AMD CPU speculation... and expert conjecture

[blackkstar]

juangra actually kepler cards high end gpus take a lot less power then previous fermi graphics cards. I have downclocked my gtx670 100mhz below base clock and with i5 4430 it takes like 210w and plays the game I want at 60fps. My q9300 and 6850 took on similar game 220w. Heck a gtx780 takes slightly less then a gtx580 to run and its like just over double its performance.

I just realized something, the performance per dollar has really gone to hell with both the Radeon cards and nVidia cards. The 470 (350-380ish) and 5870 (3$90-420ish) had far more performance per dollar than per say today's 680/770 or 7970 at launch. When the Kepler and Southern Islands 680/7970 cards come along after Fermi and Evergreen/Northern Islands, they were at laughable prices ($600 7970 and 4GB 680, I am looking @ YOU!) Even more now, with the 780 and especially the TITAN are such terrible value for money, after all, their predecessor, the GTX 580 was only $500...

Nvidia has been leading the "GPU price creep" for a while.

It was the entire purpose of Titan. To set an astronomical price on a single GPU card, and then to release GTX 780 as "LE" version of Titan at a price much higher than what high end cards would normally be.

It worked wonders too. No one went "hey, GTX 780 is $150 more than what GTX 580 cost!" they went, "Oh this is a great deal for $650 I get a Titan that costs $999!"

I suspect this is what AMD is doing with FX 9590 and we will see AMD release desktop parts to fill in the $200 to $499 range which will look like absolutely fantastic values compared to FX 9590.

It is a marketing strategy that works absolutely fantastically. Intel does it too. 3930k would look pretty expensive, but when you compare it to $999+ 3970x which is just a little faster clock speed and more cache, 3930k suddenly looks like a great value.

It is evil and it works so well on people that it isn't even funny.

 

[noob2222]

Why did you get a 1000 watt psu if your only going to be pushing a max of 200W? Thats like totally inefficient, OMG your just ruined your PPW with a cheap psu.

 

[juanrga]

juangra actually kepler cards high end gpus take a lot less power then previous fermi graphics cards. I have downclocked my gtx670 100mhz below base clock and with i5 4430 it takes like 210w and plays the game I want at 60fps. My q9300 and 6850 took on similar game 220w. Heck a gtx780 takes slightly less then a gtx580 to run and its like just over double its performance.

I have actually discussed efficiency. Improved efficiency means that you can obtain the same performance consuming less power or that you can obtain more performance consuming the same power.

 

[lithium6]

lol my psu is not cheap. I just got it cheap hence I got it for my new rig. 5 star review.
http://www.overclockers.co.uk/showproduct.php?prodid=CA-233-SV
http://www.kitguru.net/components/power-supplies/zardon/silverstone-strider-plus-1000w-review-sst-st1000-p/
At 250w it is 87.74% efficient.
http://www.kitguru.net/components/power-supplies/zardon/silverstone-strider-plus-1000w-review-sst-st1000-p/5/

The problem is that you bought an overkill of a PSU because you got it cheap. The Seasonic G360 is over 91% efficient at the wattage you aim for: http://www.silentpcreview.com/article1297-page3.html 

By getting a big PSU cheap you will waste about 4% more electricity, that's more than 8W! 

Back to Steamroller: I seriously thought that AMD was about to unveil Athlon III FX for FM2+ when they hyped the 10 year anniversary of FX. All we got was a picture with two logos side by side. Thank you.

 

[gamerk316]

Considering Windows 8 BLOWS and this is an OS optimized for gaming. I'm happy with it. It would allow me to unclutter my desktop and use it for programs that support windows while I can have another large drive filled with games... who knows, they might even work a better cache system into there.

So you will reboot every time you run a game when you don't have to? Currently they don't seem to offer anything extra.

What do you have against windows 8 besides the tiles? The OS is perfectly functional and runs more than steamOS ever will. Will probably run most games better than steamOS since most PC devs don't use openGL and will be supporting windows as the main platform due to accessibility.

I don't think you really understand Linux performance. My Core 2 Duo 1.66ghz laptop with x1600 mobility chip runs Windows 7 horribly. I have Gentoo on it and tons of eye candy and it's fantastic.

Considering Windows 8 BLOWS and this is an OS optimized for gaming. I'm happy with it. It would allow me to unclutter my desktop and use it for programs that support windows while I can have another large drive filled with games... who knows, they might even work a better cache system into there.

So you will reboot every time you run a game when you don't have to? Currently they don't seem to offer anything extra.

What do you have against windows 8 besides the tiles? The OS is perfectly functional and runs more than steamOS ever will. Will probably run most games better than steamOS since most PC devs don't use openGL and will be supporting windows as the main platform due to accessibility.

