AMD CPU speculation... and expert conjecture

[jdwii]

http://www.macrumors.com/2014/09/29/retina-imac-27-amd/

That's a nice small edition to help Amd
Perhaps we can wish the K12 might be in all Apple iphones one day.

 

[blackkstar]

http://www.x.org/wiki/RadeonFeature/

I think you under-estimate what has been done for AMD Radeon Drivers. Like I said, they are pretty good right now. Seeing OpenCL as a WIP is very good! Radeon hybrid is coming along and still broken on Nvidia in Linux.

I tend to hang out in #gentoo IRC quite often helping and requesting help and just yesterday someone was using binary Nvidia drivers and couldn't get 3d acceleration working after an update. But to be fair that's common for anything closed source on Linux. So generally when you're on Linux you want to kick that stuff to the curb as much as possible. It might not be as bad on something like Ubuntu but when you're compiling everything yourself with "emerge" it's rather difficult at times. For example, my Asus BT adaptor is a special treat when I compile kernel properly and it actually works.

 

[8350rocks]

At least it was not made by broadcom...spent 2 days trying to get my first ubuntu install up before I found out about broadcom requiring closed drivers.

 

[gamerk316]

Technically, even the Open Source Linux AMD driver relies on closed source firmware. Do we REALLY want to devolve into this?

 

[jdwii]

8% faster at 1080P and uses 80 watts more on max and uses 61 watts more on average(compared to the 760 also both are priced the same) wtf is this the new design? No wonder the 380X is rumored to be made on 20nm. When you have to make APU's and share tdp with the CPU you would assume engineers would be more worried.
http://www.techpowerup.com/reviews/Sapphire/R9_285_Dual-X_OC/23.html

 

[UnrelatedTopic]

8% faster at 1080P and uses 80 watts more on max and uses 61 watts more on average(compared to the 760 also both are priced the same) wtf is this the new design? No wonder the 380X is rumored to be made on 20nm. When you have to make APU's and share tdp with the CPU you would assume engineers would be more worried.
http://www.techpowerup.com/reviews/Sapphire/R9_285_Dual-X_OC/23.html

yeah as the title says this is an OC card so the power consumption numbers are up.

 

[juanrga]

8% faster at 1080P and uses 80 watts more on max and uses 61 watts more on average(compared to the 760 also both are priced the same) wtf is this the new design? No wonder the 380X is rumored to be made on 20nm. When you have to make APU's and share tdp with the CPU you would assume engineers would be more worried.
http://www.techpowerup.com/reviews/Sapphire/R9_285_Dual-X_OC/23.html

This is the second time that you chose an overclocked AMD card to make your silly power consumption comparisons against Nvidia. And you were corrected the first time.

stop spreading misinformation

 

[wh3resmycar]

you can run that 285 on stock but obviously the the difference in performance will dwindle too the same as it's power consumption. 5w to 10w give or take, still miserable against nvidia if you put it into perspective.

 

[8350rocks]

Actually the 285 uses less power than my old 7870XT. By quite a bit...

 

[Embra]

ARM Cortex-A57-Based Hadoop Shown on AMD Opteron A-Series
http://www.guru3d.com/news-story/arm-cortex-a57-based-hadoop-shown-on-amd-opteron-a-series.html

 

[logainofhades]

A part of me wants a GTX 970 just because it uses nearly half the power of my HD 7970, and is decently faster. Hoping AMD answers this with a new generation, that can take on Nvidia in performance and power consumption, as I generally prefer AMD cards.

 

[cdrkf]

8% faster at 1080P and uses 80 watts more on max and uses 61 watts more on average(compared to the 760 also both are priced the same) wtf is this the new design? No wonder the 380X is rumored to be made on 20nm. When you have to make APU's and share tdp with the CPU you would assume engineers would be more worried.
http://www.techpowerup.com/reviews/Sapphire/R9_285_Dual-X_OC/23.html

http://www.tomshardware.co.uk/amd-radeon-r9-285-tonga,review-33030-6.html

Toms review showed it to be equal in efficiency to Kepler. I mean ok it's 8% faster at 1080p, if you look at your own review the gap widens as the resolution goes up (where you need the extra performance more) and it sits *just below* a GTX 770. The difference in cards at those price points is fairly minimal- the 285 nestles in there exactly where you would expect. It *isn't* as efficient as Maxwell, but it's definitely a step in the right direction compared to earlier GCN offerings. One thing I've noticed about GCN cards is their efficiency is really hurt by clock speed- the most efficient cards are usually running around the 800 - 900 mhz range, however AMD have clocked them all up to 1ghz+ to better compete in raw speed with nVidia (at the expense of efficiency). The thing is that down clocking a card can reduce the power consumption *more* than the proportional reduction in performance, it isn't linear. The same can be said for over-clocking- a 5% over-clock can result in a 10% average power consumption increase. I think that has a lot to do with how efficient the new 970 and 980 cards are- I was *very* surprised by the high clock speeds of both of them, I imagine nVidia has done lots of hand optimisation to the core design in order to facilitate high efficiency at high clock speed like that.

