FEATURED AMD ZEN Discussion (Previous Rumor Thread)

chrisjames61 said:

You don't know much about business do you? That is the perfect way to go bankrupt. Sell twice as much and lose twice as much money. How are all the high volume Android and pc manufacturers doing? The answer is they make little to NO profit. The companies that make big profits make that money by having fat margins. Apple, Intel etc..... Number sold are secondary to the amount of profit per unit.

Click to expand...

I'll admit to not having a great amount of business knowledge, but I think Android manufacturers are a bad example. That market is much more diluted than the proc market. And it is my understanding that keeping your competition from making a dollar is almost as good as making a dollar yourself. As long as overhead is taken care of. It's reasonable to expect that if they can keep Intel from making a massive profit, then they will by default corner the market. An making NO profit is MUCH preferred to having a large amount of unbought cpus - meaning a large amount of debt. Every cpu that AMD sells is *conceivable* one customer that wont be going to Intel, thereby denying them hundreds of dollars in profit with each sale. These kinds of things may be what is keeping me from owning a fortune 500 company though

If you aren't making money on each processor sold, then you need to be making money on accessories, licensing, etc.

That's how sony/Microsoft make money at the beginning of a console generation (they lose on the consoles at first but make it back on game licensing). AMD doesn't have that kind of an option with just selling CPUs, as the only other required piece is a motherboard that I am unsure if they recoup anything from the manufacturer for.

I don't know about zero profit per unit , but I can imagine reducing the percentage per unit to promote more units sold. That is a win-win. Fewer competitor units sold+more units sold=sound corporate strategy. Providing they sell in those numbers. It would be a gamble , but every new product is a gamble. You pays your money and you takes your chances.

Dolk said:

Even if AMD flops and is dissolved. By the time that happens, x86 architecture will either be changed and new players will enter

Click to expand...

Right, because so many new players have entered x86 in the decades it has been around... TI dropped out way back in the Pentium era, Transmeta has been dead for almost a decade now, and Via is only around because they bought IDT/Centaur, so they don't count as a new player.

petteyg359 said:

Right, because so many new players have entered x86 in the decades it has been around... TI dropped out way back in the Pentium era, Transmeta has been dead for almost a decade now, and Via is only around because they bought IDT/Centaur, so they don't count as a new player.

Click to expand...

Im still wondering when are they gonna bring back my lovely Commodore Amigas to replace this piece of crap PCs we have nowadays, everyone said it was just a games machine but did everything better then the other brands at the time.

Alright so let me clear the air on some rumors and thoughts about the future:

ARM will most likely NOT be the next architecture the PC market moves toward. x86 will stick around but how it gets better will not be due to transistor size.
Rather than ARM, people should be paying attention to Die size, and why Intel bought Altera.
AMD has given some indication that HBM will be on the new Zen CPUs. Which models is unknown. But what is clear is that the HBM will be used in a heterogeneous memory architecture (Think of a crap ton of RAM right on the CPU that everyone can use).
With the market changing to Die Size rather than Transistor Size, overclocking will be come very difficult. Power and Clock gating techniques for large dies has been handled exclusively by FPGA companies, IBM and Intel. AMD is starting to get a strong grasp on this very important technique: http://www.realworldtech.com/steamroller-clocking/

overclocking become difficult? I read that as challenging.....I like challenging.

I agree...where there's a will there's a way

Overclocking is not up to the consumer, its up to the engineer. They enable the feature, and they can easily remove it, and a simple BIOS mod will not be able to do anything.

Since die size is increasing, power needs to be carried through a more denser package than what current CPUs are used to. This means there is a higher IR drop across large areas of the CPU. This phenomena causes issues in clocking and reading voltage signals to differentiate between 0s and 1s. In order to fix this issue, several fixes can be performed: add more power, better power gating, or lower frequency/power budget. Some of these problems and solutions are linear, and some are exponential. Overall it creates a difficult problem for engineers, and the first thing they will do (as seen by Intel) is lock down frequency. If they know what the frequency will be at all times, the CPU can tweaked to work more efficiently with the increase in transistors.

Now another reminder for everyone: Moores Law is NOT dead, and will not be killed. The increase in transistors does not always mean that transistor size has to go down. You can easily increase the size of the die. This means you get more features in your CPU/Cores. Either increased IPS within specific modules of the CPU, or new features like 256-bit FP modules are simple examples that we are seeing.

Dang Dolk....when your done peeing in my cheerios let me know....AMD has always been more enthusiast friendly, you think that will fall by the wayside with the zen?

Would something like the C2Q with dual chips help with that problem? Or even multiple chips ,server style

NZKshatriya said:

@chrisjames you know what I mean.

A decent percentage less than $1100+ for a cpu.

Click to expand...

chrisjames61 said:

I hear you. W want a very good UNLOCKED cpu at a good price. Unlike the gouging that we get from Intel for an unlocked cpu.

