The AMD Execution Thread [2007 - 2017]

[UniversalTruth]

Manufacturing processes and "Microsoft" aside, the Bulldozer family remains plagued by poor per-thread performance.

Yeah, let's just say that Bulldozer family is designed for a different world, or just ahead of its time.
The design is ok, the environment didn't change accordingly.

As for Microsoft and Nokia, is there a phone from them using an Intel chip?

 

[silent_guy]

Yeah, let's just say that Bulldozer family is designed for a different world, or just ahead of its time.

It's ... it's a forward looking design!

The design is ok, the environment didn't change accordingly.

The design is noble in a world of peasants.

 

[3dilettante]

It kinda uses it, they have their Surface Pro line.

I forgot about the Surface Pro.
It's a rather steep price to pay for cost savings on the CPU component. I haven't seen sales numbers, but it was claimed that Surface had broken a billion dollars in quarterly revenue.
The most expensive models retail in the thousands of dollars, but the mix is unclear. Intel can charge over $600 for the i7 CPU used in the top tier, but Microsoft wouldn't be paying retail.
Intel's margins take some sliver of that billion dollars, which should give a ceiling to what the Intel tax may be. It may take a rather long play to recoup an acquisition that will cost billions in one form or another, not including the assumption of AMD's cost of doing business for every other product it makes and loses money on, and the legal thicket this would crash into.

 

[entity279]

Mentioned that for the sake of completeness. Personally I do find that device appealing, but it does appear not to be at the core of MS's business for the moment.

Maybe in the future, should they plan to go bigger & aggressive on the tablet front...

Further I'm not sure weather they have a strong expertise in RF. They should have some due to the Nokia purchase. If not enough, they maybe need to invest in this segment first.

 

[AlexV]

It's ... it's a forward looking design!

I know. It looked so far into the future that it caught sight of the time after the end of technology, after regression into a primordial soup. It totally rules there...as a way to maim amorphous enemies. Universal Truth, seriously, you must be trolling...

 

[Albuquerque]

In multi-threaded applications, Vishera (FX-8350) is within 10% of Haswell (i7-4770) so it's a reasonable assumption. Games and other somewhat serial applications are a very different story, although DX12 may help with that.

I think we need to be more clear here: under very rare and specific multithreaded circumstances, the 8350 might be able to come within 10% of a 4770k. If considered within the context of the "general 80/20 rule", the 8350's ability to pull that kind of scoring is firmly in the "20" category. Very low in that 20 category, at that.

References: (I'm ignoring gaming in all instances)
BitTech's review of the 8350 shows it consistently below the i7-3770k, but that's at least partially due to the benchmarks chosen.

Anand's review of Vishera shows three distinct wins for the 8350: 7zip, h264 encoding (2nd pass) and POVray. All are good indicies for multithreaded performance, but other multithreaded apps are won convincingly by the i7-3770.

Toms Hardware only shows the 8350 winning in a single benchmark, and only by a tiny margin compared (again) to the 3770k. To be fair, it's solidly a second place contender.

There are more examples online, but they follow mostly the same pattern: the 3770k was the competitor, and the FX could only claim a scant few wins of that era. Later benches of the 4770k show examples where it's beating the 3930k in multithreaded benches, continuing to pull away from the much older FX 8350: Tom's, Anand, and Bit-Tech again. The BitTech and Tom's reviews actually include the 8350, and I think I only saw a single benchmark where it could be said that it was within 10% -- and only just by rounding error.

Also remember that Intel makes 8-core, 16-thread "desktop" processors. They're not cheap, but that's really not what was asked. They also make far more reasonably priced 6-core, 12-thread "desktop" processors.

 

[Alexko]

I think we need to be more clear here: under very rare and specific multithreaded circumstances, the 8350 might be able to come within 10% of a 4770k. If considered within the context of the "general 80/20 rule", the 8350's ability to pull that kind of scoring is firmly in the "20" category. Very low in that 20 category, at that.

References: (I'm ignoring gaming in all instances)
BitTech's review of the 8350 shows it consistently below the i7-3770k, but that's at least partially due to the benchmarks chosen.

