AMD CEO Teases Zen Processor Family, Summit Ridge Makes An Appearance

[jimmysmitty]

True but if they perform similar they wont charge less and to me they need a real winner in the server market before the consumer market.

When given the choice between the under-dog brand and the leading brand from the past several years for the same price, most people will choose the leading brand and completely ignore the under-dog. If AMD wants to regain market share, they will need to under-cut Intel. At price and performance parity, most of AMD's sales would go to AMD loyalists and not expand AMD's market share by much. Same goes for server chips: given the choice between AMD servers and Intel servers that cost and perform about the same, most companies would pick Intel.

When you lack the benefit of leading reputation, trying to charge the same as leading brands for a nearly identical product is often suicidal. It is much of the same between AMD and Nvidia: ~75% of gamers are going with Nvidia despite AMD GPUs with similar performance costing $50-100 less.

They have and will. The Athlon 64 cost equal to or more than Intel at the time, they charge appropriately for their GPUs even with 75% market share the FuryX launched at 980Ti pricing and only dropped much later on.

Their marketing is not strong but that has never stopped them from pricing it where it performs.

If Zen performs equal to a 6900K it will be priced equal to a 6900K, they will not price it at a i5 6500K level.

 

[InvalidError]

They have and will. The Athlon 64 cost equal to or more than Intel at the time

Keep in mind that "at the time", Intel was desperately trying to polish the Netburst turd. There isn't much merit in achieving premium pricing when there is effectively no competition. That's basically the same situation Intel is currently in with massive premiums due to AMD being a non-concern across most of the board.

In times were both AMD and Intel were in the same performance ballpark, AMD's chips were usually priced around 10% lower.

 

[jimmysmitty]

They have and will. The Athlon 64 cost equal to or more than Intel at the time

Keep in mind that "at the time", Intel was desperately trying to polish the Netburst turd. There isn't much merit in achieving premium pricing when there is effectively no competition. That's basically the same situation Intel is currently in with massive premiums due to AMD being a non-concern across most of the board.

In times were both AMD and Intel were in the same performance ballpark, AMD's chips were usually priced around 10% lower.

Before Athlon 64, sure that was true. But even in recent history AMD has priced to start high, where they wanted it to compete, and lowered it only when the performance didn't match. The FX 8150 launched at mainstream i7 levels and only dropped when the performance was not enough to sell. The FX 9590 launched with a $1K price tag yet is now cheaper than an i7 mainstream part.

Phenom 9900 launched at $350 but was outperformed by the Q6600 that was cheaper at the time.

I am not saying it is an absolute but AMD is a business first and foremost and is what Lisa has said is true, that they don't want to be the cheaper option anymore, then I do not expect AMD to price their CPUs cheaper with near or like performance. If they have similar performance then they will price similarly, maybe undercutting and starting a price war that they really do not want to do with Intel as Intel has the headroom for it.

 

[InvalidError]

I am not saying it is an absolute but AMD is a business first and foremost and is what Lisa has said is true, that they don't want to be the cheaper option anymore

If you look at what AMD announced it is going to do on the GPU side of things, selling high clearly isn't what it is planning to do in the mainstream where the bulk of revenue comes from. The RX-480 is priced to hopefully move significant volumes instead of large gross profit margins.

Having a new architecture does you little good if you don't give people a reason to seriously consider it over the previously established leader. If you did a poll on THG asking if people would switch from AMD to Intel, Intel to AMD, stick with Intel or stick with AMD if there was price parity between the two for a given performance, I bet most would stick to whatever brand they are already using or ditch AMD for Intel as they were already been planning to prior to Zen's launch. I know it would take more than CPU price and performance parity to convince me to try Zen since my past experiences with AMD-based systems is that AMD's chipsets have ~10% worse IO performance.

 

[somebodyspecial]

I really hope AMD looked at 300mm^2 and shot for that. If it's less than 200 this whole thing is pointless and will just be a better crappy chip.

Skylake i5/i7 for LGA1151 are only 122mm^2 including the IGP. If AMD wants to be profitable, they have to beat Skylake on bang-per-buck and they have to do so at a lower silicon cost if possible to improve AMD's gross margin. If AMD went for 300mm^2 for mainstream chips, they would end up outright pricing themselves out of the market.

There is no way you'd price yourself out of the market. IF that was the case NV/AMD couldn't put out 600mm^2+ gpus and sell them on $400-600 cards right? We're talking an entire CARD, with lots of memory, PCB, design to get the whole thing working, etc so it's a pretty cheap 600mm^2 chip correct? If a 100mm^2 Tegra (insert soc name here) can be made for $10-20 and sold for $25-35 material is not much of a problem (IE, Tegra K1 sold in Xiaomi mi pad for $28 according to their whitepaper, and MS got T4 for surface for under 25, Google estimated $23 etc - T4 was ~100mm^2) . It is DESIGN that costs all the money. K1 was measured at 121mm^2 IIRC (but not sure denver or regular here, either way, CHEAP, X1 is supposedly bigger than 121 but can't prove that AFAIK and it's in $199 Shield TV). NV is making money selling it for $28 to a small company like Xiaomi (small order that is compared to someone like google/msft). The chip was made for $15-18 supposedly. Once you're done with design and pumping the crap out it's cheap as chips as they say. Sure yields come in to play etc, but it would be very tough to price yourself out of a market at 200-300mm^2 when it has been done many times before (and larger for both sides actually).

http://www.anandtech.com/show/9582/intel-skylake-mobile-desktop-launch-architecture-analysis
Note the size of Ivy, Haswell E etc (vishera too...).

http://www.anandtech.com/show/6396/the-vishera-review-amd-fx8350-fx8320-fx6300-and-fx4300-tested
Vishera 8 core debut, 315mm^2, sold for $153-$190. Umm...Do you think they sold at a loss from day one? I don't. The problem here was 8 core instead of being a 315mm^2 QUAD! This shows in the benchmarks of course as games barely use 4 efficiently for the most part. It's far easier to clock a quad higher than eight cores also as seen many times. This is mainstream correct? I could end here, but lets just run down a bunch of data so we bury this topic for good.

http://www.anandtech.com/show/8426/the-intel-haswell-e-cpu-review-core-i7-5960x-i7-5930k-i7-5820k-tested
Haswell 6core lowest price was $389 (5820). Even the step up 5930 was only $589. In order for AMD to get back into the high end they should aim higher. You can make the junkers (bang for buck crap) later, but you need a king now. When ivy 6c came out they sold for $555/990 (257mm^2). My guess is they're made (material) for well below $100 & for Intel likely for less than $60 (as they don't pay a fab like AMD/NV), or how can you explain AMD's 315mm for $153? Even if you said $153 was the actual material cost (no way), this could be sold for $300 easily right? I mean assuming you were outperforming Intel that is, which you would (for gaming at least 95% of the time) if it was a FAST quad, instead of slow 8. AMD's problem was it got smashed in almost everything vs. Intels quad, as not much uses 8 well. Even the 9590 from AMD was only going for $300-370 with the $370 having liquid cooling as shown in the article here:
http://www.anandtech.com/show/8316/amds-5-ghz-turbo-cpu-in-retail-the-fx9590-and-asrock-990fx-extreme9-review

As a buyer, of course I want a low price, but we're talking about the company and their perspective here for getting profits. We're not talking nearly identical here either. We are talking about smashing the competition, like Intel does now to AMD. There are many games where Intel is >50% faster than AMD's top stuff. As such Intel charges FAR MORE than AMD. That was kind of the point of the LARGE QUAD here. Beat them like they did before for 3yrs (and like Intel does to them now), so you can charge more and get back to 50%+ margins. Intel has 62% margins for a reason (nvidia has almost 60% now too) and neither has a die that prices them out of the same markets AMD is in. Both make PROFITS, while AMD loses money (for various reasons). So the bang for buck garbage has worked out how well for the last 13yrs? Oh, right, 7-8B is losses. Yeah, it's a great way to go 😉

