AMD's PileDriver Possibly Cancelled!! (Rumor)

[bononos]

Neither did AMD for example when it directly made a ripoff copying Intels CPUs.

Umm....you might want to read through that one again, it's not even clear what you were trying to say. If you're talking about the 286,386 and 486 "clones" they were actually licensed designs, much like (but definitely not identical to) how ARM currently runs. ......

IIRC AMD put out its own 80386 which used Intel's 80386 microcode after Intel suddenly cancelled its licensing agreement with AMD. AMD got sued but won the case and then lost again in the appeal. So AMD had to redesign its own 80486 under clean room conditions. I can't remember if my DX4 was the 'dirty' one or the 'clean' one.

I remembered reading something about Intel adding certain new instructions in SSE4 which would frustrate AMD in attempting to add the same to its own line of cpus, some technical hurdle of some sort, I don't understand the details.

 

[tommo123]

100% agree, no matter which side you're on, having Intel and AMD battling it out, and a strong AMD is good for EVERYONE!

i agree but it wouldn't be as bad as it would have been a decade ago. intel have to compete now against

1 - arm
2 - themselves.

for most people, a core2duo e6600, 4GB ram would do for vast majority of things they'd do. replace a gpu (pay a kid in the neighbourhood to do it for you) and you could gave too (hell, that would only need be done every 3-4 years really)

intels worst competitor is themselves. i have a Q6600 desktop and still does everything nice (with an SSD). unless i video encode etc i won't get that much difference in oomph out of it. unless the PC dies entirely then most people don't need to replace it IMO.

i will though, next year

 

[exar333]

There are other ways Intel is ripping you off, you just haven't been paying attention. Instead of raising prices, they are giving you less die space, (i.e., less performance than what should have been).

When Intel thought AMD was still competitive, Nehalem/Lynnfield had ~ 300mm^2 die sizes. That was 2008-2009 before they knew how terrible Bulldozer was. Since then AMD has been falling behind and Intel has been selling us less performance than it normally would have. When Pentium 4 added HT, and Athlon 64 slaughtered it, Intel was giving away HT for "free". There was no $100 premium for Pentium 4 HT chips during that era. Intel couldn't possibly be able to afford that. Now they charge $100 for basically just that and 100mhz more in the 3770K. :thumbsdown:

Also, we have had Q6600 quad-core for $300 or so around August 2007 (that's when I got mine for $300 CDN). Fast forward 6 years later and Intel will continue to sell Haswell for $300 as yet another quad-core. Now you may say that most programs don't benefit from quad-cores so why would Intel make Haswell a hexa-core for $300? Well they do sell 3930-3960 for $500-1000. There you go. If AMD was competitive, there is no way Intel would have been able to pull that off. They'd have to sell a hexa-core for $300. So we are already suffering believe it or not. Based on that graph above, Intel had no problem selling me a 300mm^2 Core i7 860 for $330. So where is my $330 Hexa-core 300mm^2 Haswell? I bet it'll be again $500+.....



Not trying to be a salesmen, but there is a legitimate way to help AMD stay in business. Instead of wasting $ on their CPUs, buy their GPUs instead. Unlike throwing $ at their slow CPUs you don't really give up GPU performance, make $ bitcoin mining on the side, and boy AMD GPUs in double precision distributed computing projects make NV cards look like prehistoric dinosaurs. I looked at your DC production the other day (164,919 BOINC points per week), it's weak! A single HD6950 would make more than that in MilkyWay@Home in 1 day! For you it could be almost a no brainer unless you must stay attached to your specific DC projects (Folding@H, Seti@H). Otherwise, you'd get a card that pays for itself and when mining fails, you still have a card that crushes NV cards in DC points. You could then still use the CPU for some of those DC project that don't benefit much from GPU speed in the first place.

Your comment assumes too much. (1) The market doesn't change (2) costs stay the same.

Both are incorrect. There is more competition from other fronts (ARM, etc) and the ASP for an entire computer has been going down quite a bit. Intel needs to cut costs to keep prices down so people will still buy their products. Node improvements are getting ridiculous, so it will become even more important to keep die sizes down.

