AMD Piledriver rumours ... and expert conjecture

Page 150 - 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.
We have had several requests for a sticky on AMD's yet to be released Piledriver architecture ... so here it is.

I want to make a few things clear though.

Post a question relevant to the topic, or information about the topic, or it will be deleted.

Post any negative personal comments about another user ... and they will be deleted.

Post flame baiting comments about the blue, red and green team and they will be deleted.

Enjoy ...
 
But make it a generic brand at least, I'd like to avoid the ibupro in the blue package, hahaha.

I'm not that hurt though, since the MoBo has been excellent in so many ways for its price and I can still hope for PD to be an upgrade. Even if it is small, I'll take it, lol. It will be cheaper than building a whole new Intel rig (mal and I already put up some numbers for that, hehe).



Wait, but weren't compilers supposed to "patch" the code for that? I mean, if you CPU doesn't have it, then use the other ASM combo for it, right? Or you're still talking about the dispatcher model? Or I just got horribly confused? hahaha.

Cheers! 😛

Ok image you write a simple program, the infamous "hello world" what displays the result of some vector calculation.

Intel compiler compiles it and creates multiple ways to execute that program, namely in the calculation part. The selection of these paths will depend on the CPU's ProductID string and only if the VenderID strong is GenuineIntel. If the VenderID is not GenuineIntel then the dispatcher will automatically use the lowest compiled code path regardless of the capabilities of the CPU. Prior to 2010 this path was i386/8087, after 2010 this path was SSE2 / MMX. Those flags you pass to the compiler can define to what level the lowest compiled path use's, thus you can "enable" it on non-Intel CPU's by making the non-Intel code path compiled with SSE3/4 (AVX currently doesn't work). The side effect is that because your lowest code path is now set to a higher level, the program will not run on CPU's that don't have those instructions defined in the lowest code path. Software makes, especially game makers, do not like to do this as it limits their market size.

How lets take your program one step further, it actually counts the number of repetitions of that vector calculation and reports that to you along with how long it took to do those repetitions. *BAM* you just made a benchmark, and if it was compiled with Intel's compiler it would be horribly biased in it's number crunching. I've always debated writing a dirty program that did something similar to this and biasing it against Intel CPU's, just to show how much of a difference instruction selection can make, never could be bothered honestly.
 
@palladin9479:
Intel compiler compiles it and creates multiple ways to execute that program, namely in the calculation part. The selection of these paths will depend on the CPU's ProductID string and only if the VenderID strong is GenuineIntel. If the VenderID is not GenuineIntel then the dispatcher will automatically use the lowest compiled code path regardless of the capabilities of the CPU. Prior to 2010 this path was i386/8087, after 2010 this path was SSE2 / MMX. Those flags you pass to the compiler can define to what level the lowest compiled path use's, thus you can "enable" it on non-Intel CPU's by making the non-Intel code path compiled with SSE3/4 (AVX currently doesn't work). The side effect is that because your lowest code path is now set to a higher level, the program will not run on CPU's that don't have those instructions defined in the lowest code path. Software makes, especially game makers, do not like to do this as it limits their market size.
post 2010, is this an illegal move from intel? seems to me that intel found a way around the court order with this:
"The side effect is that because your lowest code path is now set to a higher level, the program will not run on CPU's that don't have those instructions defined in the lowest code path."
if this is illegal, what is amd doing to remedy that and why doesn't amd or anyone from amd bring that up? i only see these brought up during this kind of debate and in news articles that favor amd.
 
undocumented problem .... 😱

just curious, is that your justification for everything being ok to take a big dump on AMD? oh, AMD sucks so bad, if anything actually tried to work, its broken, thats why it was done this way. AMD cpus are broken. AMD gpus are broken. anything AMD has is broken. ... AMD would be bankrupt if that was the case from all the lawsuits against them.

Why do you think people were thinking of doing Nvidia + AMD cards in the first place? the people actually doing it weren't idiots who didn't have a clue.

its not broken. its business, business to screw over the competition, ie AMD.

The thing is if AMD even tried some of the stuff that has been pulled on them, they would be sued so fast their head would spin.

http://www.techradar.com/news/computing-components/intel-sues-amd-over-breach-of-agreement-585553?src=rss

AMD is under a microscope all the time while Intel are sitting on mount olympus where no one is allowed to look down upon them. It is encouraged to take a leak on AMD. This mentality is complete bs. Too many people think this way.

