's CPU team and on GloFo (which really needs to get 14FF right after what happened with 20nm). While I think it would be awesome if, after all that's happened, AMD succeeds with Zen; I won't be putting any money on it.

They are already a different company. They went from competing on all CPU segments, netbooks to 4P servers to being a niche companies competing in a few brackets of the bottom market, and their current plans don't envision competition across all segments. The question is whether they will need to shrink even further or if they will somehow stabilize their business.

I think K12 postponement is a hint that AMD will cancel it if demands doesn't materialize, just like they did with Skybridge.