AMD CPUs, SoC Rumors and Speculations Temp. thread 2

[gamerk316]

EDIT

The above point actually raises an interesting question: For Zen, when AMD says "X number of cores", how are they defining X? Because I hope they learned their lesson and aren't counting SMT cores in the total again...

I'm pretty sure AMD did specify cores/threads, or if it was just some of the rumours that was been publish. Everything I have seen have been specifying cores and threads. I doubt they would make that mistake.

Considering that BD was marketed as an 8 core CPU when it was not a full 8 core CPU, they can make that mistake. Of course the marketing team for BD was pretty bad anyways.

Yeah, JF-AMD pretty much got run out of the forums after that debacle.

 

[Vogner16]

EDIT

The above point actually raises an interesting question: For Zen, when AMD says "X number of cores", how are they defining X? Because I hope they learned their lesson and aren't counting SMT cores in the total again...

I'm pretty sure AMD did specify cores/threads, or if it was just some of the rumours that was been publish. Everything I have seen have been specifying cores and threads. I doubt they would make that mistake.

Considering that BD was marketed as an 8 core CPU when it was not a full 8 core CPU, they can make that mistake. Of course the marketing team for BD was pretty bad anyways.

mainstream pc buyers don't know the difference between smt or cmt or ram speeds and clock speeds. all they know is cores gb of ram, and hard drive cap. I know people who bought a Pentium laptop with 6 gb of ram back when most came with 4 (2010) and thought that the $150 more was a huge upgrade in performance! turns out that the Pentium cant push the amount of programs 6 gb of ram can fill up but that's another story.

amd marketing went for the majority of people who look at simple specs and said WOW it has twice the cores of an intel i7! it must be so fast! I bet it poops on intel! and from that perspective it worked great.

 

[jimmysmitty]

EDIT

The above point actually raises an interesting question: For Zen, when AMD says "X number of cores", how are they defining X? Because I hope they learned their lesson and aren't counting SMT cores in the total again...

I'm pretty sure AMD did specify cores/threads, or if it was just some of the rumours that was been publish. Everything I have seen have been specifying cores and threads. I doubt they would make that mistake.

Considering that BD was marketed as an 8 core CPU when it was not a full 8 core CPU, they can make that mistake. Of course the marketing team for BD was pretty bad anyways.

mainstream pc buyers don't know the difference between smt or cmt or ram speeds and clock speeds. all they know is cores gb of ram, and hard drive cap. I know people who bought a Pentium laptop with 6 gb of ram back when most came with 4 (2010) and thought that the $150 more was a huge upgrade in performance! turns out that the Pentium cant push the amount of programs 6 gb of ram can fill up but that's another story.

amd marketing went for the majority of people who look at simple specs and said WOW it has twice the cores of an intel i7! it must be so fast! I bet it poops on intel! and from that perspective it worked great.

The marketing failed because they cherry picked every benchmark. They only compared the FX series to CPUs when it won.

Most people know Intel, very few normal consumers know AMD. The marketing team was trying to focus on the enthusiasts but they treated us like idiots as if we were going to take their benchmarks as the proper benchmarks.

Point is, AMD markets their non 4c/8t CPU as a 8 core CPU. Intel markets their 4c/8t CPU as a 4c/8t CPU. Hell Intel even uses Hyperthreading in their marketing. AMD could have done the same but their marketing team for BD was trying to pull the wool over the eyes of the consumer. There is a reason they got fired after all.

I highly doubt AMD will allow the marketing team to market Zen in any other way than what it truly is, so far we have "Up to 8 cores with 16 threads". No word on if that is on the consumer end or server end and what other configurations that will come in.

 

[Vogner16]

EDIT

The above point actually raises an interesting question: For Zen, when AMD says "X number of cores", how are they defining X? Because I hope they learned their lesson and aren't counting SMT cores in the total again...

I'm pretty sure AMD did specify cores/threads, or if it was just some of the rumours that was been publish. Everything I have seen have been specifying cores and threads. I doubt they would make that mistake.