Anand's review of Vishera shows three distinct wins for the 8350: 7zip, h264 encoding (2nd pass) and POVray. All are good indicies for multithreaded performance, but other multithreaded apps are won convincingly by the i7-3770.

Toms Hardware only shows the 8350 winning in a single benchmark, and only by a tiny margin compared (again) to the 3770k. To be fair, it's solidly a second place contender.

There are more examples online, but they follow mostly the same pattern: the 3770k was the competitor, and the FX could only claim a scant few wins of that era. Later benches of the 4770k show examples where it's beating the 3930k in multithreaded benches, continuing to pull away from the much older FX 8350: Tom's, Anand, and Bit-Tech again. The BitTech and Tom's reviews actually include the 8350, and I think I only saw a single benchmark where it could be said that it was within 10% -- and only just by rounding error.

Also remember that Intel makes 8-core, 16-thread "desktop" processors. They're not cheap, but that's really not what was asked. They also make far more reasonably priced 6-core, 12-thread "desktop" processors.

My comment was based on this Hardware.fr review: http://www.hardware.fr/focus/99/amd-fx-8370e-fx-8-coeurs-95-watts-test.html

The FX-8350 scores 115.2 while the i7-4770K gets 126.2 (9.5% higher) on the performance index, which is an average of all the applications they tested, games excluded.

More specifically : [ 3ds max 201- MR 3.12 ] [ 3ds max 2015 - Vray3 ] [ Visual Studio 2013 ][ MinGW-w64 - GCC 4.7.1 ] [ WinRAR 5.10 ] [ 7-Zip 9.20 ] [ x264 v2453 ] [ x265 v1.2+507 ][ Lightroom 5.5 ] [ DxO Optics Pro 9.5 ][ Stockfish 5 ] [ Houdini 4 Pro ]

 

[pcchen]

My comment was based on this Hardware.fr review: http://www.hardware.fr/focus/99/amd-fx-8370e-fx-8-coeurs-95-watts-test.html

The FX-8350 scores 115.2 while the i7-4770K gets 126.2 (9.5% higher) on the performance index, which is an average of all the applications they tested, games excluded.

More specifically : [ 3ds max 201- MR 3.12 ] [ 3ds max 2015 - Vray3 ] [ Visual Studio 2013 ][ MinGW-w64 - GCC 4.7.1 ] [ WinRAR 5.10 ] [ 7-Zip 9.20 ] [ x264 v2453 ] [ x265 v1.2+507 ][ Lightroom 5.5 ] [ DxO Optics Pro 9.5 ][ Stockfish 5 ] [ Houdini 4 Pro ]

One has to be very careful when using averages like this. In this case, out of 12 tests, FX-8350 is "faster or within 10% slower" in only 7 of these tests. Not exactly rare, but still.

 

[Alexko]

One has to be very careful when using averages like this. In this case, out of 12 tests, FX-8350 is "faster or within 10% slower" in only 7 of these tests. Not exactly rare, but still.

Sure, but I only said that it was within 10% (implying an average) not "faster or within 10%" which would have been misleading indeed.

 

[pcchen]

Sure, but I only said that it was within 10% (implying an average) not "faster or within 10%" which would have been misleading indeed.

It's basically the same in this case though, as FX-8350 is never more than 10% faster than 4770K in these tests.
What I wanted to say is using average can be misleading as "generally within 10%" gives an impression that the performance of both processors are similar (within 10%) in most case, but it's only the case 60% of the time.

 

[Alexko]

Fair point. Allow me to rephrase, then: the FX-8350 is usually slower, but only by about 10% on average.

 

[Rys]

Shame it's at much worse energy efficiency. I'd like to buy AMD, and I would if it was close enough in major metrics, but they have no energy efficiency story in the high end, and that's a deal breaker.

 

[pharma]

Hardware report card: Nvidia vs AMD

http://www.pcgamer.com/hardware-report-card-nvidia-vs-amd/#page-1

 

[UniversalTruth]

Shame it's at much worse energy efficiency.

But you know that the reason for this is the aging 32 nm process. AMD don't have any access neither to 20 nm GloFo, Samsung, nor to 20 nm TSMC, even less so to the manufacturing capacity of Intel.

