AMD CPU speculation... and expert conjecture

[gamerk316]

Lots of RISC designs have tried to compete with x86 and have not been able to keep up. I realize x86 itself is an ugly mess and x86 ASM over Power or SPARC or whatever is for people who like really good challenges, but people have tried throughout history to unseat x86 with RISC and it never works, at least in the long run.

Also, ARM was designed specifically to not target high performance x86 so they didn't have to take Intel on directly. Much like Via's plans with their CPUs even though they are x86.

Historically I don't see ARM being this magic bullet. In fact I recall hearing similar arguments that x86 is in trouble when PowerPC was being pushed in Macs and there were a lot of fanboys clamoring at the fact that PPC was infinitely better.

Architecturally, PPC was better. As was the 6800 before it. But Intel knows how to improve and market their chips over time.

 

[Cazalan]

Kaveri 7850K ($179)
Athlon X4 860K + Radeon HD 7750 ($89 + $89)

APUs still don't make sense for desktops. Since the GPU is now 1/2 the die they could be making a true 860K for less than half the price. AMD was a bit too early with the APU push.

 

[con635]

Kaveri 7850K ($179)
Athlon X4 860K + Radeon HD 7750 ($89 + $89)

APUs still don't make sense for desktops. Since the GPU is now 1/2 the die they could be making a true 860K for less than half the price. AMD was a bit too early with the APU push.

a8 7600 $103, 65w and 90% of the performance of 7850K

 

[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...

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

 

[8350rocks]

When we find out that ARM scaled up won't beat X86, I will be in full "I told you so" mode. Shame 8350 is under NDA and can't give us a reveal timeline.

Emphasis mine...

 

[juanrga]

I think that at ~100W level ARM is about 20--30% more efficient than x86. There are people who agrees on this and others who disagree, e.g. D. Kanter think that K12 will only be 10% better than Zen.

 

[jdwii]

Almost everyone in the industry applaud AMD strategy of migrating from x86 to ARM to stole market server to Intel. We also know that half-dozen of people that dislike/hate AMD will continue posting nonsense and lies in forums (like their "ARM cannot scale up because isn't CISC", "Denver is quad-core", "ARM cannot do OoO", "ARM only can run Android"...) and they will continue attacking and lying about everyone who don't accept their silly viewpoints.

Of course, no amount of your trolling/lying will change the facts but you can think otherwise :-D

juan just stop it, of course it was a good move for Amd why continue a battle they can't win in? Dumb IMO
WTF who said arm can only run on andoird is this another lie?
DO you even understand why CISC has its places in the market?
Also its not attacking its when you cherry pick results something i never go for you have to use the majority.
Arm can scale up why do you insist on saying people think otherwise? Its just going to take a lot of time to do it and it will require more complicated designs

 

[jdwii]

Regarding performance

HD 7750 DDR3 version << Kaveri iGP << HD 7750 GDDR5 version

I think that at ~100W level ARM is about 20--30% more efficient than x86. There are people who agrees on this and others who disagree, e.g. D. Kanter think that K12 will only be 10% better than Zen.

Please explain your definition on efficiency? Also please show us where you got that Nvidia quote about dgpu's dying i asked this before i want to see the direct quote when you show it or you made it up. Not saying you made it up but please i want to read it myself, as stated before i looked for the quote already and i couldn't find it.

 

[palladin9479]

Kaveri 7850K ($179)
Athlon X4 860K + Radeon HD 7750 ($89 + $89)

APUs still don't make sense for desktops. Since the GPU is now 1/2 the die they could be making a true 860K for less than half the price. AMD was a bit too early with the APU push.

As con just said the 7600 is a far better cost / value then the 7850K, even the 6800K is a better cost / value then the 7850K. The 7850K really has no place in the market, it's too expensive to compete in the value segment when someone can take a 750/860 and marry it to a low end GDDR5 dGPU. The price has to be under $150 USD for it to be competitive and preferably at 65W or less. That being said, my 7600 is kicking a$$ right now. There is only one problem, you can't really OC the north-bridge which removes a sizable performance tweak for knowledgeable people. Doesn't effect graphics performance though which is what your buying that chip for anyway.