Considering that BD was marketed as an 8 core CPU when it was not a full 8 core CPU, they can make that mistake. Of course the marketing team for BD was pretty bad anyways.

mainstream pc buyers don't know the difference between smt or cmt or ram speeds and clock speeds. all they know is cores gb of ram, and hard drive cap. I know people who bought a Pentium laptop with 6 gb of ram back when most came with 4 (2010) and thought that the $150 more was a huge upgrade in performance! turns out that the Pentium cant push the amount of programs 6 gb of ram can fill up but that's another story.

amd marketing went for the majority of people who look at simple specs and said WOW it has twice the cores of an intel i7! it must be so fast! I bet it poops on intel! and from that perspective it worked great.

The marketing failed because they cherry picked every benchmark. They only compared the FX series to CPUs when it won.

Most people know Intel, very few normal consumers know AMD. The marketing team was trying to focus on the enthusiasts but they treated us like idiots as if we were going to take their benchmarks as the proper benchmarks.

Point is, AMD markets their non 4c/8t CPU as a 8 core CPU. Intel markets their 4c/8t CPU as a 4c/8t CPU. Hell Intel even uses Hyperthreading in their marketing. AMD could have done the same but their marketing team for BD was trying to pull the wool over the eyes of the consumer. There is a reason they got fired after all.

I highly doubt AMD will allow the marketing team to market Zen in any other way than what it truly is, so far we have "Up to 8 cores with 16 threads". No word on if that is on the consumer end or server end and what other configurations that will come in.

you know my argument on this one its been said over and over, but intel markets 4c 8t because there are only quad cores. physically only 4 exist the other 4 are simulated. while amd used modules the cores are physically real. shared cashes and fpu's schedulers etc make the good argument that a core needs to have these as a part of it to count as a core but I don't think so.

amd had 8 physical cores and 4 units to feed these cores. making it an octa core cpu. their marketing cherry picking benchmarks was garbage but that's why we have independent reviewers. I knew exactly what BD was capable of when I bought it Q1 2012 and was happy with my purchase

 

[-Fran-]

you know my argument on this one its been said over and over, but intel markets 4c 8t because there are only quad cores. physically only 4 exist the other 4 are simulated. while amd used modules the cores are physically real. shared cashes and fpu's schedulers etc make the good argument that a core needs to have these as a part of it to count as a core but I don't think so.

amd had 8 physical cores and 4 units to feed these cores. making it an octa core cpu. their marketing cherry picking benchmarks was garbage but that's why we have independent reviewers. I knew exactly what BD was capable of when I bought it Q1 2012 and was happy with my purchase

We already had a looooong discussion on what a "core" or "CPU" actually is/was by definition, and it was something along these lines: registers + decoder + AGU + ALU = 1 "core". The FPU (x87) and other fixed function pipelines or SIMD units were added in time after the term was coined. So, for all the hate even I have towards marketing BD as an 8 core, they weren't technically incorrect in calling it an 8 core CPU. In any case, no need to deepen this discussion. This is a holy battle we shouldn't fight again IMO. No winners, only losers.

--

In regards to "Zen" coming out of TSMC's line. I would imagine that will be the APU incarnation of it, not the CPU. I would imagine AMD wants Zen based CPUs to come out of GF. Will that make a big difference? I think nope. APUs will be denser and CPUs will be tuned for speed. I'm sure of that. If they manage to make a decent tradeoff for the CPU, I have my doubts. So far there hasn't been any sign of GF improving the process towards speed from low power. Or maybe they don't need over 4Ghz stock.

Cheers!

 

[juanrga]

Zen has been delayed to 2017.

Any sources to that?

http://www.mystatesman.com/news/business/amid-challenges-chipmnaker-amd-sees-a-way-forward/nngdf/

It is called Zen — a microprocessor design project whose results are expected to start showing up in a family of new products starting in 2017.