I don't think you really understand Linux performance. My Core 2 Duo 1.66ghz laptop with x1600 mobility chip runs Windows 7 horribly. I have Gentoo on it and tons of eye candy and it's fantastic.

what does that have to do with games which run in runtimes like openGL which no dev is going to optimize for except for valve.

Windows does a LOT of stuff under the hood that 99% of people don't care about. But since MSFT aims at mass market appeal, they include those features by default anyways. As a result, a barebone Windows install requires a lot more to run then a barebones Linux install.

As for OGL, remember Value is basically using a D3D to OGL shim to get support up and running. As a result, I'd suspect a lot of games are going to lose some graphical features, simply because some stuff does not port over to OGL well this way.

BTW, OGL is a HORRID graphical API these days. Doesn't even reflect how the HW actually works anymore. The entire API needs to be replaced as far as I'm concerned. That's part of the reason no one targets it anymore.

 

[blackkstar]

http://www.xbitlabs.com/news/graphics/display/2013090230253_Nvidia_Is_Preparing_the_Most_Amazing_Thing_Ever_Made_Mark_Rein.html#discussion_last

I have been waiting for nvidia shadowplay for a long time. Now with a gtx 670 and 650m this will be great. I wonder if amd release something like this as shadowplay uses built in h264 encoder built into the kepler gpu.

Valve already did, it's a feature of SteamOS and it doesn't require proprietary hardware. It also is free as in beer and free as in freedom. Once again Nvidia is going to end up bringing a proprietary, locked down platform to compete against something free and open.

Look how well that worked out for PhysX.

 

[gamerk316]

Hell just froze over:

http://www.phoronix.com/scan.php?page=news_item&px=MTQ2NzY

NVIDIA just open-sourced its GPU documentation.

 

[gamerk316]

http://www.xbitlabs.com/news/graphics/display/2013090230253_Nvidia_Is_Preparing_the_Most_Amazing_Thing_Ever_Made_Mark_Rein.html#discussion_last

I have been waiting for nvidia shadowplay for a long time. Now with a gtx 670 and 650m this will be great. I wonder if amd release something like this as shadowplay uses built in h264 encoder built into the kepler gpu.

Valve already did, it's a feature of SteamOS and it doesn't require proprietary hardware. It also is free as in beer and free as in freedom. Once again Nvidia is going to end up bringing a proprietary, locked down platform to compete against something free and open.

Look how well that worked out for PhysX.

Bullet: Free and open, almost never used
PhysX: Free and closed, moderately used
Havok: Licensed and closed, heavily used

So...what point were you trying to make again?

 

[hcl123]

I think just the contrary. Intel is at the peak of the x86, whereas ARM is starting to release its true potential. Look at AMD, they have benchmarked the new A57 core and found it is better than its own jaguar core, which was already an optimal peak in x86 _64 bit_ space.

Better at what? On what software? Maybe the X86 code is REALLY unoptimized compared to the ARM code. Etc, etc, etc.

Again people: X86 isn't going anywhere because of the 20 years worth of software. So please, this whole "ARM is going to take over" is just nonsense.

Better at raw performance and performance per watt. I have no reason to believe that x86 code is not optimized.

ARM and Apple will be switching to an ARM world. Nvidia is preparing an entire range of ARM-based products from phone to supercomputers, including servers, tablets, desktops... and several of us are convinced than next gen game consoles will be ARM based. This gen was close to being it.

Ive already proven to you that ARM is no where near raw performance of X86 and now your back at it again, except this time its both raw and PPW?!

just admit your wrong and leave it at that. Oh, thats right, you can't, you know more than anyone working for AMD and Intel combined.

ARM ~= Atom << Ivy.

As for Nvidia's supercomputer, the ARM core is only there to turn on the quaddro gpus so they can do ALL the work. ARM is aimed to take on microservers, thats hardly considered supercomputers.

How can you prove anything if there isn't yet *any* ARM chip to compete with the likes of I7 and FX DTs ?

**Clock** counts a lot, and we don't know how ARM uarch evolution scales with clock

**DRAM speed and bandwidth** also counts and we don't know how it affects the ARM uarch

** Large sophisticated cache systems** how it will impact ARM uarch (v8).

That ARM has potentially (**already proven**) the best Perf/Watt around is unquestionable, and yet without sophisticated power management (which Intel and AMD have for DT and else).

