AMD's Hawaii Is Officially The Most Efficient GPGPU In The World To Date, Tops Green500 List

[harrynowl]

makes sense, the hawaii gpu adjusts power draw 100,000 times per second based on load, drawing a sporadic current from the capacitors, while actual power draw through the PSU appears stable. Its a cpu technology that's more than a few years old, and Nvidia through their work on Tegra finally figured out how to do the same with Maxwell as AMD does with Power Tune.... even better perhaps because of the secrets Nvidia gleaned from working with ARM technology.

Potentially. Although the ARM portion is just a pure stock Cortex design from ARM so power management is just handled by ARMs kernel code and then they add their own GPU, for example K1 is 4 A15 cores (or 2 A57s in the 64bit one) and then a single Kepler SMX.

[harrynowl]

That is one part of it, but there are many more sides to the story than just the pipeline. you have the hardware engines, the design of the cores (shaders, cudas, stream processors), the manufacturing process (32nm on the same arch would be different to 14nm FinFET+)

Yep. Doing a little reading on it, but I'm not going to pretend to be an expert here and chime in too much

[LukaP]

Potentially. Although the ARM portion is just a pure stock Cortex design from ARM so power management is just handled by ARMs kernel code and then they add their own GPU, for example K1 is 4 A15 cores (or 2 A57s in the 64bit one) and then a single Kepler SMX.

isnt denver (the A57 ones) slightly redesigned, with higher IPC and power consumption?

[LukaP]

Yep. Doing a little reading on it, but I'm not going to pretend to be an expert here and chime in too much

well im no expert either (havent spent a whole semester studying it *kappa*) but i have a slight idea how it all works

[harrynowl]

isnt denver (the A57 ones) slightly redesigned, with higher IPC and power consumption?

Denver might be a slightly custom version of the A57, as they called it something else rather than "Quad A15s" or the old Cortex-A9 designs in Tegra 2 (no NEON) and 3.

 

I think nVidia might have just made it scale at higher clocks in comparison to other A57 designs.

[LukaP]

Denver might be a slightly custom version of the A57, as they called it something else rather than "Quad A15s" or the old Cortex-A9 designs in Tegra 2 (no NEON) and 3.

 

I think nVidia might have just made it scale at higher clocks in comparison to other A57 designs.

i havent looked into it, but ive heard it considerably higher IPC than normal A57. ill look into it tomorrow if ill have the energy... pneumonia is a bitch

[harrynowl]

i havent looked into it, but ive heard it considerably higher IPC than normal A57. ill look into it tomorrow if ill have the energy... pneumonia is a bitch

Well using Geekbench its twice as fast as the A15 quad core version (single threaded), I think that lines up about right with ARMs slides for the new ARMv8 arch.

 

I believe cyclone is currently the highest IPC mobile chip though

[LukaP]

Well using Geekbench its twice as fast as the A15 quad core version (single threaded), I think that lines up about right with ARMs slides for the new ARMv8 arch.

 

I believe cyclone is currently the highest IPC mobile chip though

OH yes, cyclone is pretty much in the range of Atoms or higher. Damn now you got me all excited and i wont be able to sleep

[harrynowl]

OH yes, cyclone is pretty much in the range of Atoms or higher. Damn now you got me all excited and i wont be able to sleep

If apple licensed cyclone out ...

 

 

I'd find a android tab with it and run some overclocked benchmarks against the bay trail SoCs I have available

[LukaP]

If apple licensed cyclone out ...

 

 

I'd find a android tab with it and run some overclocked benchmarks against the bay trail SoCs I have available

you could just run a single core benchmark of a cyclone core, and then clock a baytrail core at the same frequency and do the benchmark there, im pretty sure the cyclone would be ahead

[harrynowl]

you could just run a single core benchmark of a cyclone core, and then clock a baytrail core at the same frequency and do the benchmark there, im pretty sure the cyclone would be ahead

Just checked a geekbench thing for that. An E3845 bay trail loses by 2x in single core at 1.9ghz vs 1.5 and has double the power draw.

 

And I'm pretty sure the SGX6850 rogue will beat the 6 EU intel hd graphics too.

[LukaP]

Just checked a geekbench thing for that. An E3845 bay trail loses by 2x in single core at 1.9ghz vs 1.5 and has double the power draw.

 

And I'm pretty sure the SGX6850 rogue will beat the 6 EU intel hd graphics too.

oh wow, its that worse? doesnt that come really close to Bw-Y though? wow...

 

amd yeah, the 6850 is about a 730GT iirc

[patrickjp93]

Gravity is theorethical and unproven. i dont see you jumping out of a 10 story window to test it (sadly)

 

I genuinly hate you retards who say something is just a theory. no. its A THEORY. and a theory is the best explanation we have so far for the effect. and one that has been challenged and stood op to the challenge. 

see the theory is an explanation for an effect. this means quantum mechanics are our best explanation for effects that we OBSERVE and use in real life. everything you FUCKING TALK ABOUT is based on quantum mechanics and mostly tunneling...

 

in quantum mechanics there are "proven" (nothing is ever proven, not even your constants, one miscase, and theyre dead) constants and there is stuff you can evaluate with 1 semester. but damn sure it doesnt mean that youre an expert in the field. further on, id say even when you do your PhD youre not even close to an expert.