I don't think being efficient with space = 'ripping anyone off'

I do wish more options were available, sure, but the place on the curve where we are today is part of the progression of die space that Intel has had for a while. GPU space is growing quite a bit, and probably will continue for the next few products. Hopefully more cores to come as well, but why put more than 4 on there when 6 or 8 of your competitors are still slower?

 

[Lonyo]

Your comment assumes too much. (1) The market doesn't change (2) costs stay the same.

Both are incorrect. There is more competition from other fronts (ARM, etc) and the ASP for an entire computer has been going down quite a bit. Intel needs to cut costs to keep prices down so people will still buy their products. Node improvements are getting ridiculous, so it will become even more important to keep die sizes down.

I don't think being efficient with space = 'ripping anyone off'

I do wish more options were available, sure, but the place on the curve where we are today is part of the progression of die space that Intel has had for a while. GPU space is growing quite a bit, and probably will continue for the next few products. Hopefully more cores to come as well, but why put more than 4 on there when 6 or 8 of your competitors are still slower?

The easier and more accurate way to work out whether consumers are being "ripped off" is to look at margins. If Intel's margins have massively increase, they are "ripping us off". If they are consistent, then not so much.
(That ignores things like product mix etc, but is a better starting point than die size).

 

[Idontcare]

IIRC AMD put out its own 80386 which used Intel's 80386 microcode after Intel suddenly cancelled its licensing agreement with AMD. AMD got sued but won the case and then lost again in the appeal. So AMD had to redesign its own 80486 under clean room conditions. I can't remember if my DX4 was the 'dirty' one or the 'clean' one.

I remembered reading something about Intel adding certain new instructions in SSE4 which would frustrate AMD in attempting to add the same to its own line of cpus, some technical hurdle of some sort, I don't understand the details.

Those were indeed interesting days in the x86 world. TI did similar to Cyrix, who in turn got screwed over by IBM.

At that time the dram market was in similar cloak-and-dagger corporate maneuvering as well.

 

[Idontcare]

The easier and more accurate way to work out whether consumers are being "ripped off" is to look at margins. If Intel's margins have massively increase, they are "ripping us off". If they are consistent, then not so much.
(That ignores things like product mix etc, but is a better starting point than die size).

< thread needs a car analogy >

The rational here can easily be applied to the auto industry as well.

For years now the auto industry has focused on producing lighter-weight cars with smaller and smaller displacement engines even though horsepower and torque have been increasing. (the cpu die is getting smaller but performance goes up)

And yet the fuel economy of those cars has increased (the cpu's performance/watt goes up).

But, the price of the cars has increased as well (gross margins have held steady or possibly improved).

So did the auto industry rip-off the consumer by selling us more expensive but better performing cars that have ever smaller engines?

Has Intel really been ripping off the consumer by increasing clockspeeds, increasing IPC, and decreasing power-consumption while making ever more silicon-efficient die's?

I'd argue the answer in both cases is "no". The bottom line for Intel is that if they ripped off their customers then those customers would go out and by a computing solution that wasn't a ripoff.

Either buy last year's Intel CPU that was "less of a ripoff" if you are to believe Russian's post, or buy an AMD or Via CPU which must be less of a ripoff unless those two companies are ripping off their customers even more so than Intel.

(I admit that my analogy breaks down when you consider the situation between the Prius and the Volt, being a Prius owner myself I can't imagine what would convince me to pay another $15k for a Volt just to have what I already have in my Prius, the Volt is a ripoff at its current price when you factor in performance...and its market numbers reflect this reality)

 

[ShintaiDK]

The easier and more accurate way to work out whether consumers are being "ripped off" is to look at margins. If Intel's margins have massively increase, they are "ripping us off". If they are consistent, then not so much.
(That ignores things like product mix etc, but is a better starting point than die size).

Guess what margins pays for R&D and node shrinks costs. You know, the rapidly increasing ones.

Either you pay or you will soon end up like nVidia/AMD with 3 years+ between nodes.

 

[MLSCrow]

Fact: Bulldozer sucks.