 

[juanrga]

zen is back on q4 2016. process node may be either made on TSMC 16nm FF or GF 14nm FF+
http://wccftech.com/amd-contracts-tsmc-produce-zen-16nm-woes-14nm-process-troubles-globalfoundries/

regarding the debate about multiplier of fx chips and only working well on highly multithreaded workloads that is what I was getting at. for example handbrake encoding fx chips were just as fast as the 8 thread chips from intel when all cores enabled while significantly slower when one core enabled.

count of cores dosent matter to program. only the threads that can be ran at the same time. if you have CMT 8 thread it gave amd a big advantage to a SMT 8 thread. the point here is 8 cores vs 4 core 8 thread looks the same to program and therefor when each core from amd is 30% slower than each core from intel we find this ratio to be important. where it physically gets its improvement could be said from the shared L2 or fpu's I don't know that part, but I do know that 8 amd cores(threads) that each do 1 unit of work are roughly equal to 8 intel threads that each do 1.3 unit of work.

this is my point its hard to grasp what is so amazing about BD until you understand this key point, and makes me still believe in the CMT design. if only amd had a way to improve that single core performance they would have blown intel out of the water on multithreaded workloads.

@juanrga yes cmt scaled poorly loosing up to 20%, but look at what smt lost! up to 45%!

(im comparing to i7 3770k the top end cpu that intel had to compete with piledriver. its unfair to compare to haswell or broadwell or skylake as amd droped development of BD and intel didn't.)

SMT doesn't lose anything. SMT is a way to use core resources that otherwise would be idle in a wide core when there is no enough ILP on the code. On the other hand CMT left lots of unused resources when there is only one thread independently of the amount of ILP on the code: e.g. ALUs on the second core are never accessible to the first core. There are also deficiencies on the whole CMT concept that add latency and reduce performance. That is why SMT is standard in industry, whereas CMT is only used by AMD bulldozer family (Jaguar/Puma are not CMT) and even AMD is abandoning it for Zen, which uses SMT.

 

[juanrga]

EDIT

The above point actually raises an interesting question: For Zen, when AMD says "X number of cores", how are they defining X? Because I hope they learned their lesson and aren't counting SMT cores in the total again...

I'm pretty sure AMD did specify cores/threads, or if it was just some of the rumours that was been publish. Everything I have seen have been specifying cores and threads. I doubt they would make that mistake.

Considering that BD was marketed as an 8 core CPU when it was not a full 8 core CPU, they can make that mistake. Of course the marketing team for BD was pretty bad anyways.

Bulldozer uses conjoined cores and has 8 of them. Zen uses CMP cores with SMT2 support. Thus when AMD says that Zen FX CPUs have 8 cores they mean 8 cores and 16 threads, like Intel.

P.S.: Multiquote continues without working to me now.

 

[Vogner16]

We already had a looooong discussion on what a "core" or "CPU" actually is/was by definition, and it was something along these lines: registers + decoder + AGU + ALU = 1 "core". The FPU (x87) and other fixed function pipelines or SIMD units were added in time after the term was coined. So, for all the hate even I have towards marketing BD as an 8 core, they weren't technically incorrect in calling it an 8 core CPU. In any case, no need to deepen this discussion.

I had not seen this thanks for the clarification!

TSMC makes amd gpu's currently and I believe got the bid for NVidia pascal as well, however im unsure that even if GF isn't ready for amd's zen core by Q3 2016 I don't think that tsmc would have the combined resources to do all of these projects. they would be making all gpu's and all of amd cpu's that's a lot for a single foundry to handle. we would see shortages on a whole new scale and intel would get another timing win. GF needs to get its <mod edit> together.

on the other hand <mod edit> those arabs. they screwed us over too many times with poor yields and timings. amd products fail to deliver when they are ready because of GF every time it seems. maybe amd give the full contract to TSMC and let them expand to meet demand. if they start now they should have facility ready for q4 2016 and really be able to pump out some chips.

<Let's be watching the language>

 

[vmN]

Any sources to that?

http://www.mystatesman.com/news/business/amid-challenges-chipmnaker-amd-sees-a-way-forward/nngdf/

It is called Zen — a microprocessor design project whose results are expected to start showing up in a family of new products starting in 2017.

That is an questionable source, I don't see it as an confirmation on a delay.

 

[gamerk316]

Not strictly AMD related, but I'm posting because of the implications:

http://www.anandtech.com/show/9659/fable-legends-directx-12-benchmark-analysis/2

Notice the i3 versus i5 versus i7 numbers at 4k. They're exactly the same. It looks like, as I predicted, DX12 is going to be very CPU agnostic, which means that cheap dual cores may be the way to go for cheap gaming rigs going forward. It looks like my prediction DX12 was going to be a major benefit to the Core i3 lineup is going true.

