AMD Ryzen (Summit Ridge) Benchmarks Thread (use new thread)

[unseenmorbidity]

Intel throws >$12B/year at R&D (and growing), more than 10x what AMD does. And that R&D is spent in service of building products that are better than what the competition is building so that they can reap the financial benefits.

Intel also spends a ton on sales and marketing (nearly as much as it does on R&D annually) -- just building a decent chip isn't enough to win and support the many system designs in the market. There is a lot of work that goes into winning/supporting designs and then a ton of effort that goes into working with channel/PC OEM partners to get these systems into the hands of consumers.

Really, it's just amazingly complex, and I think individuals whose perspective is that of an enthusiast who builds her or his own computers tend to miss the overarching business picture.

I don't think the argument that Intel is just letting AMD come in and rip away wads of profitable revenue really stands up to scrutiny. If AMD wins market share from Intel, it won't be because Intel let AMD have it. The published financials simply don't tell that tale.

Intel cannot eclipse AMD technologically speaking, because it's impossible. They have hit a tar pit that comes before the brick wall, that is the end of moore's Law. It doesn't matter how much money they toss at the problem, it won't help.

Intel could have easily destroyed AMD anytime they wanted. They don't do it, because it would in violation of antitrust laws. It's in their best interest if AMD sticks around. It just so happens that not competing with AMD is also more profitable in the short run as well.

 

[Arachnotronic]

Intel cannot eclipse AMD technologically speaking, because it's impossible. They have hit a tar pit that comes before the brick wall, that is the end of moore's Law. It doesn't matter how much money they toss at the problem, it won't help.

Intel could have easily destroyed AMD anytime they wanted. They don't do it, because it would in violation of antitrust laws. It's in their best interest if AMD sticks around. It just so happens that not competing with AMD is also more profitable in the short run as well.

Intel hasn't been trying to keep AMD around by holding back if that is what you are implying. They knew AMD's strong suit in PCs was the low end with its highly integrated Cat-core SoCs, and they still built an entire family of Atom-based low-cost SoCs for this market and drained away significant unit/revenue share from AMD as a result.

If Intel wanted to "help" AMD out, they would have just let AMD have the low-end PC market with Jaguar, Puma, and so on.

Intel's duty is to its shareholders, just as AMD's is to its own.

One more thing...anti-trust laws don't say that being a monopoly is illegal. If AMD suddenly disappears tomorrow, or decides that it doesn't want to make x86 processors, there's nothing illegal about Intel being a monopoly if it achieves this status through legitimate means:

In 1994, the U.S. government accused Microsoft of using its significant market share in the PC operating systems market to prevent competition and maintain its monopoly. The complaint, filed on July 15, 1994, stated that "The United States of America, acting under the direction of the Attorney General of the United States, brings this civil action to prevent and restrain the defendant Microsoft Corporation from using exclusionary and anticompetitive contracts to market its personal computer operating system software. By these contracts, Microsoft has unlawfully maintained its monopoly of personal computer operating systems and has unreasonably restrained trade." A federal district judge ruled that Microsoft was to be broken into several technology companies, but the decision was later reversed by a higher court. The eventual, and controversial, outcome of United States v. Microsoft was that Microsoft had not dominated the market through unfair practices, but through innovation, successful marketing and desirable products. Microsoft was ordered to make some changes but was otherwise free to maintain its operating system, application development and marketing methods.

http://www.investopedia.com/ask/answers/13/regulations-against-monopolies.asp

 

[realibrad]

Intel throws >$12B/year at R&D (and growing), more than 10x what AMD does. And that R&D is spent in service of building products that are better than what the competition is building so that they can reap the financial benefits.

Intel also spends a ton on sales and marketing (nearly as much as it does on R&D annually) -- just building a decent chip isn't enough to win and support the many system designs in the market. There is a lot of work that goes into winning/supporting designs and then a ton of effort that goes into working with channel/PC OEM partners to get these systems into the hands of consumers.

Really, it's just amazingly complex, and I think individuals whose perspective is that of an enthusiast who builds her or his own computers tend to miss the overarching business picture.

I don't think the argument that Intel is just letting AMD come in and rip away wads of profitable revenue really stands up to scrutiny. If AMD wins market share from Intel, it won't be because Intel let AMD have it. The published financials simply don't tell that tale.