An ARM at >3Ghz with cyclos RCM (as example) turbo and digital+temperature power monitoring, extensive clock and power gating with a sophisticated power micro-contoller with pertinent independent voltage regulators on FD-SOI with Dynamic Back Biasing Voltage and Frequency Scaling (BB-DVFS) scheme, is what is needed to compare with i7 in all metrics... a terrible beating is what i anticipate lol

x86 could contra-point with extensive speculation, Hardware Transactional Memory... speculative multi threading(spMT) (increase ILP by on-the-fly speculative threads since the simpler form of ILP is dead for x86) and high orders of memory parallelism and or *data-flow*, resorting to extensive vector crunching... but this also means the current iterations of x86 are more than obsolete, as bad as the ARM High Performance SKUs that don't exist yet


Yes... considering the ISA (instruction set arch), where x86 can resist is in sophisticated crunching (spMT,HTM etc) for vectors like instructions, meaning x86 will be more about "128 or better 256bit instructions" than anything else... 64bit integer x86 already lost... so please don't talk IPC, which traditional definition is exactly about 32/64 integer...

 

[-Fran-]

Hell just froze over:

http://www.phoronix.com/scan.php?page=news_item&px=MTQ2NzY

NVIDIA just open-sourced its GPU documentation.

Well, it is a very good step in the right direction IMO. AMD did it a long time ago and the open source drivers became really good.

And yes, it took everyone by surprise, hahaha. Darn.

Cheers!

 

[8350rocks]

http://www.xbitlabs.com/news/graphics/display/2013090230253_Nvidia_Is_Preparing_the_Most_Amazing_Thing_Ever_Made_Mark_Rein.html#discussion_last

I have been waiting for nvidia shadowplay for a long time. Now with a gtx 670 and 650m this will be great. I wonder if amd release something like this as shadowplay uses built in h264 encoder built into the kepler gpu.

Valve already did, it's a feature of SteamOS and it doesn't require proprietary hardware. It also is free as in beer and free as in freedom. Once again Nvidia is going to end up bringing a proprietary, locked down platform to compete against something free and open.

Look how well that worked out for PhysX.

Bullet: Free and open, almost never used
PhysX: Free and closed, moderately used
Havok: Licensed and closed, heavily used

So...what point were you trying to make again?

Bullet is used as much or more than PhysX.

Though, I will grant you Havok is used more than both of them...how that happens is mind boggling honestly. Though I guess it's brand recognition that some developers look for...Havok offers less than both of the others...

 

[esrever]

Hell just froze over:

http://www.phoronix.com/scan.php?page=news_item&px=MTQ2NzY

NVIDIA just open-sourced its GPU documentation.

They want in on the valve steamOS push

 

[Cazalan]

The day is done and still nothing, they were either going to announce something and went back or they were just trying to troll and piss some fans.

Reading it was kinda sad. 10 years to go from 3.2Ghz to 4.7Ghz.

How many thought we'd be at 10+ Ghz by now?

 

[noob2222]

http://www.xbitlabs.com/news/graphics/display/2013090230253_Nvidia_Is_Preparing_the_Most_Amazing_Thing_Ever_Made_Mark_Rein.html#discussion_last

I have been waiting for nvidia shadowplay for a long time. Now with a gtx 670 and 650m this will be great. I wonder if amd release something like this as shadowplay uses built in h264 encoder built into the kepler gpu.

Valve already did, it's a feature of SteamOS and it doesn't require proprietary hardware. It also is free as in beer and free as in freedom. Once again Nvidia is going to end up bringing a proprietary, locked down platform to compete against something free and open.

Look how well that worked out for PhysX.

Bullet: Free and open, almost never used
PhysX: Free and closed, moderately used
Havok: Licensed and closed, heavily used

So...what point were you trying to make again?

Bullet is used as much or more than PhysX.

Though, I will grant you Havok is used more than both of them...how that happens is mind boggling honestly. Though I guess it's brand recognition that some developers look for...Havok offers less than both of the others...

Intel pushes Havok hard, Im wondering if they included their "GenuineIntel" flags into the programming. It seems since they took over AMD cpus suffer in titles that utilize the Havok engine.

 

[8350rocks]

Intel pushes Havok hard, Im wondering if they included their "GenuineIntel" flags into the programming. It seems since they took over AMD cpus suffer in titles that utilize the Havok engine.

Honestly, that would not *at all* surprise me at this point...considering everything else they've done.

Why anyone would pay money to use an inferior physics engine, when a superior engine is available for free, still blows my mind.

 

[8350rocks]

There is interest in ARM simply because X86 does not scale downward particularly well. This is nothing new. That being said, Windows itself doesn't scale downward very well either. X86/Windows is going to remain the desktop platform of choice, and ARM/Android will remain the mobile platform of choice. The end.

