I guess sales could be interpreted as shipments or value of those shipments? My guess is at least one of'em will come true since AMD isn't releasing anything compelling for much of the notebook market this year or coming one.
I guess sales could be interpreted as shipments or value of those shipments? My guess is at least one of'em will come true since AMD isn't releasing anything compelling for much of the notebook market this year or coming one.If you stop chasing high volume/low margin products its surely basically axiomatic that you lose sales? Might gain profit though
If you stop chasing high volume/low margin products its surely basically axiomatic that you lose sales? Might gain profit though
Yes I understand that but we don't know if the 8 core Zen will be more efficient than an 8 core Intel, sure it'll be more efficient than AMD's very own BD but that wasn't my point
That's a deduction we can make but one that can't be proven, since there's no FX on 22nm FinFET. That's why I usually try to steer away from mass hypotheses that can't be proven, even if they're true.A lot of people are talking efficency without knowing what they are talking about, not speaking of you, but let s look at the facts before branding BD innefficient, let s see what is the part between design and process that is the most influencial in this respect.
Measurements, that i can provide if necessary, show that Intel s 4770K need 1.0V at 3.5GHz, extrapolating from AMD s FX8370E that need 1.16V at 3.3 we can easily compute (knowing the voltage/frequency caracteristic) that at the same 3.5 a FX will need 1.195V.
Consequences are straightforward, the FX has to dissipate roughly 40% more due to a higher voltage requirement, so ultimately Haswel lhas no perf/Watt advantage when it comes to uarch, quite the contrary, all the advantage is in the process, at 100%, without this process advantage their chip would be beaten hand down perf/Watt wise in anything compute intensive by a regular Piledriver based FX, let alone with an Excavator base FX.
A lot of people are talking efficency without knowing what they are talking about, not speaking of you, but let s look at the facts before branding BD innefficient, let s see what is the part between design and process that is the most influencial in this respect.
Measurements, that i can provide if necessary, show that Intel s 4770K need 1.0V at 3.5GHz, extrapolating from AMD s FX8370E that need 1.16V at 3.3 we can easily compute (knowing the voltage/frequency caracteristic) that at the same 3.5 a FX will need 1.195V.
Consequences are straightforward, the FX has to dissipate roughly 40% more due to a higher voltage requirement, so ultimately Haswel lhas no perf/Watt advantage when it comes to uarch, quite the contrary, all the advantage is in the process, at 100%, without this process advantage their chip would be beaten hand down perf/Watt wise in anything compute intensive by a regular Piledriver based FX, let alone with an Excavator base FX.
:whiste:
http://www.anandtech.com/show/9146/the-samsung-galaxy-s6-and-s6-edge-review/2
Voltage is ultimately meaningless without other information.
15W TDP Core i5 Broadwell will go against Carrizo not Kaveri in 2015.
The other information is comsumption in function of frequency, and this info is available as well, isnt it, so you re just assuming that still you dont understand how things works it s mandatory than no one can understand..
Once you have this parameter you have the dynamic behaviour of the chip, if a voltage is increased by 20% and the frequency kept constant TDP will mechanicaly increase by 44% over this value assuming the chip is not working at an axtreme of its curve.
In the exemple i gave the FX8370E consume 80W at stock with Prime 95, voltage is 1.16V, increase it by 10% and consumption will rise to 97W, reduce it to 1V, if the chip could be stable, and it will be down to 59.5W.
So all that is missing in your list are the power comsumption numbers for frequencies and voltages, all infos available for the FX.
What is useless actualy is to post irrelevancies that have no ressemblance with what i stated when it comes to the infos at disposal.
Is incorrect.Measurements, that i can provide if necessary, show that Intel s 4770K need 1.0V at 3.5GHz, extrapolating from AMD s FX8370E that need 1.16V at 3.3 we can easily compute (knowing the voltage/frequency caracteristic) that at the same 3.5 a FX will need 1.195V.
Consequences are straightforward, the FX has to dissipate roughly 40% more due to a higher voltage requirement, so ultimately Haswel lhas no perf/Watt advantage when it comes to uarch, quite the contrary, all the advantage is in the process, at 100%, without this process advantage their chip would be beaten hand down perf/Watt wise in anything compute intensive by a regular Piledriver based FX, let alone with an Excavator base FX.