If Intel wanted, they could throw money at a market and dominate it. As you said, Intel has heaps more resources to do things. If AMD takes market share Intel can choose to take it back.

Look at what Intel now spends it's money on. They are not focusing on desktops like they used to.

 

[Arachnotronic]

Look at what Intel now spends it's money on. They are not focusing on desktops like they used to.

I guess the same would apply to AMD:

Computing and Graphics operating loss was $502 million in 2015 compared to $76 million in 2014. The decline in operating results was primarily due to the decrease in net revenue referenced above, partially offset by a $696 million decrease in cost of sales, a $120 million decrease in research and development expenses and an $84 million decrease in marketing, general and administrative expenses. Cost of sales decreased primarily due to lower unit shipments in 2015 compared to 2014, partially offset by an inventory write-down of $52 million as a result of lower anticipated demand for primarily oldergeneration APU products.

 

[sirmo]

Intel hasn't been trying to keep AMD around by holding back if that is what you are implying. They knew AMD's strong suit in PCs was the low end with its highly integrated Cat-core SoCs, and they still built an entire family of Atom-based low-cost SoCs for this market and drained away significant unit/revenue share from AMD as a result.

If Intel wanted to "help" AMD out, they would have just let AMD have the low-end PC market with Jaguar, Puma, and so on.

I honestly think Intel hasn't even had AMD on the radar for awhile. Atom was Intel's desperate desire to break into tablets and mobile. Remember they started in briefly successful netbooks and then tablets killed that market. Intel has had Nvidia, and all the other ARM guys to compete with all this time and it hasn't been able to crack that nut.

Enthusiast market for Intel is an afterthought. It's where mobile chips get binned for higher clocks and server chips sell as HEDT parts. Which is why all their quads have iGPU in them, and why all the high end multi core parts have quad channel memory. Intel could have easily had a sub $200 quad core sans iGPU, under $200 but why bother? Skylake and Broadwell-E's mainly generate revenues from the mobile and server markets. I highly doubt Broadwell-E sells to many enthusiast, especially not the 10 core $1700 part. It's basically Intel saying we don't really need to sell these because we like to mark them up for the server market anyways.

None of the Intel chips are designed specifically for the enthusiast crowd. Not like Summit Ridge. I do think Skylake stepping Kaby Lake is a response to Zen however.

 

[Arachnotronic]

None of the Intel chips are designed specifically for the enthusiast crowd. Not like Summit Ridge. I do think Skylake stepping Kaby Lake is a response to Zen however.

Summit Ridge is a server die (Zeppelin) mounted onto a package that slots into AM4. It's exactly the Intel HEDT playbook, but instead of offering a separate higher end socket, it sticks the chips into the mainstream socket

I honestly think Intel hasn't even had AMD on the radar for awhile. Atom was Intel's desperate desire to break into tablets and mobile. Remember they started in briefly successful netbooks and then tablets killed that market. Intel has had Nvidia, and all the other ARM guys to compete with all this time and it hasn't been able to crack that nut.

Nah, Atom has very broad applicability across Intel's product line -- low cost notebook/netbook (Bay Trail, Apollo Lake, etc.), network appliances (Atom is actually big there), embedded/IoT (Atom is also the big volume runner there, and this is a $2B+ biz for Intel), etc.

It wasn't just about mobile although mobile was certainly the highest profile new market for it, mainly because it was thought that tablets/phones would eat into the traditional PC

Enthusiast market for Intel is an afterthought. It's where mobile chips get binned for higher clocks and server chips sell as HEDT parts. Which is why all their quads have iGPU in them, and why all the high end multi core parts have quad channel memory. Intel could have easily had a sub $200 quad core sans iGPU, under $200 but why bother? Skylake and Broadwell-E's mainly generate revenues from the mobile and server markets. I highly doubt Broadwell-E sells to many enthusiast, especially not the 10 core $1700 part. It's basically Intel saying we don't really need to sell these because we like to mark them up for the server market anyways.