It is stuck from everywhere.

What can they do?

Let's only pray that the transition to 14/16 nm next year at the latest will happen according to some sane justice in this world.

 

[Alexko]

But you know that the reason for this is the aging 32 nm process. AMD don't have any access neither to 20 nm GloFo, Samsung, nor to 20 nm TSMC, even less so to the manufacturing capacity of Intel.

It is stuck from everywhere.

What can they do?

Let's only pray that the transition to 14/16 nm next year at the latest will happen according to some sane justice in this world.

It's far from the only reason, as a comparison between late Phenoms and Bulldozer will illustrate.

 

[Albuquerque]

Well, anyway,

Fair point. Allow me to rephrase, then: the FX-8350 is usually slower, but only by about 10% on average.

Well, but that too is only specific to incredibly multithreaded scenarios. In scenarios where you're not balls-to-every-wall multithreaded (all cores, all the time) the Intel side wins by an epic margin due to IPC. It's a sad state, really.

 

[pharma]

ASMedia Technology, which has received orders from AMD for its southbridge chip R&D and manufacturing, may feel the impact of AMD's poor peformance in the PC industry recently, according to sources from the upstream supply chain.

To save costs, AMD decided to cooperate with ASMedia in 2014 to license ASMedia's SATA Express patents and outsources its chip R&D and manufacturing to the chip designer.
In 2014, ASMedia achieved net profits of NT$228 million (US$7.38 million) and EPS of NT$4.04, and expected its revenues and profits to be driven by USB 3.1 demand in 2015.

However, for the first quarter, ASMedia only achieved EPS of NT$0.64 and combined consolidated revenues for the first five months of 2015 were down 11.2% on year to reach only NT$585 million and had a monthly EPS of NT$0.12 in May.

http://www.digitimes.com/news/a20150703PD209.html

 

[3dilettante]

Is ASMedia's mix that heavily leaning towards AMD?
One of AMD's problems is a decline in the overall markets it is present in, in addition to the fact it is doing badly in them. My first reaction was wondering how AMD's reduced presence wasn't already factored in, but I suppose the market they overlap in could be in the lower end.
The quiet mention at the end that Intel is aggressively working to make ASMedia unnecessary seems important.

 

[Silent_Buddha]

One of AMD's problems is a decline in the overall markets it is present in, in addition to the fact it is doing badly in them. My first reaction was wondering how AMD's reduced presence wasn't already factored in, but I suppose the market they overlap in could be in the lower end.

Its major market segments currently are also markets with little recovery possibilities. As people looking at low performance devices increasingly look to Tablet devices, even for users that remain with the Windows platform. That means that low performance at a good price isn't as important as low performance combined with good power use characteristics.

Intel basically owns the mid level and higher market segments. AMD's best hope is that Zen can offer significant competition in the mid level market and/or have stellar power characteristics to be competitive in the changing landscape of the low performance market. I'm rather doubtful they can be competitive in the high end market.

A darkhorse potential is whether HBM could enable a niche market for gaming centric all-in-ones enabled by an AMD SOC with a high performance GPU core coupled with games designed for DX12 (not just DX 9/10/11 with DX12 tacked on).

Regards,
SB

 

[A1xLLcqAgt0qc2RyMz0y]

AMD Warns on 2015Q2.

http://www.anandtech.com/show/9418/amd-confirms-20nm-products-moved-to-finfet-warns-on-q215-earnings

AMD is looking to brace investors for a worse than expected Q2, after an already difficult Q1. Soft APU sales are being blamed for dragging down both revenue and gross margins, with AMD now expecting Q2 revenue to be down 8% sequentially, or around $950M, while the non-GAAP gross margin will be just 28%.

Also reports another one-time-charge of $33 million on moving (delaying) products from 20nm to 16FF.

AMD had previously announced their intentions to bring out some products at 20nm – these were most likely just APUs, with the only one we explicitly know about being the now-canceled Skybridge. In any case, AMD is now confirming that they have moved several of their 20nm designs to a “leading-edge FinFET node,” and as far as we know AMD no longer has any further 20nm projects in the pipeline.