AMD CPUs, SoC Rumors and Speculations Temp. thread 2

[-Fran-]

Linking the rumor of MS thinking of buying AMD and this rumor from Charlie, it might be MS injecting cash into AMD. It would make more sense for MS to invest into AMD instead of buying it. They have some very strong partnerships going on, so it would make some sense.

You also have the Chinese investors from some while ago. Those are confirmed, right? But nothing has been in the news ever since.

Then, to a lesser degree and less likely, Apple injecting cash is not that far fetched either. Intel as well, but I'd put it at 4th in terms of likeliness. After that is just wild guesses: IBM, Oracle, Qualcomm, Samsung, etc...

Cheers!

 

[gamerk316]

http://www.fudzilla.com/news/processors/38694-private-equity-to-get-20-percent-of-amd

20% of AMD being purchased by private equity firm. Figured something like this was brewing; AMD didn't have the cash to make it to Zen if their losses stayed where they are now.

Zen is now officially AMDs last chance.

 

[-Fran-]

http://www.fudzilla.com/news/processors/38694-private-equity-to-get-20-percent-of-amd

20% of AMD being purchased by private equity firm. Figured something like this was brewing; AMD didn't have the cash to make it to Zen if their losses stayed where they are now.

Zen is now officially AMDs last chance.

It's not just a "private equity firm". It's Silver Lake.

Those guys are not known to help companies without a big way to make money back even if the company sinks. I shrug at this... XD

Cheers!

 

[gamerk316]

But from my perspective, there's a very easy way to turn a quick payoff: Spin the old ATI unit, take the cash, then bail.

I'm more sure then ever then within the next few years, AMD gets broken up. I simply don't see them getting out of their debt situation.

EDIT

By my count, the total cost of this purchase was somewhere around $533 Million in change. I'm interested to see AMDs losses this quarter...

 

[cdrkf]

Gamer, I wouldn't count AMD out just yet 😛

They've been about to go out of business, broken up or purchased for as long as I can remember, yet here we are...

It occurs to me that a 'underdog' but ultimately stable amd suits the market quite well. I have a suspicious Zen might miraculously do better than expected and stabilise things (i mean I think Intel have been holding back to allow it- they could have made 6 or 8 cores standard already with their process advantage).

 

[gamerk316]

^^ Even AMD has admitted Zen won't be faster then Haswell, let alone Skylake.

 

[jimmysmitty]

Gamer, I wouldn't count AMD out just yet 😛

They've been about to go out of business, broken up or purchased for as long as I can remember, yet here we are...

It occurs to me that a 'underdog' but ultimately stable amd suits the market quite well. I have a suspicious Zen might miraculously do better than expected and stabilise things (i mean I think Intel have been holding back to allow it- they could have made 6 or 8 cores standard already with their process advantage).

But this is bad. A company can only go so long with losses before they are done for and AMD has been going for a long time only surviving due to investments, such as that from the FAB spin off, or the settlement from Intel.

20% is a pretty large chunk and they will have some weight to throw around. I guess we can just hope they don't try to push AMD into something of a worse condition.

I doubt they could survive without the ATI division TBH. That is the only division that has been competitive, even though their more recent entries are not as competitive in every aspect as before they still sell better than their CPU division does.

 

[cdrkf]

Gamer, I wouldn't count AMD out just yet 😛

They've been about to go out of business, broken up or purchased for as long as I can remember, yet here we are...

It occurs to me that a 'underdog' but ultimately stable amd suits the market quite well. I have a suspicious Zen might miraculously do better than expected and stabilise things (i mean I think Intel have been holding back to allow it- they could have made 6 or 8 cores standard already with their process advantage).

But this is bad. A company can only go so long with losses before they are done for and AMD has been going for a long time only surviving due to investments, such as that from the FAB spin off, or the settlement from Intel.

20% is a pretty large chunk and they will have some weight to throw around. I guess we can just hope they don't try to push AMD into something of a worse condition.

I doubt they could survive without the ATI division TBH. That is the only division that has been competitive, even though their more recent entries are not as competitive in every aspect as before they still sell better than their CPU division does.

@Gamer, what has been confirmed by AMD is that Zen won't be faster in terms of IPC than Haswell (so Intel will still have the per core performance advantage based on the fact we don't think the new process will let them clock that high). Still, if the efficiency is there- then AMD could still offer a very nice multi thread processor with higher overall performance in highly threaded workloads.

This is the position they were aiming for with the FX processors (unfortunately they missed the mark, offering comparable performance in multi thread and flat out worse in everything else). I'd certainly consider Zen if it offers reasonable single thread and excellent multi thread performance.

