[News] Moore?s Law ain?t dead claims AMD

[StyM]

http://www.fudzilla.com/news/process...ead-claims-amd

AMD CTO, Mark Papermaster, has told the world+dog that despite claims to the contrary Moore?s Law is not dead, it is getting better, and be going Vroom any day now.

Intel co-founder Gordon Moore said in 1965 that computing power would double every two years thanks to developments in technology over time leading to shrinking transistor sizes and while he has been right until now, Chipzilla said it would move away from the prediction for the first time, shifting transistor size from two to 2.5 years.

Papermaster said that Moore?s Law is alive and well, and said only narrow-minded people think its evolution is just about transistor size.

?It?s not just about the transistor anymore; we can?t just have transistors improving every cycle. It does take semiconductor transistor improvements, but the elements that we do in design in architecture, and how we put solutions together, also keep in line with a Moore?s Law pace.

He said that AMD had adopted an idea called Moore?s Law Plus. This means you stay in a Moore?s Law pace of computing improvement. So you can keep in with a Moore?s Law cycle but you don?t rely on just semiconductor chips, you do it with a combination of other techniques.

These include design changes and how you architect those system solutions that will keep on the Moore's Law pace.

You can mix and match combinations of CPU and GPU, other accelerators, different memory configurations, or how they are pieced together ? there is room for lots of innovation at the next level.

[vario]

http://www.fudzilla.com/news/process...ead-claims-amd

Thats funny coming from an amd guy, while their chips have gone DOWN in performance, ATM their newer cpus and gpus are less powerful than the old ones :P.
On the Cpu side its really pathethic, last performance chip was vishera, it was almost 4 years ago, and im talking even per core basis, their apus couldnt clock as high, and /or had reduced cache sizes so in reality, they went backwards.

[AliG]

Thats funny coming from an amd guy, while their chips have gone DOWN in performance, ATM their newer cpus and gpus are less powerful than the old ones :P.
On the Cpu side its really pathethic, last performance chip was vishera, it was almost 4 years ago, and im talking even per core basis, their apus couldnt clock as high, and /or had reduced cache sizes so in reality, they went backwards.

I really don't know where you get this stuff. The RX 480 is significantly more powerful than the R9 380X, and their APUs have shown pretty sizable improvements in efficiency the last 2 years (along with a few performance bumps here and there).

[zanzabar]

I really don't know where you get this stuff. The RX 480 is significantly more powerful than the R9 380X, and their APUs have shown pretty sizable improvements in efficiency the last 2 years (along with a few performance bumps here and there).

the thulban and fm1 are better IPC cpus than the current lineup, the apu did get that gpu bump though. the rx is the replacement for the r7 so in that way it is doing great.

[Darakian]

the thulban and fm1 are better IPC cpus than the current lineup, the apu did get that gpu bump though. the rx is the replacement for the r7 so in that way it is doing great.

Better IPC with lower clocks, slower ram and slower storage options. I don't want to say that bulldozer was a good arch, but amd has raised absolute performance since the phenom days. That said, AMD is also a behind intel on process tech so they know for certain that there's room to do better there and that moores law isn't dead (for them)

[zanzabar]

Better IPC with lower clocks, slower ram and slower storage options. I don't want to say that bulldozer was a good arch, but amd has raised absolute performance since the phenom days. That said, AMD is also a behind intel on process tech so they know for certain that there's room to do better there and that moores law isn't dead (for them)

the per watt went down as well with BD and initial clocks were similar. the excavator (?) did fix things with power so it could clock better but if the thulban got a die shrink to be on the same process as the BD i would bet amd would be sitting better.

the 990 and 970 chipsets also work with the thulban and i ran my 1090t when i had it with 2000mhz 10-11-11 ram, so i dont see how storage and ram were slower.

[Darakian]

the per watt went down as well with BD and initial clocks were similar. the excavator (?) did fix things with power so it could clock better but if the thulban got a die shrink to be on the same process as the BD i would bet amd would be sitting better.