And gamer memory is a huge issue for iGPU's. You can see this going from DDR3-1600 to DDR3-1866, 2133 and 2400. Overclocking memory has a much larger impact on the APU's graphics performance then overclocking the core clock. Latency means nothing, it needs sheer bandwidth to reach it's full potential. And folks the iGPU doesn't really use the IMC, not in the same way the iCPU does. That's why AMD can have relatively poor performance on memory access yet the iGPU's get full speed.

 

[de5_Roy]

TressFX Hair: Cross-platform and v2.0
http://community.amd.com/community/amd-blogs/amd-gaming/blog/2014/09/23/tressfx-hair-cross-platform-and-v20
Full AMD Tonga GPU might feature 384-bit memory interface
http://videocardz.com/52834/full-amd-tonga-gpu-might-feature-384-bit-memory-interface
http://pc.watch.impress.co.jp/docs/column/kaigai/20140926_668620.html

PCWatch also reports that Tonga is HSA-compatible, and unlike Kaveri (which is in third stage), Tonga would be in the final fourth state of HSA implementation.
What it means is that Tonga already features GPU Compute Context Switch and GPU Graphics Preemption.

AMD Is Still Moving Towards A Unified Open-Source Driver
http://www.phoronix.com/scan.php?page=news_item&px=MTc5ODE
The Next AMD Catalyst Linux Driver Is Much More Exciting
http://www.phoronix.com/scan.php?page=news_item&px=MTc5NzI
SUSE Publishes More GCC Patches For AMD's HSA Architecture
http://www.phoronix.com/scan.php?page=news_item&px=MTc5NzA

edit: fixed the wrong link. thanks blackkstar.

A Future Catalyst Update Will Improve CS:GO's Linux Performance
http://www.phoronix.com/scan.php?page=news_item&px=MTc5ODQ

 

[blackkstar]

TressFX Hair: Cross-platform and v2.0
http://community.amd.com/community/amd-blogs/amd-gaming/blog/2014/09/23/tressfx-hair-cross-platform-and-v20
Full AMD Tonga GPU might feature 384-bit memory interface
http://videocardz.com/52834/full-amd-tonga-gpu-might-feature-384-bit-memory-interface
http://pc.watch.impress.co.jp/docs/column/kaigai/20140926_668620.html

PCWatch also reports that Tonga is HSA-compatible, and unlike Kaveri (which is in third stage), Tonga would be in the final fourth state of HSA implementation.
What it means is that Tonga already features GPU Compute Context Switch and GPU Graphics Preemption.

AMD Is Still Moving Towards A Unified Open-Source Driver
http://www.phoronix.com/scan.php?page=home
The Next AMD Catalyst Linux Driver Is Much More Exciting
http://www.phoronix.com/scan.php?page=news_item&px=MTc5ODE
SUSE Publishes More GCC Patches For AMD's HSA Architecture
http://www.phoronix.com/scan.php?page=news_item&px=MTc5NzA

Such good Linux news. I am using Radeon driver on my 7970 under Gentoo, couldn't be happier. Other than lack of OpenCL I am really happy with it, specially compare to fglrx. Haven't used Nvidia closed Linux drivers since 8800GTS 320mb so I can't comment there, but from what I have heard Nvidia closed source has lost quality a bit and open source Nvidia drivers are trash. And if you're using Linux, you want everything as open as possible because closed source things are almost always a pain in the butt (like Wifi drivers)

Identifying OpenMP loops? Are they talking about just making anything with OpenMP use HSA? My mouth is drooling and my body is ready. An HSA Gentoo system is going to stomp so hard if that's the case.

Glorious Gentoo days are ahead. As soon as this stuff is available on Linux I'm going to buy it all. Blender has OpenMP so I'll probably get HSA out of that right out of the box. Absolutely amazing. And I thought talking about 80%+ performance increase with optimized Blender was sweet, this is going to be godly I hope. I just hope we can get something a little bigger than a single APU system for this. If I'm gonna spend tens of hours a week baking and rendering I wanna see some serious silicon working together.

I think you missed a link though.

 

[etayorius]

TressFX Hair: Cross-platform and v2.0
http://community.amd.com/community/amd-blogs/amd-gaming/blog/2014/09/23/tressfx-hair-cross-platform-and-v20
Full AMD Tonga GPU might feature 384-bit memory interface
http://videocardz.com/52834/full-amd-tonga-gpu-might-feature-384-bit-memory-interface
http://pc.watch.impress.co.jp/docs/column/kaigai/20140926_668620.html

PCWatch also reports that Tonga is HSA-compatible, and unlike Kaveri (which is in third stage), Tonga would be in the final fourth state of HSA implementation.
What it means is that Tonga already features GPU Compute Context Switch and GPU Graphics Preemption.

