GlobalFoundries: Only a process tech licensee?

[mrmt]

Nothing in the article says that it's due to performance reasons, only due to budget reasons.

If there was enough ROI the company would draw down debt if needed but would roll out Carrizo for desktops. The fact they chose not to is that the perceived trade off (additional debt) is forecast to be larger than the ROI of desktop Carrizo, and the reason for that is that it wouldn't be an increment in performance enough to fetch a given price and volume on the market enough to cover development costs.

In other words, Carrizo for desktop wasn't worth bothering to AMD management team, so they decided to can it.

 

[ShintaiDK]

AMDs desktop APU sales also dropped a staggering 30% in Q4. Not exactly a good market.

 

[Fjodor2001]

If there was enough ROI the company would draw down debt if needed but would roll out Carrizo for desktops. The fact they chose not to is that the perceived trade off (additional debt) is forecast to be larger than the ROI of desktop Carrizo, and the reason for that is that it wouldn't be an increment in performance enough to fetch a given price and volume on the market enough to cover development costs.

In other words, Carrizo for desktop wasn't worth bothering to AMD management team, so they decided to can it.

Yes, agreed and I'm not disputing that. AMD figure they can make more money on something else than desktop Carrizo. So if they have do downprioritize something, that's apparently it. But again, it does not say that it's because of performance reasons. There could be lots of reasons.

 

[Fjodor2001]

AMDs desktop APU sales also dropped a staggering 30% in Q4. Not exactly a good market.

No surprise. AMD's focus is on K12 and Zen 14 nm for desktop. You'll not see much improvement in AMD desktop sales until they are on the market in 2016.

 

[ShintaiDK]

I dont think AMD will make a desktop ARM chip. And its questionable if Zen will be anything more than a glorified cat core. Both are chips targetted for the server segment. I expect plenty of cores, but not so much on the performance side of each core.

 

[Fjodor2001]

I dont think AMD will make a desktop ARM chip. And its questionable if Zen will be anything more than a glorified cat core. Both are chips targetted for the server segment. I expect plenty of cores, but not so much on the performance side of each core.

Sounds like you're just guessing. Here is what AMD themselves have to say about it:

AMD hints at high-performance Zen x86 architecture

AMD has admitted that its Bulldozer microarchitecture was a misstep but claims that its next-generation replacement, Zen, will deliver the performance improvements required to become competitive with Intel once more.
[...]
The speech is the first time AMD has publicly talked about its next-generation x86 microarchitecture, now confirmed as being codenamed Zen. This will launch, it appears, alongside the K12 ARM-based desktop architecture the company has been planning, in order to 'capture [desktop] ARM before it happens.'

 

[seitur]

Well better Zen provide good single thread performance or AMD can kiss their x86 CPU business good bye.

It is kinda interesting that AMD focused on moar cores with Bulldozer when whole GPGPU idea of offloading certain tasks to GPU, will just strenghten need for strong single core and make many weak CPU cores less relevant.

 

[ShintaiDK]

You confuse the author with AMD.

Here is the goals for AMD.

 

[ShintaiDK]

Well better Zen provide good single thread performance or AMD can kiss their x86 CPU business good bye.

It is kinda interesting that AMD focused on moar cores with Bulldozer when whole GPGPU idea of offloading certain tasks to GPU, will just strenghten need for strong single core and make many weak CPU cores less relevant.

That is very ironic. but it also shows how little of value GPGPU compute really is in the grand scale. Very fast at very few things.

Looking at the R&D budget, that need to pay not only 1, but 2 uarchs that both are to compete with the best in their fields is rather...optimistic.

I would guess that Zen ends out with a singlethreaded total performance between current big cores and small cores but featuring plenty of cores and/or graphics while using HDL.

 

[mrmt]

Yes, agreed and I'm not disputing that. AMD figure they can make more money on something else than desktop Carrizo. So if they have do downprioritize something, that's apparently it. But again, it does not say that it's because of performance reasons. There could be lots of reasons.

Performance in its multiple dimensions (raw performance, perf/watt, etc) is what defines the value to consumer. If AMD is not launching the product then whatever performance gains they would be able to get wouldn't be able to pay back the investment.

It is not a question of performance per se, but performance within AMD's projected budget frame.

 

[Enigmoid]

Performance in its multiple dimensions (raw performance, perf/watt, etc) is what defines the value to consumer. If AMD is not launching the product then whatever performance gains they would be able to get wouldn't be able to pay back the investment.

It is not a question of performance per se, but performance within AMD's projected budget frame.

Exactly. You can bet that if AMD's design team stumbled upon an architecture (or some hidden 3rd party group sold one to them for peanuts) that had 50% greater IPC than haswell with no major caveats, AMD would draw down debt like crazy to get that on the market. Money would be 'found' to produce that die and market and sell it despite AMD having the same amount of funds. Its just that Carrizio isn't good enough on the desktop to justify the expenses, as I mentioned before this is because of HDL limiting clockspeed (undoubtedly carrizio will hit the clockspeed wall much sooner than kaveri). As AMD is reusing the die and IP from desktop to mobile, if the effect of HDL can be reduced to an acceptable level then desktop variants will appear.

Its a budget problem but that is why its a budget problem.

 

[Fjodor2001]

Performance in its multiple dimensions (raw performance, perf/watt, etc) is what defines the value to consumer. If AMD is not launching the product then whatever performance gains they would be able to get wouldn't be able to pay back the investment.

