The AMD Execution Thread [2007 - 2017]

[Kaotik]

AMD has just announced two more semi-custom design wins, to be launched in 2016. Obviously they can't say much about them but they did say that one in an x86 design an the other is ARM-based.

And I think they said something about $1 billion revenue over three years for those designs (presumably, combined).

Link?

 

[Alexko]

Link?

It was during their earnings conference call, I'm sure there will be a transcript available soon.

 

[Ryan Smith]

Here you go:

Relative to our new semi-custom design win pipeline, I am very pleased to announce we have secured two new wins, accomplishing our goal to close one to two new semi-custom wins this year.

These new semi-custom SoCs are expected to deliver combined total lifetime revenue of approximately $1 billion over approximately 3 years. Design work for these opportunities has started and we anticipate first silicon revenue in 2016.

Although I can’t go into details on the customers or the specific products, it is important to note that we are diversifying our semi-custom business beyond gaming. These wins also mark another significant milestone as one of the wins is our first 64-bit ARM based semi-custom design, building on our growing ARM64 momentum in the embedded and server markets.

 

[Gubbi]

Facebook or Google datacenter optimized systems ?

Cheers

 

[3dilettante]

If we go with the value Bob Feldstein's LinkedIn profile gave for the Xbox One deal, those contracts are individually on the order of 1/6 a console deal or smaller. Granted, a resume doesn't need to use the same metrics for arriving at a value.

The console deals themselves show that they haven't been able to significantly reverse the direction of AMD's finances, much less significantly smaller ones over a year away from producing revenue.

It's possible with this time frame that the ARM design could be using K12, similar to the console APUs being in development being based on Jaguar prior to that chip's release.

 

[Alexko]

If we go with the value Bob Feldstein's LinkedIn profile gave for the Xbox One deal, those contracts are individually on the order of 1/6 a console deal or smaller. Granted, a resume doesn't need to use the same metrics for arriving at a value.

The console deals themselves show that they haven't been able to significantly reverse the direction of AMD's finances, much less significantly smaller ones over a year away from producing revenue.

It's possible with this time frame that the ARM design could be using K12, similar to the console APUs being in development being based on Jaguar prior to that chip's release.

To be fair, depending on the nature of the new semi-custom deals, the margins could be significantly better.

 

[3dilettante]

At a billion over rougly three years, that's a third of a billion a year, or 1/12 of a billion a quarter.
$83 million a quarter is nice, but as we see the overall revenue picture has big money burners that can eclipse it.
AMD's projected revenue decline in 4Q could eat between 1-2 quarters of that revenue increase that won't even start until some 1-2 years from now if the wins don't get cancelled.
If AMD's board expects a lack of recovery in 2015 from a decline capable of erasing semicustom wins, it might explain why Read lost all three of his jobs.

To have semicustom wins of that magnitude rise above AMD's financial noise level, they would need to be in greater number or a big chunk of AMD's existing revenue goes away before they kick in.
Fearing the latter might get a CEO unceremoniously dumped.

 

[A1xLLcqAgt0qc2RyMz0y]

AMD's next high-end graphics delayed until the second half of 2015

From Pacific Crest Securities:

McConnell and Chung said AMD's next high-end graphics processing unit launch has been delayed until the second half of 2015.

http://news.investors.com/110314-72...etition.htm?ven=yahoocp&src=aurlled&ven=yahoo​

This seems to be a delay of at least a quarter or two (and maybe more).

By the time AMD has new offerings they will have to battle Big Maxwell (GM200/210) and soon after AMD has the new offerings out here comes Pascal.

EDIT: More color on AMD's GPU problems:

McConnell writes that price cuts of a third by AMD are not helping that company’s sales:

In line with commentary from AMD on its Q3 earnings call that management planned to competitively reposition its graphics card lineup in Q4, the company implemented 30% price cuts on its RADEON R9 290X and 290 graphics cards, with their comparable R9 290X card now priced $150 lower than NVIDIA’s GeForce GTX 980. Despite the comparable price discount in addition to game bundles, our supply chain conversations indicate minimal share shift back to AMD. Channel partners indicate that NVIDIA’s GTX 980 low-power offering (sub-200W board), which allows for high overclocking by gamers, as well as an ability to market the GTX 980’s support of Microsoft new DX12 API, are competitive advantages for NVIDIA despite premium pricing.

McConnell’s conversations with those in the supply chain world suggest the company is already taking share from AMD, he writes:

Despite sluggish overall demand for standalone graphics cards, supply-chain conversations indicate strong demand for NVIDIA’s two recently launched high-end desktop GPU cards (GeForce GTX 980/970) since their launch in September, with channel partners indicating material share gains for NVIDIA at the expense of AMD’s legacy RADEON R9 290/280 GPUs, which were launched in fall 2013. In terms of quantifying high-end share gains and magnitude, our supply-chain conversations indicate that NVIDIA’s GeForce GTX 980/970 has comprised over 80% of high-end card shipments to channel partners since mid-September. Mercury Research estimated that NVIDIA captured 67% unit share in the $300+ desktop graphics card market in calendar Q2. While the high end of the desktop GPU market (also referred to as the gaming enthusiast segment) is approximately ~10% of desktop discrete GPU market units, it garners 60% to 65% gross margin. We estimate the desktop discrete GPU segment comprises 30% of NVIDIA overall sales.

http://blogs.barrons.com/techtrader...re-from-amd-in-gpu/?mod=yahoobarrons&ru=yahoo

 

[Kaotik]

Those who don't click the link, McConnell and Chung are analysts, not working for AMD

 

[A1xLLcqAgt0qc2RyMz0y]

Those who don't click the link, McConnell and Chung are analysts, not working for AMD

You really think there are the Analysts working for AMD.