Lower the res to 720, and the CPU again matters (GPU not bottleneck anymore), but at least at higher resolutions, it doesn't look like the CPU is a major factor going forward, since the GPU remains the bottleneck.

 

[juanrga]

Any sources to that?

http://www.mystatesman.com/news/business/amid-challenges-chipmnaker-amd-sees-a-way-forward/nngdf/

It is called Zen — a microprocessor design project whose results are expected to start showing up in a family of new products starting in 2017.

That is an questionable source, I don't see it as an confirmation on a delay.

It agrees with what AMD said during Keller's departure: "sampling" in 2016 and revenue in 2017.

 

[cvjhch]

Any Carrizo reviews yet?

 

[-Fran-]

Not strictly AMD related, but I'm posting because of the implications:

http://www.anandtech.com/show/9659/fable-legends-directx-12-benchmark-analysis/2

Notice the i3 versus i5 versus i7 numbers at 4k. They're exactly the same. It looks like, as I predicted, DX12 is going to be very CPU agnostic, which means that cheap dual cores may be the way to go for cheap gaming rigs going forward. It looks like my prediction DX12 was going to be a major benefit to the Core i3 lineup is going true.

Lower the res to 720, and the CPU again matters (GPU not bottleneck anymore), but at least at higher resolutions, it doesn't look like the CPU is a major factor going forward, since the GPU remains the bottleneck.

Or Fable is using the same backend code with a touch of UE4? I mean, given history of all UE based games, hardly any scales beyond 2 cores decently. I don't think UE4 is going to break that suddenly.

I liked TechReport's core/thread analysis better as well: http://techreport.com/review/29090/fable-legends-directx-12-performance-revealed/4

You can see how different AMD's driver implementation is from nVidia's.

Cheers!

 

[8350rocks]

All SMT implementations simply come down to how much of the pipeline you want to bother duplicating. Pretty much that simple.

BDs problem is this, you have about a 30% per-core performance deficit, plus an additional 20% or so hit when both cores of a BD module are used in non-Integer exclusive workloads. Combine those two factors, and you can see why BD, even in programs that did scale to all cores, would lose to an i7. If AMD had marketed the chip as a quad core with 2-way SMT, it would have been received better, but marketed as an 8-core CPU, people expected more. That's how we ended up at "Intel quads beat AMD's eight core chips".

And I note again: MSFT wouldn't have needed to patch Windows for BD derived chips if AMD just set the HTT CPUID bit, since the Windows patch basically forces BD to be treated the same way HTT is.

EDIT

The above point actually raises an interesting question: For Zen, when AMD says "X number of cores", how are they defining X? Because I hope they learned their lesson and aren't counting SMT cores in the total again...

Core counts are physical cores for Zen, logical cores will be expressed in thread counts.

 

[8350rocks]

EDIT

The above point actually raises an interesting question: For Zen, when AMD says "X number of cores", how are they defining X? Because I hope they learned their lesson and aren't counting SMT cores in the total again...

I'm pretty sure AMD did specify cores/threads, or if it was just some of the rumours that was been publish. Everything I have seen have been specifying cores and threads. I doubt they would make that mistake.

Considering that BD was marketed as an 8 core CPU when it was not a full 8 core CPU, they can make that mistake. Of course the marketing team for BD was pretty bad anyways.

mainstream pc buyers don't know the difference between smt or cmt or ram speeds and clock speeds. all they know is cores gb of ram, and hard drive cap. I know people who bought a Pentium laptop with 6 gb of ram back when most came with 4 (2010) and thought that the $150 more was a huge upgrade in performance! turns out that the Pentium cant push the amount of programs 6 gb of ram can fill up but that's another story.

amd marketing went for the majority of people who look at simple specs and said WOW it has twice the cores of an intel i7! it must be so fast! I bet it poops on intel! and from that perspective it worked great.

The marketing failed because they cherry picked every benchmark. They only compared the FX series to CPUs when it won.

