Samsung Will Begin Manufacturing AMD's 14nm Zen Core By Year's End

[Physc]

well i dont see them smashing intel, but i can see them coming up to intels league again. i mean apparently they are doing an SMT design. on 14nm. just from that, they should achieve a similar performance as intel. so i hope this makes intel do something and play an ace they have hidden...

 

Or gain a huge lead in APU section of the market, like at least current R9 series performance.

[LukaP]

Or gain a huge lead in APU section of the market, like at least current R9 series performance.

well shrinking to 14nm would give them loads more space on die. leaving 4 cores for the CPU (real cores mind you) we could see doubling of the GPU units (if both were to be shrunk to 14nm)

[vm'N]

Except haswell has a bigger CPI than Westmere.

I'm not so sure about that.

However if I remember correctly westmere have 16 pipeline stages, where haswell only have 14 pipeline stages.

More pipeline stages will increase the CPI (loop detection, prediction is used to shorten the pipeline).

Also, no, frequency scaling does not decrease ipc inherently. The only time this occurs is when the clock rate is moving beyond the speed at which electricity can flow through the path of an instruction, causing an overlap/miss. You can have perfect IPC and high frequency.

If we don't take things like die-budget and heat into consideration, then yes, frequency does not degrease the IPC.

However because we are limited to a die-budget and a theoretical TDP, we cannot have both.

IPC requires alot of diespace. Increasing heat and powerusage means you will have to run at lower frequencies.

Multiple workers on the same data set is instruction-level parallelism and falls right in line with what I said above. There is no data race condition if you build it correctly.

My example was about software, not hardware. Multiple workers, would be considered threads.

Parallel code, the threads should never intevene or share elements with each other.

Good code will have less data races, however some algorithms are doomed to data races. Not because of bad coding, but the nature of how you potentially solve a problem.

And, it would be considered thread-level parallelism not instruction-level parallelism.

[Swndlr]

amd, why try homie why try

[dalekphalm]

amd, why try homie why try

Would you prefer the alternative of Intel becoming a complete Monopoly? No... No you shouldn't want that. That would be bad for EVERYONE except Intel.

[GzeroD]

If we don't take things like die-budget and heat into consideration, then yes, frequency does not degrease the IPC.

However because we are limited to a die-budget and a theoretical TDP, we cannot have both.

IPC requires alot of diespace. Increasing heat and powerusage means you will have to run at lower frequencies.

 

 

If your on the same architecture then yes but every time they make an efficiency improvement they gain the ability to push them both up also the TDP on most parts is pretty conservative.

[TheRagingGamer]

Hopefully something to compete with Intel.

 

However, knowing AMD; it probably won't.

they were ahead of intel before, what makes you think they won't be again 

[vm'N]

If your on the same architecture then yes but every time they make an efficiency improvement they gain the ability to push them both up also the TDP on most parts is pretty conservative.

I have never said it wasn't possible to increase the limit.

I'm saying that one WILL limit the other.

[Builder]

Above 50nm most of it is still governed by macro physics or classical E&M. This is because the Shrodinger wave function and Heisenberg's uncertainty have no bearing on systems above a certain scale.

Don't transistors function via quantum tunneling? Like on a basic level isn't cold emission a fundamental part of semiconductor design?

[LukaP]

Don't transistors function via quantum tunneling? Like on a basic level isn't cold emission a fundamental part of semiconductor design?

they do now. they didnt at the larger scales. the big transistors (around 100nm mark) work on the principles of classical physics

[jrfox87]

I hope to god this ends up being good! We really need AMD to be competitive so Intel gets their butt into gear again.

[Builder]

they do now. they didnt at the larger scales. the big transistors (around 100nm mark) work on the principles of classical physics

How did they used to work?

[LukaP]

How did they used to work?

http://en.wikipedia.org/wiki/Transistor

 

basically you have 3 layers. 2 inputs, one outlet. applying voltage to the control leg lets current through, and amplifies it if needed. want more indepth? its really late and i CBA to write paragraphs if ppl wont read them

[Builder]

http://en.wikipedia.org/wiki/Transistor

 

basically you have 3 layers. 2 inputs, one outlet. applying voltage to the control leg lets current through, and amplifies it if needed. want more indepth? its really late and i CBA to write paragraphs if ppl wont read them

Ahh so it's sort of like a vacuum tube right? I know how vacuum tubes work.

