Zen is faster than Broadwell-E clock for clock

[dalekphalm]

22 minutes ago, BrightCandle said:

This is a common misconception that the reason Intel hasn't delivered more performance is due to lack of competition. The issue is that isn't really the case. To be frank over Intel's long life as the largest CPU producer in the world it has almost exclusively been competing with itself. Intel sells product on the basis of having something better than its previous product that compel its customers to buy it. If Intel could deliver 2x performance on each silicon node they would do it, that would vastly improve their profits.

 

Intel's pricing at any given market position has actually stayed remarkably steady for over a decade, they are one of the more consistent silicon per $ companies out there. AMD doing anything isn't going to change what Intel does, Intel's path for now has been set for 6 years or so these things are so long lead time that there will be no reaction over than perhaps a binned higher clock CPU or a reduction in some prices to compete, that is it. Reality is the silicon market is mostly about selling the same customers compelling upgrades so competition wise they mostly compete with themselves most of the time.

I disagree with part of this.

 

If Intel could double the performance per node shrink, and had no viable competitors at the given price point, I do not think they would do it.

 

Why would they? They can maximize profits by dolling out regular and reliable performance improvements over a long time. They might research the tech to double performance, but they would hold it in their back pocket.

 

Simply doubling the performance per node (if even possible) without an external force putting pressure on them (competition), would simply make it more expensive R&D-wise. They'd have to invest even more money to make things even faster, to keep up with this artificially created need for such a huge gain.

 

While I don't think Intel is slacking off, exactly, they're certainly R&Ding a ton of stuff that would no doubt be amazing, that doesn't see the light of day for years.

[SpaceGhostC2C]

13 minutes ago, BrightCandle said:

This is a common misconception that the reason Intel hasn't delivered more performance is due to lack of competition. The issue is that isn't really the case. To be frank over Intel's long life as the largest CPU producer in the world it has almost exclusively been competing with itself. Intel sells product on the basis of having something better than its previous product that compel its customers to buy it. If Intel could deliver 2x performance on each silicon node they would do it, that would vastly improve their profits.

 

Intel's pricing at any given market position has actually stayed remarkably steady for over a decade, they are one of the more consistent silicon per $ companies out there. AMD doing anything isn't going to change what Intel does, Intel's path for now has been set for 6 years or so these things are so long lead time that there will be no reaction over than perhaps a binned higher clock CPU or a reduction in some prices to compete, that is it. Reality is the silicon market is mostly about selling the same customers compelling upgrades so competition wise they mostly compete with themselves most of the time.

You are mostly right, but a durable goods monopolist will still extract a monopoly rent, besides timing quality and pricing taking into account self-competition. So I agree with you, Intel won't suddenly panic and come up with a quantum computer desktop CPU or anything, but we can expect an impact on prices and on product design using the existing technology (like how they do the whole server-enthusiast-gamer-office segmentation, etc). Then again, competition is not just about having a decent product, but also about all the OEM agreements etc, which don't really resemble a "free market" (well, having just 2 companies doesn't either, but I digress  ).

 

What I would add to your comment is that innovating is always more difficult than imitating. Given the same engineering prowess and R&D spending, you will obtain less frequent and/or smaller performance gains when you are at the frontier than when you are catching up. Intel has to figure out how to do something nobody has done, AMD has to figure out how to get what Intel has. That's also why it is perfectly credible that, once settled for a similar architecture, AMD would get some decent catching up done, regardless whether it is latest generation or 2 generations ago intel. It's just easier to get in the same ball park if you don't need to reinvent the wheel.

 

Bulldozer, on the other hand, was truly innovation, and innovation went wrong, that's why it was such a hit for AMD, just like Intel made the biggest innovation moving into Sandy Bridge, but they can't just make a similar jump every day. Iterating in the core i lineup will give them some improvements over time, but it's much easier for AMD to make a bigger jump becoming "core-i-ish", as they "just" need to understand how intel got it to work and how to replicate it while avoiding patent roadblocks. Of course, they can tweak their design to try and improve over intel in some dimensions, but their fallback is to replicate intel within legal reason. When you are the one at the top, your fallback plan is just failing miserably and staying where you were, because there's no one to imitate, so you think it twice before getting too revolutionary if you can still milk the same cow in terms of climbing a performance ladder  

[Remixt]

1 hour ago, BrightCandle said:

This is a common misconception that the reason Intel hasn't delivered more performance is due to lack of competition. The issue is that isn't really the case. To be frank over Intel's long life as the largest CPU producer in the world it has almost exclusively been competing with itself. Intel sells product on the basis of having something better than its previous product that compel its customers to buy it. If Intel could deliver 2x performance on each silicon node they would do it, that would vastly improve their profits.

