AMD CPU speculation... and expert conjecture

Page 203 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.




>implying this cannot change
 


images


 


Oh no, its a 10-15W difference!
 

noob2222

Distinguished
Nov 19, 2007
2,722
0
20,860
why would you buy a haswell with a discrete graphics? I thought that was the entire reason haswell owns everything on the planet.

As for power consumption, you should play with your laptop unplugged and see how long it lasts and how long it takes to recharge without power. Im sure that ivy bridge cpu generates its own power by now on standby mode.

 

jdwii

Splendid


if he mentions one more thing about performance per watt we should ban him from the internet. Why is it so important? battery life is good enough its easy to plug stuff in does he live somewhere where power is hard to reach i'm so confused. Maybe he uses solar panels from radio shack or a cheap wind mill?
 

jdwii

Splendid
Noob2222 i'm still trying to figure this out i'm thinking he lives on the moon or middle of the desert somewhere where power inverters and power connections isn't reachable maybe he lives on an airplane? possibly a pilot
 

Cazalan

Distinguished
Sep 4, 2011
2,672
0
20,810


Performance per watt is important, unless you think the 3 billion people in the world who don't have computers don't deserve to use one. Imagine what life on earth would be like if everyone bought a gas guzzling SUV/Truck instead of economy cars.
 

lilcinw

Distinguished
Jan 25, 2011
833
0
19,010
:evil:

This is what happens when we have no SR news. Posters get bored.

SR APUs are due Q4 right? Maybe something interesting will happen at Hot Chips next month. AMD is scheduled to discuss HSA, Kabini and Richland.
 

jdwii

Splendid


Personally i'll always find it useless when battery life is good enough and i don't feel like its that big of an impact on the environment nor do i even find diesel trucks to be just hyped up stuff to drive new industries not that it doesn't matter but its just over raided. Also what country does he live in? in the U.S its barely an issue.
 

jdwii

Splendid
Yuka i agree 100% it's the same thing with PSU a actually good 750 watt PSU running at 50% will draw less power and last longer than a lowered PSU running at 80+%.

I guess i'll lay off of him now since i know he does live somewhere where electricity is a bigger thing than just a couple bucks more every month sometimes i forget there's a world outside of the U.S.
 


That's the issue. APU's don't pair well with dGPU's that have high bandwidth memory interfaces, the system memory bus will be what's holding the pairing back. You'll get much better pairing when you combing it with a cheaper DDR3 card, yet that kinda defeats the purpose. It's why I don't really expect much from ACF / DG, the purpose of APU's are to provide an integrated CPU + GPU solution for low power and SFF devices. Putting even a moderate performance dGPU with an APU starts to defeat its designed purpose.
 
Honestly i still don't get the one for all thing i never liked it when Amd started it the last thing i want is a laptop CPU in a desktop i wish they would do both try to improve performance a lot in the desktop market and people who want lots of performance and then make the less performance parts do 2 different things. And i will always think x86 is horrible all together in a smartphone/tablet device or other low powered devices if that's the main thing you care about Arm/Snapdragon should be in your Portable device/work machine?.

The issue with x86 in the ultra low power arena (phablet / portable devices) is that the current method of implementing the ISA does not scale down very well. Binary x86 is too clunky and non-predictive to scale to high processing speeds. To get around that limit the current x86 manufacturers do a form of dynamic recompilation / emulation (instruction decoder actually) in the hardware to translate the binary x86 into a native RISC machine language (micro-ops) that are then rapidly executed. The results are put together and returned in a manor consistent with the x86 ISA even though the actual executed instructions were not x86 at all. As bass ackwards as this sounds, it actually ends up being much faster then trying to directly execute x86 binary.

That extra circuitry is expensive from a hardware point of view and it doesn't scale down very well. Native RISC processors don't need that "emulator" design, they execute the raw binary code directly and thus they can scale down to a very low power / miserly regime. You can strip out or add excess processing resources without running into a huge component that's only purpose is to make your CPU appear to be something it's not.
 


I figured the Bandwidth is the issue as running the HD6670 by itself is silky smooth and the APU on its own is silky smooth but Dual Graphics is nothing short of a stutterfest.

I am going to try swap for the DDR3 version and run tests, while i do expect a small drop off I don't think just because its DDR5 to DDR3 you will see massive reductions probably around 10% due to bandwidth (38GB/s vs 56Gb/s) it also teases the fact how nice Kaveri would be with DDR5 as it could properly run DG with DDR5 specified cards.



 

noob2222

Distinguished
Nov 19, 2007
2,722
0
20,860

couldn't agree more. apus are an ok chip, but they are lacking in refined and high end computing. no l3 cache neuters some programs and if it slows something down for me, then I don't want it on my main system. I also don't support blatently breaking the law so don't ask me why I hate intel.

the only thing I see an apu becoming a high end device in a full desktop is when/if HSA can utilize 100% of the igp when a discrete card is detected.

that's not to say I don't like apus, quite the opposite. for my htpc, it rocks. mb, apu, memory. done. nothing else to worry about except storage and wich linux distro to install.
 