This is what I've been trying to tell people. In order to scale up you need to start adding dedicated things to the die that aren't directly related to processing power. Cache / instruction schedulers / branch predictors / ect. Those things consume power and space and thus your cost per performance starts to erode rapidly. ARM doesn't scale up nearly as well as it's proponents say. It's a horizontal architecture, you get high performance by connection dozens of ARM processors to each other and executing a ton of distributed code. Try to close an ARM CPU very high and you slam into the same problems everyone else ran into two decades ago.

As for RISC CPU's being high powered, their have been high performance RISC CPU's for a very long time. RISC is actually much easier to make *fast* due to it's incredibly predictable instruction execution time. SUN even tried making a "Desktop" SPARC machine (Sunblade 100/150), didn't work out too well. Apple was using PPC which worked out better then the Sunblade's but even they switched over to x86 eventually. x86 is just cheaper to design around and already has a huge ecosystem to draw revenue from.

So the entire ARM vs x86 is just some wishful thinking. ARM is amazing at low power (electrical usage) devices that do just enough to be useful. It doesn't scale up nearly as well as the other design's but it's ability to scale down is unmatched.

+1

I have been trying to tell them ARM doesn't scale upwardly nearly as well as they think...it just wasn't sinking in, I guess. It scales downward ridiculously well, however, to scale it upward...you're really going into larger and larger clusters of ARM CPUs to get the performance. Otherwise the PPW angle will vanish, and the hardware will have to have more and more added complexities which x86 has been integrating for 10 years now.

ARM is on the precipice of 64 bit processing, but they're not anywhere near the complexity and capability of the x86-64 ISA at this point. It doesn't mean ARM isn't viable, by any stretch, for what it's intended use is; however, it doesn't make sense from a DT OS perspective, or even a high performance server angle either.

ARM will not defeat x86-64 anytime soon. AMD will get into that market to try and grab a toe hold in ARM micro servers because of the margins. Other than that, ARM will remain THE gold standard for mobile tablets, phones, etc. It will not replace HEDT anytime soon either, as that segment of x86 DTs is clearly on the rise...by a fair margin I might add.

 

[gamerk316]

How can you prove anything if there isn't yet *any* ARM chip to compete with the likes of I7 and FX DTs ?

Historically, RISC does not scale as well per clock as X86, and RISC architectures (notably PPC and MIPS) have had issues getting clocks above 3GHz. Using those two arches as a baseline, you can guesstimate performance for ARM.

Bullet is used as much or more than PhysX.

Though, I will grant you Havok is used more than both of them...how that happens is mind boggling honestly. Though I guess it's brand recognition that some developers look for...Havok offers less than both of the others...

No, these days, PhysX (including CPU based PhysX) is used far, far more then Bullet in games. And Havok is used the most due to the fact it has the best toolchain of the bunch and the fact its licensed by just about every game engine out there.

 

[8350rocks]

How can you prove anything if there isn't yet *any* ARM chip to compete with the likes of I7 and FX DTs ?

Historically, RISC does not scale as well per clock as X86, and RISC architectures (notably PPC and MIPS) have had issues getting clocks above 3GHz. Using those two arches as a baseline, you can guesstimate performance for ARM.

Bullet is used as much or more than PhysX.

Though, I will grant you Havok is used more than both of them...how that happens is mind boggling honestly. Though I guess it's brand recognition that some developers look for...Havok offers less than both of the others...

No, these days, PhysX (including CPU based PhysX) is used far, far more then Bullet in games. And Havok is used the most due to the fact it has the best toolchain of the bunch and the fact its licensed by just about every game engine out there.

While the list of AAA titles might be shorter...games running Bullet physics are more common than you think. A lot of them are Indie games using the open source Blender Game Engine (which has Bullet physics integrated into it). Though, that doesn't detract from the fact that Bullet Physics is used in just as many games.

Additionally, Havok tools are probably the easiest to use, I will give you that. Though, frankly, I think that speaks to the laziness of developers more than anything. Bullet doesn't need to be licensed to be used by a game engine.

 

[gamerk316]

Additionally, Havok tools are probably the easiest to use, I will give you that. Though, frankly, I think that speaks to the laziness of developers more than anything

We're willing to spend more to shave months of overtime off our schedules, thank you very much. It's not like we, the people who make software, have any say over the timeline of the product release schedule...

Hence: Havok over PhysX and Bullet, DirectX over OpenGL, MSVC over GCC, and so on. We like making things REALLY simple: That's why we wrote software to write software for us for a chosen target platform (compilers).

 

[de5_Roy]

^^ said xzibit....wryly...