 

[de5_Roy]

hynix teases moar h.b.m. (high bandwidth memory)
http://techreport.com/news/27129/hynix-slides-tease-vertically-stacked-memory-with-256gb-s-of-bandwidth
256GB/s. me like.

applied micro has launched 64bit custome arm soc )afaik, applied micro co-developed arm v8 with arm)
http://www.anandtech.com/show/8580/hp-appliedmicro-and-ti-bring-new-arm-servers-to-retail
according to above, amd's seattle hasn't entered production yet.8350rocks, have you heard anything about that?

applied micro apparently has the successor lined up
http://www.tomshardware.com/news/armv8-64-bit-x-gen-hp-moonshot,27784.html

edit3: fixed wording.

 

[cdrkf]

hynix teases moar h.b.m. (high bandwidth memory)
http://techreport.com/news/27129/hynix-slides-tease-vertically-stacked-memory-with-256gb-s-of-bandwidth
256GB/s. me like.

applied micro has launched 64bit custome arm soc
http://www.anandtech.com/show/8580/hp-appliedmicro-and-ti-bring-new-arm-servers-to-retail
according to above, amd's seattle hasn't entered production yet. afaik, applied micro co-developed arm v8 with arm. 8350rocks, have you heard anything about that?

applied micro apparently has the successor lined up
http://www.tomshardware.com/news/armv8-64-bit-x-gen-hp-moonshot,27784.html

The bit about A1100 sounds odd... I mean AMD have already shipped units to developers so we *know* they've taped the things out already.

 

[de5_Roy]

microsoft pulled off a switcheroo you.will.not.believe.
http://www.tomsguide.com/us/windows-10-unveiled,news-19648.html
spoiler: it's not windows 9.

 

[jdwii]

8% faster at 1080P and uses 80 watts more on max and uses 61 watts more on average(compared to the 760 also both are priced the same) wtf is this the new design? No wonder the 380X is rumored to be made on 20nm. When you have to make APU's and share tdp with the CPU you would assume engineers would be more worried.
http://www.techpowerup.com/reviews/Sapphire/R9_285_Dual-X_OC/23.html

This is the second time that you chose an overclocked AMD card to make your silly power consumption comparisons against Nvidia. And you were corrected the first time.

stop spreading misinformation

If the card uses that much power by a simple OC then wow. Also i keep overlooking that its not silly either even a simple 20 watts extra can lead to problems when producing APU's.
http://images.anandtech.com/graphs/graph8460/67237.png
Also stop acting like you don't make the same mistakes twice or even several times like 80% of your post.

 

[jdwii]

microsoft pulled off a switcheroo you.will.not.believe.
http://www.tomsguide.com/us/windows-10-unveiled,news-19648.html
spoiler: it's not windows 9.

I was confused on the whole 10 thing

 

[gamerk316]

microsoft pulled off a switcheroo you.will.not.believe.
http://www.tomsguide.com/us/windows-10-unveiled,news-19648.html
spoiler: it's not windows 9.

I was confused on the whole 10 thing

Win(d)X lives!

 

[8350rocks]

ARM scales upward poorly.

Not because the core cannot do the work, but because by the time you get ARM to do what x86 can on the high end, you now need all kinds of additional hardware for things like OoO, and other stuff.

What is being highly misplaced is comparing cherry picked benchmarks for an in order execution chip designed for phones and tablets against an OoO capable chip for laptops.

Consider this, AnTuTu is not aimed at OoO execution, because no ARM chips to this point will do OoO with perhaps exception to 80W server chips, and even then, you are no longer in much better territory, which is what I said all along. If you want ARM to compete on that end it will not be at 20w. If their cores excel at microserver functions, that is fine, they are designed for it. However do not hold your breath watching for a wall street firm to sink big money into ARM servers...

As for HPC, everyone knows that you need just enough cores to baby sit a massive bank of GPUs and keep them fed well...let us not kid ourselves shall we?

It's my understanding that A57 is an OoO design though?

Yes, but K1 is not...

Wrong. 32bit version of K1 uses OoO cores.

A57 is also mostly for servers and such...as I stated above, a noted exception.

Wrong. It is not the exception. As mentioned above the 32bit version of K1 uses OoO cores and former Tegra 4 also used OoO cores as did the Tegra 3 before. And none of them are used in servers. Apple cyclone is another OoO core that is not used in servers.

And the use of A57 in (micro)servers will be minimal (only AMD uses them now because lacks a custom core); however, we will see the A57 is phones and tablets. For instance in those from Qualcomm

http://www.tomshardware.com/news/qualcomm-snapdragon-smartphones-mobile-processor,27464.html

Stop spreading misinformation.

Outside of the Apple Cyclone being in the iphone, etc., name one device that was actually meritorious containing one of the [strike](failures)[/strike] chips you mentioned above...?