That aside, you may well be on the money with the graphics split:
http://venturebeat.com/2015/09/09/amd-to-set-up-radeon-graphics-chips-as-a-separate-business/

Mind you, that's probably more a defensive move than confirmation of anything.

 

[salgado18]

@Gamer, what has been confirmed by AMD is that Zen won't be faster in terms of IPC than Haswell (so Intel will still have the per core performance advantage based on the fact we don't think the new process will let them clock that high). Still, if the efficiency is there- then AMD could still offer a very nice multi thread processor with higher overall performance in highly threaded workloads.

This is the position they were aiming for with the FX processors (unfortunately they missed the mark, offering comparable performance in multi thread and flat out worse in everything else). I'd certainly consider Zen if it offers reasonable single thread and excellent multi thread performance.

That is the multi-million dolar question, and one AMD has not responded. Maybe they are working hard to get the best efficiency out of the chips, and come out with competitive products, instead of aiming at unprecedented performance, like the FX 9* chips. That (and better marketing) is what they need to turn around and be profitable.

That aside, you may well be on the money with the graphics split:
http://venturebeat.com/2015/09/09/amd-to-set-up-radeon-graphics-chips-as-a-separate-business/

Mind you, that's probably more a defensive move than confirmation of anything.

That is probably a defensive move of the investing firm. They believe the CPU division is a gamble but the GPU is not, so only one breaks if Zen fails, and their money is secure.

Dam capitalists. XD

 

[gamerk316]

My take, having been through a couple of these myself, is this is a precursor to a spinoff. All the signs are there, especially if another wave of restructuring occurs.

This is normal in the corporate world: Company makes less money then some investors think it should, they purchase a large quantity of stock, force the company to split, jettisoning off the "bad" parts so they can enhance the value of the "good" parts.

http://venturebeat.com/2015/09/09/amd-to-set-up-radeon-graphics-chips-as-a-separate-business/

Called it.

 

[jimmysmitty]

My take, having been through a couple of these myself, is this is a precursor to a spinoff. All the signs are there, especially if another wave of restructuring occurs.

This is normal in the corporate world: Company makes less money then some investors think it should, they purchase a large quantity of stock, force the company to split, jettisoning off the "bad" parts so they can enhance the value of the "good" parts.

http://venturebeat.com/2015/09/09/amd-to-set-up-radeon-graphics-chips-as-a-separate-business/

Called it.

Problem I see is that AMD would not survive unless Zen is that good. I also doubt that Intel would allow for AMD to die off just because investors don;t want to try to turn it around.

Intel doesn't want to deal with a monopoly again nor do I think they would like to deal with setting up a new company to become the second x86 manufacture.

Some people think if AMD dissolved that x86-64 would become unusable to Intel but I would imagine Intel would buy it and license it to another. Not sure who, I can only imagine someone like Samsung who has the weight to actually FAB chips.

 

[gamerk316]

Problem I see is that AMD would not survive unless Zen is that good. I also doubt that Intel would allow for AMD to die off just because investors don;t want to try to turn it around.

I've been through three spins in five years working professionally. The investors couldn't give a damn about whatever company they invest in, they care about how much money they can make off it. And guess what? A quarter of them are already shorting the stock.

This is a classic setup for a breakup. I've seen it happen personally, and all the signs are there. Investors will keep the part think has the most earning potential, ride it as long as they can, then bail the second they think they maximized the stock price. And anyone thinking the investors care about AMD as a business is just fooling themselves.

 

[Cazalan]

Edit: oops didn't see page 8 😉

 

[gamerk316]

^^

EDIT

Yes, you did 😛

And I point out, isn't spinning the GPU division EXACTLY what I predicted would happen? Next 18 months are going to be interesting to watch, though maybe not so much if your an AMD fan.

 

[de5_Roy]

AMD's Raja Koduri promoted to Radeon SVP
http://www.fudzilla.com/news/graphics/38705-amd-s-raja-koduri-promoted-to-svp-of-radeon
http://semiaccurate.com/2015/09/09/amd-forms-radeon-technologies-group/

 

[jdwii]

I pretty much knew something was up back when i switched from my build i mean the moves they were making was questionable. Zen is it like everyone has been saying i'm even sure K12 has been put up on the rack to Zen is more closer to launch. Them doing this to the radeon divison while still making APUs means something to me. FuryX didn't save them while it performed basically on par meaning if Zen comes out we can't even have it compete on par we need a killer CPU something Amd already said they don't have.

Keep in mind the current design wins they do have which is basically consoles.

Also just to say it Juan said this would all happen over a year ago lol

 

[Cazalan]

I've been through three spins in five years working professionally. The investors couldn't give a damn about whatever company they invest in, they care about how much money they can make off it. And guess what? A quarter of them are already shorting the stock.

