AMD Piledriver rumours ... and expert conjecture

[Reynod]

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 ...

 

[viridiancrystal]

+1

and as for the AMD vs Intel debate.
just be like me and run both.....
😉

Deneb C3 that is... :lol:
Bulldozer =

You cannot stop me! Bulldozer seems to be my build option, but that's my loss, right?

I want to say to all of you that are in the Intel vs AMD "Intel bribes software devs, so AMD runs badly" that you should probably take a moment to understand economics. If 75% of your costumers run Intel, then you first focus would be to optimize better for Intel. AMD comes second, and in the cases that changing coding to make software run better AMD negatively effects running on Intel, then it doesn't happen, because you focus on the majority. That's called democracy.
Just to say it again, I'll say it again: By the time BD becomes the bottleneck, games are playable anyway, so why does it matter that much? You are blowing things way out of proportion.


All this because AMD is all hushed about Trinity & Piledriver.

 

[viridiancrystal]

http://i1130.photobucket.com/albums/m529/malmental/Skype-emoticons-46-punch.gif

I'll be putting on a show for all 8 cores , so it will be used how it works best. 😀
http://www.youtube.com/watch?v=Ss0E__ltHg4&list=PLCE3009B988A15B5A&index=51&feature=plpp_video
^bonus! calming music, yay!

Also, like i mentioned before, BD-->PD seems like a better upgrade than SB-->IB

 

[fazers_on_stun]

I fail to see why this compiler issue is such a big deal. Am I missing some key component that puts Intel firmly in the wrong?

Not really. This issue periodically gets dragged out by the AMD fanbois. IIRC Intel's compiler has <10% marketshare anyway, so not like it is pervasive throughout the industry.

Intel spends $$ working with the software devs while AMD does not, to any noticeable extent. Why should Intel provide free assistance to the competition? If it's as simple as faking the CPUID, why doesn't AMD just provide patches to update the most popular software, like they do with their frequent GPU driver updates? As I mentioned previously, perhaps they enjoy playing the victim a bit too much??

 

[viridiancrystal]

still behind the performance curve however.
and it's SB to Haswell, people who waited out for Ivy over Sandy really should have went ahead with Sandy anyways.

it's gonna take a performing Piledriver to make me come off my Deneb C3, I keep saying Deneb C3 and not 965BE
because I have the opportunity to get the 980BE for cheap...

P.S.
IT'S NOT 8 CORES, IT'S 4..!!!!! :lol:

Behind, absolutely. The idea of me going with BD is because I'll be taking advantage of its 8 [strike]cores[/strike] threads. (I would call them cores, because they scale over 50%, which essentially makes them more of a core than not.) I still think Intel wont push 6 core mainstream processors on Haswell, so I'd still be using 4, which I think would be limiting me.

 

[Cazalan]

I only used the money to show how easy it is for Intel do do something AMD can't afford. That will never change, hence why AMD is "no longer competing with Intel"

Well they could compete better if they hadn't done the module stuff, or at least waited until the design was more optimized. Now that Microsoft is supporting Intel and ARM, AMD is a 3rd rung afterthought for the developers. I'm surprised they actually got a patch for BD before Windows 8. AMD has to be smarter with their design changes.

That said the BD issues look solvable. I think they should take off the veil of secrecy and start shipping 10x more engineering samples to review sites. It's amazing how much free testing they could get out of the community.

Imagine where game developers would be if they didn't do the open betas.

 

[noob2222]

Not really. This issue periodically gets dragged out by the AMD fanbois. IIRC Intel's compiler has <10% marketshare anyway, so not like it is pervasive throughout the industry.

Intel spends $$ working with the software devs while AMD does not, to any noticeable extent. Why should Intel provide free assistance to the competition? If it's as simple as faking the CPUID, why doesn't AMD just provide patches to update the most popular software, like they do with their frequent GPU driver updates? As I mentioned previously, perhaps they enjoy playing the victim a bit too much??

Well also consider this. The last cross liscence agreement amd posted, they had to pay royalties for every cpu sold.