I think that Dual Graphics is just a feature that is left unrefined, notably the software level as it does work and can potentially add a lot of value to lower cost systems. I can't confirm yet but it seems like GDDR5 models are a little problematic but won't know for sure until i run a GDDR3 version.

The reality is the APU will become AMD's processor for the future and it will eventually end up in FX line, that said they will then become more APU than they are FX, right now the DT APU is still very much evolving.
 

Cazalan

Distinguished
Sep 4, 2011
2,672
0
20,810


The bulk of the power savings comes from the voltage and frequency scaling. That is incredibly complex circuitry but once developed it benefits all platforms, so why not use it. That Pentium test chip Intel made could scale from 1Mhz to 1Ghz and run on a postage stamp sized solar cell. Why make your server farms run full tilt 24/7 unless you have to?

Intel's phone chip sucked for so long because it was a 5+ year old Atom design using an old 45nm node. The 32 nm Clover Trail pretty much caught ARM on efficiency even with the older core design. The new 22nm Bay Trail with OoO will likely exceed ARM efficiency.

I got nothing against ARM. I work on ARM core chips on almost a daily basis, but the market still needs competition to thrive. Besides the x86 is an American legacy. The ARM is a British design. Tomorrow we celebrate the July 4th, 1776 Declaration of Independence from Great Britain.
 

Cazalan

Distinguished
Sep 4, 2011
2,672
0
20,810


Dual Graphics is a neat idea but the designs are quite asymmetrical. I can see why they have such problems balancing the load. You got dedicated GPU ram on the discrete and the APU is competing with the CPU for it's RAM. The latency for the APU can vary quite a bit.

As games advance they could just use the APU graphics for compute and let the discrete render full speed.
 

amdfangirl

Expert
Ambassador
In other news:
http://blog.documentfoundation.org/2013/07/03/amd-joins-the-document-foundation-advisory-board-to-accelerate-libreoffice/

Berlin, July 3rd, 2013 – The Document Foundation (TDF) announces that AMD is now a member of its Advisory Board. AMD is a leading designer and integrator of pioneering technologies that are at the heart of the digital devices people use and experience daily, pushing the boundaries of what is possible.

“It is great to work on LibreOffice with The Document Foundation to expose the raw power of AMD GPUs and APUs, initially to spreadsheet users,” said Manju Hegde, corporate vice president, Heterogeneous Solutions at AMD. “Bringing the parallelism and performance of our technology to traditional, mainstream business software users will be a welcome innovation for heavy duty spreadsheet users, particularly when combined with the compute capabilities of the upcoming generation of AMD Heterogeneous System Architecture (HSA) based products.”

“It is exciting to work together with AMD and their ecosystem to take advantage of AMD’s cutting edge innovation right inside LibreOffice,” said Michael Meeks, SUSE Distinguished Engineer and TDF Board Member, “The growth in performance and parallelism available in the GPUs of today, and particularly with AMD’s revolutionary APUs of tomorrow, is something we’re eager to expose to LibreOffice users.”

HSA is an innovative computing architecture that enables CPU, GPU and other processors to work together in harmony on a single piece of silicon by seamlessly moving the right tasks to the best suited processing element. This makes it possible for larger, more complex applications to take advantage of the power that has traditionally been reserved for more focused tasks. While the biggest impact will be for AMD APU users, supporting benefits of the work will improve the LibreOffice core data structures enabling larger spreadsheets to calculate faster for all users. This is only the start of exposing the power of the HSA enabled APU to business users providing better analytics and decision making across the board from finance to science.

With the addition of AMD, the Advisory Board of The Document Foundation now has eleven members: AMD, Google, RedHat, SUSE, Intel, Lanedo, the King Abdulaziz City of Science and Technology (KACST), the Inter-Ministry Mutalisation for an Open Productivity Suite (MIMO), the Free Software Foundation (FSF), Software in the Public Interest, and Freies Office Deutschland e.V.

I doubt the usefulness of HSA in everything but spreadsheets, but I hope this means Impress presentations stop having graphical glitches.

Hope.
 

blackkstar

Honorable
Sep 30, 2012
468
0
10,780


First time I believe I've ever heard a reason for Intel being purchased being a competitive price.

/golfclap

You've set the bar to a whole new level.

As for 20nm, if AMD puts their GPUs on GloFo's 20nm process and Nvidia is stuck on 28nm TSMC for the entirety of 2014, Nvidia is dead. It will basically be $300 AMD 300mm^2 GPU competing with $650 500mm^2 Nvidia GPU and the AMD would be faster. And it'd trickle down the entire product range as AMD's die size to performance advantage would be massive.
 
Status
Not open for further replies.