 

Intel's pricing at any given market position has actually stayed remarkably steady for over a decade, they are one of the more consistent silicon per $ companies out there. AMD doing anything isn't going to change what Intel does, Intel's path for now has been set for 6 years or so these things are so long lead time that there will be no reaction over than perhaps a binned higher clock CPU or a reduction in some prices to compete, that is it. Reality is the silicon market is mostly about selling the same customers compelling upgrades so competition wise they mostly compete with themselves most of the time.

They aren't spending a tremendous amount of effort on R&D is my point. 

[BrightCandle]

4 minutes ago, Remixt said:

They aren't spending a tremendous amount of effort on R&D is my point. 

As  an absolute number and as a percentage of revenue Intel is one of the worlds largest researchers. I am sorry but the fundamental data about Intel disagrees with this position. Its a very popular position to hold but its not the case, Intel would make money if it kept advancing performance more rapidly. Intel is hampered by the laws of Physics and the fundamental nature of the technology approach and they have tried to research a way out of this hole for a decade.

 

None of this is true, lots of people believe its true that Intel is lazy and just holding back the good stuff until AMD does something but its nonsense. AMD isn't Intel's main competitor, Samsung and ARM are.

[Remixt]

9 minutes ago, BrightCandle said:

As  an absolute number and as a percentage of revenue Intel is one of the worlds largest researchers. I am sorry but the fundamental data about Intel disagrees with this position. Its a very popular position to hold but its not the case, Intel would make money if it kept advancing performance more rapidly. Intel is hampered by the laws of Physics and the fundamental nature of the technology approach and they have tried to research a way out of this hole for a decade.

 

None of this is true, lots of people believe its true that Intel is lazy and just holding back the good stuff until AMD does something but its nonsense. AMD isn't Intel's main competitor, Samsung and ARM are.

They aren't trying to re-invent the wheel, they are making the wheel more efficient. Until AMD or any other company forces them to rethink their strategy we won't see anything new/revolutionary from them in at least a decade.

[Dabombinable]

7 hours ago, Vode said:

Kinda AMD's fault for marketing them as 8-cores though.

Its one of those technicality things.

 

What people expect in regards to the core count after every single CPU ever made previously (6 cores 6 threads):

http://www.cpu-world.com/CPUs/K10/AMD-Phenom II X6 1100T Black Edition - HDE00ZFBK6DGR (HDE00ZFBGRBOX).html

 

What AMD sold to them under the guise of 6 cores (really 3 modules 6 cores-had a bit more of a dig into CMT):

http://www.cpu-world.com/CPUs/Bulldozer/AMD-FX-Series FX-6350.html

 

TL;DR AMD would have been better off just shrinking K10 to 32nm (and adding L3 cache), its more efficient and the die size would actually be a lot smaller, even for 8 K10 cores.

[Vode]

9 hours ago, Remixt said:

They aren't spending a tremendous amount of effort on R&D is my point. 

So having the most advanced lithography process in the world comes from nothing?

 

And no Samsungs 14nm is not the same as Intels. Intels is actually smaller, better.

[BrightCandle]

1 hour ago, Vode said:

So having the most advanced lithography process in the world comes from nothing?

 

And no Samsungs 14nm is not the same as Intels. Intels is actually smaller, better.

Its more than that. We have had a decade of incremental upgrades from both AMD and Intel in terms of performance with Intel leading that charge and now suddenly AMD is producing an amazing CPU where its not even going to be faster than Intel's latest from last year? If Intel is so lazy and AMD is better then why didn't they just make their CPU 5x faster and just take the entire market easily? Maybe its a bit more fundamental than that?!

 

Its beyond ridiculous to even be having this discussion, it has no basis in reality there is literally no evidence supporting it. Its so common to hear this chain of thinking however, that some how companies and people don't want to do better unless someone is pointing a gun at them, its actually the opposite of how this works in reality. Its a really dumb argument and the position that is held so strongly by those most ignorant of how CPUs/silicon products are made.

[SpaceGhostC2C]

10 hours ago, BrightCandle said:

Its more than that. We have had a decade of incremental upgrades from both AMD and Intel in terms of performance with Intel leading that charge and now suddenly AMD is producing an amazing CPU where its not even going to be faster than Intel's latest from last year? If Intel is so lazy and AMD is better then why didn't they just make their CPU 5x faster and just take the entire market easily? Maybe its a bit more fundamental than that?!

 

Its beyond ridiculous to even be having this discussion, it has no basis in reality there is literally no evidence supporting it. Its so common to hear this chain of thinking however, that some how companies and people don't want to do better unless someone is pointing a gun at them, its actually the opposite of how this works in reality. Its a really dumb argument and the position that is held so strongly by those most ignorant of how CPUs/silicon products are made.