If you think about it, intel is taking amd's money and giving it to devs to hinder performance on amd chips. Not literally paying directly but giving free "help".

Why hasn't another cpu company even tried to come? If they look at what happens to amd, they don't even want to.

Don't get stuck on the fact that I said hinder. Optomize for one essentially hinders the opposite.

 

[triny]

You cannot stop me! Bulldozer seems to be my build option, but that's my loss, right?

I want to say to all of you that are in the Intel vs AMD "Intel bribes software devs, so AMD runs badly" that you should probably take a moment to understand economics. If 75% of your costumers run Intel, then you first focus would be to optimize better for Intel. AMD comes second, and in the cases that changing coding to make software run better AMD negatively effects running on Intel, then it doesn't happen, because you focus on the majority. That's called democracy.
Just to say it again, I'll say it again: By the time BD becomes the bottleneck, games are playable anyway, so why does it matter that much? You are blowing things way out of proportion.


All this because AMD is all hushed about Trinity & Piledriver.

It's your money buy what you want ,don't listen to those guys.you can upgrade to steamroller down the line.

 

[Cazalan]

AMD Pat Patla jumps ship..."

http://www.fudzilla.com/home/item/26544-amd-corp-vp-and-gm-of-server-products-leaves

Not too surprising.

I'm guessing he objected to the SeaMicro purchase. That 350 mil could have been spent ramping PileDriver/Steamroller quicker.

Budget constraints stuck them on Bulldozer B2 stepping, and yet 350mil can get shot out to buy a company doing fancy stuff with PCIe.

They should be doing more IP purchases like the RCM. Immediate cost and effect. Whoever spearheaded that purchase should get a big raise.

 

[triny]

steamroller will be a different socket..
I see your still too busy talking before you have knowledge of or you should make yourself more clear.
but in this case you were just talking again.
as usual.

no i don't talk to myself when I write

 

[triny]

Originally had nothing to do with Intel optimizing for their own CPU's, they released it without telling anyone that it crippled code running on AMD CPUs. We only found out because VIA CPU's allow you to change the CPUID. When changed to "Intel" suddenly benchmarks jumped higher, no hardware changing or OS reconfiguration requires. Just had to tell programs you were running an Intel CPU and they performed better. Tell them you were running an AMD CPU and they crippled themselves. One of the stipulations of the AMD vs Intel lawsuit was that Intel had to stop the practice of coding software to cripple AMD CPU's. Before it would disable SIMD on AMD CPU's entirely, now it just limits them to SSE2.

I happen to have a Via Nano, I've actually done testing on it and yes you can still get more performance by telling programs that your running Intel. AMD should make their CPUID's changeable in software so we can get a better handle on it. It's also why I recommend everyone to use gcc or MS's compiler instead of Intels.

And before anyone says something, it's not "optimizing" at all, it does a simple check for the CPUID's VenderID and if it detects "AMD" it puts the code down a limited core path, end of story. If it's "Intel" then it'll do a Flag check to determine capabilities and use that to determine what instructions to execute. This is done on the clients system during run time not compile time.

From Agner,

http://www.agner.org/optimize/blog/read.php?i=49

Program to get reimbursed for having to recompile your programs with non-intel compilers.

http://www.compilerreimbursementprogram.com/

Program to fake your CPUID
http://www.agner.org/optimize/#cpuidfake

Intel's compiler is still limiting anything non-Intel to SSE2, at least now their forced to tell their clients that their compiler does that.

Gangster activity from Intel is well known ,once a gangster always a gangster.

 

[Cazalan]

I think AMD's server marketshare is going to drop well below the current 5% with the new E5 SB-based Xeons, and that is the most profitable segment by far. It will be interesting to compare Intel's and AMD's Q1 earnings reports due out in about 3 weeks...

Agreed on the server end. If the Sony rumors are true then the server loss could be offset by the PS4 gain.

The APUs are still selling well, but Intel did start shipping 32nm Atoms (2.13Ghz 2C/4T) which compete with the E-350 chips (1.6Ghz 2C/2T). Prior to this AMD had the upper hand competing against 45nm chips.