It is not a question of performance per se, but performance within AMD's projected budget frame.

You're drawing the conclusions too far. All has said is that they assumes they will get better ROI from other investments, and have to down-prioritize Carrizo desktop due to budget reasons. Anything else is just speculation.

 

[Fjodor2001]

You confuse the author with AMD.

No, the author of the article summed up the statements and quotes made by AMD.

 

[Fjodor2001]

--

 

[mrmt]

You're drawing the conclusions too far. All has said is that they assumes they will get better ROI from other investments, and have to down-prioritize Carrizo desktop due to budget reasons. Anything else is just speculation.

Desktop Carrizo was a cancelled project, not a project that was in conception competing with another projects for funding.

 

[AtenRa]

You know, all they have to do is take the same (literally) Carrizo chip they will use for the mobile platform and just put it in a FM2+ socket package in order to have a desktop SKU.

We may see FM2+ carrizo when Skylake appears

edit : Desktop Broadwell is not a thread to Kaveri but Skylake Core i3 and bellow.

 

[ShintaiDK]

You know, all they have to do is take the same (literally) Carrizo chip they will use for the mobile platform and just put it in a FM2+ socket package in order to have a desktop SKU.

We may see FM2+ carrizo when Skylake appears

edit : Desktop Broadwell is not a thread to Kaveri but Skylake Core i3 and bellow.

You know they cant due to the integrated FCH.

Also Carrizo on the desktop would run slower due to HDL.

The fact that AMD makes a Kaveri refresh pretty much seals the deal about Carrizo for the desktop.

 

[mrmt]

You're drawing the conclusions too far. All has said is that they assumes they will get better ROI from other investments, and have to down-prioritize Carrizo desktop due to budget reasons. Anything else is just speculation.

No, what was said is that the ROI to finish developing Carrizo was smaller than the cost of funding AMD would have to bear to finish it, and the only single reason for that is that the product would be a flop. Budget is by no means an unsurmountable restriction.

 

[AtenRa]

You know they cant due to the integrated FCH.

Also Carrizo on the desktop would run slower due to HDL.

The fact that AMD makes a Kaveri refresh pretty much seals the deal about Carrizo for the desktop.

1: Do you have any tech info that prohibits an SoC to be paired with a FCH ??

2: It may have lower clocks at 100W TDP but i dont think it will be slower at 45-65W TDP. Also Carrizo have more features and higher iGPU perf than Kaveri.
At 45W TDP a Carrizo SKU will be as fast or faster than 65W TDP Kaveri both in CPU and iGPU. So i dont see why people bring this HDL/low frequency myth all the time.

3. I am not convinced yet about the authenticity of that slide, there are SKUs in that slide that have lower frequencies than current Kaveri SKUs at the same segment, something you dont do on a refresh.

 

[Fjodor2001]

No, what was said is that the ROI to finish developing Carrizo was smaller than the cost of funding AMD would have to bear to finish it, and the only single reason for that is that the product would be a flop. Budget is by no means an unsurmountable restriction.

Let's say a company has 3 potential projects. Each will require $100M funding to complete. The company only has $200M, so it can only select 2 of the projects. Does that mean that one of the projects (the one not selected) has to have a negative ROI? No, it only means the other two are expected to have higher ROI.

So I repeat: All we know is that AMD find better ROI from other projects than desktop Carrizo currently, so they have to prioritize them due to budget reasons. Everything else is pure speculation.

 

[mrmt]

Let's say a company has 3 potential projects. Each will require $100M funding to complete. The company only has $200M, so it can only select 2 of the projects. Does that mean that one of the projects (the one not selected) has to have a negative ROI? No, it only means the other two are expected to have higher ROI.

If the three potential projects are good projects and have ROI rates higher than WACC and/or debt rates it will go to the debt market or to a bank, get a loan and complete the three. The fact that AMD has chosen to not to proceed with the desktop Carrizo tells us that forecast return rates are lower than whatever funding rate AMD could use.

 

[Fjodor2001]

If the three potential projects are good projects and have ROI rates higher than WACC and/or debt rates it will go to the debt market or to a bank, get a loan and complete the three. The fact that AMD has chosen to not to proceed with the desktop Carrizo tells us that forecast return rates are lower than whatever funding rate AMD could use.

It's not as simple as that. You never know for sure in advance what the actual turnout and ROI will be in the end. There are lots of companies that cannot get funding for their projects despite being potentially profitable.

Also, as was stated in the article, AMD is still debating whether to go ahead with a desktop Carrizo or not. So if the budget situation improves or they get additional funding, we might see Carrizo on desktop after all:

Interestingly he also mentioned that Carrizo still might end up on the desktop segment and AMD is currently debating this very question.

 

[ShintaiDK]

wccftech article with a lot of homemade speculation. Love the might, maybe, debating and so on.

Nothing but a clickbait site.

 

[Phynaz]

It's not as simple as that. You never know for sure in advance what the actual turnout and ROI will be in the end. There are lots of companies that cannot get funding for their projects despite being potentially profitable.

Monte Carlo Simulation

That's how companies decide to spend billions of dollars on a program.

 

[mrmt]

It's not as simple as that. You never know for sure in advance what the actual turnout and ROI will be in the end.

What made you think that the procedures I just described are remotely simple?