Actually AMDs GPUS since the 4000 series and up have been great. The HD2K/3K were meh, although I liked my HD2900 Pro. The HD7K is wayyyy too over priced honestly but then again I got used to $300 dollar high end GPUs.

As for the screw over, nVidia screwed everyone over in many ways. The SLI on their chipset only crap during Core 2 Quad/Phenom was pretty crappy. Then they got poed when they didn't get a new license for Intels chipsets but I understand why not. Intel, and AMD, are moving everything to the CPU. There is not chipset for PCIe anymore, the south bridge is SATA/USB mainly as PCIe is on the CPU die. So it would have been a pointless license in the end.

As for the PhysX crap, I can understand why nVidia did it as they cannot gurantee it works properly with every possible configuration of AMD GPU and one nVidia GPU (be it single Radeon or CFX). Its much like how PC game devs cannot gurantee that upon release any persons PC will not have an issue with said game as its impossible to test every single hardware configuration out there possible.

Still PhysX is over rated. I like the idea but its nothing needed as it doesn't add to gameplay which I find much more important.

looks like intel (like nvidia) gets deep into game development with other companies. amd doesn't do that, usually.
i noticed that in this specific case, starcraft uses havok - owned by intel. it's kind of a no-brainer that intel would push software optimization for their own hardware. say, intel put the compiler fix in havok but forgets to turn it on. i don't think they're really breaking the law or explicitly sabotaging amd. according to court order, intel put in the compiler fix in the software (which is turned off by default) and leaves as it is, since havok belongs to intel.
i've noticed that it almost always turns to games but there are other softwares people use.

Intel leaves it off as they, again, cannot gurantee how it will work or if it will even benefit the AMD CPUs at all. They do have a disclaimer on the website.

Stalker may be optimized for AMD, but was it crippled for Intel systems in the same way SC2 was crippled for AMD?

http://downloadsquad.switched.com/2010/01/04/intel-forced-to-provide-a-compiler-that-isnt-crippled-for-amd-processors/

What software vendor in any industry would cripple their software for 80% of the market. When you make the program run equally on both sets of hardware, guess what? Intel is only ~10% faster.

But no one wants to see that, they want to see massive differences that SC2 show. Ultimately thats Intel's agenda. Ignore all the programs that run similar, and only look at software that Intel pushed, either with their compiler or with the Havok engine.

Stalker and SCII, again as I said before, two completley different game types. RTS performance cannot be compared to a FPS games performance. A FPS does not have nearly as much AI going as a RTS which utilizes CPU power, not GPU power. Thats why the older CPU tests in 3DMark had a ton of robots running around and not intensive graphics.

but yet the fastest amd cpu is 28% slower than the slowest intel system in that test. and 40% on the top end with both cpus being clocked slower.

It is possible. Athlon 64 was quite a bit faster than Pentium 4 at lower speeds. So why couldn't Intel have the same lead in performance considering in the same time frame Intel has introduced 6 new CPUs, all with performance gains over the previous, while AMD has introduced 3 (4 if you count Thuban when it was just a 6 core Deneb), some that didn't have performance gains (or veryy little with Athlon 64 -> Phenom).

So it would seem to me that it is within the relm of possibility that Intel could have a decent performance lead on AMD.

bit tech ... rofl.

Note: the AMD chips were tested in an ATX motherboard, while the Intel LGA1155 chips were tested in a micro-ATX board. This difference can account for up to 20W,

and their overclock ... rofl ... lets see how far we can push power draw, CRANK THE VOLTAGES, ALL OF THEM. lets burn this thing up.

AMDs CPUs have almost always used more voltage than Intel. But still the Intel CPU was clocked higher than the FX and it was probably pushing around 1.40-1.45v which is pretty high for Intels 32nm process.

Still most sites noted the same inefficient design in the FX series when overclocking.

I think you guys are not understanding each others point and are arguing your own points over and over again, lol.

Intel side: AMD does poorly -> benchmarks to prove it -> fact put and accepted (at least by me, lol).
AMD side: Intel does not give Devs/OEMs room to make AMD look better and cripple them in very dubious ways altering fair and square competition -> shows several FTC proof and explain the benchmarks of why AMD is behind by a wide margin -> does not accept (or is it accepted but still pushing the other argument).

That makes this thread go around in circles till a new tid bit of information about Trinity or PD comes around and we start it over again, lol.