Another point, the king isn't mainstream is it? Although $300 and under to me these days is mainstream...LOL. Enthusiast chips to me are above this usually. It's top end for me (my budget can afford more, but I just OC to get there anyway - I'd rather put extra cash into gpu), but not for enthusiasts. I'd expect AMD to put a price of $400-1000 (roughly, you get the idea) on the top two models, and down from there from there just like Intel. As production ramps and yields go up release lower models for less of course (and defective ones end up as tri-core etc as usual).

http://hothardware.com/reviews/intel-core-i7-6700k-and-z170-chipset-review-skylake-for-enthusiasts?page=8
If they were doing this to Intel, instead of the other way around, enthusiast would pay, just as they do now for Intel. I'm sure someone will argue over the res they used to show the difference, but as vid cards increase in perf, that is exactly what you'd be seeing down the road as all cpus end up being your limiter at some point in their 5-7yr lifetime (for most people, which is why a lot of us buy a new one right?).

https://www.guru3d.com/articles_pages/core_i7_6700k_processor_review_desktop_skylake,16.html
Same story here vs. 8370 AMD.

https://www.guru3d.com/articles_pages/core_i7_6950x_6900k_6850k_and_6800k_processor_review,16.html
For the whiners about res...1080p, still 135fps for i7-6700 to AMD 100fps 8370 with a GTX 1080, illustrating my point already. DX12 didn't change much in this case and certainly isn't going to save slow cpus. You have to go above 1080p to bring Intel back down to earth here and 95% of us play 1920x1200 or below still. Toms article on this chip shows Ashes and F1 2015 use the cores, so there is hope for more cores I guess. So far, most programmers have taken the easy route. Apple's die is always larger than anyone for socs, and fewer cores, so I still say a huge 4 is best today at least and will always do pretty well regardless of how the programmer goes. I'd rather buy what wins 95% of today's stuff rather than someone claiming "future proof will win at some point". People buy what the review says TODAY.

From Toms article on the 6950x:
"And indeed, we know that many games simply cannot utilize as many cores as Ashes and F1."

"We measure precisely the way you play your game, with proper image quality settings. As history has proven, faster cores matter more than more cores. Basically, once you pass four cores you'll be hard-pressed to find substantial enough differences."

The above quote from guru3d says it all right? 1fps for 6800 vs. 6700 and only 1fps extra for 6850. Yeah, wasted on us gamers. The 6900/6950 (8 and 10 cores...LOL) actually slow things down due to overhead or something (in some stuff at least), losing a few fps. OUCH. I'm shocked AMD didn't learn from Vishera etc. Superfast QUAD rules gaming (for now). Hitman on the next page shows the same for dx12, but worse. 6700 wins @1080p. Hitman everything scores the same from 4-10 core from Intel at 2560. AMD pretty much lost me when they said 8 core. 🙁 I'd probably go for a 6 core 200mm+ but...AMD seems to have blown it again going 8 when gamers need a FAST 4 core.

Forgot Intel was 122mm^2 now but you're talking 4 core here while Intel has 6-10 core on 246mm^2 sans gpu (these are the 8 core AMD competition IMHO), but I'd still say put out a 200mm+ or bust. But alas, it looks like AMD went 8 core so it won't make a dime anyway most likely (should have listened to DIRK, but they let him go in 2011...LOL). If it's "good enough" Intel will just price it to death. At 8 core, it is unlikely to smash Intel due to games just not using them well and Intel likely putting out Kaby by then (Q1?). IF it was a big quad, they would have made some real money probably. Gamers flocked to AMD before and paid enough one year for AMD to make a billion in profits (in 2000 - almost 1B that is). It's funny Dirk was fired for not wanting to chase tablet/phone (felt it would divert R&D from CORE CPU/GPU), and we see how well it worked so far for Intel with 4B+ losses per year and NV with tegra still not making profit yet. Though I think NV will soon make profit on Tegra as Cars blow up big time and die shrinks allow a modem from someone else (samsung/qcom/Intel) to be included without hurting power leading to NV going back into mobile with success.

https://www.sec.gov/Archives/edgar/data/2488/000101287001001228/0001012870-01-001228-0013.txt

My how times have changed since the started chasing the low end stuff. Lost 2B doing this in the last 5yrs alone. Chasing consoles was just as dumb since they had no margins for such low production. AMD said single digits for a long time, and mid teens according to AMD now so consoles were the reason they couldn't R&D a winning cpu (bullsnozer delay etc). The chips were $100, so a single digit margin is $10 tops. That won't make you rich over 20mil/yr sold basically. It barely pays the interest on AMD's debt yearly as we've seen in the Q reports. Any profits they had recently came from layoffs, not really consoles.

Another point on cost: NV's GTX 1060 gpu is ~200mm^2 and is sold on cards that appear to be priced $180-$215. Again, this includes your PCB, mem, etc etc. Also note unlike selling a CPU direct, the gpu has a slice going to that vid card maker too. So a 200mm^2 die can be sold for profit by AMD/NV at ~$180-220 and a vid card maker can make a profit on that too. So how cheap is that chip after you take out all the other parts of the vid card, the vid car makers profit etc? See the point. Same story for their ~310mm^2 GTX 1080/1070 gpus. The 1070 will go for $380 or so if not the founders early version. So you can sell a ~310 gpu with all the crap on a vid card, the vid card maker's profit, and still have NV make good money on the chip itself. I hope you get the point. There is no way in the world you're going to price yourself out of the market with a chip this large. Note GM204 was 294mm^2 and sold for $329 on GTX970 cards, again with all the profit a vid card maker has to make, all the card crap etc. Chips are cheap to PUMP out (maybe not designing them), once the design is done. AMD's Trinity was ~246mm^2 and sold for $120 😉 Just thought of another. Tegra's etc go into tablets that are $199-299 also and pretty much everyone is ~100mm^2+ now at top end of socs.

http://www.anandtech.com/show/7677/amd-kaveri-review-a8-7600-a10-7850k/4
Kaveri 245mm^2. How cheap were those? $119/152/173. IF this isn't enough examples to make the point...They must be making those for under $100 right? Even if it cost you another $50 for a larger chip (~250-300 today) who cares? You add that on top of Intel's prices? Brute force for profits works in this case, but AMD keeps trying to be the CHEAP guy instead of the KING. Intel/NV make money. AMD doesn't. Who would you rather be? Go big, or go HOME. If you can't out engineer the other guy, out size him and brute force it I think I've covered multiple processes, vendors etc here. Your point is moot IMHO, or none of these guys would be selling at these prices. The data says you're incorrect sir. Do larger dies cost more, and get less per wafer?...Of course, but that wasn't the point here. It seems clear you can make anything from ~100-300mm^2 and sell for under $200 (cpu or gpu). Remember, the last time AMD made real money (a Billion), they had the KING across all ranges with pricing to match. They had 50% of the home desktop market, and 20% overall IIRC (basically none of Enterprise), but that limit was due to a production limit on the fabs. AMD could use multiple sources for fabbing today and blow that away severely hitting Intel profits over 3yrs (longer?) for a REAL intel answer. Trinity in that story above also at 245mm^2. Again cheap as chips. As you can see from the charts AMD made 5 cpu/apu's from 212-315mm^2 recently and they all sold dirt cheap. What they heck are you talking about?