Enthusiast market leverages silicon primarily developed for other higher volume markets, but clearly they do work with the mobo vendors to build these Z170/X99 boards, and they enable overclocking, add things like XMP profiles, etc. There is more focus on enthusiast than you might think at Intel, and now they've been calling enthusiast out specifically in their earnings calls and even in its restructuring announcement:

While making the company more efficient, Intel plans to increase investments in the products and technologies that that will fuel revenue growth, and drive more profitable mobile and PC businesses. Through this comprehensive initiative, the company plans to increase investments in its data center, IoT, memory and connectivity businesses, as well as growing client segments such as 2-in-1s, gaming and home gateways.

 

[sirmo]

Summit Ridge is a server die (Zeppelin) mounted onto a package that slots into AM4. It's exactly the Intel HEDT playbook, but instead of offering a separate higher end socket, it sticks the chips into the mainstream socket

I think they are different, different sockets and different memory configuration at least.

 

[sirmo]

It wasn't just about mobile although mobile was certainly the highest profile new market for it, mainly because it was thought that tablets/phones would eat into the traditional PC

Well there is always going to be overlap and business opportunities, but your initial theory was that Atom was intended to shut AMD out of the market, when in reality, Intel was chasing a much bigger prize.

 

[Arachnotronic]

Well there is always going to be overlap and business opportunities, but your initial theory was that Atom was intended to shut AMD out of the market, when in reality, Intel was chasing a much bigger prize.

Well, Atom-based Pentium/Celeron chips are still being used to fend off ARM tablets, in the form of low-cost Windows-based 2-in-1 computers, Chromebooks (remember, ARM was widely seen as a threat to Intel via Chromebooks...Bay Trail and its progeny took care of that real quick), and so on.

I mean, why do you think the tablet market is cratering while PC sales -- while not great -- are no longer cratering? Part of it is phablets, but part of it is the fact that you can now buy low cost, fanless, Atom-based Windows notebooks/convertibles with decent components and industrial design for relatively cheaply. Go to Best Buy and you will see what I am talking about here.

Atom is basically Intel's response to any company that says, "we can do cheaper products than you to try to take share from you in your core markets." Atom wasn't good enough, for a variety of reasons, to successfully mount profitable attacks on the core ARM markets like tablets and phones, though, as we are now seeing. It didn't have the cost structure to compete at the low end against the teensy A7s and A53s of the world, and it didn't have the performance to fight the A72/A73s of the world. ARM is clearly better at designing these economical, low-power/high performance cores than Intel is.

 

[Arachnotronic]

I think they are different, different sockets and different memory configuration at least.

Socket is a matter of packaging not die design. With the Skylake gen, Intel will have three platforms: LGA 1151 socket for consumer, LGA 2066 for workstation/HEDT, and LGA 3647 for the EP/EX chips.

For Zen, AMD seems to have PGA 1331 for Bristol Ridge/Summit Ridge APU/CPU and LGA 4xxx for the Naples chip. I don't know if there is a middle ground socket, maybe The Stilt can chime in since he is the expert on these things.

 

[sirmo]

Well, Atom-based Pentium/Celeron chips are still being used to fend off ARM tablets, in the form of low-cost Windows-based 2-in-1 computers, Chromebooks (remember, ARM was widely seen as a threat to Intel via Chromebooks...Bay Trail and its progeny took care of that real quick), and so on.

I mean, why do you think the tablet market is cratering while PC sales -- while not great -- are no longer cratering? Part of it is phablets, but part of it is the fact that you can now buy low cost, fanless, Atom-based Windows notebooks/convertibles with decent components and industrial design for relatively cheaply. Go to Best Buy and you will see what I am talking about here.

Atom is basically Intel's response to any company that says, "we can do cheaper products than you to try to take share from you in your core markets." Atom wasn't good enough, for a variety of reasons, to successfully mount profitable attacks on the core ARM markets like tablets and phones, though, as we are now seeing. It didn't have the cost structure to compete at the low end against the teensy A7s and A53s of the world, and it didn't have the performance to fight the A72/A73s of the world. ARM is clearly better at designing these economical, low-power/high performance cores than Intel is.