Facts and Speculation: Vishera is mostly aimed at power efficiency improvements rather than IPC or processing performance improvements based on information we have so far from AMD and Trinity reviews.

Fact: AMD has been completely silent with regards to Vishera and has not advertised it much, if at all.

Fact: AMD has already started advertising Steamroller.

Speculation: AMD knows that Vishera's performance is going to be disappointing, so they are keeping that on the DL so as not to create another marketing debacle the way they did with BD. They know that Steamroller will be the first major processing performance increase in their Bulldozer architecture and therefore they want to advertise that CPU over Vishera.

Conslusion: Will AMD release Vishera? In my opinion, yes, but silently, while they try to draw all attention to Steamroller. Will Vishera be worth it? That depends on what you have already? In my opinion, if you have a Thuban, no. Anything earlier than a Thuban, maybe, but we all know that Bulldozer arch isn't very good for gaming, and therefore, if anyone were to ask me what CPU they should get for a gaming rig right now, I'd have to give the award to the Intel 2500K (unless they wanted to mess with the IHS and then I'd say 3570K). I personally have a 955BE@4.0 and am waiting to see what this 8350 can pull off in terms of gaming. I'm skeptical that it will be a worthy upgrade, but we'll see soon enough.

 

[DeeDot78]

Fact: Bulldozer sucks.

Facts and Speculation: Vishera is mostly aimed at power efficiency improvements rather than IPC or processing performance improvements based on information we have so far from AMD and Trinity reviews.

Fact: AMD has been completely silent with regards to Vishera and has not advertised it much, if at all.

Fact: AMD has already started advertising Steamroller.

Speculation: AMD knows that Vishera's performance is going to be disappointing, so they are keeping that on the DL so as not to create another marketing debacle the way they did with BD. They know that Steamroller will be the first major processing performance increase in their Bulldozer architecture and therefore they want to advertise that CPU over Vishera.

Conslusion: Will AMD release Vishera? In my opinion, yes, but silently, while they try to draw all attention to Steamroller. Will Vishera be worth it? That depends on what you have already? In my opinion, if you have a Thuban, no. Anything earlier than a Thuban, maybe, but we all know that Bulldozer arch isn't very good for gaming, and therefore, if anyone were to ask me what CPU they should get for a gaming rig right now, I'd have to give the award to the Intel 2500K (unless they wanted to mess with the IHS and then I'd say 3570K). I personally have a 955BE@4.0 and am waiting to see what this 8350 can pull off in terms of gaming. I'm skeptical that it will be a worthy upgrade, but we'll see soon enough.

I had the same feeling. How come these slides highlighting Steamroller already, even though Vishera isnt out? Its not gonna be the Sandy Bridge competition we hoped for, hopefuly we are both wrong.

 

[Eeqmcsq]

I had the same feeling. How come these slides highlighting Steamroller already, even though Vishera isnt out? Its not gonna be the Sandy Bridge competition we hoped for, hopefuly we are both wrong.

It could be that Piledriver cores have already been released, talked about, and benched for Trinity APUs, so it would be more exciting to talk about the new Steamroller cores instead.

Edit - I thought about it some more, and the Vishera silence is pretty deafening.

 

[exar333]

Fact: Bulldozer sucks.

Facts and Speculation: Vishera is mostly aimed at power efficiency improvements rather than IPC or processing performance improvements based on information we have so far from AMD and Trinity reviews.

Fact: AMD has been completely silent with regards to Vishera and has not advertised it much, if at all.

Fact: AMD has already started advertising Steamroller.

Speculation: AMD knows that Vishera's performance is going to be disappointing, so they are keeping that on the DL so as not to create another marketing debacle the way they did with BD. They know that Steamroller will be the first major processing performance increase in their Bulldozer architecture and therefore they want to advertise that CPU over Vishera.