Anyway, I still consider BD to be a very expensive side grade for me, but still recognize the boldness of the new design. I still think AMD screwed up trusting Intel in the FMA4 and not PUSHING early development of software that actually supported BD out of the door. Hell, not even their own bloody compiler supported BD fully at launch (not optimized as it should been). That's just dumb.

Cheers!

It does seem to go around and around, doesn't it?

And it is bold. But bold doesn't always mean better. NetBurst was a bold change from the Coppermine Pentium II and much like Bulldozer, it failed to impress until a few generations down the line. Still with even that it was outperformed by Athlon 64 which used less power and performed better at a lower clock speed.

Of course it seems that should stand as impossible by Intel, only AMD can do it.

I say BD is fine, its just nothing amazing to get super hyped about. PD may be or it may not be. Hard to say. Still it wont be up against IB very long. It will meet Haswell and if Intel can get their 22nm in line by then (I am sure they will) and the changes they make are good enough, PD may just be Phenom II. It catches up only to get left behind agai.n
 
moving my Z68 mobo over to replace the AMD (AM3+) unit.

going to grab a Z77 (ATX) and CPU; then move the Z77 mobo and pair it with the 2500K in my 'gamer' unit.
then the Z68 (mATX) will then be my 'daily unit' with the new CPU.

bye AMD and will be 'greener' with the utilities..

That's what I did minus the AMD stuff :lol:
 
I doubt we will see AMD go down the crapper like that. too much to lose. They just need a better strategy in how they approach innovation. Like, BD was a great move...just a bad execution. I also blame the marketing 'guru's' and their circus performance associated with BD. Fire them, I would if I were on the BOD! Hopefully this somewhat new CEO will steer the S.S.AMD in the right direction and hire the best of the best to get the job done. That is what they need...is new ideas that work. Fresh meat. Dang...I'm hungry!
 
@palladin9479:
Intel compiler compiles it and creates multiple ways to execute that program, namely in the calculation part. The selection of these paths will depend on the CPU's ProductID string and only if the VenderID strong is GenuineIntel. If the VenderID is not GenuineIntel then the dispatcher will automatically use the lowest compiled code path regardless of the capabilities of the CPU. Prior to 2010 this path was i386/8087, after 2010 this path was SSE2 / MMX. Those flags you pass to the compiler can define to what level the lowest compiled path use's, thus you can "enable" it on non-Intel CPU's by making the non-Intel code path compiled with SSE3/4 (AVX currently doesn't work). The side effect is that because your lowest code path is now set to a higher level, the program will not run on CPU's that don't have those instructions defined in the lowest code path. Software makes, especially game makers, do not like to do this as it limits their market size.
post 2010, is this an illegal move from intel? seems to me that intel found a way around the court order with this:
"The side effect is that because your lowest code path is now set to a higher level, the program will not run on CPU's that don't have those instructions defined in the lowest code path."
if this is illegal, what is amd doing to remedy that and why doesn't amd or anyone from amd bring that up? i only see these brought up during this kind of debate and in news articles that favor amd.


Depends on the exact definition of "Artificial performance impairments" AMD / Intel want to use. Prior the ICC wouldn't run SSE / MMX code on non-Intel CPU's, period end of story. They marketed their compiler to software companies on the premise that would compile diverse auto-detecting code that would self-optimize, aka the dispatcher would detect the CPU and run the appropriate code. This is kind of a holy-grail for software companies, they can get the benefits of newer CPU technology (SIMD) without needing to compile multiple libraries and binary's. What Intel didn't tell them was that the ICC would create code that completely disabled x86 extensions on anything not Intel, it would self-optimize but only on Intel CPU's, anything else was forced to 80386 level code. That is what the FTC got onto them about, and thus Intel was forced to not only offer monetary compensation to any customer who needed to recompile their code but also to adjust their marketing material and offer a method to force optimizations on their product.

Honestly it's not much of an issue anymore now that everyone knows about it. The difference from SSE2 -> SSE4 is not nearly as large as the difference from i386 -> SSE2, so while AMD / VIA CPU's would technically be impaired, it wouldn't be some great margin unless the developer purposely did something shady. There is an issue that if you specific your compile target as i586 or less, then the code generated will not run SSE2 on non-Intel CPUs (lowest code path is now i586 and that doesn't include SSE instructions) while Intel CPU's will still get the benefit of SSE4.2. That is more the fault of the developer for not anticipating unintentional consequences.

I just really don't like historical revisionism, altering facts about the past in an attempt to change present day perceptions. All companies act evil at some point in time, they don't need us trying to sugarcoat or obfuscate their actions.
 