I am not saying Atom doesn't have it's place. You misunderstood me. I am saying Atom started off as a very lucrative business for the Netbook market which exploded until Apple crashed the party with the iPad and later Android. Since then Intel has seen the shrinkage of the PC market and an explosion in mobile based ARM devices, and has used Atom to try and fight it. It's still an important product for them I agree, but they have also realized they can't break into the mobile, hence the reason for cancellation of Sofia and Broxton for instance.

My point was only that Atom wasn't a response to AMD.

 

[sirmo]

Socket is a matter of packaging not die design. With the Skylake gen, Intel will have three platforms: LGA 1151 socket for consumer, LGA 2066 for workstation/HEDT, and LGA 3647 for the EP/EX chips.

For Zen, AMD seems to have PGA 1331 for Bristol Ridge/Summit Ridge APU/CPU and LGA 4xxx for the Naples chip. I don't know if there is a middle ground socket, maybe The Stilt can chime in since he is the expert on these things.

Right but Summit Ridge is not a bin of Naples. They have a different memory controller (dual/quad). Intel's HEDT chips are binned Xeons on the other hand.

 

[Arachnotronic]

I am not saying Atom doesn't have it's place. You misunderstood me. I am saying Atom started off as a very lucrative business for the Netbook market which exploded until Apple crashed the party with the iPad and later Android.

This is true, but even when Intel had initial success in netbooks, AMD was fielding arguably superior products based on the Bobcat core. Intel basically had to fight the Bobcat level of performance with cut-down/defeatured Core chips and frankly the die costs (CPU + PCH) as well as the platform BoM was at a huge disadvantage to what AMD was doing. Platform power was worse, too.

That is why Intel built the Bay Trail-M/D, targeted platforms/SoCs based on the same basic Atom cores to try to improve its cost/margin structure in this segment to better compete with AMD on performance/power.

AMD might not have been the only factor driving the development of the Atom cores, but I do think it was a non-trivial factor. Intel was getting slapped around silly by the Cat cores, and AMD was raking in $$$ from them.

Since then Intel has seen the shrinkage of the PC market and an explosion in mobile based ARM devices, and has used Atom to try and fight it. It's still an important product for them I agree, but they have also realized they can't break into the mobile, hence the reason for cancellation of Sofia and Broxton for instance.

Murthy recently said that they cancelled SoFIA and Broxton (for tablets/phones...obviously Broxton lives on as Apollo Lake-M/D/I) because they weren't competitive, but said Intel would try to do mobile platforms again in the future. If they want to be competitive here, I would hope that they use ARM-licensed CPU cores, Atom just isn't going to cut it.

My point was only that Atom wasn't a response to AMD.

I would say that AMD was a factor in building the platforms/SoCs around the basic Atom cores for these markets, but yeah, Atom would have probably been developed anyway.

 

[Arachnotronic]

Right but Summit Ridge is not a bin of Naples. They have a different memory controller (dual/quad). Intel's HEDT chips are binned Xeons on the other hand.

Naples is an MCM of the Zeppelin dies. Four Zeppelins (8 core/dual channel) times four gets you to Naples (32 cores/octal-channel). The bad news for AMD is that you are going to have some overhead in trying to have all of these dies communicate to each other (even if the fabric used to stitch them together is nice). The good news is that this allows AMD to do products like Summit Ridge which are easily compatible with the other AM4 chips (dual channel memory controller, likely the same integrated IOs as Bristol Ridge, etc.), and it might also be easier to build four smaller dies than one humongous die (though there is risk on the packaging side...you try to put together four Zeppelins to make a Naples and your packaging fails...can you even recover the dice?)

 

[StrangerGuy]

Mobile is just a huge cut-throat money trap, the industry was so blinded by the sheer volume of the market and the obscene profits by Apple whose success has little to do with the SoC per se. The real surprise was how long and much losses Intel took before calling it quits; they shouldn't have even bothered.

 

[sirmo]

Naples is an MCM of the Zeppelin dies. Four Zeppelins (8 core/dual channel) times four gets you to Naples (32 cores/octal-channel). The bad news for AMD is that you are going to have some overhead in trying to have all of these dies communicate to each other (even if the fabric used to stitch them together is nice). The good news is that this allows AMD to do products like Summit Ridge which are easily compatible with the other AM4 chips (dual channel memory controller, likely the same integrated IOs as Bristol Ridge, etc.), and it might also be easier to build four smaller dies than one humongous die (though there is risk on the packaging side...you try to put together four Zeppelins to make a Naples and your packaging fails...can you even recover the dice?)