This is a classic setup for a breakup. I've seen it happen personally, and all the signs are there. Investors will keep the part think has the most earning potential, ride it as long as they can, then bail the second they think they maximized the stock price. And anyone thinking the investors care about AMD as a business is just fooling themselves.

Anything is possible but it is also confirming what many have said was a gamble. Focusing too much on what they used to call Fusion. Greater emphasis on APUs when the memory interfaces were lacking and the technology nodes weren't good enough.

I think this has a lot to do with VR. One Fiji can run 2 Oculus HMDs. https://twitter.com/missquickstep/status/637478111957815296

Imagine the implications for entertainment/health services/education purposes.

 

[jimmysmitty]

I pretty much knew something was up back when i switched from my build i mean the moves they were making was questionable. Zen is it like everyone has been saying i'm even sure K12 has been put up on the rack to Zen is more closer to launch. Them doing this to the radeon divison while still making APUs means something to me. FuryX didn't save them while it performed basically on par meaning if Zen comes out we can't even have it compete on par we need a killer CPU something Amd already said they don't have.

Keep in mind the current design wins they do have which is basically consoles.

Also just to say it Juan said this would all happen over a year ago lol

It is interesting how spot on Juan can be even though he gets a bit to harsh sometimes.

I guess the next few months will tell. I doubt they will let AMD just die off though. That would be a jerk of a move.

 

[Cazalan]

Spot on about what exactly? All this is is another reorg to build better products. A reorg plus cash injection being a good thing. Not the usual doom and gloom spread here.

http://semiaccurate.com/2015/09/09/amd-forms-radeon-technologies-group/

 

[-Fran-]

I don't think they will sell the GPU BU just yet. What they are doing is making it independent of the CPU division. Putting that man at the top is recognizing they have been taking very crappy business decisions, and we all can agree to that.

Think about it this way: all of the money the GPU division could have brought in after the mild success VLIW4 had with the 6970 series and their potential evolution, has been hampered by the Fusion vision/dogma, where GPUs are meant to be slapped into the CPU and call it a day. Like Cazalan points out, it's a bet they lost some time ago when BD screwed up every single plan they had to fund a proper road map on each division (they were tied form that point on) and had to make do with the evolution of their APUs. To their good luck, nVidia has not been making major leaps even though they have a strong advantage in several fronts: scaled designs, optimized designs for different markets and single development costs. For all the "jack of all trades" GCN is, it just doesn't scale very good and it's tied to the "fusion" dogma. I bet they will now get rid of the fat and put a proper design GPU wise putting the "APU" in second place.

I don't really know if it's good or bad long term, but at least is a change that will bring money in the short term from the GPU division.

Also, this could mean HSA will go bye bye, isn't it? Like officially.

Cheers!

 

[jimmysmitty]

Spot on about what exactly? All this is is another reorg to build better products. A reorg plus cash injection being a good thing. Not the usual doom and gloom spread here.

http://semiaccurate.com/2015/09/09/amd-forms-radeon-technologies-group/

Not the doom and gloom but he talks about this kind of stuff quite a bit.

I also would not say doom or gloom but AMD has been making a lot of bad choices and focusing on a lot of the wrong areas. While ATI was good for their APU idea it didn't help and the cost heavily weighed them down.

I don't think they will sell the GPU BU just yet. What they are doing is making it independent of the CPU division. Putting that man at the top is recognizing they have been taking very crappy business decisions, and we all can agree to that.

Think about it this way: all of the money the GPU division could have brought in after the mild success VLIW4 had with the 6970 series and their potential evolution, has been hampered by the Fusion vision/dogma, where GPUs are meant to be slapped into the CPU and call it a day. Like Cazalan points out, it's a bet they lost some time ago when BD screwed up every single plan they had to fund a proper road map on each division (they were tied form that point on) and had to make do with the evolution of their APUs. To their good luck, nVidia has not been making major leaps even though they have a strong advantage in several fronts: scaled designs, optimized designs for different markets and single development costs. For all the "jack of all trades" GCN is, it just doesn't scale very good and it's tied to the "fusion" dogma. I bet they will now get rid of the fat and put a proper design GPU wise putting the "APU" in second place.

I don't really know if it's good or bad long term, but at least is a change that will bring money in the short term from the GPU division.

Also, this could mean HSA will go bye bye, isn't it? Like officially.

Cheers!

It was not just BD that screwed everything up. The first big stumble was K10 and their very leaky 65nm. That was the start. BD was worse by far but they have been stumbling with a few saves here and there.

