The AMD Execution Thread [2007 - 2017]

[Babel-17]

http://sample-files.com/video/hevc-ts-video-test-sample-files.html
"Beauty" kills my VLC Player, in software mode, but MPC HC handles it with aplomb, using only around 25% of my i7 2600. Afterburner shows no gpu usage, though it shows one short spike. I have a HD 7950. It's possible the anandtech video was using a tough to handle profile, instead of "Main" like "Beauty". Does my HD 7950 have a video decoder that Afterburner won't measure?
Edit: I found a tougher video with a higher bitrate and profile. It uses around 20% of my gpu and 40 - 50% of my cpu. Fifa_WorldCup2014_Uruguay-Colombia_4K-x265.mp4

 

[A1xLLcqAgt0qc2RyMz0y]

Another quarter in the books for AMD and another earnings disaster reported.

Advanced Micro Devices, Inc. (NASDAQ: AMD) has now released its first quarter earnings results. Maybe that should read “losses” rather than earnings, as the report still looks soft against estimates — and revenue guidance is not favorable either.

AMD’s non-GAAP operating net loss was $73 million, or -$0.09 per share, on revenue of $1.03 billion. The consensus estimates for the first quarter were -$0.05 in EPS and $1.05 billion in revenue. The same period in the previous year had $0.02 in EPS and $1.40 billion in revenue.

...

AMD went on to say that its Computing and Graphics segment revenue decreased 20% sequentially and 38% from a year ago. Its Enterprise, Embedded and Semi-Custom segment revenue decreased 14% sequentially and 7% from a year ago.

The company did release some general guidance for the coming quarter, and this is hurting worse than the lagging first quarter numbers. AMD sees revenues falling by 3% sequentially, plus or minus 3 percent. This implies that revenues would be roughly $999 million at the mid-point of -3%, or as low as $970 million or as much as $1.03 billion. AMD’s consensus analyst estimates for the quarter ahead are -$0.01 EPS and $1.13 billion in revenues. For a year over year comparison, AMD had a small gain of $0.02 EPS and $1.44 billion in revenues for the second quarter of 2014.

http://247wallst.com/technology-3/2...to-escape-earlier-shadow-of-intel-on-earnings

 

[A1xLLcqAgt0qc2RyMz0y]

AMD also has dumped SeaMicro.

http://www.anandtech.com/show/9170/amd-exits-dense-microserver-business-ends-seamicro-brand

As a part of the strategy to simplify and sharpen the company's investment focus, AMD is exiting the dense server systems business, formerly SeaMicro, effective immediately.

AMD recorded $75 million of special charges in Q1 2015 primarily related to impairment of previously acquired intangible assets.

Another $334 million wasted.

http://www.pcmag.com/article2/0,2817,2402063,00.asp

 

[3dilettante]

I wonder if the R&D that was going into that portion of the company was significant, and if that was part of the R&D drop that was reported earlier.
AMD exiting the business of selling the actual dense servers makes sense, although I think it wouldn't stop making the chips that would go into them. There's not much point to any likely chips coming from AMD if weak single-socket server is off the table. Whether that means the IP from Seamicro is dead, I don't know. It hasn't yielded much in terms of synergy (basically dead silicon in Seattle) or any further news, which seems grim.

The WSA with GF has apparently also been agreed upon for 2015, at a billion dollars.

 

[Florin]

Hmm, I wonder what happened to the widely touted Verizon deal that Seamicro had. For a while it sounded as if AMD was on track to become a large Intel OEM.

 

[3dilettante]

Being in that business seemed full of conflicts. Besides selling competing hardware, it put AMD at odds with potential customers for its own hardware.

Verizon's cloud efforts seem troubled, so perhaps we should keep an eye out for Verizon making a change there. Regardless, there are new products out in that space, and AMD has not shown up.

 

[silent_guy]

On the CFO commentary slides, they give a Q2 inventory guidance that's $100M higher than the current $688M. So instead of winding down, they're actually increasing inventory. What's up with that?

Nvidia inventories last quarter where $483M and that's higher due to ramping Maxwell products

Unless they're ramping up new products, I see another write-down coming...

 

[3dilettante]

Why not both? Lisa Su is promising a stronger second half on the back of new products that would start ramping earlier, and the aged inventory that AMD has failed to clear by now could be written down.

Per a quote on AMD's announcement:

"Under the backdrop of a challenging PC environment, we are focused on improving our near-term financial results and delivering a stronger second half of the year based on completing our work to rebalance channel inventories and shipping strong new products.”

 

[A1xLLcqAgt0qc2RyMz0y]

Lisa Su: "on completing our work to rebalance channel inventories”

Did she really say that again in the latest conference call?

If so that could explain the lower revenue outlook as old stale inventory has to be marked down to move.

 

[3dilettante]

It was from AMD's press release.
I thought they were already pricing things to move, with things like the drop in 290 prices earlier in the year.

 

[A1xLLcqAgt0qc2RyMz0y]

It was from AMD's press release.
I thought they were already pricing things to move, with things like the drop in 290 prices earlier in the year.

Which clearly did not work as seen by this quarters results.