I don't disagree, just that not sure how Naples or Zeppelin is related to my point. My point is that Summit Ridge is its own architecture exclusively intended for the desktop. Naples and Zeppelin by proxy are server parts.

 

[Arachnotronic]

I don't disagree, just that not sure how Naples or Zeppelin is related to my point. My point is that Summit Ridge is its own architecture exclusively intended for the desktop. Naples and Zeppelin by proxy are server parts.

Summit Ridge is just a Zeppelin die mounted on the PGA 1331 package Intel will be going the other way with Kaby Lake-X/Coffee Lake-X, mounting the mainstream dies onto an HEDT-compatible package.

Neither AMD nor Intel is designing dice specifically targeted at the enthusiast desktop market, but they are designing products/platforms for it. Important distinction to make.

 

[sirmo]

Summit Ridge is just a Zeppelin die mounted on the PGA 1331 package Intel will be going the other way with Kaby Lake-X/Coffee Lake-X, mounting the mainstream dies onto an HEDT-compatible package.

Neither AMD nor Intel is designing dice specifically targeted at the enthusiast desktop market, but they are designing products/platforms for it. Important distinction to make.

No it's the other way round. I missed this in your previous post. Naples is the 16c part. Zeppelin is the 32c MCM of Naples.

The name is fitting too because: https://en.wikipedia.org/wiki/Zeppelin had a raid on Naples.

Summit Ridge is it's own SoC specifically designed for the desktop.

 

[Arachnotronic]

No it's the other way round. I missed this in your previous post. Naples is the 16c part. Zeppelin is the 32c MCM of Naples.

The name is fitting too because: https://en.wikipedia.org/wiki/Zeppelin had an raid on Naples.

Nope!

Zeppelin is the 8 core/2 channel die ("ZP"), Naples is the 32 core/64 thread MCM.

 

[sirmo]

Nope!

Zeppelin is the 8 core/2 channel die ("ZP"), Naples is the 32 core/64 thread MCM.

Hmm.. I am confused now, there has been so many conflicting news on this. Zeppelin was one of the early leaks from the Linux kernel too which had 32 cores.

 

[AtenRa]

I believe ZEN is smaller than Broadwell-E die (246mm2). ZEN its made to compete in three markets, Server, Mainstream Desktops and up to $400-500 HEDT segment.
This approach has advantages and disadvantages. You sell a bigger die in Mainstream Desktop segment (4C 8T) but you also spend a once time R&D for a single die for three markets.

AMD is focusing on Desktop because they dont have a desktop market at this time (less than 10% ??). On the other hand Intel already have 90% or more and they know Desktop is not growing so they have put Desktop development on life support at the moment.
Skylake 14nm 2015, Kabylake 14nm 2017, Coffelake 14nm 2018, first time in the last 6-7 years Intel will use the same process (14nm) for more than two CPU generations for Desktop.

 

[tajoh111]

Well the thing is that today AMD can use half the ZEN die and make a huge profit by selling 4C 8T SKU at $199, they cannot do the same with Polaris 10 GPU because BOM is way higher than selling a CPU alone.

For example, if you use half the Polaris 10 die you end up close to Polaris 11 performance territory and because of the higher BOM (memory + PCB) you cannot sell it at the same price as RX 460. You cannot sell it higher because your competition has higher perf product at same price.

Also, the same 200-220mm2 ZEN die will scale from $199 for the 4C 8T SKU and up to $400-600 for the 8C 16T, something no GPU die at the same size can do.

I did a much more thorough analysis here.

RTG and Intel Team Up

In summary, because the board partner margin, the free games, the PCB,memory and board(which is about 50 dollars) is not build into the selling price like a rx480 and rx470, a zen sales is vastly more profitable.

Even the cheapest zen has a margin atleast 3x times the margin of a polaris chip. AMD GPU division barely made any money when they had the highly competitive. 63 million dollars and the 5870 were more competitive, AMD had a whole lineup from top to bottom and 44% marketshare and they were selling at higher prices on a cheaper node. Also AMD was selling 8 million graphics cards, twice what they sell now. So that 63 million dollars, just won't happen in todays market. The same cannot be said for zen.

