AMD CPU speculation... and expert conjecture

[juanrga]

You guys really need to learn how to ignore you know who.

He has gone out of his way to actually private message me to try and convince me Intel is superior to AMD in every way.

He is either getting money for forum sliding or he's got some sort of personal issues that I am not going to touch on in this thread.

Why "EITHER... OR"? why not "BOTH... AND"?

He has also PM me... twice.

Also, I found it kind of interesting, but I was looking back at Steamroller news from 2012 and there were two prominent rumors floating around. The first was that there was going to be a Steamroller that came out in early 2013 (which clearly didn't happen).

The second was that we'd see DT SR in mid to late 2014.

Oddly enough, it seems like the release info for Steamroller has been around for a year and a half, just seems like everyone forgot about it.

Read: https://www.google.com/search?q=amd+steamroller+roadmap+2012

Tons of conflicting issues, but if we look at what was supposed to happen and what has actually happened, things might become a little more clear.

http://www.xbitlabs.com/news/cpu/display/20121105000202_AMD_s_Roadmap_Slide_Does_Not_Predict_Steamroller_in_2013.html

That one is quite interesting to me. Yes, they were wrong about no SR in 2013, however they were right about Richland.

So, I guess what I am asking is, how do you guys feel about there being two versions of Steamroller (like seronx has mentioned previously) considering how delays and new products have crept up? I am kind of feeling that Steamroller 1 was supposed to be out now if it wasn't scrapped and it was then replaced with a Steamroller 2.

It makes sense to me, the old slides had Steamroller marked as "greater parallelism", which is something AMD doesn't need right now. It needs better IPC, and if they can do it while keeping the module design AND keeping the module die sizes about the same, they will have a massive winner of a product on their hand.

I think that maybe this is what has happened. We might see the "greater parallelism" show up in some sort of PD refresh with more cores, and then SR will release a year later than what we would have normally expected (October release) and it will have a far different goal than "greater parallelism".

Just my 2 cents, but it sounds about right. And I am a strong believer that AMD saying they are done competing with Intel means that AMD will still go to HEDT market, that is no longer something Intel is concerned with and it's why it's taken to strongly to internet campaigns against power usage and spreading FUD related to it, when power consumption is irrelevant for desktop CPUs in gaming PCs. (and as a side note, I had to tell some guy who was all excited about Broadwell doing 4.8ghz at an estimated 100w or so that he had three 170w GTX 760s in that very same computer, and that saving 80w on your CPU when you have 510w of graphics is a waste).

Fudzilla calls them Steamroller A core and B core. Apparently the A core was cancelled and Kaveri will use the B core.

Which is the difference between A and B? Nobody outside AMD knows for sure. One option could be that A core didn't had HSA enabled and was aimed to FX Steamroller/Warsaw, but was cancelled by migrating to APU/HSA approach Kaveri/Berlin.

Another option is that A core is what was presented at Hot Chips 2012 and that B core is the rumoured 3 ALUs per core that appeared in a recent AMD slide.

The "greater parallelism" of Steamroller refers to the doubled decoder, which can now execute two threads at once. The shared decoder in Bulldozer/Piledriver only can feed one core in the module at once.

 

[Ags1]

AMD Earnings Call (quoted from http://seekingalpha.com/article/1752782-advanced-micro-devices-ceo-discusses-q3-2013-results-earnings-call-transcript?part=single):

Rory Read: Toshiba and HP announced the first Temash powered convertibles and we expect more designs from other OEMs to launch this quarter. Our work to strengthen our desktop business is gaining momentum.... We delivered the second straight quarter of channel revenue growth based on strong demand for our high-end A8 and A10 APUs and FX CPUs. We expect this trend to continue.

Computing Solutions segment revenue was $790 million, down 6% sequentially, due to lower notebook and chipset unit shipments, partially offset by higher desktop unit shipments.

We remain on track to begin shipping our next generation Kaveri APU for the channel this quarter. And key motherboard partners are already offering new product in anticipation of this launch.

We need to move stronger into desktop and as we talked about a year ago, we worked on the inventory on the desktop segment and we built and repaired that and we have seen two quarters of consistent revenue growth in that segment and we believe that we have the right product stack to continue to make progress and that’s part of our business as well.

the clarification around SeaMicro servers. As I said the Sea -- the Verizon engagement has lasted over the past two years. So some of the initial deployments were with the Intel processors but we do have significant deployments with AMD Opteron as well.