The role of the analyst is to analyze information about the company from the outside to come to proper conclusions as to where the company is headed. They dig out data external to the company and in doing so they are not influenced by company bias.

 

[Kaotik]

You really think there are the Analysts working for AMD.

The role of the analyst is to analyze information about the company from the outside to come to proper conclusions as to where the company is headed. They dig out data external to the company and in doing so they are not influenced by company bias.

I know, but without clicking the link, your post didn't in any way say they're analysts

 

[A1xLLcqAgt0qc2RyMz0y]

I know, but without clicking the link, your post didn't in any way say they're analysts

I did have the line "From Pacific Crest Securities" before the quote.

 

[Grall]

They dig out data external to the company and in doing so they are not influenced by company bias.

Well, in theory, anyway. Analysts are very often completely clueless and just speculate blindly based on their own set of bias and expectations (which are often nothing but echo chamber rumors.)

Makes one think if they're really nothing but veiled attempts at stock market manipulation...

 

[A1xLLcqAgt0qc2RyMz0y]

Well, in theory, anyway. Analysts are very often completely clueless and just speculate blindly based on their own set of bias and expectations ...

What's really funny is that these Pacific Crest Securities analysts were doing exactly that before "using their own set of bias and expectations" on Nvidia.

They were the longest running bears on Nvidia and for them to come clean (just before earnings) with their own data showing how wrong they were is astounding. Usually analysts who are proven wrong almost never change their direction and issue a major reversal. To do so must mean that the data they saw was without a doubt real and could no longer be ignored.

 

[lanek]

What's really funny is that these Pacific Crest Securities analysts were doing exactly that before "using their own set of bias and expectations" on Nvidia.

They were the longest running bears on Nvidia and for them to come clean (just before earnings) with their own data showing how wrong they were is astounding. Usually analysts who are proven wrong almost never change their direction and issue a major reversal. To do so must mean that the data they saw was without a doubt real and could no longer be ignored.

Is their report was about AMD situation ? for what i have seen they was initially done a report forecast about Nvidia .

Maybe they have got good sources and the info is right.. i dont know... As we have no information, no leaks, absolutely nothing, we cant say it is right or not.

 

[A1xLLcqAgt0qc2RyMz0y]

Is their report was about AMD situation?

Yes it also covers AMD's situation.

McConnell writes that price cuts of a third by AMD are not helping that company’s sales:

In line with commentary from AMD on its Q3 earnings call that management planned to competitively reposition its graphics card lineup in Q4, the company implemented 30% price cuts on its RADEON R9 290X and 290 graphics cards, with their comparable R9 290X card now priced $150 lower than NVIDIA’s GeForce GTX 980. Despite the comparable price discount in addition to game bundles, our supply chain conversations indicate minimal share shift back to AMD. Channel partners indicate that NVIDIA’s GTX 980 low-power offering (sub-200W board), which allows for high overclocking by gamers, as well as an ability to market the GTX 980’s support of Microsoft new DX12 API, are competitive advantages for NVIDIA despite premium pricing.

McConnell’s conversations with those in the supply chain world suggest the company is already taking share from AMD, he writes:

Despite sluggish overall demand for standalone graphics cards, supply-chain conversations indicate strong demand for NVIDIA’s two recently launched high-end desktop GPU cards (GeForce GTX 980/970) since their launch in September, with channel partners indicating material share gains for NVIDIA at the expense of AMD’s legacy RADEON R9 290/280 GPUs, which were launched in fall 2013. In terms of quantifying high-end share gains and magnitude, our supply-chain conversations indicate that NVIDIA’s GeForce GTX 980/970 has comprised over 80% of high-end card shipments to channel partners since mid-September. Mercury Research estimated that NVIDIA captured 67% unit share in the $300+ desktop graphics card market in calendar Q2. While the high end of the desktop GPU market (also referred to as the gaming enthusiast segment) is approximately ~10% of desktop discrete GPU market units, it garners 60% to 65% gross margin. We estimate the desktop discrete GPU segment comprises 30% of NVIDIA overall sales.

http://blogs.barrons.com/techtrader...re-from-amd-in-gpu/?mod=yahoobarrons&ru=yahoo

 

[lanek]

Ok my bad, i had only the Nvidia part .

 

[A1xLLcqAgt0qc2RyMz0y]

Ok my bad, i had only the Nvidia part .

No problem.

It can be a real pain to get the whole story (original analyst report is not public) since the financial press cut-n-pastes from the initial report sometimes leaving out whole sections. That was why you only saw the Nvidia part before.

 

[Kaotik]

The analysts seem to either have a bias, or lack of knowledge, as they're talking about new NV GPU's DX12 support, while in reality everything since Fermi from NV, GCN from AMD and Gen7 (or 7.5?) from Intel can declare the same thing as it is. Whether Maxwell will support full DX12 or not remains to be seen, but at this point it can't be declared.

 

[silent_guy]

The analysts seem to either have a bias, or lack of knowledge, as they're talking about new NV GPU's DX12 support, while in reality everything since Fermi from NV, GCN from AMD and Gen7 (or 7.5?) from Intel can declare the same thing as it is. Whether Maxwell will support full DX12 or not remains to be seen, but at this point it can't be declared.

The full DX12 requirements are not known yet, but when Microsoft talks up the new DX12 features during the Maxwell introduction, you can't blame analysts (or anyone, really) for making not of it.

Calling that bias says more about your's than that of the analysts...

It's a great rule of thumb actually: when looking for bias, start with the one who claims it's there. It rarely fails...