Thanks for making me do a refresher. Due to a lack of time in the last year and not being that active recently (reading reviews to death like I used to), I'd forgotten some of this stuff

One last thought: This isn't hate for AMD, it's hate for their management! Then again, I could be wrong and maybe they'll shock us with a 300mm^2 8 core, which would mean 4 cores would be around 150mm^2 in there (not far behind and Intel 8 core here) so maybe really fast right? There is no gpu, so I'll happily buy AMD if this is the case...LOL. But dollars to doughnuts this thing is well under 200mm^2 and what a bummer that will be. I say that because the 10core Intel is 246mm^2 (6950x-no gpu), so 8 core 6900 from Intel must be ~200 if it was MADE to be 8 core (not a 10 with 2 disabled like now)? But I've never heard AMD say they'd beat Intel regarding the ZEN chip so...I should paste this into word and see if it all makes sense, but no time and this already was more effort than desired...ROFL. It is what it is, and the data seems clear to me.

After finding the 246mm^2 for 6950x 10 core (no gpu on that or 6900K 8 core, or the 6 core models as all 4 come from the same chip), now I believe AMD needs a 300mm^2 8 core again to win...LOL. I don't think TSMC 16nm process (or 14nm GF/Samsung) will be able to shove what Intel does into the same space at 14nm but I could be wrong. IF that is correct, it would mean AMD would have to be well above 200mm^2 to match a GPU-less Intel 6900 right? It will likely be under 200mm^2, have far less cache and lose everything IMHO, but I really hope to be incorrect. You should see why it must be bigger with Intel's 4 top chips coming from the same die at 246. Mind you, Intel keeps 60%+ margins while giving away 4B+ in losses in mobile (though that should change soon as they stop doing this crap). Seeing as AMD themselves have had multiple cpu designs at 240mm^2 under $200 I really don't see your point. A GPU-LESS cpu is sold to more serious crowds (as you must buy a gpu too) right? The top dollar earned from these is supposed to pay for that bang for buck lower margin crap (just like titan's/tesla's etc do for gpus, Xeons/extreme editions etc do for cpus). For your statement to be correct (200-300mm^2 cpu pricing AMD out of the market), AMD should never have been able to make all those cpus I pointed out ABOVE 212mm^2 sold for peanuts ($119-200 is peanuts for people who are discrete gpu buyers). People buying mid-top gpus are not looking for $50-100 cpus IMHO. I really only needed to list the AMD chips, as they tell everything you needed to know but who doesn't like more data? Don't forget when AMD was winning the cpu war they had $1000 cpus (athlon K7 1ghz as an example, and 4-5 models from $250-1000). Circa 1999-2001 People seem to want AMD to act like a charity, and their profits show this. If you want AMD to get back to yearly profits, they need to re-enter the top end. PERIOD. Doesn't mean I'll buy a $1000 chip, just that AMD needs to return to competing on that end or continue losing money. Margins are ridiculous at the top. IE Titan profit margins were estimated to be 85%, with Titan Z it was estimated 90% (no doubt same for tesla/quadro types on top end). Check the margin chart for an idea of top end.
http://www.nextplatform.com/2015/05/08/tesla-gpu-accelerator-grows-fast-for-nvidia/

I digress.

 

[InvalidError]

Mind you, Intel keeps 60%+ margins while giving away 4B+ in losses in mobile (though that should change soon as they stop doing this crap).

Intel announced that they canceled Broxton (Airmont's successor) and SoFIA (licensing x86 to third-party SoC designers) mobile development back in April to focus on IoT instead. Until further notice, mobile x86 is dead, unless AMD decides to try its hand.

If you look at AMD's financials where their CPU division is making losses, their current large die size handicap and inability to command decent prices is clearly not profitable. Keep in mind that about half of the retail price is distributor and vendor markups, not AMD gross profit.

 

[somebodyspecial]

They have and will. The Athlon 64 cost equal to or more than Intel at the time

Keep in mind that "at the time", Intel was desperately trying to polish the Netburst turd. There isn't much merit in achieving premium pricing when there is effectively no competition. That's basically the same situation Intel is currently in with massive premiums due to AMD being a non-concern across most of the board.

In times were both AMD and Intel were in the same performance ballpark, AMD's chips were usually priced around 10% lower.

Umm, no. Keep in mind it took 5yrs back then to make a cpu design and get it out there. As they said "back then" making cpus was like playing russian roulette, only you don't find out if your dead or not for 5yrs. Intel spent a LOT of money to speed that up to around 3yrs and at the time had many designs in the works to work from/incorporate features from etc, as opposed to today where they no longer practice that (otherwise it would have been closer to 5yrs). The whole idea is to make the other guy ALWAYS look like a turd so you can price like a king across your product line. Why be satisfied with 2nd place and 2nd place pricing/margins? Who cares about whether there is MERIT to your profit or not anyway? Intel just collectively laughed as they upped their top end pricing with HEDT chips. They even gave the idea a new name (HEDT...LOL "Hey, all your chips are turds AMD, welcome to High End DT joe public). Intel only cares that they sell and seek to charge the highest the market will pay at all times possible, and that is their JOB. Based on the market, and no competition they clearly believe there is room ABOVE their own pricing for another tier. The best business charges exactly what the market will bear and being the leader makes sure you can do this as much as possible. Business is not a charity. Top end prices aren't worth it to me, but they are to many, hence selling out the first 100K titans in days etc. I'll note I LOVE the fact those buyers and products exist, so I can buy something that rocks that is reasonably affordable. AMD shouldn't want to be in the "ballpark", they should desire to LEAD and price accordingly. Every company's dream is ZERO competition or as near as they can get without FCC chasing them 😉 Don't get me wrong, I love competition as a buyer, but as a company I'd hate it just like any company with half a brain would

People go with NV in many cases just for drivers (longer support in most cases, usually monthly updates or more) as AMD has had a lot of issues for years. Well duh, it's hard to win a war with a less than a Billion R&D vs. say 1.35B NV R&D (AMD R&D dropping 130mil/yr now - never mind Intel R&D in there too). You won't win customers taking 6 months between drivers or abandoning an entire DX11 group way too early either. I could go on and back further, but you should get the point. There is more to a buyer than cost. IE your own admission that you hate the chipset (low R&D no doubt hitting here) and will go Intel due to that on a cpu. If it's not drivers for NV, then it's the watts/heat/noise in other cases (R&D again), or even perceived value from stuff like gameworks, gsync etc.

AMD charged a mint for the 3yrs they blew away Intel, and nobody had any problem buying them even though Intel's reputation was top notch. In those days AMD sold everything they could make. It was not suicidal because they were better (about 10% in apps, and ~20% in gaming) and gamers knew it and we tell our friends If AMD puts out a King cpu, they'll have no trouble selling to the same people again (those people buy Intel right now, myself included) and it's a big enough audience to make AMD a billion in profit easily. PC gaming is larger than it has ever been & far bigger than console revenue now. Due to the ability to make more than 20% of the market as opposed to the last time they led for 3yrs (fab limited to ~20% last time), they could easily rack up billions per year. You could literally have TSMC/GF/Samsung ALL making ZEN cpus now. This is why it is very important AMD aimed for a KING this time. They could take over the bulk of the high-end and more this time. But if they aimed for "good enough", well expect bargain pricing to be your max AMD and people like me will continue to pay a premium for FASTER Intel chips. Today far more people read reviews than 15yrs ago (they're everywhere now, and the internet isn't much of a secret...LOL), so AMD could wreak havoc very quickly. The general public is nowhere near as ignorant as 15yrs ago. It's far too easy to get info today. IE, a news story 15yrs ago used to be LOCAL and that's about it. But today, a story goes national in minutes. It's not that fast maybe with PC hardware reviews, but anyone can easily get to "amd zen review" or whatever long before their purchase and if desired read a dozen reviews on product X.