The rough part will be Q2 and later with Ivy and more 32nm Atoms. There's no direct competitor for Trinity (if leaks are true) but the other APUs could get knocked off the radar (C50,E-350,E-450). This would leave Trinity as the only star for AMD this year.

 

[Cazalan]

If you think about it, intel is taking amd's money and giving it to devs to hinder performance on amd chips. Not literally paying directly but giving free "help".

Anyone making a product will have a somewhat deeper relationship with a few customers. When it's a 100bil company they have more of those relationships.

And it's not just a 1 way street. Yes Intel will help optimize code or the SDK's for the company but they also get more hands on experience with the bottlenecks of the architecture,.

Making a product is one thing. Releasing that into the wild and having thousands of software engineers do things you never thought of is quite another.

Keep in mind that cost is probably just a fraction of a fraction of how much Intel spends on marketing.

 

[jimmysmitty]

What do you think would happen if intel convinced every software developer to use their compiler? It would be no different than bribing vendors not to sell AMD products.

How is that line of thinking wrong?

I only used the money to show how easy it is for Intel do do something AMD can't afford. That will never change, hence why AMD is "no longer competing with Intel"

There is nothign wrong with it. Intel optimizes for their CPUs, just like AMD optimizes for their CPUs/GPUs.

A lot of games are optimized for nVidia, a lot of the very popular ones, and they run better on nVidia. Thats just how it goes.

Fanbase tends to help drive this loyalty. as i stated, if a dev is building a game for 95% to 5% marketshare ... its obvious who they will program for. AMD needs to be able to do their part as well.

The problem is still the review sites that cherry pick what games they use in their articles.

Most review sites pick the most popular games, unless you didn't notice, to benchmark the CPUs/GPUs, not based on what CPU/GPU they are backed by.

Thats why the games change over time, goign with popularity and of course feature set (DX11 etc)

 

[viridiancrystal]

logical reasoning...
one question for you and I'm not trying to persuade you to Intel, what you do I have confidence will be a good unit from you..
but with that being said, Intel and 4 stronger cores plus hyper-threading won't do.?
like the i7-2600K or is that just out of the price range.?

The 2600k did seem like a worthy option, but it, as you said, put me outside of my budget. I have about $1200 to build a system from scratch completely. Compiling the costs of windows and a monitor with all the necessary parts, the 2600k would be a bit much. I'm set on a 8120 + Radeon 7850. I've made all the parts solid, don't want to sell anything short. I'm going to be doing video editing/exporting, along with small server hosting.
The 2500k could find its way into my budget, but i have a gut feeling that having the extra cores would be of greater benefit.

The advice is well taken, there are not many people who own Intel and AMD systems, so insight from those who do is nice to have.

I'm not sure if I would be better off going i3 for now and upgrading to a better Ivy bridge down the road, or staying AMD and upgrading to PD. I've had no experience with high end systems from either, so I cant say the benefits of either. I know that Intel has better IPC, and scores better in benchmarks almost every time, but I read reviews of people saying how much better BD does multitasking and multithreading than SB (which it very well may). I felt that I fit into that place where BD does really well, but almost everyone tells me to go Intel anyway.

I'm not sure which team to choose.

 

[r0ck3tm@n]

The 2600k did seem like a worthy option, but it, as you said, put me outside of my budget. I have about $1200 to build a system from scratch completely. Compiling the costs of windows and a monitor with all the necessary parts, the 2600k would be a bit much. I'm set on a 8120 + Radeon 7850. I've made all the parts solid, don't want to sell anything short. I'm going to be doing video editing/exporting, along with small server hosting.
The 2500k could find its way into my budget, but i have a gut feeling that having the extra cores would be of greater benefit.

The advice is well taken, there are not many people who own Intel and AMD systems, so insight from those who do is nice to have.

I'm not sure if I would be better off going i3 for now and upgrading to a better Ivy bridge down the road, or staying AMD and upgrading to PD. I've had no experience with high end systems from either, so I cant say the benefits of either. I know that Intel has better IPC, and scores better in benchmarks almost every time, but I read reviews of people saying how much better BD does multitasking and multithreading than SB (which it very well may). I felt that I fit into that place where BD does really well, but almost everyone tells me to go Intel anyway.