This has nothing do with what my post. The bolded is true however, my point is your conclusion and comparison between haswell and the FX series is incorrect because you cannot make any conclusions between chips on different architectures and processes.
The A57 and A53 run at nearly the same voltages at 1.9/1.3 ghz yet the A57 runs at a greater clockspeed and uses more power.
Architecture has a tremendous impact on power consumption so the statement and conclusion.
You are attributing the entire voltage difference to process without any concern to architecture.
If you stop chasing high volume/low margin products its surely basically axiomatic that you lose sales? Might gain profit though
We can, because we are talking of electrical parameters of the basic building blocks, transistors.
This is pretty much why I'm going to stop this conversation. Can't argue with someone who thinks like this.
They said they will lose sales. And if the best case scenario is a slightly growth in sales, coupled with the low starting number, that means AMD isn't having much hopes for its products on the consumer market.
Indeed, why bother to argue when being totaly cluless about what is a transistor, let alone how it works...?.
You seems also to be unaware that whatever the designs the parts of a CPU that works at actual clock frequencies are the same, otherwise you would have IPC differences of one order of magnitude between designs...
This is thoses parts that are dependant of the process, other parts are much less critical generaly.
You seems also to be unaware that whatever the designs the parts of a CPU that works at actual clock frequencies are the same, otherwise you would have IPC differences of one order of magnitude between designs...
This is thoses parts that are dependant of the process, other parts are much less critical generaly.
Indeed, why bother to argue when being totaly cluless about what is a transistor, let alone how it works...?.
Yes because voltage caracteristic of a CPU is a consequence of the transistors electrical parameters and nothing else, no new uarch can make a transistor switch faster and/or conduct better than what is allowed by its intrinsical electrical characteristics.
They will lose sales but so will everyone else. They made no statements about market share.
You're modelling Haswell and Bulldozer as one big transistor. That's all your analysis is good for.
P = V^2 *f is only good for approximating power between the exact SAME design running the SAME code.
What about capacitance from layout and transistor properties? Device sizing? Switching activity from glitches (bulldozer was FULL of these!)? Power gating granularity? Or just good design from not having a decoder bottleneck while the execution units spin their wheels?
This makes me sad really. You love AMD so much that you spend time researching on the Internet to support them, but confirmation bias makes it impossible for you to learn how things actually work.
The Voltage-Frequency curve of a CPU is heavily affected by:
1) Pre-silicon frequency targets
2) Wire RC values and its delay proportion to device delay
3) And last but not least, implementation quality
(not saying that better transistors don't help, but other things can help or hinder the final VF curve)
A mere quote from a post on here will do.Oh so you want a signed statement is it, guess what no does that on AT
Looks like he quoted that "news" article to kick along discussion, for no were in that thread, does he state he believes it.& if the OP wasn't expecting such results (even for a sec) he would've stated his position in the first few pages itself wouldn't he?
By the same token no one here has stated explicitly that Zen will be x times better than BD or y times more efficient, will sell z multiples more than what opterons are selling today.
Oh hey look here now what would you call that - prediction, statement whatever else?
That you are cluless
That you are cluless of the curves doesnt mean that everybody is cluless as well, do your homework, as for me i m talking out of real figures, no need to flood your post with assumptions that give an impression of argumentation, for instance what is the relevance of Power gating granularity when we are talking of designs working at full throughput..?..
The end result is a device whose TDP generaly scale as a square of frequency in the utilised part of the curve.
When frequency is increased voltage is increased in a square root proportion, this lead to said square law, this can be observed with Broadwell as well as with the FX for whom made the little effort to check enough reviews of decent quality.
So i m not talking of theorical cases as are doing all post that i m adressed but of real figures, for instance Broadwell need 1.148V at 3.1Ghz, 1.114V at 2.9Ghz and 0.926V at 2.2Ghz, do the maths, the power/frequency curve is about a square in function of frequency, for the FX it is even more easier to compute the number since there s also the power comsumption measurements available and they correlate totaly.
As for designs the capabilities are the same, AMD manage high frequencies without the benefit of a smaller node, so this point is irrelevant, what matters is the end result.
If anybody cares at this point, I confirmed with AMD directly that the 7th gen APUs will not be based on Zen.
If anybody cares at this point, I confirmed with AMD directly that the 7th gen APUs will not be based on Zen.
If anybody cares at this point, I confirmed with AMD directly that the 7th gen APUs will not be based on Zen.