One more point: Athlon 64 FX-60 vs. Intel 955 EE
http://www.anandtech.com/show/1910
$999 for Intel
http://www.anandtech.com/show/1920
Over a $1000 in trays of 1K. Basically when in the same ballpark AMD priced equal back then. AMD won all the games, and split much of the rest with Intel. FX series used to be price parity with Intel not below it for many versions.
"As with all FX series processors, the FX-60 debuts at $1031 in quantities of 1000, so you can expect street pricing to be at or around that number. The FX-57 will drop to $827 mark as it will co-exist with the FX-60."

http://www.cpu-world.com/CPUs/K8/AMD-Athlon%2064%20X2%204800%2B%20-%20ADA4800DAA6CD%20%28ADA4800CDBOX%29.html
May 2005 $1000 x2 4800+.

AMD had $1000 chips for years. People today forget that or just don't know it. AMD needs to go back to this or they'll go bankrupt at some point. Intel just released $1700+ DT chip...LOL. AMD won't erase their debt with <$160 chips vs. this stuff.

http://www.cpu-world.com/CPUs/K8/AMD-Athlon%2064%20FX-74%20-%20ADAFX74GAA6DI%20%28ADAFX74DIBOX%29.html
FX-74 Dec2006. Still $999.

http://www.cpu-world.com/CPUs/K7/AMD-Athlon%20950%20-%20AMD-K7950MNR53B%20A.html
950mhz athlon $999 March 2000.
http://www.cpu-world.com/CPUs/K7/AMD-Athlon%201000%20-%20AMD-K7100MNR53B%20A.html
Athlon 1ghz, $1299. March 2000. Chips like these pay for AMD R&D, not $100 chips. When they exited the top end, the exited profits. Buying ATI at 3x it's cost didn't help but you get the point (this was another major mistake back in the day). When they ruled, price clearly showed it

http://www.cpu-world.com/CPUs/K7/AMD-Duron%20700%20-%20D700AUT1B.html
Lowly Duron 700 went for $192 in 1K pricing. That is above AMD top chips today...LOL. This was the poor man's model. I could go on.

 

[InvalidError]

Athlon 1ghz, $1299. March 2000. Chips like these pay for AMD R&D, not $100 chips.

$1000 desktop chips don't pay for R&D when they account for less than 1% of sales. The $200-300 ones which account for 80-90% of sales with a $50-100 margin does.

 

[somebodyspecial]

Athlon 1ghz, $1299. March 2000. Chips like these pay for AMD R&D, not $100 chips.

$1000 desktop chips don't pay for R&D when they account for less than 1% of sales. The $200-300 ones which account for 80-90% of sales with a $50-100 margin does.

http://www.tomsguide.com/us/amd-intel-financial,news-935.html
2008 97% of processor sales were under $200.

http://www.forbes.com/sites/patrickmoorhead/2012/05/10/discrete-graphics-the-sky-is-not-falling/#49959bb87f1e
2012, Patrick Moorhead says Chips $250+ account for 10% of the market.

http://hwinsights.com/press-release/features-not-frequency-key-to-greater-profitability-in-hedt-cpu-market-says-hwinsights/
HWinsights shows value of all the HEDT cpus from 2008 and up. Note the EXPLOSION of the value of chips OVER $300. After 2008, as moorhead says also, the top end just keeps getting bigger. Also note this is happening while PC sales are plummeting for years. We've went from 365mil units (2011) down to 250mil this year.

https://www.google.com/imgres?imgurl=https%3A%2F%2Fstaticseekingalpha.a.ssl.fastly.net%2Fuploads%2F2016%2F2%2F8%2F9511331-1454909144575552.png&imgrefurl=http%3A%2F%2Fseekingalpha.com%2Farticle%2F3875926-deep-dive-intels-2015-results-2016-much-worse&docid=uFEGXT-iyGFIzM&tbnid=ubqLEjZ6ZalgeM%3A&w=539&h=394&bih=537&biw=1184&ved=0ahUKEwjIoMTRxrDNAhUB6WMKHXtbDCIQMwgeKAIwAg&iact=mrc&uact=8
(hopefully that link comes through, you have to sign up for seekingalpha so used google instead). Main point here is while sales of PCCG has plummeted, profits in that biz segment haven't, nearly as much as they should. So selling fewer, and clearly from hwinsight/moorhead things on the top end have exploded since 2008 to 2012 to 2014 and as data for ocing shows it's exploding more and more on the top end.

Overclocking info:
http://hwinsights.com/press-release/demand-for-overclocking-surges-in-the-face-of-declining-desktop-market-says-hwinsights/
Ocing grew 50% 2013-2014, and up another 100% for 2015! As they say Intel and GPU got 66% of the $1151 avg PC cost here and further Intel cpus pocket 42% or an avg of $483 each. How much growth for 2016? Trend is up, and Clearly Intel sees it or you wouldn't plop 4 new chips ON TOP of your old ones. The value of the high end vs. mainstream components shows the same (cpu+gpu+mboard+gpu+psu). Sales price have doubled in the last year in high-end Ocing, while mainstream has been flat priced for 2yrs. People are spending that on cpu/gpu.

http://hwinsights.com/press-release/silicon-providers-main-beneficiaries-of-enthusiast-pc-growth-opportunities-exist-for-vendors-to-raise-profits-with-innovative-products-says-hwinsights/
Info on the money all going to cpu/gpu as noted above.

I'm guessing if Moorhead wrote a 2016 article it would say $300+ for the top 10% of the market now based on HWinsight's ocing info and value of the HEDT segment (rather than his old $250+ for that top 10%). Those chip values of >300 chips are massive compared to UNDER $300 right? That chart doesn't lie about where things have shifted (in a massive slump) over the years.

Same story at NV by the way, as gpu sales tumbled, their margins have went quite a bit higher quarter after quarter setting records. Hmm...So selling fewer, and clearly getting more as revenue, margins hit records.

More data: What do reviews show is selling?
http://www.newegg.com/Product/ProductList.aspx?Submit=ENE&N=100007671%20600565702%208000&IsNodeId=1&bop=And&Order=PRICED&PageSize=30
Socket 1151 - TWO chips above $319 have 386 reviews total with most being the top chip i6700 at 357 reviews! Only 29 of the cheaper $315 model. The rest of the models are below $229 with 363 reviews between that and $200. So the bulk is ABOVE $315 judging newegg reviews right? Reviews below $200 are basically 120 or so total.

https://www.amazon.com/dp/B012M8LXQW/ref=psdc_229189_t1_B00MMLXIHM
Note amazon showing i7-6700K @ $360 (393 reviews) outselling i7-6700 reg ($297) and i5-6600K ($229) combined (296 total reviews between them) by 100 reviews. Again, the trend is toward by higher and probably never upgrading until the next chipset/gen like me. I have another PC to build and will be going up myself to 6700K or better and dumping my 4790K on my dad (ROFL - he won't mind that upgrade from his dual core xeon). Note the 4790K here also has another 1226 reviews! WOW. But then, it's had just over 2yrs to sell. The new chip has another 14 months to go, and xmas/back to school sales will likely pump it up obviously, but 393 already for 6700K is already a good sign and dwarfing the two below it.