 

[8350rocks]

hynix teases moar h.b.m. (high bandwidth memory)
http://techreport.com/news/27129/hynix-slides-tease-vertically-stacked-memory-with-256gb-s-of-bandwidth
256GB/s. me like.

applied micro has launched 64bit custome arm soc )afaik, applied micro co-developed arm v8 with arm)
http://www.anandtech.com/show/8580/hp-appliedmicro-and-ti-bring-new-arm-servers-to-retail
according to above, amd's seattle hasn't entered production yet.8350rocks, have you heard anything about that?

applied micro apparently has the successor lined up
http://www.tomshardware.com/news/armv8-64-bit-x-gen-hp-moonshot,27784.html

edit3: fixed wording.

SDKs are already out there for developers. Meaning that the silicon is basically in final stage, but they want some "launch" software availability, so while they finish tuning, etc. they are allowing some to go ahead and write/convert software over.

See the bit about hadoop someone referenced earlier...similar type marketing ploys will come out until they officially launch.

 

[juanrga]

you can run that 285 on stock but obviously the the difference in performance will dwindle too the same as it's power consumption. 5w to 10w give or take, still miserable against nvidia if you put it into perspective.

No, because

Performance ~ f
Power ~ f^3

Thus increasing frequency reduces efficiency.

Stop favoring Nvidia cards by repetitively comparing the efficiency and power consumption of overclocked AMD cards to stock Nvidia cards. Using your own words: it is "miserable".

 

[juanrga]

hynix teases moar h.b.m. (high bandwidth memory)
http://techreport.com/news/27129/hynix-slides-tease-vertically-stacked-memory-with-256gb-s-of-bandwidth
256GB/s. me like.

applied micro has launched 64bit custome arm soc )afaik, applied micro co-developed arm v8 with arm)
http://www.anandtech.com/show/8580/hp-appliedmicro-and-ti-bring-new-arm-servers-to-retail
according to above, amd's seattle hasn't entered production yet.8350rocks, have you heard anything about that?

applied micro apparently has the successor lined up
http://www.tomshardware.com/news/armv8-64-bit-x-gen-hp-moonshot,27784.html

edit3: fixed wording.

I never understood why AMD chose HBM. ARM, Intel, Nvidia, and others chose HMC, which is much faster. Intel will be selling the announced CPU with >500GB/s soon, and both Intel and Nvidia are working in future processors/cards with access to terabyte of bandwith.

AppliedMicro chip will be using a custom core that is about 38% faster than the A57 used in AMD Seattle. And efficiency will be about the same despite AppliedMicro will be using an ancient 40nm node for X-Gene v1. Their v2 will use 28nm and v3 will be using 16nm FinfFET.

Moreover the X-Gene SoC is paired with Nvidia Tesla cards for HPC applications

http://www.apm.com/news/apm-demonstrates-x-gene-for-hpc-environments-at-isc-14/
http://www.hpcwire.com/2014/06/23/gpus-add-arm-chips-hpc/
http://ragingbull.com/forum/topic/1004796

AMD did need Seattle for yesterday. I hope that will launch K12 on time.

 

[wh3resmycar]

you can run that 285 on stock but obviously the the difference in performance will dwindle too the same as it's power consumption. 5w to 10w give or take, still miserable against nvidia if you put it into perspective.

No, because

Performance ~ f
Power ~ f^3

Thus increasing frequency reduces efficiency.

Stop favoring Nvidia cards by repetitively comparing the efficiency and power consumption of overclocked AMD cards to stock Nvidia cards. Using your own words: it is "miserable".

that's cute, but count how many AIBs are actually selling the 285 at stock? then count the difference in $$$ between a stock version and an OC version.

newegg

so you're banking on those who bought the 285 to underclock their cards just to appease you? :lol:

P.S. i love the pseudo-math.

 

[de5_Roy]

AMD Launches A-Series Performance Mobile APUs (Kaveri) in Certain Markets
http://wccftech.com/amd-launches-aseries-performance-mobile-apus-markets/
lots of marketing promo slides.

HBM 3D Stacked Memory is up to 9X Faster Than GDDR5 – Coming With AMD R9 390X and R9 380X 20nm Pirate Islands
http://wccftech.com/amd-20nm-r9-390x-feautres-20nm-hbm-9x-faster-than-gddr5/

 

[wh3resmycar]

AMD Launches A-Series Performance Mobile APUs (Kaveri) in Certain Markets
http://wccftech.com/amd-launches-aseries-performance-mobile-apus-markets/
lots of marketing promo slides.

HBM 3D Stacked Memory is up to 9X Faster Than GDDR5 – Coming With AMD R9 390X and R9 380X 20nm Pirate Islands
http://wccftech.com/amd-20nm-r9-390x-feautres-20nm-hbm-9x-faster-than-gddr5/

i live >10km away from the venue and i wasn't aware that they're officially launching these chips here in the philippines. quite sad really, AMD don't have any marketing presence here, no TV/radio/even local youtube advert (which will be dirt cheap compared to the AMD logo you used to see on a ferrari f1 car) . i've stated this before, they hold a monopoly with MOBA-capable APUs, they're not aware of it unfortunately. The whole Southeast Asia region is glued to the MOBA scene and rarely you'll see AMD sponsored events (see dota2/LoL usage stats).

the a6-6400k is THE BEST CHIP for anything LoL related over here.