Conslusion: Will AMD release Vishera? In my opinion, yes, but silently, while they try to draw all attention to Steamroller. Will Vishera be worth it? That depends on what you have already? In my opinion, if you have a Thuban, no. Anything earlier than a Thuban, maybe, but we all know that Bulldozer arch isn't very good for gaming, and therefore, if anyone were to ask me what CPU they should get for a gaming rig right now, I'd have to give the award to the Intel 2500K (unless they wanted to mess with the IHS and then I'd say 3570K). I personally have a 955BE@4.0 and am waiting to see what this 8350 can pull off in terms of gaming. I'm skeptical that it will be a worthy upgrade, but we'll see soon enough.

Eventually folks will pick-up on the fact that AMD is always touting the 'around the corner' product over the imminent offering. IMHO it hurts their image and poisons excitement, rightfully so, for the upcoming offering. Also kills confidence in them overall...

 

[AtenRa]

AMD reveals their next x86 architecture every August at the HotChips Symposium. (Bulldozer/bobcat 2010, Trinity/PileDriver 2011, SteamRoller 2012)

Intel is doing the same thing every September at the Intel Developer Forum(IDF) (SandyBridge 2010, IvyBridge 2011, Haswell 2012 ??)

I dont see whats the problem with the SteamRoller reveal at Hotchips 2012.

 

[Phynaz]

Simple, AMD is always saying "Just wait, there's this wonderful thing coming." And it turns out that thing isn't so wonderful.

Intel on the other hand doesn't do it so much. And coming out with the details during IDF makes sense, it gets the developers up to speed. Get the difference?

 

[intangir]

AMD reveals their next x86 architecture every August at the HotChips Symposium. (Bulldozer/bobcat 2010, Trinity/PileDriver 2011, SteamRoller 2012)

I dont see whats the problem with the SteamRoller reveal at Hotchips 2012.

Fact check: AMD did not reveal anything about Trinity or Piledriver at HotChips 2011... in fact, they presented more about Llano and Bulldozer there. I notice you failed to provide a link for HotChips 2011, so here it is: http://www.hotchips.org/archives/hc23

So yes, AMD seems to be jumping the gun on Steamroller, considering it's not supposed to be out until 2H/2013.

 

[Homeles]

Fact: Bulldozer sucks.

Facts and Speculation: Vishera is mostly aimed at power efficiency improvements rather than IPC or processing performance improvements based on information we have so far from AMD and Trinity reviews.

Fact: AMD has been completely silent with regards to Vishera and has not advertised it much, if at all.

Fact: AMD has already started advertising Steamroller.

Speculation: AMD knows that Vishera's performance is going to be disappointing, so they are keeping that on the DL so as not to create another marketing debacle the way they did with BD. They know that Steamroller will be the first major processing performance increase in their Bulldozer architecture and therefore they want to advertise that CPU over Vishera.

Conslusion: Will AMD release Vishera? In my opinion, yes, but silently, while they try to draw all attention to Steamroller. Will Vishera be worth it? That depends on what you have already? In my opinion, if you have a Thuban, no. Anything earlier than a Thuban, maybe, but we all know that Bulldozer arch isn't very good for gaming, and therefore, if anyone were to ask me what CPU they should get for a gaming rig right now, I'd have to give the award to the Intel 2500K (unless they wanted to mess with the IHS and then I'd say 3570K). I personally have a 955BE@4.0 and am waiting to see what this 8350 can pull off in terms of gaming. I'm skeptical that it will be a worthy upgrade, but we'll see soon enough.

This is completely illogical.

If AMD's ready to go public with their Steamroller details, why shouldn't they? Just because you have an apparent selfish need for Vishera to launch prior to talking about Steamroller does not mean that everybody else that is curious about Steamroller, like myself, should have to wait to hear about it.

Piledriver's already out in the form of Trinity APUs. But no, a very small percentage of the desktop market is somehow more important than AMD's mainstream products. You seem to be suffering from entitlement issues.

Homeles, you seem to be suffering from delusions of grandeur. Specifically that you can get away with breaking our rules about not attacking other members.
-ViRGE

 

[Ajay]

I'm really wondering how Steamroller will work out with the new HD libraries @ 28nm:

Still, AFAIK, GF only has an HP node that Steamroller can be built on, I haven't heard any news of a SHP node. If steamroller is built on 28nm HP, my guess is that we are more likely to get substantial power reductions than much in the way of net performance increases.