More: How did socket 1150 do? SAME story.
http://www.newegg.com/Product/ProductList.aspx?Submit=ENE&N=100007671%208000%20600436886&IsNodeId=1&bop=And&ActiveSearchResult=True&Order=PRICED&PageSize=30
Tom selling by far is my chip, i7-4790 at $339. A whopping 1036 reviews, which is equal to the total of i5-4690K (629 K+reg+S models)+i5-4590 (128 reg+S model)+4570 (reg+S 145 total)+4670 (94) which is 996 total. You have to go below 190 to find another chip (i5-4440 with 125) to add up to more than a SINGLE chips total ABOVE $300. The i7-4790K was $349-339 and there's over 120 other's above $300 is S1150, but what's the point in mentioning those with a single chip at $339-349 doing 1036 reviews? LOL. So massive buying above $300 even before we tack on all the ones above it that show massive buying in $389-999 HEDT as shown in HWinsights, where everything has exploded to the top end except for the outlier in 2008 of 284 (the 920 chip). But even back then in that year and the next 6 months the four above the 920 (940/950/965/975), all being $560-999 ate the 920 sales for breakfast and from there every year a chip <$310 got swamped by $389+ chips, getting worse yearly. The 2014 totals were huge showing ASP going up. $999 does well every year in the chart vs. the bottom.

Socket 2011 v3? :
http://www.newegg.com/Product/ProductList.aspx?Submit=ENE&N=100007671%20600535697%208000&IsNodeId=1&bop=And&page=1
Another ~540 reviews, and all $389+. So these are yet more on top of 4790, and barely any of the latest chip (they just hit of course) so most are comparable in gen with 4790. The new chips have 15 reviews total, so all of these reviews are in the same timeframe for 4790 chips. I'd lump them in with it which again, overall shows a massive jump in the top end. We'd have to discuss the latest 4 in a year I guess. But the trend is easy to see.

New models have only been selling for <3 weeks, so again, we'll have to wait to see, but all signs point UP trends. Sorry, but I'll take HWinsight, Patrick Moorhead, and review data at amazon/newegg (basically verifying what these two industry vets are saying) over your word every day Of note, so far at amazon the 6950x has 14 reviews ($1735), more than the other 3 from this new class combined so far (which have 5 total...LOL - and it showed up 1 day later than the others at amazon). Same trend 😉 IF you have the money to buy in this class (along with the $200-550 board that goes with them), $400-1735 seems to be the choice rather than whatever the bottom rung is in any gen. The 5960x had far less reviews (1/3) but value wise, added up to more than 5820 as shown by hwinsight (never mind adding the 5930x on top of it, upping their combined value to dwarf the 5820). The 5820 was $389 anyway though in the chart (still $373 at amazon). Of course none of this is even discussing server chips prices (which I'd include in that comment I made, all high-end stuff allows them to give us peons better stuff at decent prices), which blow up value even more. Again, all told, the top end of crap pays for a ton of R&D.

Same thing on Nvidia's side where Tesla has ~80-90% margins vs. quadro in the ~mid 50's and less for home market (but their selling top end at home very well too).
http://www.pcgamer.com/nvidias-surprised-the-titan-outsold-the-year-old-gtx-690-in-just-3-months-were-not/
According to NV they sell every one they can make of their first Titans (not sure since, Jen hasn't said AFAIK).
http://seekingalpha.com/article/1243621-why-nvidias-titan-keeps-selling-out
Or I guess sold out in an hour repeatedly? 😉 Clearly high end gpus do well now too as HWinsights said also.

Oh, and not to mention how much Intel would have blown up profits if not tossing 4B+ a year on mobile while this high end cpu stuff blew up. It will be interesting to see what happens to Intel's Q reports as that stops. I think we're done here.

 

[InvalidError]

http://hwinsights.com/press-release/features-not-frequency-key-to-greater-profitability-in-hedt-cpu-market-says-hwinsights/
HWinsights shows value of all the HEDT cpus from 2008 and up. Note the EXPLOSION of the value of chips OVER $300.

Keyword here: by VALUE.

With most chips in the category costing twice to thrice as much, that 10% by value translates to an optimistic 5% by volume. There aren't any sane businesses that get carried by 5% of their sales.

 

[somebodyspecial]

http://hwinsights.com/press-release/features-not-frequency-key-to-greater-profitability-in-hedt-cpu-market-says-hwinsights/
HWinsights shows value of all the HEDT cpus from 2008 and up. Note the EXPLOSION of the value of chips OVER $300.

Keyword here: by VALUE.

With most chips in the category costing twice to thrice as much, that 10% by value translates to an optimistic 5% by volume. There aren't any sane businesses that get carried by 5% of their sales.

Umm, you're not quite doing the math right. Margins on these 246mm^2's, which Intel makes no doubt for well under $100 (considering AMD makes chips for $100 that are far larger than 300mm^2 for xbox1, ps4, and sells them for $100-115 AFTER paying a fab to make them, which Intel doesn't do), are 2x-15x higher. Not 2-3x. All four new chips come from the same die and are $434/617/1089/1733. The value of these will be huge, and almost all margin. Like I already showed, people who can afford this crap are going top end. Even preliminary sales show 16 reviews for $1733 model and only 5 for the other 3 models, repeating 2014, 2013, 2011, 2010, 2009. The only gens that didn't sell well had no features that prompt sales and in those years it affected all chip prices (low to high). Really chips between years (some were release 6 months after successful chips with features/chipsets etc), with no reason to succeed.

As usual you go from extreme to extreme ignoring the reality that exists between. It's like explaining something to a 10yr old that keeps saying "why, why, why," to every response you give them. With your one liners it’s like you’re saying “You've soundly destroyed my argument so here comes a silly straw man for ya”. Nice. Are you trolling me or what? Your 1% comment was proven silly in my previous post, and I guess we'll have to show how silly the 5% comment is too. Nobody said they will carry the business on 5% of the sales, but it's not 5% anyway. But the value and more importantly profit/margin on that 10% pays for a LOT of R&D. Did you look at the value of the 5960x+5930x on that chart? The value of the 5960x was 3rd out of everything in the chart and barely 3rd to the value of all the 920's that sold for $284 and was a massive success while NONE of the others performed like that under $300 and the #1 overall value was another $999 chip (980x..ROFL, WOW ate everything for breakfast by FAR). Now consider the reviews showing the 4790K/6700K on top of the HEDT stuff, add in the server chips too, which make prices on desktops seem puny (Intel has server chips at $6800+), and you should get the point. Then again you seem to completely miss the point of all of my posts, or just choose to ignore it with a one liner at best. Either do the homework to make a valid argument or sit in silence please. I expect more from a "moderator" than just dismissing the data I provided.

http://ark.intel.com/products/93791/Intel-Xeon-Processor-E7-8893-v4-60M-Cache-3_20-GHz
Servers for instance, just launched June 6. $6840 quad core with 60MB.

http://www.anandtech.com/show/10401/intel-launches-4s-and-8s-broadwellex-xeons-e74800-v4-and-e78800-v4-families-up-to-24-cores
Rundown on all of them, one over 7K. Die sizes from 246mm^2/306/456.
"The super-high-end of Intel’s Xeon CPU range, based on servers with as many cores and as much memory as you can throw at them, represent a good part of Intel’s business with the potential to offer large margins: some customers want the most, the best, the powerful, and are willing to pay for it."