Click to expand...

Kind of an odd coupling of quotes above. If a good percentage less than $1,100 is your benchmark for a "very good UNLOCKED cpu at a good price", then Intel's selection isn't that bad. Here's current Microcenter prices for unlocked Intel CPUs for various sockets from $50 to $500:

Celeron G3258 (LGA 1150) - $50
i5 4690K (LGA 1150) - $200
i5 6600K (LGA 1151) - $230
i7 4790K (LGA 1150) - $280
i7 5820K (LGA 2011-v3) - $300
i7 6700K (LGA 1151) - $350
i7 5930K (LGA 2011-v3) - $500

I was looking forward to Zen for my first AMD build is some time. But the most I ever paid for a CPU was $300 for an i7 5820K a while back. So, for me, Zen can't be anywhere near $1,100. The low end unlocked version needs to be $300 or less. They can charge what ever they want for the server chips, I don't really care. I was also interested in Fury until the pricing came out. Except for my current GTX 980, 9 of the last 10 GPUs I purchased were AMD going up the chain from the HD 7750, 7770, 7850, 7950, 7970, R9 270, 280X, 290 up to the 290X. But the GTX 980 @ $425 vs. base Fury @ $570 made the decision to back to Nvidia easy. AMD needs to be able to come in low on Zen pricing or the game is over for them. Just my opinion.

Zen's overclocking abilities are still unknown. Will it be as good as the Thubans? I don't know. To early to say. My guess is that it will be just like everyone expects: Another AMD CPU that is easy to OC and fun to have.

When I say that OCing will go away, I'm talking about near future CPUs. This would be after 10nm and 7nm are vetted through consumers/foundry testing.

Dolk said:

Zen's overclocking abilities are still unknown. Will it be as good as the Thubans? I don't know. To early to say. My guess is that it will be just like everyone expects: Another AMD CPU that is easy to OC and fun to have.

When I say that OCing will go away, I'm talking about near future CPUs. This would be after 10nm and 7nm are vetted through consumers/foundry testing.

Click to expand...

whew.....for gods sake man, dont scare me like that again

Well, as with modchips on consoles, Ill bet somewhere down the line, someone will find a way to modify something somewhere, to make it overclockable, even if its not designed to be.

Isn't that the entire purpose of machines? To use them in ways they weren't meant to be?

- - - Updated - - -

@DaveB

I still say Nay to EvilInside!

NZKshatriya said:

Well, as with modchips on consoles, Ill bet somewhere down the line, someone will find a way to modify something somewhere, to make it overclockable, even if its not designed to be.

Isn't that the entire purpose of machines? To use them in ways they weren't meant to be?

- - - Updated - - -

Click to expand...

not even close.

Tell that to the idiots at rodeos who do souped up tractor-pulls....

NZKshatriya said:

Tell that to the idiots at rodeos who do souped up tractor-pulls....

Click to expand...

Easy there.....ya know what kind of engineering it takes to shoehorn 4 aries hemis in an allis chalmers chassis?

Alaric said:

I don't know about zero profit per unit , but I can imagine reducing the percentage per unit to promote more units sold. That is a win-win. Fewer competitor units sold+more units sold=sound corporate strategy. Providing they sell in those numbers. It would be a gamble , but every new product is a gamble. You pays your money and you takes your chances.

Click to expand...

That isn't how it works. AMD needs to make decent money on each processor sold. Volume or marketshare don't equate to profits. Margins are what is important.

- - - Updated - - -

petteyg359 said:

Right, because so many new players have entered x86 in the decades it has been around... TI dropped out way back in the Pentium era, Transmeta has been dead for almost a decade now, and Via is only around because they bought IDT/Centaur, so they don't count as a new player.

Click to expand...

Yeah he is dead wrong. The barrier to entry is way, way too much money for new players in the mainstream cpu market. It cost billions and billions of dollars then you get your lunch eaten by Intel. Obviously no one wants a part of this.

chrisjames61 said:

That isn't how it works. AMD needs to make decent money on each processor sold. Volume or marketshare don't equate to profits. Margins are what is important.

- - - Updated - - -



Yeah he is dead wrong. The barrier to entry is way, way too much money for new players in the mainstream cpu market. It cost billions and billions of dollars then you get your lunch eaten by Intel. Obviously no one wants a part of this.

Click to expand...

In many cases that IS how it works. Smaller per unit margins and high volume sales. That is a WalMart supplier contract in a nutshell. That is why Chevy parts are cheaper than Honda parts. The manufacturing cost is amortized over a larger production run. I have a relative , who as CEO of an international corporation , built factories in France , Germany , and China. The China factory has the smallest per unit margin and by far the highest profit in dollars. To quote Stalin , quantity has a quality all it's own.