Most people know Intel, very few normal consumers know AMD. The marketing team was trying to focus on the enthusiasts but they treated us like idiots as if we were going to take their benchmarks as the proper benchmarks.

Point is, AMD markets their non 4c/8t CPU as a 8 core CPU. Intel markets their 4c/8t CPU as a 4c/8t CPU. Hell Intel even uses Hyperthreading in their marketing. AMD could have done the same but their marketing team for BD was trying to pull the wool over the eyes of the consumer. There is a reason they got fired after all.

I highly doubt AMD will allow the marketing team to market Zen in any other way than what it truly is, so far we have "Up to 8 cores with 16 threads". No word on if that is on the consumer end or server end and what other configurations that will come in.

8c/16t will be the flagship consumer part...servers will likely offer somewhere in the neighborhood of double that. Whether they do it with MCM or monolithic die will remain to be seen until launch...I would expect MCM, though I hope for a monolithic 16c die.

 

[vmN]

Not strictly AMD related, but I'm posting because of the implications:

http://www.anandtech.com/show/9659/fable-legends-directx-12-benchmark-analysis/2

Notice the i3 versus i5 versus i7 numbers at 4k. They're exactly the same. It looks like, as I predicted, DX12 is going to be very CPU agnostic, which means that cheap dual cores may be the way to go for cheap gaming rigs going forward. It looks like my prediction DX12 was going to be a major benefit to the Core i3 lineup is going true.

Lower the res to 720, and the CPU again matters (GPU not bottleneck anymore), but at least at higher resolutions, it doesn't look like the CPU is a major factor going forward, since the GPU remains the bottleneck.

The benchmark is limited to showing off landscape isn't it?
So there might be some more CPU activity in the actual gameplay.

Haswell or newer I3 are doable gaming processors no doubt!

 

[gamerk316]

But here's the concern: If Intel undercuts AMD with a lower-power lower-cost i3 chip, they will continue to hold that market tight. Likewise, if there's no performance advantage to 8/16 from, say, 4/8 or even 2/4, how is AMD expecting to sell their highest-margin products?

See the problem? AMD is making a performance CPU at the same time new APIs (that AMD ironically helped spur the development of) are lessening the CPU impact in gaming. With Mantle, AMD may have accidentally ensured Zen's commercial failure.

 

[vmN]

But here's the concern: If Intel undercuts AMD with a lower-power lower-cost i3 chip, they will continue to hold that market tight. Likewise, if there's no performance advantage to 8/16 from, say, 4/8 or even 2/4, how is AMD expecting to sell their highest-margin products?

See the problem? AMD is making a performance CPU at the same time new APIs (that AMD ironically helped spur the development of) are lessening the CPU impact in gaming. With Mantle, AMD may have accidentally ensured Zen's commercial failure.

I dont think the majority off Intel extreme CPU is using it only for gaming (or atleast not its sole purpose). Certainly there are some, but even today there are minor gains past 4 cores on an Intel processor in gaming.

 

[juanrga]

Some guys in the whole internet, also here, have suggested that Keller left AMD because the work on Zen is done. They draw parallelism with K8. They claim that Keller left AMD after doing K8 work and much before AMD did launch K8.

AMD has also tried to sell that myth. Drew Prairie, director of corporate communications at AMD, said:

Architecture decisions are many years in advance of products. As you may remember, he was a primary architect on K8. We launched that product in 2003 and he had left AMD in 1999.

http://www.kitguru.net/components/cpu/anton-shilov/legendary-microprocessor-developer-jim-keller-leaves-amd/

Well, this is what Kanter said me, when discussing Keller's departure:

Keller was never the guy behind the K8. He was the chief architect early on, but left AMD well before the project converged on a design. Dirk Meyer, Fred Weber, Kevin McGrath, and a couple of others ended up picking the pieces. I think the original idea was for something that was fairly different from the K7...but due to time-to-market and the disappearance of the original chief architect, they scaled back their ambitions.

We got a K7-like architecture with 64-bit, on-die memory controller, and Hypertransport (obviously, there were other changes, but those were the big ones).

Thus I extend my hypothesis (no theory) to account for this info. My hypothesis is that Keller left AMD due to discrepancies with Lisa Su due to cancellation/delays of projects for which he was hired by Rory Read. And that the Zen that will be released by AMD is different than the Zen that Keller initially developed.