More testing on these obviously, but margins are also obviously ridiculous. A "good part of Intel's business", when added on top of all the previous data should make the point. Stuff well over $300 (jeez, at $6800+ for a few models these things dwarf HEDT 6950x $1733 pricing! 4x! All probably made for under $100) pays for a LOT of R&D as I said. Consider the 5% you now acknowledge for second - and really we're talking 10% according to Moorhead so DOUBLE everything here for chips/profit/margin/revenue info: Intel sells 220mil chips of the ~270mil that sold TTM. So ~11mil chips (really ~22mil? @10%). Knowing the OC market is avg $486, add to that the server market, the rest of the HEDT market, Sales of chips over $300 (ie, massive sales of 4790K, seeming soon to be massive 6700K etc) you get a lot of profit from this stuff. Lets just say it was all $500ea on that 11mil (stupidly low with server, and on top of HEDT stuff showing such high margin sales and bulk of it over $500). That's 5.5B total. Knowing they make these for <$100 over them all (die size 246/306/456) that means ~4.4B profit on 5.5B revenue. Note Intel made 14B operating income in 2015 on 55B revenue. 4.4B of that 14B came from the top stuff? So 1/10 of the revenue is giving a >1/3 of the income? The point is you can't LIVE without the top end! Stupidly high profits comes from "I want it now and first so I can brag" and "I'm enterprise and time is money, just get it done faster". Thank god these two groups exist. This is why AMD hasn't made squat since they lost the race (well 7B losses in the last 10yrs and about 8B in the last 15yrs). The lower margin stuff only works for Intel because they own 80+% of the market and fab their own stuff. The king chips is where massive money is made. IE, ARM IP is in 2B+ units every year but they only make ~600mil yearly...ROFL. Imagination IP is in every apple mobile device pulling down billions in chips but they only get ~10-60mil yearly or lose money. As hwinsight showed the money is in the chips and not the card that's built around gpu chips either, as card vendors have crap margins too (asus, msi, gigabyte etc ~12% across their whole markets).

Same as I said for NV. In their case Margins on Telsa's are 80-90%, with Quadro well above 60% (overall 56%, gaming looks ~56%). I think tegra pulls everything else down some. Look at the value of their revenue in each segment:
Desktop/gaming etc (all but tesla/quadro/grid) was 687mil of 1.305B. Grid/Tesla was 143mil (up 63%), and quadro was 189mil. GPU total 1.079B (the rest is tegra, auto, intel payment). So the pro market value (counting grid/tesla/quadro) was 332mil of 1.079B gpu revenue. It is no small amount and it's 60-90% margin on that crap. Total operating income was 322mil. Even if pro market is just 55% margin, you are talking over 1/2 of their operating income from >1/4 of revenue. Since we know tesla portion is WAY over 55% (as avg quadro margin is above this already and tesla is just pure evil margins), they are getting more than 1/2 of that 332mil of pro market in profit. ROFL. Intel gives them 66mil direct profit too...LOL. Add that to pro profit and not much coming from the gaming side right? You still think it's 5% for Intel or even NV?...LOL. I think not. Read some Q reports, channel data etc and get a reality check because I'm having a hard time believing you're anywhere near correct 😉 I may be off a bit one way or the other & likely doing you a favor here based on all the numbers we can already see here, but the general idea here is solid for both Intel and NV.

At this point I'm really hoping ZEN (8core) is roughly the size of xbox1's 363mm^2 since we know AMD makes it for $100 and sells for $110-115 and I'd say upper end there now as margins have hit mid-teens according to AMD which I guess means <15 or you'd say 15% not mid-teens (production of the first runs at least was single digits). PS4 is slightly cheaper at 328mm^2 and according to IHS is $100 (so AMD makes it for $90 at 328mm^2?). ZEN needs to be ~350mm^2 to rule for a few years. With 3 fabs able to easily tool up (tsmc, and GF/Samsung sharing process info), they could take far more of the market in a few years vs. ages ago when they couldn't produce more than ~20% no matter how fast their chips were back then. Intel had that working for them before, but NOT this time. This time those 3 fabbers can make way over 50% if AMD had the perf to take the sales and clearly enthusiasts would bite at the high end and so would servers, HEDT. You could put a massive dent in Intel's HIGH margin stuff. They already had to dump mobile because they can't afford to keep up with fabs losing 4.1B, and it was growing yearly, hence operating income dropped from 15.3B to 14B (2014/2015) even with higher margins. This is probably AMD's last "PERFECT STORM" if they get the chip size right to dominate by brute force. IF Intel profits dropped drastically, margins dumped, reviews everywhere said "AMD wins again - it's 1999 again", you can bet shareholders would leave to at least some extent on top. Share price goes down, cash flow down, etc leaving far less cash to work with etc. I won't hold my breath, but I'm praying they hit xbox1 size I say 275-300mm^2 catches Intel (Intel likely can shove slightly more into the same area due to process), and xbox1 size wins all the gaming benchmarks by 10-20% (like before, then Intel did it back to them...ROFL). Knowing management (they fired Dirk for saying KING baby KING ME), they probably shot for smaller/cheaper.

http://ir.amd.com/phoenix.zhtml?c=74093&p=irol-newsArticle&ID=884787
One more comment on AMD: Margins were all time high in 2006 and actually topped Intel for a Q. 58%, 58.5 & 56% here for 3 Q's. Wow, and then they slowly gave up the high end stuff since. Margins, revenue and profit all went with it. AMD seems to be aiming at Kaby Lake, so maybe they'll be back in the high end shortly. I just wish it was a huge 4core, or 6 core tops and 300mm^2+ (now that we know what Intel just did, 246mm^2 for all the top four 6-10 core models). They needed to WIN again, not be ~competitive. Just a guess here, but if it's an 8core Zen at ~250, Intel will likely be able to shove more in the same size chip than AMD on GF 14nm. I doubt GF's 14nm is better than Intel's process, which again is why I say it needs to be BIG to win and price accordingly. A tie just means Intel prices you down to ensure you get no share gains, or worse, not enough profit on whatever you sell to make a meaningful change in the bottom line.

http://www.anandtech.com/show/10353/investigating-cavium-thunderx-48-arm-cores/19
JohanAnandtech - Thursday, June 16, 2016 - link
“More than 30% of Intel's revenue, and the most profitable area for years, and for years to come...”
SERVERS! Comment section of the article, A repeat of their previous comment but with actual % this time. Now consider the margins on these chips and realized a HUGE portion of profit comes from that puny 10% (LOL, you said 1%, then said 5%) you so eagerly attempted to diminish. If you add that and the other side (desktop, HEDT, etc we’ve been discussing for the most part) you should realize how far off your statement really is. It’s even worse than I thought, though I did say I was being conservative IMHO. Clearly a gave you way too much leeway. Obviously, to me anyway, it’s 40%+ of the profit from this top stuff when all added together and I’m again being conservative (IMHO by quite a bit). So again, while you can “live” without it, it would be INSANE. AMD seeming, really CAN’T live without it…LOL. Thanks for making me feel even safer in my stock positions. Remember it’s 10% of the CHIPS sold are above $250. The VALUE of them is FAR above this and margins make a HUGE portion of their profit. No more straw man comments please.

 

[InvalidError]

Umm, you're not quite doing the math right. Margins on these 246mm^2's, which Intel makes no doubt for well under $100 (considering AMD makes chips for $100 that are far larger than 300mm^2 for xbox1, ps4, and sells them for $100-115 AFTER paying a fab to make them, which Intel doesn't do), are 2x-15x higher.

And AMD is bleeding money instead of making a profit.

 

[somebodyspecial]

Umm, you're not quite doing the math right. Margins on these 246mm^2's, which Intel makes no doubt for well under $100 (considering AMD makes chips for $100 that are far larger than 300mm^2 for xbox1, ps4, and sells them for $100-115 AFTER paying a fab to make them, which Intel doesn't do), are 2x-15x higher.

And AMD is bleeding money instead of making a profit.