We do see the percentage of Opteron processors increasing because that’s what we’d like to do.

 

[juanrga]

AMD official statement about MANTLE

http://community.amd.com/community/amd-blogs/amd-gaming/blog/2013/10/17/the-four-core-principles-of-amd-s-mantle

Some relevant quotes:

It’s not that Mantle is the initial language with which developers are writing their games on each platform, as some have surmised; the point of Mantle is that it’s easy to reuse, in whole or in part, the development effort expended on the next-generation consoles when bringing the same game to life on the PC. This is because Mantle allows developers to use the same features and programming techniques they are already utilizing for next-gen game consoles. And while the initial iteration of Mantle is intended specifically for PCs, it has been designed from the beginning to be extensible to other platforms as well.

Aka the "this API is DoA because no developer will be learning a different API" claim is... DoA :lol:

As a company, we do not chase one or two percentage points of performance when we set our minds to a project as elaborate and historic as Mantle. While it remains too early to disclose performance figures, we expect you will be pleasantly surprised when Battlefield™ 4 is patched to support Mantle in December. And with a Mantle back-end baked into Frostbite™ 3 you can expect other games based on this engine to reap all of its benefits.

Aka the "it will only provide a 5% gain" claim is DoA.

While Mantle is uniquely optimized for PCs containing the Graphics Core Next architecture, we recognize there other architectures in the market. Gamers with these architectures deserve good gameplay just as much as anyone else, and we have t designed Mantle in a way that respects their right to game well.

Aka the "it will break everything" claim is DoA.

 

[juanrga]

And more good news http://www.zdnet.com/amd-q3-beats-expectations-returns-to-profitability-7000022098/

 

[8350rocks]

You guys really need to learn how to ignore you know who.

He has gone out of his way to actually private message me to try and convince me Intel is superior to AMD in every way.

He is either getting money for forum sliding or he's got some sort of personal issues that I am not going to touch on in this thread.

Why "EITHER... OR"? why not "BOTH... AND"?

He has also PM me... twice.

Also, I found it kind of interesting, but I was looking back at Steamroller news from 2012 and there were two prominent rumors floating around. The first was that there was going to be a Steamroller that came out in early 2013 (which clearly didn't happen).

The second was that we'd see DT SR in mid to late 2014.

Oddly enough, it seems like the release info for Steamroller has been around for a year and a half, just seems like everyone forgot about it.

Read: https://www.google.com/search?q=amd+steamroller+roadmap+2012

Tons of conflicting issues, but if we look at what was supposed to happen and what has actually happened, things might become a little more clear.

http://www.xbitlabs.com/news/cpu/display/20121105000202_AMD_s_Roadmap_Slide_Does_Not_Predict_Steamroller_in_2013.html

That one is quite interesting to me. Yes, they were wrong about no SR in 2013, however they were right about Richland.

So, I guess what I am asking is, how do you guys feel about there being two versions of Steamroller (like seronx has mentioned previously) considering how delays and new products have crept up? I am kind of feeling that Steamroller 1 was supposed to be out now if it wasn't scrapped and it was then replaced with a Steamroller 2.

It makes sense to me, the old slides had Steamroller marked as "greater parallelism", which is something AMD doesn't need right now. It needs better IPC, and if they can do it while keeping the module design AND keeping the module die sizes about the same, they will have a massive winner of a product on their hand.

I think that maybe this is what has happened. We might see the "greater parallelism" show up in some sort of PD refresh with more cores, and then SR will release a year later than what we would have normally expected (October release) and it will have a far different goal than "greater parallelism".

Just my 2 cents, but it sounds about right. And I am a strong believer that AMD saying they are done competing with Intel means that AMD will still go to HEDT market, that is no longer something Intel is concerned with and it's why it's taken to strongly to internet campaigns against power usage and spreading FUD related to it, when power consumption is irrelevant for desktop CPUs in gaming PCs. (and as a side note, I had to tell some guy who was all excited about Broadwell doing 4.8ghz at an estimated 100w or so that he had three 170w GTX 760s in that very same computer, and that saving 80w on your CPU when you have 510w of graphics is a waste).