[harrynowl]

they were ahead of intel before, what makes you think they won't be again 

Intel were caught with their pants off with the P68 microarch

 

Intel probably wont be beat by AMD again in regular CPUs, they're just too far ahead unless AMD buy a licence to Haswell

[LukaP]

Ahh so it's sort of like a vacuum tube right? I know how vacuum tubes work.

yes almost exactly like that. its based on it. but the transistors in chips nowadays are called field effect transistors. they use quantum field theory to their advantage, and its consequences, like tunnelling and annihilation

[landon418]

I do not understand fanboys. I have been using nvidia GPUs since 2001 (Built my first rig this year)and I chose an r9 290x. Why? I got it for 450 bucks during the mining craze, which was better than the 780 at the time. I buy whatever meets my expectations and price point. I could care less about the brand. I've never used AMD processors, but if zen performs better than Intel at a significantly lower price, I would sure as hell get it. Also, I think AMD is doing all they can considering they do not have near the amount of money to burn as Intel.

[Builder]

yes almost exactly like that. its based on it. but the transistors in chips nowadays are called field effect transistors. they use quantum field theory to their advantage, and its consequences, like tunnelling and annihilation

Yeah I understand how FETs work I just have a bit of an understanding gap between vacuum tubes and FETs. Given that's what? 60 years? I suppose it's more of a gaping void than a gap  

[LukaP]

Yeah I understand how FETs work I just have a bit of an understanding gap between vacuum tubes and FETs. Given that's what? 60 years? I suppose it's more of a gaping void than a gap

well nobel prize for a transistor was given in 52 iirc. QFT originates in the 20s. FET is about 90s. so some 40 years of gaping holes

[TheRagingGamer]

Intel were caught with their pants off with the P68 microarch

 

Intel probably wont be beat by AMD again in regular CPUs, they're just too far ahead unless AMD buy a licence to Haswell

14nm tho, you know what this means. Moar power, Moar core, Moar gigaherzz 

[harrynowl]

14nm tho, you know what this means. Moar power, Moar core, Moar gigaherzz 

technically it just means a higher transistor:die size ratio

 

I don't expect to see a higher TDP, any more than 6 cores and any more than 3.8ghz

[TheRagingGamer]

technically it just means a higher transistor:die size ratio

 

I don't expect to see a higher TDP, any more than 6 cores and any more than 3.8ghz

AMD does things the ghetto way. Brute force. AMD has an 8core cpu that goes over 5ghz stable so you better expect higher tdp 

[harrynowl]

AMD does things the ghetto way. Brute force. AMD has an 8core cpu that goes over 5ghz stable so you better expect higher tdp 

yea but I don't care about 5468ghz 8 cores if my ~4ghz quad will floor it in most tasks

[TheRagingGamer]

yea but I don't care about 5468ghz 8 cores if my ~4ghz quad will floor it in most tasks

STFU you haven't seen the new cpu, you can't say anything about it. All i am saying is amd has been beating intel in past more so than they are beating amd now so i don't see why its not possible. 

Streamers love this stuff and i like it to because i can have more games/programs running at the same time. 

 

[harrynowl]

STFU you haven't seen the new cpu, you can't say anything about it. All i am saying is amd has been beating intel in past more so than they are beating amd now so i don't see why its not possible. 

Streamers love this stuff and i like it to because i can have more games/programs running at the same time. 

 

chill.. you're taking what I've said out of context. and for the note neither have you so... moot point. zen could be good yes, but realistically AMD have no arch to base it on so it's not like netburst -> core which was a derivative of the pentium-m arch so they'd already got work done... 90% ipc deficit probably not going to disappear in 1 gen, paricuarly on 14nm finfet which whill not clock to 556564ghz like 32nm soi.

 

I wasn't even referring to the 'zen' cpu when I said that.

 

 

 

AMD has an 8core cpu that goes over 5ghz stable so you better expect higher tdp 

that's an fx9590 you're referring to yes? there you go then.