Which has nothing to do with the size of xbox1/ps4 chips (if that is what you're implying, you don't seem to get the message, Intel's $7K chips cross into 400 territory) and everything to do with the fact that they chose low margin consoles for those 320-360mm^2 dies (and R&D). What they should have done instead was the same size in a QUAD core which would have smacked around Intel for 3-5yrs in desktop and charged accordingly with a 2nd die for servers (much like Intel) with 6/8/10 (or whatever) cores that would do the same, of course with no way for Intel to price them down just like years ago. Management made a poor decision on margins to win a deal, much like years ago on cards they got on walmart shelves (did that for nothing). Nvidia passed for a reason on consoles, and AMD had x86 which is the direction they all seem to want to go for various reasons (good reasons IMHO), with a far better gpu than Intel. What would sony/msft do if AMD has asked for more? Go to Intel? ROFL. This gen would be even weaker than they are already, hence mid season rev2 already coming...ROFL.

You're making my point. Get back into the high-end or bleed to death on low margin crap that can barely pay interest on your debt yearly, GF fines for take or pay crap etc etc. I sincerely hope they told MSFT/Sony we won't bother for your next model without 20%+ and if it sells less than 15mil/yr (or whatever) you owe another 10% margin payment. We are talking $10-15 per console for crying out loud and even this is lower than I'd accept already to waste R&D on. They seem to have the negotiating skills of Obama/Kerry...LOL It's clear Intel charges $1733-7K per chip for the same size as AMD makes in consoles basically. Should you have spent your R&D on making those, or dumb low margin consoles? You can't be as wit-less as your one-liners suggest. I'd say ignorant, but you don't seem to be able to learn the point. As I said, the last time AMD made a Billion they had 58% margins, just like Intel. Cost of silicon is moot or NV wouldn't be making 600mm^2 gpus ever and making great margins on them. Giving away your products at margins you can't live on is stupid too.

I find it hilarious you still don't seem to understand these simple concepts, and instead opt for straw man after straw man comments in spite of all the data I used in my attempts to overcome your edumacation 😉 Whatever...Good to see you didn't argue with the data in any of my posts as usual. No time to bury you again anyway (too busy with brexit stuff...YAY!). I commend you on your nick choice, it fits. Reboot your brain and maybe you'll stop getting errors Thanks for my morning laugh :lol:

 

[InvalidError]

Which has nothing to do with the size of xbox1/ps4 chips (if that is what you're implying, you don't seem to get the message, Intel's $7K chips cross into 400 territory) and everything to do with the fact that they chose low margin consoles for those 320-360mm^2 dies (and R&D). What they should have done instead was the same size in a QUAD core which would have smacked around Intel for 3-5yrs in desktop

Throwing die area at a CPU core does not necessarily make it faster: the more transistors you need to coordinate across the pipeline, the more propagation delays you introduce and that lowers achievable clock frequencies. You can offset that by cutting operations into more pipeline stages but deeper pipelines have higher execution latencies, increased misprediction penalties and more frequent pipeline stalls. This stuff has practical limits and peak performance requires striking the correct balance for currently available technology, which is where Intel is at.

The reason why AMD is where it is because nobody wants their chips, so they aren't generating much sales regardless of how low their margins might be. SoC designers like Rockchip, Spreadtrum and Mediatek manage to make a decent profit from selling mostly $5-10 chips because they have volume to compensate for their lower per-chip margin.

 

[somebodyspecial]

Which has nothing to do with the size of xbox1/ps4 chips (if that is what you're implying, you don't seem to get the message, Intel's $7K chips cross into 400 territory) and everything to do with the fact that they chose low margin consoles for those 320-360mm^2 dies (and R&D). What they should have done instead was the same size in a QUAD core which would have smacked around Intel for 3-5yrs in desktop

Throwing die area at a CPU core does not necessarily make it faster: the more transistors you need to coordinate across the pipeline, the more propagation delays you introduce and that lowers achievable clock frequencies. You can offset that by cutting operations into more pipeline stages but deeper pipelines have higher execution latencies, increased misprediction penalties and more frequent pipeline stalls. This stuff has practical limits and peak performance requires striking the correct balance for currently available technology, which is where Intel is at.

The reason why AMD is where it is because nobody wants their chips, so they aren't generating much sales regardless of how low their margins might be. SoC designers like Rockchip, Spreadtrum and Mediatek manage to make a decent profit from selling mostly $5-10 chips because they have volume to compensate for their lower per-chip margin.

Oh jeez, here we go again...ROFL

Your bottom P is pretty much what I said. Low margin only works for huge volume which AMD doesn't have (pc I mean) and consoles don't give. For ref, mediatek makes around 800mil/yr and terrible net margins (~6.6B in revenue). Again, it only works due to volume, which AMD doesn’t have on consoles or market share on gpu/cpu. Your argument now is simply to repeat my comments? Also note mediatek made the same on 4.2B revenue in 2013. So margins seem to be tanking as everyone competes in the crap end now. NV, who notably left (along with Intel) will come back IMHO when Vulkan is ubiquitous, and a Samsung/qcom/intel modem can be slapped on without power costs blowing out due to die shrinks. It will happen at 14nm coming up for tegra or 10nm. No point though until Vulkan pushes better gaming (pushing the need for NV gpus and 20yrs of working with devs), and the modem is easy to do without it being in house. They want the GPU to become the selling point, thus fetching the high end market and until then auto will suffice. I think they’ll also produce a 50-100w chip at some point too, for a desktop once google gets a decent OS for them out (merging of chrome/android I guess?). Just like consoles were passed, no point in fighting for $200 phones. NV would rather be apple than mediatek etc. Even mediatek is trying to add stuff to up margins and set themselves apart (VR, Dual-cam etc). Qcom is seeing the pain of being low-ended to death too with operating profit dropping ~75% for the chip business (wow). They must feel like AMD these days with Intel giving away stuff (dropping that crap now though, but damage is done), mediatek flogging dirt cheap stuff and Samsung taking on their modems with intel right behind them. NV should get a great price on the modem soon with the war between the 3 for dominance. QCT unit only dropped 25% in sales, yet the whopping 75% hit to profits showing margins tanking. Spreadtrum makes peanuts compared to the rest (<150mil AFAIK but I don’t track their stock really, just the tech they make). Rockchip is private last I checked and china is slumping.

Regarding the chip design, smarter people hit a better balance and there is no proof of what you say. In fact I’d say we have ample proof you can throw more at the die and achieve more (transistors, die size, or both). NV/Intel amaze us often with what they can do. Apple throws more at fewer cores/larger die every gen and always beats everyone for a while even though everyone else has more cores. Sure you can make a crappy large chip, but then you must have crappy engineers or they simply make a wrong bet on where software is headed etc, like AMD betting on 8 core while games just ignored it for years (fx chips recently, it’s their fault nobody wants their chips). I don’t think papermaster/keller etc would fail here. They’ve been involved in a lot of great chips (apple and AMD winners for example). While gamers might not see 6-10core (for a while?), some stuff does in a big way. The flipside as shown by Intel, is a fast quad keeps up with them in much of the stuff we play (but that’s game devs fault, not crappy Intel 10core designs), which is exactly why AMD should have had a big quad and then later a second die for servers with more cores like Intel, or servers first, then quad desktop. Either way works, just get in the top end for a change. LOL if you think intel has hit a limit with a quad at 122mm^2. They are where they are because AMD isn't anywhere close so have no need to do anything but add gpu or simply take the easy way out now with more cores/cache and larger dies. You can mitigate much of what you say with many things (more cache for instance, faster cache, better prediction etc – basically all the stuff you say works against you can also work FOR you). It is also easier to clock a quad higher than a 6-10 core chip. Again a fast large quad would win most stuff vs a same sized 8 for quite some time to come it seems as not many games seem to take advantage of more than 4 cores. See AMD FX chips vs all of Intel's chips and Intel’s weren't too big, just fast and fit for gaming. Now ripping, rar, par, pro apps etc? More cores please.