 

[8350rocks]

Additionally, Havok tools are probably the easiest to use, I will give you that. Though, frankly, I think that speaks to the laziness of developers more than anything

We're willing to spend more to shave months of overtime off our schedules, thank you very much. It's not like we, the people who make software, have any say over the timeline of the product release schedule...

Hence: Havok over PhysX and Bullet, DirectX over OpenGL, MSVC over GCC, and so on. We like making things REALLY simple: That's why we wrote software to write software for us for a chosen target platform (compilers).

I am not disputing ease of use being a factor; however, there is something to be said for being able to make the engine do specifically what you want it to as well.

DX over OGL, I totally empathize with...believe me, I have been complaining about OGL for some time. When we discussed the D3D "wrapper" being used to convert D3D calls to OGL, I was somewhat surprised someone had taken the time to do it, but it would make life much easier on that end, as DX tools are clearly better setup.

As far as compilers go...GCC isn't really all that bad to use. MSVC is moderately easier, but it's not like running a compiler is rocket science anyway.

I think some of that can be lumped into "creature of habit" tendencies, and some of that is genuinely true of the quality of the tool interfaces and functionality.

 

[The Q6660 Inside]

juangra actually kepler cards high end gpus take a lot less power then previous fermi graphics cards. I have downclocked my gtx670 100mhz below base clock and with i5 4430 it takes like 210w and plays the game I want at 60fps. My q9300 and 6850 took on similar game 220w. Heck a gtx780 takes slightly less then a gtx580 to run and its like just over double its performance.

I just realized something, the performance per dollar has really gone to hell with both the Radeon cards and nVidia cards. The 470 (350-380ish) and 5870 (3$90-420ish) had far more performance per dollar than per say today's 680/770 or 7970 at launch. When the Kepler and Southern Islands 680/7970 cards come along after Fermi and Evergreen/Northern Islands, they were at laughable prices ($600 7970 and 4GB 680, I am looking @ YOU!) Even more now, with the 780 and especially the TITAN are such terrible value for money, after all, their predecessor, the GTX 580 was only $500...

Nvidia has been leading the "GPU price creep" for a while.

It was the entire purpose of Titan. To set an astronomical price on a single GPU card, and then to release GTX 780 as "LE" version of Titan at a price much higher than what high end cards would normally be.

It worked wonders too. No one went "hey, GTX 780 is $150 more than what GTX 580 cost!" they went, "Oh this is a great deal for $650 I get a Titan that costs $999!"

I suspect this is what AMD is doing with FX 9590 and we will see AMD release desktop parts to fill in the $200 to $499 range which will look like absolutely fantastic values compared to FX 9590.

It is a marketing strategy that works absolutely fantastically. Intel does it too. 3930k would look pretty expensive, but when you compare it to $999+ 3970x which is just a little faster clock speed and more cache, 3930k suddenly looks like a great value.

It is evil and it works so well on people that it isn't even funny.

Same applied somewhat to the 7970. inb4 GTX 880 Ultra. We can't really stop the "ERMEHGAWD ITS LIAK SHO MUCH VALUE FOR MERNEH COMPAHURRED TO THE TITAN AND HAS LIAK 13 PERCENT INCREASE OVER DEH 680 FOAR LIAK TWO HUNDRED BAWX MOAR" logic. Even the somewhat expensive back then and sexy looking GTX 480 curbstomps today's GTX 780 in value for money for their respective eras, don't even get me started on GTX 470 vs GTX 770.. If only we went back to 2010, oh 5870 how much I loved you....

Anyways, SteamOS will likely be nothing large for the PC gaming community, aside from perhaps a few Source games... There will still be a huge community of Windows Vista/7 gamers out there.

 

[8350rocks]

Anyways, SteamOS will likely be nothing large for the PC gaming community, aside from perhaps a few Source games... There will still be a huge community of Windows Vista/7 gamers out there.

Yes, but we chip away at Winblows, 1 piece at a time...one small crack expands over time until it begins to become a larger problem.

Then, eventually, WE RULE THE WORLD!!! *cough* err...I mean...yeah. <.<...>.>...nobody heard that last bit right?

 

[The Q6660 Inside]

Anyways, SteamOS will likely be nothing large for the PC gaming community, aside from perhaps a few Source games... There will still be a huge community of Windows Vista/7 gamers out there.

Yes, but we chip away at Winblows, 1 piece at a time...one small crack expands over time until it begins to become a larger problem.

Then, eventually, WE RULE THE WORLD!!! *cough* err...I mean...yeah. <.<...>.>...nobody heard that last bit right?

TIS UBUNTU TIAM!

 

[montosaurous]

I'm pretty sure there aren't any Vista gamers left out there, or very, very few.