I think you start to derail, mate. The incentives faced by a durable goods monopolist to release certain products at a certain time, as opposed to a different combination, is one thing. "Intel is so lazy and AMD is better" is nothing anyone has said except you in your last post. Strategy is not laziness. AMD being better or worse doesn't even enter the picture. You had a good point, but seem to be falling to fanboyism now. Not trying to judge too quick, though.

The point is, if you develop haswell, you don't release skylake 3 months later, whether your R&D has it ready or not. You do it late enough to allow for Haswell sales, but soon enough for it to stay relevant. You have 6 core chips, but you don't make mid-range versions of them. You simply choose how to play your cards. If competition increases, you choice set narrows, and you do less what you want or what maximizes your profit, and more what you have to.

Still, it's not like one AMD launch is going to turn this into Adam Smith's paradise... 

 

[Dash Lambda]

12 hours ago, BrightCandle said:

-snip-

Imagine both you and your neighbor across the street have a lemonade stand. You're both using the same lemonade mix, and you're both trying to attract the other's customers by changing up the amount of mix and sugar you use. Eventually, you decide to invest in a better, more expensive mix that tastes better. After a while, you've attracted most of the customers on the block, and your neighbor's stand is almost deserted. He can improve his lemonade by using a little more sugar, but his mix just isn't as good as yours, and everyone knows it.

 

So now you've won over the lemonade market, and the lemonade stand across the street is deserted while he keeps changing the sugar and hanging banners that say "improved formula!" and "50% more sugar!" while you have all the customers. At that point, what motive do you have to put even more money into a new mix?

[Guest]

9 minutes ago, Dash Lambda said:

Imagine both you and your neighbor across the street have a lemonade stand. You're both using the same lemonade mix, and you're both trying to attract the other's customers by changing up the amount of mix and sugar you use. Eventually, you decide to invest in a better, more expensive mix that tastes better. After a while, you've attracted most of the customers on the block, and your neighbor's stand is almost deserted. He can improve his lemonade by using a little more sugar, but his mix just isn't as good as yours, and everyone knows it.

 

So now you've won over the lemonade market, and the lemonade stand across the street is deserted while he keeps changing the sugar and hanging banners that say "improved formula!" and "50% more sugar!" while you have all the customers. At that point, what motive do you have to put even more money into a new mix?

This is an absolutely retarded analogy. Competition is not the only reason to innovate. There are always bigger and better goals that need to be attained. 

 

In this case, improving CPU technology affects the entire landscape of technology as it is a quintessential piece of hardware necessary for all technology to work. Artificial Intelligence, machine learning, etc. are all for the taking. 

 

This is far too simplistic and is not even true for the field of technology. 

[Dash Lambda]

5 minutes ago, Alexokan said:

This is an absolutely retarded analogy. Competition is not the only reason to innovate. There are always bigger and better goals that need to be attained. 

 

In this case, improving CPU technology affects the entire landscape of technology as it is a quintessential piece of hardware necessary for all technology to work. Artificial Intelligence, machine learning, etc. are all for the taking. 

 

This is far too simplistic and is not even true for the field of technology. 

I agree that it's simplistic, that's the point. Inaccurate? Not really, for what it's supposed to say.

What you're saying is akin to saying that the 'a joke is like a frog' analogy is stupid because a frog is an animal and a joke is a nonphysical concept.

 

Intel and AMD are companies, they run a business. They exist to make money. Technological progress is not enough in and of itself to motivate companies to invest heavily in R&D.

 

Competition isn't the only reason to innovate, in fact it's not directly a reason at all. For a company like Intel, AMD, or Nvidia, the reason to innovate is money. Competition leads to a system where innovation is usually the best way to get more money, so companies invest more into R&D and progress faster. No competition can lead to a system where big innovation isn't very profitable, so companies invest less in it and instead make tweaks and shift focus to more profitable options (Those options can be innovation in other areas, though, like efficiency over performance).

 

There are a lot of factors that change the landscape and there are a lot of components to making money that change with that landscape. Competition just happens to have a considerable effect on competition.

 

So, basically, it's not that a lack of competition means a lack on innovation. It's that competition gives greater incentive for innovation, and therefore innovation receives more focus.

 

If competition were the sole impetus for innovation, we wouldn't have Skylake. But at the same time, if there was competition, we'd have something better.

[BrightCandle]

1 hour ago, Dash Lambda said:

Imagine both you and your neighbor across the street have a lemonade stand. You're both using the same lemonade mix, and you're both trying to attract the other's customers by changing up the amount of mix and sugar you use. Eventually, you decide to invest in a better, more expensive mix that tastes better. After a while, you've attracted most of the customers on the block, and your neighbor's stand is almost deserted. He can improve his lemonade by using a little more sugar, but his mix just isn't as good as yours, and everyone knows it.

 

So now you've won over the lemonade market, and the lemonade stand across the street is deserted while he keeps changing the sugar and hanging banners that say "improved formula!" and "50% more sugar!" while you have all the customers. At that point, what motive do you have to put even more money into a new mix?