And i wouldn't put VLIW4 or the HD6970 that high up. There is a reason why I didn't jump from a HD5870 to a HD6970, it just didn't offer enough of a performance jump. Hut the HD7970 did and was a vastly better GPU.

As I said we will see what happens in the next few months.

But remember, a lot of people were sure that GloFlo would still be somewhat owned by AMD but now they are independent. It could very well be that Radeon Technologies Group will end up becoming the same. I kind of wish that AMD kept the ATI name so if that does happen they just become ATI again.

 

[Cazalan]

Also, this could mean HSA will go bye bye, isn't it? Like officially.

Cheers!

Not sure where you got that from. There was no mention of ending semi-custom or APU development.

This means the GPU division won't be held back by APU decisions. The APU side is still going to need the IP for products too.

 

[-Fran-]

It was not just BD that screwed everything up. The first big stumble was K10 and their very leaky 65nm. That was the start. BD was worse by far but they have been stumbling with a few saves here and there.

And i wouldn't put VLIW4 or the HD6970 that high up. There is a reason why I didn't jump from a HD5870 to a HD6970, it just didn't offer enough of a performance jump. Hut the HD7970 did and was a vastly better GPU.

As I said we will see what happens in the next few months.

But remember, a lot of people were sure that GloFlo would still be somewhat owned by AMD but now they are independent. It could very well be that Radeon Technologies Group will end up becoming the same. I kind of wish that AMD kept the ATI name so if that does happen they just become ATI again.

K10 was not a "flop" by any meaning of the word. And even if you want to call it that, I'd say the scales are totally different.

In regards to the jump from VLIW4 to GCN 1.0, don't forget they also had a process shrink and a huge bump in speeds. I'm talking about the evolutionary steps *that could have been* keeping the work with a VLIW arch. Maybe it was a stop gap, but imagine they kept the VLIW4 as mid and low range and focused on getting a polished GPU only GCN arch? I'm just theorizing about the lost opportunity by abiding to the Fusion dogma. To give you an example: 7970 based on GCN optimized and 7870 VLIW4+.

EDIT2: Forgot the second part XD

Spinning their fabs was a sell from the beginning. I don't remember them saying "we're looking to make the Fab it's own business unit". They went directly for the sell, didn't they?

Not sure where you got that from. There was no mention of ending semi-custom or APU development.

This means the GPU division won't be held back by APU decisions. The APU side is still going to need the IP for products too.

It's mostly a shot in the dark. Or intuition.

HSA has only been living up because they *need* to make APUs a viable computing platform. No one has picked it up and most of their APU and GPU work has been HSA centric for some time. If they either scrap HSA or re-work the spec, it would make more sense to me to keep it. In other words, I think, at this point, HSA is a money sink that has not brought anything to the table really. Or at least, I haven't seen anything yet. Hell, even Intel has hybrid decoding for flocks sake. AMD, having ways to make hybrid computing a thing, have done squat in one of the *easy* areas to actually promote the tech using HSA.

Cheers!

EDIT: Typos.

 

[jimmysmitty]

I never said K10 was a flop but it was a bad CPU. It ran at lower clocks, was beat by a CPU that was2 years older and MCM vs monolithic and with a chipset MC vs IMC and lower clocks (i.e. a Q6600 beat the Phenom 9800). It could not overclock and the original had the TLB bug which be it that you would run into it or not most every firmware got updated to patch it which caused a big performance drop.

Phenom II, K10.5 on 45nm, was much better but still lagged. It was one of the places where AMD caught themselves before they smashed face first into the ground. But K10 itself was a bad chip. Again not as bad as BD, which is a flop in some ways, but still it was by no means a good chip either.

 

[-Fran-]

I never said K10 was a flop but it was a bad CPU. It ran at lower clocks, was beat by a CPU that was2 years older and MCM vs monolithic and with a chipset MC vs IMC and lower clocks (i.e. a Q6600 beat the Phenom 9800). It could not overclock and the original had the TLB bug which be it that you would run into it or not most every firmware got updated to patch it which caused a big performance drop.

Phenom II, K10.5 on 45nm, was much better but still lagged. It was one of the places where AMD caught themselves before they smashed face first into the ground. But K10 itself was a bad chip. Again not as bad as BD, which is a flop in some ways, but still it was by no means a good chip either.

But you're mixing process with uArch. K10 was not bad in any way (yes, "flop" was kind of overstated on my interpretation of what you said). Also, the TLB bug was so stupidly overrated at the time I don't even want to go there, please... And no, K10.5 was not Phenom II (K10 all the way up to Thuban). K10.5 is STARS, which is Llano. That was the last evolution of K10 before BD came around pooping the party.

Cheers!

EDIT: Typos.