Zen can turn things around big time. No matter how successful AMD's GPU division is, it will never make AMD enough money where they aren't underfunded. Their CPU division can change that. Bigger market + higher margins = several hundred millions in profit if Zen is decent, billions in profit if zen is good.

@wit
First you base the asumption of a die if 220mm2. Meaning same size as bwe 8c at 224mm2. Can you say with a straight face that thats what you will have used if it wasnt just posted?
I dont think so and with all due respect thats a bit hard for me to accept when i think you use numbers in your calculation you dont quite think yourself is right. You migh help me here.

I think its fair to asume its a bit smaller than p10 going by Dresdenboys estimates but also that there can be more defects and perhaps 4c and 6c will make up for it plenty. But at the end of the day its more or less the same ballpark. By all means - We have a similar business case here for our intends and purposes..

Now if we look at amd portfolio we have the consoles making a profit and seriously i think the rest sans polaris is selling at basement prices not making any money. But somewhere some of the money is comming from and polaris is a good guess.

We have 460 selling at sub 99usd retail. Make profit. Take markup at oem and retail and look at bom for eg. Board vrm ram. There is no way a polaris p10 die cost more than 50usd. And 25usd for the p11. I would asume an upper estimate for zen is those 50usd after 2 or 3 months startup time.

Not even close. a9x which is a much smaller chip costs 37.20 dollars and apples has much better engineers and equipment(or atleast better funded) to get the yields where they want to because of the sheer volume. Polaris is probably closer to 75-80 dollars to make. The larger the die, the more costs go up.

If chip costs were as low as you thought, Nvidia's margins would be even higher, but we still have a 59% margin for their products and this is being inflated by their professional stuff which has 80% margins which makes up a third of their sales.

 

[itsmydamnation]

Hmm.. I am confused now, there has been so many conflicting news on this. Zeppelin was one of the early leaks from the Linux kernel too which had 32 cores.

its not conflicting or confusing at all.

The die is called Zeppelin ( cuz it look like a damn Zeppelin)

1 Zeppelin in AM4 skin is called summit ridge ( upto 8 cores, 2 memory channels)
1 Zeppelin in server sku == no idea
2 Zeppelin in MCM in server sku == snowy owl ( upto 16 cores, 4 memory channels)
4 Zeppelin in MCM in server sku == naples (upto 32 cores 8 memory channels)

I believe ZEN is smaller than Broadwell-E die (246mm2). ZEN its made to compete in three markets, Server, Mainstream Desktops and up to $400-500 HEDT segment.
This approach has advantages and disadvantages. You sell a bigger die in Mainstream Desktop segment (4C 8T) but you also spend a once time R&D for a single die for three markets.

technically summit ridge doesn't compete with skylake thats raven ridges job but the fact most people running with an i5/i7 aren't using the on soc gpu means thats dead weight that it is carrying.
I think most people are overestimating the size of Zepplin so that easily allows amd to tackle a ~120-130mm part with a 180-200mm part without to much issue.

 

[KTE]

These measures are useless: they just show what percentage of the events happened in blender vs the whole system, or in which function they happen; this does not tell how many such events happened

I don't think you use that tool correctly.

Not when bound to process.

I worked the figuee out myself using the instr totals.

Didn't go off percentage from the tool.

Sent from HTC 10
(Opinions are own)

 

[Ancalagon44]

Socket is a matter of packaging not die design. With the Skylake gen, Intel will have three platforms: LGA 1151 socket for consumer, LGA 2066 for workstation/HEDT, and LGA 3647 for the EP/EX chips.

For Zen, AMD seems to have PGA 1331 for Bristol Ridge/Summit Ridge APU/CPU and LGA 4xxx for the Naples chip. I don't know if there is a middle ground socket, maybe The Stilt can chime in since he is the expert on these things.

Not correct. Die design does matter in socket choice, since the memory controller is on the CPU. Therefore, even if you could design a motherboard that somehow had DDR2 slots and an AM4 slot, an AM4 CPU would not work on it since no AM4 CPUS support DDR2.

Its not a matter of simply changing the pins.