I'm not sure which team to choose.

I have the i7-2600K so you know I think a lot of it. However, and I just learned this, hyperthreading is helpful when one of the two threads running within one core is waiting (blocked) on a resource outside the core. When both threads are contending for resources within the core, hyperthreading then will lead to a (slight) loss of processing power. So it's not always going to help although I generally like it, myself.

Why not wait for Piledriver? I honestly think, after reading that guy's blog that Palladin referred to, that AMD could fix the issues that Bulldozer has shown. Then I will be happy.

 

[Chad Boga]

It's your money buy what you want ,don't listen to those guys.you can upgrade to steamroller down the line.

How is he going to upgrade any easy to SteamLoller than he would an Intel CPU, when he will have to change motherboards either way?

 

[Chad Boga]

What do you think would happen if intel convinced every software developer to use their compiler? It would be no different than bribing vendors not to sell AMD products.

How is that line of thinking wrong?

That line of "thinking" is about as meritorious as asking what would happen if Intel kidnapped the family of every games developer. It just ain't going to happen.

That will never change, hence why AMD is "no longer competing with Intel"

Amd is no longer competing with Intel because they are about to be even more disadvantaged on manufacturing process, than has been the case in the last 15 years.

Also they stuffed up the design of what will be their base for the next 4 or more years, so they are trying to make lemonade out of the lemons they got, and present it as though the course their charting is their preferred one, not one that was forced upon them.

But do you really think that TH should ignore 3 of the biggest titles because they are optmized for some set of hardware?

Yes he does. Impressive huh? TH reviews aren't supposed to be about doing reviews for consumers, they are meant to be apologists for those lagging behind.

Do you really think the reason amd sucks at game benchmarks is solely because of the cpu and not the software code? That's the conclusion everyone uses, but what is the more realistic reason?

The more realistic reason is that AMD sucks at game benchmarks mainly because of the CPU and the software code is just the cherry on top.

Amd cpus rune perfectly fine against sb in any game that's not "optomized for intel" so the problem must be _________.

No, AMD cpus run fine in a massively GPU limited situation.

Thanks to Consolitis, game engines are trying to get away with as weak a CPU as they can. AMD benefits enormously from this.

Once the game engines for next gen consoles kick off, AMD's CPU shortcomings will be exposed even more for a couple of years.

But the conclusion is "AMD sucks"

Pretty much as far as gaming goes.

Sure you can get away with it on most titles with little problem, but you are going to be forced to update quicker down the track for next gen games, if you have a BD today instead of an SB or IB.

Why hasn't another cpu company even tried to come? If they look at what happens to amd, they don't even want to.

Because competing for the riches of the X86 money pit, has enormous financial and technological barriers to entry, with two well entrenched competitors.

IBM use to sell X86 CPU's before they realised they couldn't keep up with Intel's relentless process & design march and got out.

Why would anyone after seeing IBM quit like that, think they will do better?

 

[sonoran]

I'm going to be doing video editing/exporting, along with small server hosting.
...
I'm not sure which team to choose.

Despite being an Intel employee, my take is it's your money - buy the one you want. There's just one more factor I would suggest you take a look at, if you haven't already. Can the software you're using utilize QuickSync? I don't think it affects video editing speed(?), but it can have a significant impact on video encode/decode performance.

 

[viridiancrystal]

Because competing for the riches of the X86 money pit, has enormous financial and technological barriers to entry, with two well entrenched competitors.

IBM use to sell X86 CPU's before they realised they couldn't keep up with Intel's relentless process & design march and got out.

Why would anyone after seeing IBM quit like that, think they will do better?

I used to think that IBM just wasn't really good in the tech department, but looking at all the news articles that crops up about these amazing things they've done, I can tell I was misinformed. The fact that IBM felt they could not keep up shows how dominant Intel has been, basically since the dawn of cpus.