Fudzilla calls them Steamroller A core and B core. Apparently the A core was cancelled and Kaveri will use the B core.

Which is the difference between A and B? Nobody outside AMD knows for sure. One option could be that A core didn't had HSA enabled and was aimed to FX Steamroller/Warsaw, but was cancelled by migrating to APU/HSA approach Kaveri/Berlin.

Another option is that A core is what was presented at Hot Chips 2012 and that B core is the rumoured 3 ALUs per core that appeared in a recent AMD slide.

The "greater parallelism" of Steamroller refers to the doubled decoder, which can now execute two threads at once. The shared decoder in Bulldozer/Piledriver only can feed one core in the module at once.

I have no less than 4 PMs in my inbox from him in the last 5 days.

Additionally...this is just speculation mind you...

Based on the early ES's, it seemed that the first batch of Kaveri ES's were doing poorly against what expectations would have been. Whether this stems from bulk process or design itself, I have no idea. I can, however, draw conclusions from those poor results that they were not going to move forward with that product, or we would have seen more advanced ES's emerge beyond something akin to an "alpha" format.

This leads me to believe 1 of 2 things is going on, either (A) they are changing process to retain clockspeed and thermal envelope performance, or (B) they are tweaking the core design further beyond what we know at this point.

I am inclined to think it could easily be either one, or even both, potentially. However, with the alpha stage ES results, they may have taken Jim Keller aside and said something to the effect of..."Hey, this doesn't look right...it needs your attention..."

So, at this point, who knows what we will see when it does launch.

However, if Kaveri 2M APUs match i5 performance with an added ALU per core...look out. That would easily be a world beater for the average DT user. Price would likely slide up a notch because your design cost would likely be a bit more; however, if you could buy a 4 GHz Kaveri that outruns or matches a 4670k with a better iGPU for ~$200, they will still sell a ton of chips and take loads of market share back in DT.

Of course, it is possible that all of that is speculation based on improper conjectures and none of it happens.

 

[Ags1]

However, if Kaveri 2M APUs match i5 performance with an added ALU per core...look out. That would easily be a world beater for the average DT user.

Something to get excited about!

In the earnings call AMD stated they had actual growth in the desktop sales, driven by A10/FX, and that is pretty impressive given the desktop is declining generally.

From what I read of the call transcript, AMD stated several times they are focussing on improving their position in desktop, and that would be at the performance end. So I have high hopes now for a next generation of FX based on Piledriver.

 

[8350rocks]

However, if Kaveri 2M APUs match i5 performance with an added ALU per core...look out. That would easily be a world beater for the average DT user.

Something to get excited about!

In the earnings call AMD stated they had actual growth in the desktop sales, driven by A10/FX, and that is pretty impressive given the desktop is declining generally.

From what I read of the call transcript, AMD stated several times they are focussing on improving their position in desktop, and that would be at the performance end. So I have high hopes now for a next generation of FX based on Piledriver.

Gaming desktops are the only segment growing in DT (i.e. HEDT is growing as fast or faster than the DT market is shrinking)

 

[griptwister]

Sometimes, I ask myself, what the hell are these people thinking. And I found myself asking that very question again when I found out Nvidia is planning on releasing the GTX 780 Ti. -_- It's friggin moronic! The Titan is only 10% faster!

Anyhow, hafijur reminds me of a kid at Fry's I met when I was in the Motherboard's section. The kid was autistic and when he saw me holding onto my MSI 970a Motherboard, the kid tried to convince me Intel was better, that AMD can't possibly beat intel. I tried to tell the kid I couldn't afford a z77 platform, but the kid persisted Intel was better. The kid was somewhat Intelligent, but he could not get past the fact that Intel does NOT hold the best value for money.

 

[juanrga]

Fudzilla calls them Steamroller A core and B core. Apparently the A core was cancelled and Kaveri will use the B core.

Which is the difference between A and B? Nobody outside AMD knows for sure. One option could be that A core didn't had HSA enabled and was aimed to FX Steamroller/Warsaw, but was cancelled by migrating to APU/HSA approach Kaveri/Berlin.

Another option is that A core is what was presented at Hot Chips 2012 and that B core is the rumoured 3 ALUs per core that appeared in a recent AMD slide.