 

[cytg111]

seems to me that AMD should consider teaming up with samsung..

 

[AtenRa]

Fact check: AMD did not reveal anything about Trinity or Piledriver at HotChips 2011... in fact, they presented more about Llano and Bulldozer there. I notice you failed to provide a link for HotChips 2011, so here it is: http://www.hotchips.org/archives/hc23

That is because, first, they talked about Trinity (By CEO Rick Bergman) in June at AFDS 2011 (AMD Fusion Developer Summit).

Second, PileDriver is not a new x86 architecture, it is a refined Bulldozer module.

So yes, AMD seems to be jumping the gun on Steamroller, considering it's not supposed to be out until 2H/2013.

That is only your opinion. Most of the world doesn't see it this way.

 

[AtenRa]

seems to me that AMD should consider teaming up with samsung..

They did that by joining HSA.

Expanding the HSA Foundation

HSA Foundation Announces Six New Members

 

[sandorski]

seems to me that AMD should consider teaming up with samsung..

:hmm:

 

[Arkadrel]

@Sandorski

Yes we both interpreated that the same way.
He fully well knew of samsung joining the HSA,
and had read this part:

IIRC AMD put out its own 80386 which used Intel's 80386 microcode after Intel suddenly cancelled its licensing agreement with AMD.

So it was a Innuendo.
Oh well "Haters gonna hate".

Personally dont think Samsung got a fair trail, or where really copying Apple.
I think its retarded that you can copyright the shape of a phone, or colour.

And the icons? well... only really so many ways to depict a thing.
Thus if everyone had to use differnt depictions for the same things, no one would understand the depictions going from 1 phone to another
(then what is the point of the depiction?).

Its just plain silly. Its certainly not worth Billions of $ dollars.

Apple are mis-useing the legal system to defeat fair competition.

 

[AtenRa]

 

[sequoia464]

So yes, AMD seems to be jumping the gun on Steamroller, considering it's not supposed to be out until 2H/2013.

Not arguing the point here because I have lost track on this lately; I thought AMD has the release of SR as H1 2013, but because of recent events everyone is speculating H2 2013??

 

[intangir]

That is because, first, they talked about Trinity (By CEO Rick Bergman) in June at AFDS 2011 (AMD Fusion Developer Summit).

If you listen to the keynote, it was also all about Llano, and GCN. Trinity was announced, the way Steamroller and Excavator have been announced, but nothing about its microarchitecture was revealed there.

Second, PileDriver is not a new x86 architecture, it is a refined Bulldozer module.

That's not what you were saying before.

AMD reveals their next x86 architecture every August at the HotChips Symposium. (Bulldozer/bobcat 2010, Trinity/PileDriver 2011, SteamRoller 2012)

And the Steamroller slides also describe it as a tweaked modular core, giving it the same emphasis as Piledriver. In fact, they explicitly mention IPC with Piledriver, but not with Steamroller.

That is only your opinion. Most of the world doesn't see it this way.

It's not only my opinion. Others obviously hold it as well. Don't get me wrong, I'm happy to see any details of what architecture tweaks AMD is doing. But there seems to be a really long lead time this time until we get real measurements of what actual impact these changes will have.

 

[intangir]

Not arguing the point here because I have lost track on this lately; I thought AMD has the release of SR as H1 2013, but because of recent events everyone is speculating H2 2013??

No one knows when Steamroller for the desktop will be released. Those chips haven't been announced. We don't even know if it will use the same socket AM3+ as Vishera. The desktop chip roadmaps list Vishera lasting through 2013.

The speculation is that the first we'll see SR is in Kaveri the APU, similar to Trinity being the first incarnation of PD. The best information we have on Kaveri is from Charlie at semiaccurate.com: http://semiaccurate.com/2012/08/06/amd-to-update-trinity-to-trinity-2-0/

Being a late Q1 or early Q2 update leaves Kaveri for 2H/2013, something that SemiAccurate has been hearing was the schedule for a while now. The update to Trinity pretty much nails down Kaveri as later 2013, not the usual late spring release like v1.0 and Llano before that.