 

and all of this applies to making supercomputer... yes you can know some technical aspects of it, but after one semester of some subject, you are not even close to knowing enough about it to be able to challenge people who have worked in teh field for years. so dont do it, it pisses everyone off. if you came to me when i was making a model, and started pissing on my technique, id not tell you to get lost. i would personally kick your ass out of the place. and i can tell you thats exactly what those people would do to you if you were to go spit your BS into their face. but maybe they would do it more politely, since they dont want a PR disaster

How it works...Jesus Christ. We all know Relativity works on the macrocosmic scale but that it breaks down in the quantum scales, and we have 3 competing quantum gravity theories, all unproven.

 

Yes, theories are the best explanations we have. Some have strong mathematical backing, and others don't. The only experts in the quantum physics fields are delving new ideas without yet knowing the math to prove or disprove them. I'm not at all ragging on physics or you. Sheesh would it kill you to not instantly get pissed just because I post?

[Godlygamer23]

Except in your case you are just repeating words you heard me say in THE EXACT ORDER, and before you were a lover of oCL... Go sleep, its late for a 14 y/o to stay up so long

 

except its not. its within 10% of the GK110, and probably lower than GK210

 

Dont bother, he is a troll that knows a few words and then uses them in sentences that appear to make sense so he can start arguments. I havent seen him be right once. and when he gets bashed in the head by like 5 people at the time, he just takes what they have been saying, and starts using those words as well, having no idea what they mean

I'm not going to argue about this, because frankly it's a waste of my time. However, you said that the temps are within 10% of GK110. According to Anandtech, the difference is more like 11.8%. Granted, I am kind of nitpicking, but I believe accuracy is important. Both are reference cards in the review.

http://www.anandtech.com/show/7492/the-geforce-gtx-780-ti-review/15

[harrynowl]

oh wow, its that worse? doesnt that come really close to Bw-Y though? wow...

 

amd yeah, the 6850 is about a 730GT iirc

Although... BW-Y has the 96 EU Iris pro 2 solution on the high end I believe.

[LukaP]

How it works...Jesus Christ. We all know Relativity works on the macrocosmic scale but that it breaks down in the quantum scales, and we have 3 competing quantum gravity theories, all unproven.

General relativity is not proven. nothing is proven. congratulations, youve just proven yourself unworthy of any more of my attention. you speak of science, yet you dont understand the basic concept of it.

 

and no, there are not 3 competing theories, all unproven. there is one theory right now, that is being worked on. Its not really even a theory yet, its just a bunch of hypothesis' that were not shot down with our observations yet. its called the M Theory and it attempts to unify and translate each of the string theories and supergravity as the infinite limits of a bigger theory. http://en.wikipedia.org/wiki/M-theory

 

I'm not going to argue about this, because frankly it's a waste of my time. However, you said that the temps are within 10% of GK110. According to Anandtech, the difference is more like 11.8%. Granted, I am kind of nitpicking, but I believe accuracy is important. Both are reference cards in the review.

http://www.anandtech.com/show/7492/the-geforce-gtx-780-ti-review/15

seriously? youre gonna nitpick 2%? im a physicist. to me pi^2 is 10, and a year has pi * 10^7 seconds. you wont make me accept 11.8 to be different from 10

[LukaP]

Although... BW-Y has the 96 EU Iris pro 2 solution on the high end I believe.

oooh nonono im talking CPU only, not gpu, broadwells gpu is basically magic for that form factor

[harrynowl]

oooh nonono im talking CPU only, not gpu, broadwells gpu is basically magic for that form factor

Well if CPU is even then its down the GPU I guess, particularly for apple.

 

I really think apple would switch to x86 over ARM designs as they like to push GPU performance.

[LukaP]

Well if CPU is even then its down the GPU I guess, particularly for apple.

 

I really think apple would switch to x86 over ARM designs as they like to push GPU performance.

GPUs have nothing to do with x86 im thinking they will be doing a custom gpu with a licence from either power or nvidia very soon. maybe A10X even. 

[harrynowl]

GPUs have nothing to do with x86 im thinking they will be doing a custom gpu with a licence from either power or nvidia very soon. maybe A10X even. 

Well, I don't see Intel giving their designs away to ARM/Imagination Tech/Qualcomm anytime soon so if Apple really want the GPU performance and Intel is offering it (and only Intel perhaps), I think they'd go for it. That's what I meant

[LukaP]

Well, I don't see Intel giving their designs away to ARM/Imagination Tech/Qualcomm anytime soon so if Apple really want the GPU performance and Intel is offering it (and only Intel perhaps), I think they'd go for it. That's what I meant

no i dont think they would take intels GPU. first of all, because they like their in house solutions, and second of all, intel is trying to shove their D into the mobile market. just giving all that performance to a competitor is the equivalent of chopping that D off

[PhantomChevron]

More efficient then a 980? Maybe they didn't test a 980 yet?

[LukaP]

More efficient then a 980? Maybe they didn't test a 980 yet?

in compute. a 980 is quite bad at DPFP compute

[wrath]

nice

[harrynowl]

More efficient then a 980? Maybe they didn't test a 980 yet?

I'm pretty sure maxwell cut compute even further

 

At least at double precision