The high end today is better than the high end of yesterday. It's hard to say if a die shrink would have helped thulban or not. It's possible that the phenom arch had hit some other wall, without being a cpu architect I don't think I can make a meaningful comment on that. The upshot of BD being such a bad arch is that zen will look much better even if it is a return to a thulban like arch

[drmrlordx]

Sorry, but the idea that Thuban has better IPC than XV is a bit far-fetched. Go bench some SuperPi, see what happens.

[zanzabar]

Sorry, but the idea that Thuban has better IPC than XV is a bit far-fetched. Go bench some SuperPi, see what happens.

go bench wprime. if one module is one core then the BD/XV is better, if one thread is one core then the phenom is better in any multithreaded bench.

[Darakian]

go bench wprime. if one module is one core then the BD/XV is better, if one thread is one core then the phenom is better in any multithreaded bench.

That's not true at all. Heavy integer workloads run like a champ on the bulldozer design.

Edit: Example taken from the anandtech 8150 review

[zanzabar]

That's not true at all. Heavy integer workloads run like a champ on the bulldozer design.

Edit: Example taken from the anandtech 8150 review
http://images.anandtech.com/graphs/graph4955/41698.png

that shows lower IPC. anyways lets hope amd gets good with zen.

[FlanK3r]

the thulban and fm1 are better IPC cpus than the current lineup, the apu did get that gpu bump though. the rx is the replacement for the r7 so in that way it is doing great.

No, Excavator destroy both easy clock to clock.
Superpi at stock clocks:

[vario]

I really don't know where you get this stuff. The RX 480 is significantly more powerful than the R9 380X, and their APUs have shown pretty sizable improvements in efficiency the last 2 years (along with a few performance bumps here and there).

It is pretty evident i was talking about absolute performance .
Vishera is more powerful than any apu today(per module), 390X fury are more powerful then the 480X.
And even with their apus its pretty weak showing (besides the gpu part) efficiency got higher clocks got lower, cache was slashed with the excavator parts.All in all performance hardly went up.
And even Thuban vs bulldozer/vishera wasnt a clear victory , i changed 4ghz clocking thuban to a 4.6ghz vishera, and in the tests it wasnt always faster...
On the GPU side its more like bad planning, we will get vega. But the CPU side for like 4-5 years, a pretty bleak picture

[AliG]

Give it a little bit, I think Vega will change your mind on absolute performance

[drmrlordx]

go bench wprime.

Hmm okay, here's a guy who did wprime with an 845 @ 4.1 GHz:

http://hwbot.org/hardware/processor/athlon_x4_845/

wow, that's fast for only two modules. It beats the hell out of my 7870k @ 4.7 GHz!

[zalbard]

Give it a little bit, I think Vega will change your mind on absolute performance

At this rate, it may have to compete with Volta...

[BeepBeep2]

that shows lower IPC. anyways lets hope amd gets good with zen.

Excavator has surpassed Phenom II X6 in pretty much every benchmark when at the same frequency. Remember that Deneb had better IPC than Thuban despite worse memory bandwidth due to L3 structure as well

I don't think desktop AM4 Bristol Ridge parts will blow anyone away but I think they will be a pleasant surprise in terms of overall improvements in both power consumption and performance that AMD has made - impressive for the 28nm node. Laptop parts (Carrizo) were castrated by single-channel DDR3 and the Athlon II 845 has locked multiplier and limited overclocking.

[vario]

Excavator has surpassed Phenom II X6 in pretty much every benchmark when at the same frequency....

woohooo ...
So six years, or more counting phenom x4.
INSERT CASTRATED DARth VADER VOICE - IMPRESSIVE, MOST IMPRESSIVE

Maybe im too harsh, are there any serious reviews of the AM4 parts ? There are some x845 but they look bad.