AMD Is Still Moving Towards A Unified Open-Source Driver
http://www.phoronix.com/scan.php?page=news_item&px=MTc5ODE
The Next AMD Catalyst Linux Driver Is Much More Exciting
http://www.phoronix.com/scan.php?page=news_item&px=MTc5NzI
SUSE Publishes More GCC Patches For AMD's HSA Architecture
http://www.phoronix.com/scan.php?page=news_item&px=MTc5NzA

edit: fixed the wrong link. thanks blackkstar.

A Future Catalyst Update Will Improve CS:GO's Linux Performance
http://www.phoronix.com/scan.php?page=news_item&px=MTc5ODQ

Awesome news, anything related to MANTLE, TressFX, HSA, OpenCL, Havok and Bullet excites me, it was great to see a comparison between TressFX2.0 and HairWorks.

 

[juanrga]

Neither Nvidia linux drivers are "trash" nor AMD linux drivers are the best.

A pair of weeks ago Valve opened a special thread in forums for discussing problems with AMD hardware on SteamOS machines. No similar thread exist for Intel or Nvidia hardware which gives a idea of how good is their support of Linux.

The news
http://www.phoronix.com/scan.php?page=news_item&px=MTc5MDI
http://www.muycomputer.com/2014/09/23/soporte-linux-gpus-amd

The thread
http://steamcommunity.com/groups/steamuniverse/discussions/1/616187839533548192/#

 

[juanrga]

AMD Catalyst Satisfaction Feedback

http://www.amdsurveys.com/se.ashx?s=5A1E27D26C60C805&spMailingID=21506523&spUserID=NzI2NzcyNDE4NjQS1&spJobID=401628065&spReportId=NDAxNjI4MDY1S0

 

[gamerk316]

Yeah, the NVIDIA binary Linux driver is considered very good. NVIDIA typically wins easily over AMD on linux.

 

[jdwii]

I noticed i had a smoother experience when testing mint my favorite distro using nvidia. But i think Both are stable now but tbh i don't use linux enough to really claim what is smoother.

 

[juanrga]

I use AMD and Mint and I am satisfied, but I recognize that Nvidia/Intel support for Linux is better.

 

[juanrga]

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.

 

[blackkstar]

I noticed i had a smoother experience when testing mint my favorite distro using nvidia. But i think Both are stable now but tbh i don't use linux enough to really claim what is smoother.

I am mostly talking about effort required to keep system up to date. With Radeon drivers I can stay on latest xorg and such with no problems. There is no reinstalling drivers after updates, things breaking, etc. In fact I never have done an update to my system with Radeon drivers and had no 3d acceleration after rebooting the next day.

I have had all of that with fglrx and nvidia drivers (but nvidia was some time ago). But given my experience with closed source on Linux that's expected. Then again I am a bleeding edge Gentoo user and I want the latest and greatest everything.

If you are having problems with smoothness you can tweak compositor target frame rate or play with vsync. It is one of the big perks of open source.

Regardless AMD has been making a lot of really good strides in Linux. I remember in the 00s going into a forum and saying "I'm having problems with fglrx" was universally met with "buy Nvidia, the ATI is beyond salvation" and now that's simply not the case.

All I really want to see is OpenCL compiler be able to handle large compilers so I can use my 7970 with Blender.

 

[de5_Roy]

I use AMD and Mint and I am satisfied, but I recognize that Nvidia/Intel support for Linux is better.

amd what, gpu or cpu/apu or both? can you be more specific?

 

[cdrkf]

As has been shown: ISA doesn't matter when it comes to performance. So at best, you'll get ARM equaling X86, which won't be enough to get people to switch over, due to lack of SW.

Oh wait a sec, ARM won't have AVX or SSE. So even if ARM is competitive at the ISA level, it will lose due to CPU extensions.