It could be a quasi-repetition of the K8 issue again! Keller is hired to do something. There are discrepancies, then he left AMD and engineers that remain at AMD finish a different design is less ambitious than the one originally made by Keller.

This hypothesis could explain why Keller left before the design of Zen is finished. It is not his baby! It could explain why AMD only "expects" Keller's departure will not affect products/roadmaps. It could explain the rough departure and the lack of replacement. It coiuld explain why AMD gives the press a photo of the Zen team without Keller. It could explain why Suzzane didn't mention him a single time. It could also explain the discrepancies between the performance claimed by Papermaster at FAD 2015 for Zen and the performance claimed by Keller himself for his Zen [1].

It is a hypothesis, but the irony here is that all people has been saying that Keller leaving AMD again was a copy of what happened when Keller left AMD in 1999, could be correct, but just by the contrary reasons that they thought!

[1] Sure some of you recall the 100% higher IPC figure that I posted in the old thread. That figure was coming from Keller himself!

 

[-Fran-]

Some guys in the whole internet, also here, have suggested that Keller left AMD because the work on Zen is done. They draw parallelism with K8. They claim that Keller left AMD after doing K8 work and much before AMD did launch K8.

AMD has also tried to sell that myth. Drew Prairie, director of corporate communications at AMD, said:

Architecture decisions are many years in advance of products. As you may remember, he was a primary architect on K8. We launched that product in 2003 and he had left AMD in 1999.

http://www.kitguru.net/components/cpu/anton-shilov/legendary-microprocessor-developer-jim-keller-leaves-amd/

Well, this is what Kanter said me, when discussing Keller's departure:

Keller was never the guy behind the K8. He was the chief architect early on, but left AMD well before the project converged on a design. Dirk Meyer, Fred Weber, Kevin McGrath, and a couple of others ended up picking the pieces. I think the original idea was for something that was fairly different from the K7...but due to time-to-market and the disappearance of the original chief architect, they scaled back their ambitions.

We got a K7-like architecture with 64-bit, on-die memory controller, and Hypertransport (obviously, there were other changes, but those were the big ones).

Thus I extend my hypothesis (no theory) to account for this info. My hypothesis is that Keller left AMD due to discrepancies with Lisa Su due to cancellation/delays of projects for which he was hired by Rory Read. And that the Zen that will be released by AMD is different than the Zen that Keller initially developed.

It could be a quasi-repetition of the K8 issue again! Keller is hired to do something. There are discrepancies, then he left AMD and engineers that remain at AMD finish a different design is less ambitious than the one originally made by Keller.

This hypothesis could explain why Keller left before the design of Zen is finished. It is not his baby! It could explain why AMD only "expects" Keller's departure will not affect products/roadmaps. It could explain the rough departure and the lack of replacement. It coiuld explain why AMD gives the press a photo of the Zen team without Keller. It could explain why Suzzane didn't mention him a single time. It could also explain the discrepancies between the performance claimed by Papermaster at FAD 2015 for Zen and the performance claimed by Keller himself for his Zen [1].

It is a hypothesis, but the irony here is that all people has been saying that Keller leaving AMD again was a copy of what happened when Keller left AMD in 1999, could be correct, but just by the contrary reasons that they thought!

[1] Sure some of you recall the 100% higher IPC figure that I posted in the old thread. That figure was coming from Keller himself!

So, it is even less relevant that he left 😛

Also, thanks for finding that information, but the meat of the subject here is that Keller, being the good architect he is, just called the external design shots, which would make sense to a certain degree given the position he had. Then that implies, the reason why he left is just chitchat, girl talk and/or gossip, and not worthy of discussion anymore. We can close that topic for good then, isn't it? 😛

Cheers!

 

[8350rocks]

Some guys in the whole internet, also here, have suggested that Keller left AMD because the work on Zen is done. They draw parallelism with K8. They claim that Keller left AMD after doing K8 work and much before AMD did launch K8.