@palladin9479: i too am against historical revisionism, sugarcoating doesn't make anything go away. the facts are right there for anyone to look into them.
i didn't really understand what amd is doing now i.e. 2011 and later. i assume that phenom and phenom ii cpus had to face the compiler sabotaging back then (2009, 2010~). i don't know how much time it takes to rework the compiler to comply with the court order intel was given, but would current softwares (2011 and later) still run slower on amd cpus because of the sabotaging? i mean, does intel still mess with the compiler (other than restricting using only one compilation level and later when the "/QxO" is used) so that current software run slower on amd's current cpus e.g. phenom ii, llano, bulldozer, brazos etc. i am only focusing on the compiler aspect disregarding architectural and other aspects.
imo, now, if the devs don't setup their software to run optimally on all cpus, it's their fault. i still don't rule out intel secretly/passively/passive-aggressively pursuading devs to leave the compiler switch as it is even now... but that's a conspiracy theory without evidence.
 
BD is performing badly because there are serious issues with it's design. You can't rely on clock speed to make your CPU faster, there is a very real physics problem once you go past 5Ghz. Electron tunneling, superposition and QM really start messing with transistors, heck even minute cosmic radiation will start to mess with you. Those aren't so bad a problem at very low levels of ambient temperature (phase change or straight up super conductors), but for now their big issues with "go faster" mentality.

PII was good for it's time but it's old and needs updating.

My personal thoughts are that circa-steamroller we'll see PII integer cores being bonded to a GPU SIMD array and have their own dedicated L2 cache. Sharing the front end decode isn't a bad idea, just don't skimp on the pipelines otherwise it'll create issues.
 
311 watts vs 583watts.....
Which side of the funny farm did you pull your above post from?

Bit tech are a UK reviewer, and feature in PC magazines, such as PC Format and Custom PC.

Well reputable 😉
The fact that bit tech's numbers were well over 100Watts above and beyond any other test out there.

http://techreport.com/articles.x/21848/2 - 348W at an insane 1.55V they didn't show intel's oc power draw.
http://hexus.net/tech/reviews/cpu/32155-amd-fx-8150-47ghz-does-stand-tall/?page=5 - 325W ??? V
http://www.rage3d.com/reviews/cpu/amd_fx_8150/index.php?p=7 415 and 446 (2 diff ocs), managed to reach 545 while pushing the video card with it.
http://www.kitguru.net/components/cpu/zardon/power-consumption-fx-8150-v-i5-2500k-v-i7-2600k/ 406W

Nothing even close to 580W other than the one running the video card with the oc cpu. How reliable are they?

http://www.overclockersclub.com/reviews/amd_fx8150/4.htm
At 4940MHz (229x21.5 @ 1.435V), I was able to navigate through the OS and game for hours, though the system was not fully stable due to the thermal load. To get Prime95 stable, I had to drop the clock speeds down to 4640MHz (232x20 @1.325V)[/qote]

I know their numbers are accurate, thats nearly where I am at with my 8120. you don't need 1.5v to hit 4.7+ ghz. unfortunately they didn't show power draws, would love to see 4.65ghz at 1.325V instead of 1.45+V

Its easy to just crank the voltage to show artificially inflated numbers.
 
Phenom II K8/K10 vs Modular

I am in the same boat as Yuka and I am assuming Palladin, in that the newer architecture is rather promising if but for the implemenation of the first attempt, Bulldozer is not completely without positives and my experiences with things, something new and radical doesn't always work out at the first throw of the dice, I don't think many are under any illusions that AMD's architecture is in need of fine tuning and revision, but the idea of hetrogeneous computing is a rather interesting arch which AMD are persuing. I also reject the arguement that AMD should have followed on with the Phenom II architecture which was by and large already 10 years old and inefficient.

10% + 12% + 3%.

I have seen this come up with a degree of regularity now and it is somewhat distressing, if first gen was deemed around 10% faster than PhenomII/Zambezi and SB is another 12% on top followed by 3% by IB, that makes a cumulative difference of 20-30% factoring in margin of error. In non of the benches does a contemporary Intel Chip at a similar price point (even the high end chips) beat the Zambezi by anything close to 25%, if we take BF3 and or Skyrim as the defacto big titles if SB hypothetically gets 85FPS that would mean that the Zambezi would get around 60FPS, when most would show there is in a overwhelming majority of cases as little as 5-10FPS, which is rather that SB is about 10% faster than Zambezi.