The "greater parallelism" of Steamroller refers to the doubled decoder, which can now execute two threads at once. The shared decoder in Bulldozer/Piledriver only can feed one core in the module at once.

I have no less than 4 PMs in my inbox from him in the last 5 days.

Additionally...this is just speculation mind you...

Based on the early ES's, it seemed that the first batch of Kaveri ES's were doing poorly against what expectations would have been. Whether this stems from bulk process or design itself, I have no idea. I can, however, draw conclusions from those poor results that they were not going to move forward with that product, or we would have seen more advanced ES's emerge beyond something akin to an "alpha" format.

This leads me to believe 1 of 2 things is going on, either (A) they are changing process to retain clockspeed and thermal envelope performance, or (B) they are tweaking the core design further beyond what we know at this point.

I am inclined to think it could easily be either one, or even both, potentially. However, with the alpha stage ES results, they may have taken Jim Keller aside and said something to the effect of..."Hey, this doesn't look right...it needs your attention..."

So, at this point, who knows what we will see when it does launch.

However, if Kaveri 2M APUs match i5 performance with an added ALU per core...look out. That would easily be a world beater for the average DT user. Price would likely slide up a notch because your design cost would likely be a bit more; however, if you could buy a 4 GHz Kaveri that outruns or matches a 4670k with a better iGPU for ~$200, they will still sell a ton of chips and take loads of market share back in DT.

Of course, it is possible that all of that is speculation based on improper conjectures and none of it happens.

I don't have any information/rumour about they changing the process. So far as I know Kaveri is made at Glofo on 28nm bulk. If they are going FD-SOI I don't know.

However, Hot Chips 2012 was describing a Steamroller module before the HSA approach was launched. And we know, from the HSA specification, that CPUs have to be modified. Therefore the idea that Steamroller B includes HSA modifications seems legit.

Moreover, the slide representing a module with 3 ALUs per core was given by AMD during a recent talk about programming Kaveri. There are only three options:
- That is Steamroller
- That is Excavator and was included by error in the slide
- That is nothing and was included by error in the slide

 

[blackkstar]

However, if Kaveri 2M APUs match i5 performance with an added ALU per core...look out. That would easily be a world beater for the average DT user.

Something to get excited about!

In the earnings call AMD stated they had actual growth in the desktop sales, driven by A10/FX, and that is pretty impressive given the desktop is declining generally.

From what I read of the call transcript, AMD stated several times they are focussing on improving their position in desktop, and that would be at the performance end. So I have high hopes now for a next generation of FX based on Piledriver.

Gaming desktops are the only segment growing in DT (i.e. HEDT is growing as fast or faster than the DT market is shrinking)

While I'm here, I'd also like to bring up the point that this is FUD released by Intel that DT is dying. They are constantly performing worse in their high end "mainstream" chips (2600k > 3770k > 4770k) when it comes to IPC + maximum OC headroom.

Not to mention Intel's main agenda is a world where everyone buys Ultrabooks and throws them away in a year. Right now, Intel sells you a chip and it is so far ahead of most user's software that they end up buying new graphics chips from Nvidia or AMD and not giving Intel sales.

Meanwhile, Apple is making a killing selling disposable garbage products that get thrown away after a year.

It is in Intel's best interest to ensure that users think that:

1. Desktop is dying, you need to go mobile

2. Intel's desktop performance is doing great by making power consumption a central focal point to distract from the fact that IPC gains have not made up for total loss of overclocking headroom

3. Gaming PCs are dying: Intel makes far less money off of this market than Nvidia or AMD do. Nvidia and AMD can see people spend $500+ easily on their products in a single gaming PC, which will probably get replaced in a year or two. Intel sells you a 2600k and 3 years later it's still a good chip and the customer has no reason to give Intel more money.

Intel loves it when kids run around saying DT is dead, gaming PCs are over, and power consumption is the primary performance metric you should be looking at while buying a HEDT CPU.

And it works wonders, too. Look at Hajifur. He constantly brings up mobile parts and their power consumption. It's all part of the marketing plan.

 

[-Fran-]

Well, a lot of you pointed out that BULK wasn't a very good move for the BD uArch, since it needs high clocks and BULK is not very good at that to keep the same TDPs, so keeping high clocks and low power would have been almost impossible even moving into (early) modified 28nm. Maybe that's the reason and in all fairness, it makes sense to me.