 

[eastmen]

To be fair if the new amd flagship turns out as rumored ( new titan performance at $600) then a 290x at $360 and 290 at $300 are not good deals. I'd imagine they'd need the 290x at $200 to sell at that point and who knows the 290x may be replaced by something faster that uses less power and makes less heat.

 

[3dilettante]

I'd imagine they'd need the 290x at $200 to sell at that point and who knows the 290x may be replaced by something faster that uses less power and makes less heat.

It already has, which is why AMD's still working inventory down.
A massive chunk of the market is locked up for another upgrade cycle already. Even if Fiji is a strong solution, it will be an isolated success.

 

[sheepdogexpress]

To be fair if the new amd flagship turns out as rumored ( new titan performance at $600) then a 290x at $360 and 290 at $300 are not good deals. I'd imagine they'd need the 290x at $200 to sell at that point and who knows the 290x may be replaced by something faster that uses less power and makes less heat.

At the current pricing, even if it performs as rumors, the 300 dollar pricing of the 290x at 300 or so currently still makes sense with a 390x launching now.

E.g 390x at $600 will cost 100% more with 50-60 percent higher performance than a 290x. Heck if it offers titan x beating performance and a water cooler and 8 gb of hbm, 700 would not be out of the question, which would leave the 290x completely intact.

At this point, with all the financial woes and shrinking opportunity for AMD, all signs point this to be a bum launch like the gtx 480 or 2900xt where something is wrong right now because if they had the potential to give benchmarks in January(the leaks everyone is getting hyped about), they should have been able to launch the card alot sooner. There is a good chance the performance leaks were just FUD leaked by AMD to prevent Nvidia card sales.

This is turning into a bigger misfire than the gtx 480 launch. Atleast with that launch, 16nm cards were not shortly on the horizon. AMD now has a very short time frame with the need to sell alot of cards in a very competitive market where much of the market has already purchased cards. Heck the launch of a new windows won't help much this time around because of MS decision to give everyone with windows 7 and 8 free windows. This means far less new computer and upgrades. AMD has so poorly forecasted the market that it makes Nvidia look like they had an oracle at their disposal.

 

[eastmen]

Well hey everyone on tehse forums have been saying that the next node is on its way and they have been saying it for 4 years now. Who knows is 16nm will even hit in 2015 at this point.

 

[sheepdogexpress]

Well hey everyone on tehse forums have been saying that the next node is on its way and they have been saying it for 4 years now. Who knows is 16nm will even hit in 2015 at this point.

I don't think 16nm will hit this year, but it should next year.

What should happen I am guessing is 390x cards launch in mid July to August. That gives it solid 5 month period to sell. Then January 2016 rolls around and people having less disposable income after christmas, the buying rate of card lowers. Add in the Spark of rumors of 16nm cards coming about with a march launch, and AMD has a 5-6 months to really sell these cards before people stop caring about it because of other issues.

Add in its about time for a product refresh from Nvidia by the time AMD cards do come about. I.E 1080 series(same cores but higher clock) and AMD is in a situation where delaying the product never made sense and it's likely the 390x cards are simply not ready right now.

 

[silent_guy]

Why not both? Lisa Su is promising a stronger second half on the back of new products that would start ramping earlier, and the aged inventory that AMD has failed to clear by now could be written down.

Per a quote on AMD's announcement:

You're right. I missed that in the announcement and only just noticed it in the conf call transcript. (They're publishing the transcript on their own website. At least something where they are better than others.)

Still, $788M seems like a whole lot, yet they claim that they "remain on track to return to normal inventory levels by the end of the second quarter." I guess they are much more diverse than Nvidia and that this makes it harder to keep inventories low.

Edit: I should learn to read the whole thing before commenting: the Q2 increase in inventory is apparently a seasonal thing for semi-custom (read XB and PS).

Another interesting point is where they are focussing their R&D: "we are, in a very targeted manner, investing in the enterprise, embedded, and semi-custom area". AFAIK consumer CPU and GPU are not part of that, but maybe Fiji is as a compute part?

 

[pTmdfx]

interesting point is where they are focussing their R&D: "we are, in a very targeted manner, investing in the enterprise, embedded, and semi-custom area". AFAIK consumer CPU and GPU are not part of that, but maybe Fiji is as a compute part?

If they cannot monetise enough in those segments (or want more profits), they can always "fallback" and sell a consumer-focused SKU of those so-called "EESC investments" after all. Moreover, investing into a profiting business group always sounds better than that into a business group with so far endless loss. So this sounds just language arts. But it might also have an implication that they would get chip designs that matters more in server and embedded out first, rather than plain consumer mainstream APUs.

 

[silent_guy]

I doubt that it's just PR speak. They also said that they are exiting segments with low margins, but didn't elaborate which segments exactly. (Low end discrete GPU?)

I don't think they're exiting consumer desktop GPUs altogether but with a market share that's starting to resemble their professional GPU market share (which they claim keeps increasing) it makes sense that they're starting to cut back.

So how much how much R&D out of $350M pie do they get? Against an Nvidia that spends way more on just GPUs alone.

 

[Kaotik]

They're definitely not exiting desktop GPU's, since it walks hand in hand with both APUs and OEM GPUs, and GPU's is the one thing that's bringing money in when something is