It also is rather confusing then when people claim that a 15% improvement on Zambezi will put it around first gen performance. In personal experiences with first gen i3 and i5, most later core2 chips matched up well there, and certainly Phenom II was aptly matched with first gen up until the i7 920/950 region. Experiences with a FX 8120 seem to be more responsive in general computing and notably memory performance over Phenom II, in reality I would be more inclined to believe that Zambezi sits somewhere between the Phenom II / first Gen core I and SB, probably closer to the former ie: the 10% differential.

Thanks To

I would like to say I have enjoyed the posts from Noob, Palladin, Yuka (and a few others for which I apologise I forgot the user handle, but you will know who you are.) in regard to the compilers. It is the reason why this forum is a great place to be, you learn from those that know more.
 
I have seen this come up with a degree of regularity now and it is somewhat distressing, if first gen was deemed around 10% faster than PhenomII/Zambezi and SB is another 12% on top followed by 3% by IB, that makes a cumulative difference of 20-30% factoring in margin of error. In non of the benches does a contemporary Intel Chip at a similar price point (even the high end chips) beat the Zambezi by anything close to 25%, if we take BF3 and or Skyrim as the defacto big titles if SB hypothetically gets 85FPS that would mean that the Zambezi would get around 60FPS, when most would show there is in a overwhelming majority of cases as little as 5-10FPS, which is rather that SB is about 10% faster than Zambezi.

Again, beware of GPU limitations when using games to benchmark CPU performance figures.
 
Well most synthetics relied on are from Toms which by and large are highly reputable, but then the bells go off and the synthetics arguement comes up again. Synthetics is software, it is at the behest of the writers biases or opinion on how something can be numerically quantified so as to represent performance, sure its a watered down definition but the point is that it is human engineered and subject to the interests of that person.

That said when I look for a chip, I will pay modest attention to synthetics, I will ask around and if lucky get to try out first hand, either that or I will be given a chip and asked to test it, strain it and give a opinion/review on its performance, either way to much hay is being made over exactly where AMD's performance is relative to Intel, granted FX is not what anyone expected as for general peformance it seems right. AMD's net capital is like a tength of Intels R&D alone, at the end money dictates a lot of things. Where money is the issue, this is where you can factor in the intel dirty tactics and snowballing and the profound effects it has had on AMD as competitor.
 
my reasons are more to it than just that.
I have a 2500K unit already.

but if you want me to break down all the pros and cons then I will be happy to,
basically AMD is behind in performance AND power consumption.

We are to our own, why do you feel like you need to justify your decision, you buy what you want to buy not for any affiliation or alliegences.





 
because you dare to say it's an unwise decision.
and I'm trying to show you the error in your thinking.
😛

Well I guess I will say that I get given them to test and bin at my own behest so that said, on synthetics the intel chips are great, but if one is a minimalist then they are content with that, I just get the same feeling when I run a QX9650, 2500K and 3570K, it is quintessentially the same feeling with no new flavour, its just Intel giving the consumer the bear minimum and reaping the profits.
 
Again, you fail to differentiate between the optimization and code gen stages of compilation. I read the settlement as meaning a failure to optimize.

Really simple to test though: take a piece of code, manually insert an SSE instruction, and compile with all optimizations disabled. Then compile with optmizations enabled.

If what you argue is true, then with no optimizations, the only code path would be the SSE instruction, and the app would instantly crash within the CPU dispatcher.

If I recall correctly Palladin won the compiler debate last time ... no more discussion in that regard thanks.

We are sick to death of that ongoing debate ...

:)
 
again I'm not being too serious, more in a playfully jousting manner...
and to be honest bro, I'm tired of AMD.
maybe when Piledriver shows some of being a worthy upgrade from Deneb C3 I'll be back..
(I gotta sell mine first, or would like to anyways)

in the words of my man 'yuka'...
cheers.

Don't worry, it is not like I don't enjoy the discussions/debates/maybe odd arguement and disagreement, like I said you have your reasons, and they don't need to be justified, fortunately I understand why.
 
rekon-uk - is now playing with his modded SLi GTX 480's that can out perform SLi GTX 680's...
or so he thinks... 😛

GTX 480 is Fermi firestarters, someone is trying to convince me that the GTX 480 is the best value for money card, claims a minor clock bump gets you GTX 580 performance, if it matches a 6970 or 570 I will be shocked.
 
Status
Not open for further replies.