Unlike lemonade processors last for decades. The only reason to upgrade a processor is for more performance, as I stated before Intel's primary competition is its prior products. There is a big difference between selling a product that spoils and is fully consumed and one that doesn't have either property.

[Dash Lambda]

9 minutes ago, BrightCandle said:

Unlike lemonade processors last for decades. The only reason to upgrade a processor is for more performance, as I stated before Intel's primary competition is its prior products. There is a big difference between selling a product that spoils and is fully consumed and one that doesn't have either property.

Which is why we have new processor generations that make small, incremental improvements upon the last.

Like I said, innovation doesn't just stop, it accelerates and decelerates.

Making that nitpick on the analogy is missing the point of the analogy.

[BrightCandle]

33 minutes ago, Dash Lambda said:

Which is why we have new processor generations that make small, incremental improvements upon the last.

Like I said, innovation doesn't just stop, it accelerates and decelerates.

Making that nitpick on the analogy is missing the point of the analogy.

And there are 10+ top tier silicon companies competing for this business in the world against each other and none of them is able to drastically out perform the other. Intel isn't just sitting in a market on its own, its fighting a myriad of companies on all areas of its business its a long long way from being a monopoly. But all of these companies are facing the same problem, somehow I think that is more likely what they are telling us ("its physics") than some macro economics theory about lack of competition.

[SpaceGhostC2C]

1 hour ago, BrightCandle said:

 somehow I think that is more likely what they are telling us ("its physics") than some macro economics theory about lack of competition.

Actually, that's micro  

[Dash Lambda]

1 hour ago, BrightCandle said:

And there are 10+ top tier silicon companies competing for this business in the world against each other and none of them is able to drastically out perform the other. Intel isn't just sitting in a market on its own, its fighting a myriad of companies on all areas of its business its a long long way from being a monopoly. But all of these companies are facing the same problem, somehow I think that is more likely what they are telling us ("its physics") than some macro economics theory about lack of competition.

I'm talking about the x86 chips in desktops and most laptops. Aside from a company called "VIA," which I've never heard of before today, Intel's only competition there is AMD.

 

The limited scaling capability of silicon is definitely a roadblock, but there are plenty of potential ways to still improve without just trying to make a silicon chip smaller. We still would've seen more progress if there were competition.

[dalekphalm]

3 minutes ago, Dash Lambda said:

I'm talking about the x86 chips in desktops and most laptops. Aside from a company called "VIA," which I've never heard of before today, Intel's only competition there is AMD.

 

The limited scaling capability of silicon is definitely a roadblock, but there are plenty of potential ways to still improve without just trying to make a silicon chip smaller. We still would've seen more progress if there were competition.

VIA mostly does custom embedded SoC's, so chips for billboards or whatever else. They used to be huge in the PC Space, they mainly made chipsets for motherboards back in the 90's.

[Dabombinable]

4 minutes ago, dalekphalm said:

VIA mostly does custom embedded SoC's, so chips for billboards or whatever else. They used to be huge in the PC Space, they mainly made chipsets for motherboards back in the 90's.

Apollo Pro 133A-you'll be missed.

[SpaceGhostC2C]

19 minutes ago, Dash Lambda said:

I'm talking about the x86 chips in desktops and most laptops. Aside from a company called "VIA," which I've never heard of before today, Intel's only competition there is AMD.

 

The limited scaling capability of silicon is definitely a roadblock, but there are plenty of potential ways to still improve without just trying to make a silicon chip smaller. We still would've seen more progress if there were competition.

I think competition would be most noticeably in ways that aren't necessarily about technology improvement. For example, intel divide CPUs between locked and unlocked, and further divide chipsets between those that allow overclocking and those that don't (even though changing cpu settings in BIOS isn't really a chipset feature). So, you must have a Z chipset and a K CPU or no multiplier OC for you, even if your need for PCIe lanes or SATA ports is satisfied by lower end chipsets. This is called "bundling" (more microeconomics ), and is a way to perform price discrimmination across consumer types.

AMD, having lower performance, tried to exploit its OC reputation making all CPUs unlocked. Intel, having a wide performance edge, doesn't need to react to this, leaves low end to AMD if they want it (which is different from HTPC or low TDP,  I mean low end here) and keeps its bundling practices.

Now, if a competitor was able to get near intel performance (doesn't matter it's not exactly the same), competition in lesser features becomes more relevant. If the competitor chooses an "always unlocked" policy as well, it makes it much harder to continue with price discrimination through bundling, and would possibly lead to a reduction in the number of CPU tiers in their offering.

Having only two companies, though, it could always happen that AMD imitates intel practices with its new architecture, putting little pressure in anything but prices.

 

In the end, it's not so much, or not only, about technology development, but about how to use the developed technology depending on your market position.