Seriously, AMD couldn't beat intel when making a x86-64 chip, and they're the ones who created the bloody ISA in the first place. So why should we suddenly believe AMD could take a foreign CPU architecture, add in 20 years worth of HW improvements, minus the CPU extensions, and somehow come out with a chip that out-competes a high end X86-64 chip from Intel? Because when you put it that way, you realize how silly a prospect what AMD is trying to do really is.

I use AMD and Mint and I am satisfied, but I recognize that Nvidia/Intel support for Linux is better.

amd what, gpu or cpu/apu or both? can you be more specific?

I think given the context of the discussion he's referring to GPU... AMD *open source* Linux drivers are much better than those available for Nvidia. I'm not sure why that would be, maybe Nvidia isn't keen to support the development of 3rd part drivers for their products? I don't know though...

 

[blackkstar]

http://yosoygames.com.ar/wp/2014/09/is-everything-ok-nvidia/

Found it on S|A, couldn't find it again though so I can't give credit where it's due. But from that it sounds like Nvidia wants to keep as much as possible top secret and AMD is more concerned about users.

 

[gamerk316]

As has been shown: ISA doesn't matter when it comes to performance. So at best, you'll get ARM equaling X86, which won't be enough to get people to switch over, due to lack of SW.

Oh wait a sec, ARM won't have AVX or SSE. So even if ARM is competitive at the ISA level, it will lose due to CPU extensions.

Seriously, AMD couldn't beat intel when making a x86-64 chip, and they're the ones who created the bloody ISA in the first place. So why should we suddenly believe AMD could take a foreign CPU architecture, add in 20 years worth of HW improvements, minus the CPU extensions, and somehow come out with a chip that out-competes a high end X86-64 chip from Intel? Because when you put it that way, you realize how silly a prospect what AMD is trying to do really is.

I use AMD and Mint and I am satisfied, but I recognize that Nvidia/Intel support for Linux is better.

amd what, gpu or cpu/apu or both? can you be more specific?

I think given the context of the discussion he's referring to GPU... AMD *open source* Linux drivers are much better than those available for Nvidia. I'm not sure why that would be, maybe Nvidia isn't keen to support the development of 3rd part drivers for their products? I don't know though...

Open Source side, it isn't close: The NVIDIA OS Drivers suck, mainly because of all the undocumented things NVIDIA has put in the HW quietly over the years to make things work better. That being said, the state of OS GPU Drivers is sad, since none of them have even basic reclocking support yet, making them unusable in mobile settings, due to power draw.

In terms of all GPU drivers though? The NVIDIA blog blows everything else out of the water, performance wise. On Windows, you can say the AMD 290 is slightly better then the NVIDIA 770, but on Linux, its a beatdown in the other direction.

 

[cdrkf]

As has been shown: ISA doesn't matter when it comes to performance. So at best, you'll get ARM equaling X86, which won't be enough to get people to switch over, due to lack of SW.

Oh wait a sec, ARM won't have AVX or SSE. So even if ARM is competitive at the ISA level, it will lose due to CPU extensions.

Seriously, AMD couldn't beat intel when making a x86-64 chip, and they're the ones who created the bloody ISA in the first place. So why should we suddenly believe AMD could take a foreign CPU architecture, add in 20 years worth of HW improvements, minus the CPU extensions, and somehow come out with a chip that out-competes a high end X86-64 chip from Intel? Because when you put it that way, you realize how silly a prospect what AMD is trying to do really is.

I use AMD and Mint and I am satisfied, but I recognize that Nvidia/Intel support for Linux is better.

amd what, gpu or cpu/apu or both? can you be more specific?

I think given the context of the discussion he's referring to GPU... AMD *open source* Linux drivers are much better than those available for Nvidia. I'm not sure why that would be, maybe Nvidia isn't keen to support the development of 3rd part drivers for their products? I don't know though...

Open Source side, it isn't close: The NVIDIA OS Drivers suck, mainly because of all the undocumented things NVIDIA has put in the HW quietly over the years to make things work better. That being said, the state of OS GPU Drivers is sad, since none of them have even basic reclocking support yet, making them unusable in mobile settings, due to power draw.

In terms of all GPU drivers though? The NVIDIA blog blows everything else out of the water, performance wise. On Windows, you can say the AMD 290 is slightly better then the NVIDIA 770, but on Linux, its a beatdown in the other direction.

770? I think you mean 780 But agreed on Linux the 770 is probably faster

 

[juanrga]

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