AMD has also tried to sell that myth. Drew Prairie, director of corporate communications at AMD, said:

Architecture decisions are many years in advance of products. As you may remember, he was a primary architect on K8. We launched that product in 2003 and he had left AMD in 1999.

http://www.kitguru.net/components/cpu/anton-shilov/legendary-microprocessor-developer-jim-keller-leaves-amd/

Well, this is what Kanter said me, when discussing Keller's departure:

Keller was never the guy behind the K8. He was the chief architect early on, but left AMD well before the project converged on a design. Dirk Meyer, Fred Weber, Kevin McGrath, and a couple of others ended up picking the pieces. I think the original idea was for something that was fairly different from the K7...but due to time-to-market and the disappearance of the original chief architect, they scaled back their ambitions.

We got a K7-like architecture with 64-bit, on-die memory controller, and Hypertransport (obviously, there were other changes, but those were the big ones).

Thus I extend my hypothesis (no theory) to account for this info. My hypothesis is that Keller left AMD due to discrepancies with Lisa Su due to cancellation/delays of projects for which he was hired by Rory Read. And that the Zen that will be released by AMD is different than the Zen that Keller initially developed.

It could be a quasi-repetition of the K8 issue again! Keller is hired to do something. There are discrepancies, then he left AMD and engineers that remain at AMD finish a different design is less ambitious than the one originally made by Keller.

This hypothesis could explain why Keller left before the design of Zen is finished. It is not his baby! It could explain why AMD only "expects" Keller's departure will not affect products/roadmaps. It could explain the rough departure and the lack of replacement. It coiuld explain why AMD gives the press a photo of the Zen team without Keller. It could explain why Suzzane didn't mention him a single time. It could also explain the discrepancies between the performance claimed by Papermaster at FAD 2015 for Zen and the performance claimed by Keller himself for his Zen [1].

It is a hypothesis, but the irony here is that all people has been saying that Keller leaving AMD again was a copy of what happened when Keller left AMD in 1999, could be correct, but just by the contrary reasons that they thought!

[1] Sure some of you recall the 100% higher IPC figure that I posted in the old thread. That figure was coming from Keller himself!

I am going to squash this pretty quickly here.

Having had the opportunity to discuss this in depth with someone I trust...Keller never worked on any of the excavator designs, those changes were delegated to a team handling APUs and he rubber stamped them. In fact, when he came in, it was his call to outright scrap everything past PD on HEDT and just do APUs because he wanted to focus the R&D budget on Zen/K12 for HEDT/Server. The executives agreed, as the best margins are in servers anyway, and even a 5% uptick in server sales would bring windfall revenue to AMD.

The actual designs behind Zen/K12 and, more specifically the cores and "uncore", were all well designed out, thoroughly tested in simulations, and settled on as the final course of action. Keller left because he felt confident that the engineers left behind could refine the initial design, and his passion is the creation of new architectures...not refining his own existing designs.

As much as AMD would love to have him do another uarch (and they may when that time once again comes...), at this point they cannot guarantee him that opportunity in the near future.

 

[Cazalan]

If Zen is close to their targets then they can ride it with tweaks for 4+ years. There is no immediate need to design a brand new core.

 

[Cazalan]

But here's the concern: If Intel undercuts AMD with a lower-power lower-cost i3 chip, they will continue to hold that market tight. Likewise, if there's no performance advantage to 8/16 from, say, 4/8 or even 2/4, how is AMD expecting to sell their highest-margin products?

See the problem? AMD is making a performance CPU at the same time new APIs (that AMD ironically helped spur the development of) are lessening the CPU impact in gaming. With Mantle, AMD may have accidentally ensured Zen's commercial failure.

The benchmark in its current state is mostly the environment. There is no gameplay/AI/units yet so it is not indicative of final performance. As others said it is a "grass simulator" at this point.

 

[salgado18]

Both juanrga and 8350rocks make valid points, and both seem to have good personal sources. But none of you can validade the exact reason he left, just get clues from information both of you get elsewhere. I even think that both facts align with each other, but the leaving AMD part is just speculation.

Having said that, I think that him not working on K8 in the past doesn't exclude him working on Zen, especially since the company (supposedly) learned that he is a talent that shines when he follows his own rules. Then, either he was responsible for the entire design of Zen, or he started it before being swaped to K12.