http://www.fool.com/investing/general/2014/03/31/nvidias-1000-beast-is-selling-out.aspx
Another point on revenue of the high-end stuff.
"The fundamental NVIDIA thesis is that even though the low-end/OEM discrete GPU market seems to be contracting with the general PC market, the high-end gaming GPUs (which according to NVIDIA's analyst day made up 61% of the company's revenue base during the last fiscal year) market is growing enough to not just offset that decline, but is actually enough to drive mid-single-digit growth"

Considering NV has pushed up top end pricing (just like Intel), it is likely higher 2yrs later as shown by shrinking sales of gpus, yet great margins/revenue/income at NV. Discrete hit a high of 100mil per year (Q2 2007-Q1 2008) and NV pretty much made the same as NV today.
http://www.kitguru.net/components/graphic-cards/anton-shilov/sales-of-desktop-graphics-cards-hit-10-year-low-in-q2-2015/
Insane margins on the high end caused this as shown by NV’s 61% of revenue coming from HIGH-END gaming cards. The rest is all Tesla/quadro. Not much of anything is made on the bottom clearly! Not sure why AMD went $200 and below first. Yet again chasing bottom with new tech. 61% of NV revenue from high-end gaming gpus can't be ignored for such stuff and this is just massive. Vega not good enough to compete or something? With the GTX1060 around the block (could hit before custom 480's hit end of july), AMD won't have much time to grab share or margins. I had saved this last week to finish later (oops took a while to get back to this), and now see 1060’s coming 14th with 6GB. So no time for market share grabs or margins. AMD 480 sold 2000 at launch at OCuk, while NV 1080 sold 1000 at launch at the same store. Who would you rather be? I’ll take NV based on the numbers, which do not lie, even with a larger die. $20 says NV has a GREAT quarter and AMD has a crap one. All current cards for 480 and probably at least the next few weeks have chips that were sold to OEMS before June26th which was likely the end of the Quarter for them. Shipping from AMD to OEM, OEM to put a board together, box it up, ship it and on store shelves takes more than a week. 😉 So we’ll have a pretty full view of what that $200 card did for AMD in 2 weeks in the quarterly report. We’ll get the same for NV and I suspect margins will be very close to 60%. Top end blows out the mid-bottom or NV’s quarterly reports and JPR/Hwinsights etc are all liars I guess  More to the point though, a $200 card only sold twice at much as the 1080! Again, blowing your “With most chips in the category costing twice to thrice as much, that 10% by value translates to an optimistic 5% by volume.” Comment completely out of the water. IT’s 50% of the volume it seems for top vs. $200 right? Jeez. Just keep digging your hole.

http://fudzilla.com/news/graphics/40947-geforce-gtx-1080-sold-out-selling-excellent
Fastest selling high-end card ever according to multiple sources. AMD should have been there to take whatever they could that NV couldn't supply and might have been able to supply quite a few if coming from a totally different fab (GF Vega's?). I wonder if they'll hit 70% of the revenue this year with the top end selling so well at $700 and shortly adding $999 cards again. You might want to rethink your 'sane business doesn't live on this' comment huh? Apparently you can live on it, and quite well. Intel would argue the same as previously shown. AMD should take lessons from these two right? People are moving up, as I myself have for both cpu and gpu. I now buy $300+ for both, where I used to buy $200 last upgrade for each. I'm pretty much in line with NV/Intel data at hwinsight. My next gpu will likely be 1070 (maybe 1080 once ti hits) and probably kaby/cannonlake while tossing my 4790K to my dad. He just games, I need more cores though so I could end up even higher than expected for both cpu+gpu. I'm guessing pascal v2 (like maxwell) will blow our minds again before the next shrink probably. It’s likely the GP102 and possibly not based on GP100 or GP104/106. Confused as to how you think throwing more size doesn’t equal more perf when cpu/gpu do it all the time with massive results. Guessing, GP102 will be ~475-525mm^2 and be about 50% faster than 1080 for $999. Though if I was them, I’d slap a $1099-1199 price on it to see how well it sells with $999 cards selling like mad previously. The market is paying more on avg over the last 3yrs and nothing is on the shelves, so why not?

https://www.nowinstock.net/computers/videocards/nvidia/gtx1080/
Nothing in stock except on Ebay for 1080…LOL. They should have charged more clearly. As a buyer of course that would suck, but as a company, if you can’t keep your stuff on the shelf (and they’re getting sold on Ebay for far more) you have room to price higher. Supply and demand. Not to mention they still haven’t made 2007 profits if you take out Intel’s 266mil/yr (lower profits for the last decade!).

http://wccftech.com/nvidia-geforce-gtx-1080-success/
It's not just NV enjoying more high-end sales. According to JPR both AMD and NV reporting rising high-end sales to gamers, and of course Intel said the same and plopped 4 HEDT chips on top of i7-6700.

"Jon Peddie Research also highlighted the increased migration of PC gamers to the high-end graphics cards. In Q1 2016, NVIDIA and AMD both reported an increased high-end discrete graphics board shipments while the PC market was declining in general."

http://jonpeddie.com/press-releases/details/high-end-add-in-boards-shipments-increased-in-q1-2016-while-all-else-declin
JPR goes on to say even game PC's are soaring, not just the AIB market.
"The gaming PC (system) market is as vibrant as the stand alone AIB market."

It’s comic you think Intel is at a limit when they just put out 6-24 core chips that are massive, and if you have an app that can use them, SCORE. You can always grow the chip until it is cost prohibitive to do so. Yields are good enough at some point in the process that NV can make ~600mm^2 chips and make tons of cash repeatedly on cards from $330 to $700. Then of course shrink by dumping features like DP etc for gen 2 and aiming your chip more directly at specific users. Intel’s producing chips the 2x the size of skylake and approaching Console socs and making a mint on them at $434-1730 while AMD makes them for $90-100 and makes $10-15. I’d rather be Intel with HEDT as opposed to console chips that are larger with no margin. Zen would have been out AGES ago if AMD hadn’t wasted R&D on junk margin consoles and would have beaten Intel if it was as big as console chips now. Keller/Papermaster and company could have gotten that balance right. But who knows, maybe ZEN is 400mm^2+ and will rock from $300-1700. That would dwarf Intels 6-10 core 246mm^2 and dominate. AMD can make an xbox1 for $100 and sell it for $115 to MSFT at 363mm^2 (stupid margin deal, but you get the point, or maybe not…LOL). ZEN can easily be 400mm^2 for $300-$1700 chips basically matching Intel pricing and making GOBS of revenue/margins as already shown.

I could bury your “Keep in mind that about half of the retail price is distributor and vendor markups, not AMD gross profit.” comment too as a PC business owner for 8yrs before (jeez look at tray pricing on 1k/10k/100k etc volumes and check someone with channel info), but I’ve already wasted enough time on someone who doesn’t seem to justify even being called ignorant. You don’t seem to learn despite being buried comment after comment with data that soundly proves every one of them wrong. Queue another ad hominem attack 😉 Enjoy your day.

 

[boringpie]

Wonder if it will beat my Intel 10 Core Xeon CPU, however, I use mine more for work than games. Loading a 500MB Excel Financial Model is still difficult even with my CPU. Hope AMD has a better alternative.