AMD made some noise around the sleeping giant, and showed that it wasn't the best if it did nothing. We all know how that [strike]has ended[/strike] is ending.

People pushing ARM to enter desktop space. For them, that is the worst idea possible. The fact that Intel is going to their home, and using so much of its might to get there, shows that they have to win in what they have now, not compete in a place they've never been.
Funny thing is, if you believe the benchmarks, Intel already has them beat.

Through and through, AMD has done an amazing job of keeping up with what they have, graphics and cpus, and they should probably get more praise for it. A company thats 1/25th the size of its only cpu competitor, 1/2 the size of its only gpu competitor. Remember that AMD encompasses both its cpu & gpu sectors. They do a great job givin the circumstances.

AMD is 1/3 the size of ARM, for pete's sake.

 

[viridiancrystal]

Despite being an Intel employee, my take is it's your money - buy the one you want. There's just one more factor I would suggest you take a look at, if you haven't already. Can the software you're using utilize QuickSync? I don't think it affects video editing speed(?), but it can have a significant impact on video encode/decode performance.

I'm going to have discrete graphics(gaming 😀), so that would disable QuickSync anyway.

 

[viridiancrystal]

I have the i7-2600K so you know I think a lot of it. However, and I just learned this, hyperthreading is helpful when one of the two threads running within one core is waiting (blocked) on a resource outside the core. When both threads are contending for resources within the core, hyperthreading then will lead to a (slight) loss of processing power. So it's not always going to help although I generally like it, myself.

Why not wait for Piledriver? I honestly think, after reading that guy's blog that Palladin referred to, that AMD could fix the issues that Bulldozer has shown. Then I will be happy.

I agree. AMD said, no joking around, that they messed up. They haven't mentioned once the performance of Bulldozer, they have only said they are going to make it better.

 

[fazers_on_stun]

Well also consider this. The last cross liscence agreement amd posted, they had to pay royalties for every cpu sold.

IIRC the last cross-licensing agreement AMD and Intel signed (where Intel also settled the lawsuit) let each company use the other's IP for free. IOW, no fees paid to Intel.

If you think about it, intel is taking amd's money and giving it to devs to hinder performance on amd chips. Not literally paying directly but giving free "help".

Why hasn't another cpu company even tried to come? If they look at what happens to amd, they don't even want to.

Don't get stuck on the fact that I said hinder. Optomize for one essentially hinders the opposite.

AMD also had significantly more marketshare before they bought ATI - they might have used some of the $5.4Bn on (1) better advertising (2) more assistance incl. financial to the OEMs and software devs, and (3) more R&D. In fact, if they had waited 6 months they could have picked up ATI at around half the price from what I've read, and thus been able to do all the above plus acquire ATI.

But Hector Ruiiz probably did not foresee how much improvement Intel made with Conroe vs. Netburst, and thus did not foresee the staggering losses AMD would suffer as marketshare dropped and the heavy financial burden was not offset by continued by being able to maintain those $800+ prices AMD was getting for K8 previously. He undertook a gamble that was far riskier than he thought and it undermined the health of AMD. Of course he also trashed Motorola before coming to AMD so why the Board hired him, I dunno..

 

[fazers_on_stun]

Not too surprising.

I'm guessing he objected to the SeaMicro purchase. That 350 mil could have been spent ramping PileDriver/Steamroller quicker.

Budget constraints stuck them on Bulldozer B2 stepping, and yet 350mil can get shot out to buy a company doing fancy stuff with PCIe.

They should be doing more IP purchases like the RCM. Immediate cost and effect. Whoever spearheaded that purchase should get a big raise.

Since he was GM of the server division according to the article, I'm betting that is a harbinger of not-too-good news in the next quarterly earnings report..

 

[fazers_on_stun]

How is he going to upgrade any easy to SteamLoller than he would an Intel CPU, when he will have to change motherboards either way?

Perhaps Triny means "easy" as in Triny will buy him a new mobo and install it for free?? 😀

 

[viridiancrystal]

Perhaps Triny means "easy" as in Triny will buy him a new mobo and install it for free?? 😀

I'll take it! 😀