I'm glad to know (or think) they decided to can another BD possible catastrophe instead of moving ahead with it.

In regards to the market movement. It's really hard to tell from where I'm sitting, but if marketing is of any indication, I agree that Intel is trying to move folks into mobile. They've said it before and they're doing everything in their power to shift processes and products into lower TDPs and FF (form factors).

Cheers!

 

[griptwister]

Got the PM from hafijur as to how the i3 is a better value than a A10. Know what, I'll use some circular reasoning with this one :lol:

The i3 has 2 core that are 100% faster than AMD per core because of more IPC. However, because the FX 6350 has 4 more cores the FX 6350 value is better because it offers 200% more performance at the same price point!

 

[etayorius]

So Haffijur is the JML12 guy all over AMD articles on the web?

I know one thing... i am grabbing an AMD CPU and GPU or Possibly Kaveri APU as my next upgrade, that`s for sure.

 

[tracker45]

so when are they coming out ?

 

[con635]

Sooooo back ot......when wee cut through the arm debate and the ramblings of mad men, the current info/rumor on kaveri is:
Close to i5 quad core cpu performance with built in 7790 ddr3 equivalent??? xfire or even dual graphics with gcn card eg 7790 ddr5??? With mantel titles on the way if amd price this similar or even close to richland that would make an unbeatable budget gaming machine or steam box.

>Mod Edit: Lets not call people out and make drama.

 

[Ags1]

A10 is already at mobile i5 levels, if you believe Anandtech (and you pick your benchmarks carefully):

Taken from: http://www.anandtech.com/show/7428/asus-transformer-book-t100-review/4

 

[Sigmanick]

Oh, Yeah! Can you feel it, baby? I hope to see more collaboration between AMD and OS Coders.
http:// AMD Marketing
We all know that the APU graphics are already stronger than the Intel Solution. But I have no problems with it getting better.

Now let's assume AMD is correct about future performance and add 20% for Steamroller to show how Kaveri will fare this January.

 

[tracker45]

what are the a10 range ? If they are already better than i5, should I buy them instead,

and switch my motherboard to an Fm2+ ?

 

[Ags1]

The desktop A10 is the top APU from AMD, but as a CPU it is weaker than the FX43xx chips, which are the weakest FX chips.

As it currently takes an FX63xx or FX83xx to compete with or beat an i5, I don't think the current APUs are equal to i5s.

 

[Sigmanick]

It's always a question of What are you Doing with your PC?

Gaming with Integrated Graphics? AMD is your winner

Gaming with Discrete cards? Either chip. Especially if you want to Hybrid Crossfire

Running SuperPi Benchmarks? Intel is your winner

HSA enabled programs = AMD

Single core performance = Intel
More cores for Cheaper = AMD

 

[tracker45]

On paper, is the Am3 socket even capable of beating Intel ?

Is the socket even powerful enough to do this ?

 

[The Q6660 Inside]

The desktop A10 is the top APU from AMD, but as a CPU it is weaker than the FX43xx chips, which are the weakest FX chips.

As it currently takes an FX63xx or FX83xx to compete with or beat an i5, I don't think the current APUs are equal to i5s.

That i5 was basically an i3 tbh. They have been capable of upping AM3 from the 965BE to the 8350, so yes it is capable, chances are, SR FX will be electrically compatible with AM3+, in a Phenom II AM3 in an AM2+ socket sense. The socket pretty much just seats the CPU...

 

[Ags1]

True, in the mobile segment Intel only really has i3s and i7s. It just happens to call a random subset of the i3s 'i5s'. An i5 should be a quad core without HT.

 

[8350rocks]

True, in the mobile segment Intel only really has i3s and i7s. It just happens to call a random subset of the i3s 'i5s'. An i5 should be a quad core without HT.

I thought all the i7's except the MX series in mobile were just quads with no HTT... (essentially an i5)?

 

[-Fran-]

True, in the mobile segment Intel only really has i3s and i7s. It just happens to call a random subset of the i3s 'i5s'. An i5 should be a quad core without HT.

I thought all the i7's except the MX series in mobile were just quads with no HTT... (essentially an i5)?

Nope. I have Notebook with an i7 IB which is 4c/8t.

I have to do a lot of core parking/affinity to let turbo boost kick in, lol.

Cheers!