AMD CPUs, SoC Rumors and Speculations Temp. thread 2

[jimmysmitty]

I have seen FX8350 versus Zen ST cinebench scores...Zen is 50% higher than 8350.

Let us assume for an instant that is true. Take the next graph

multiply the FX-8350 score and you obtain 144 points, which puts Zen just where many of us predicted or expected years ago. Where are your promises now?

Forget haswell, on par with skylake is coming...

JK says they will be better by skylake. My source says even if they fall short of predictions they will be "on par".

Jim Keller has had tons of time in the industry working on projects since K7...if he says it will be better, I have no reason to doubt. Though, like those I know at AMD, I consider it a win if they are within 5%.

Jim Keller said, "AMD are on track to catch Intel in high performance cores"

I said: Jim Keller expects them to be ahead by skylake, my source says he (conservatively) expects they will be about even by skylake, though he expected there would be within 5% difference in some workloads just because of differences in designs, and that would be fine...

I would be absolutely fine with scores in line with the 4960x or 5960x, especially considering the fact that the higher intel chips are there because of their increased clock speeds.

That is probably going to fall in line with Broadwell-E 8 core parts as well...which is just fine, too.

The 4790K yes but the 4770K has a lower base and turbo clock speed than the 4960X, 3.5GHz base 3.9GHz boost vs 3.6GHz base 4GHz boost.

As well the cache speed of the i7 4960X is much faster than a i7 4770K so in all reality the 4770K is at a disadvantage short of its better process (32nm vs 22nm) and newer uArch with IPC improvements, which is pretty obvious.

Since the 4960X is a Ivy Bridge-E part we can assume that if Zen is matching it then it is where a lot of predictions were, which is a CPU that matches or beats out Sandy Bridge by a bit.

A lot of this is all conjecture mind you but I would hope for a Haswell competitive part at least.

 

[cdrkf]

I have seen FX8350 versus Zen ST cinebench scores...Zen is 50% higher than 8350.

Let us assume for an instant that is true. Take the next graph

multiply the FX-8350 score and you obtain 144 points, which puts Zen just where many of us predicted or expected years ago. Where are your promises now?

Forget haswell, on par with skylake is coming...

JK says they will be better by skylake. My source says even if they fall short of predictions they will be "on par".

Jim Keller has had tons of time in the industry working on projects since K7...if he says it will be better, I have no reason to doubt. Though, like those I know at AMD, I consider it a win if they are within 5%.

Jim Keller said, "AMD are on track to catch Intel in high performance cores"

I said: Jim Keller expects them to be ahead by skylake, my source says he (conservatively) expects they will be about even by skylake, though he expected there would be within 5% difference in some workloads just because of differences in designs, and that would be fine...

I would be absolutely fine with scores in line with the 4960x or 5960x, especially considering the fact that the higher intel chips are there because of their increased clock speeds.

That is probably going to fall in line with Broadwell-E 8 core parts as well...which is just fine, too.

The 4790K yes but the 4770K has a lower base and turbo clock speed than the 4960X, 3.5GHz base 3.9GHz boost vs 3.6GHz base 4GHz boost.

As well the cache speed of the i7 4960X is much faster than a i7 4770K so in all reality the 4770K is at a disadvantage short of its better process (32nm vs 22nm) and newer uArch with IPC improvements, which is pretty obvious.

Since the 4960X is a Ivy Bridge-E part we can assume that if Zen is matching it then it is where a lot of predictions were, which is a CPU that matches or beats out Sandy Bridge by a bit.

A lot of this is all conjecture mind you but I would hope for a Haswell competitive part at least.

Yeah but remember if current info proves true zen would be operating at lower clocks than the Ivy based parts. So it could be Haswell + ipc, but lower clock speeds. That wouldn't be the worst situation ever, as clock speeds are something relative straight forward to improve generation to generation, and process improvements are likely to help. They would be in a much better position imo having a very strong IPC part and having to focus on improving clocks, than the position they were in with BD with high clocks but poor efficiency.

 

[jimmysmitty]

Problem is in the process then. How high of clocks can they get if what they are getting now and it is something they have been working on for a while?

I would assume that a process change would take place for Zen+ to get better clocks than Zen.

Again everything is conjecture and will only be proven true or false in, well who knows when.

 

[juanrga]

I have seen FX8350 versus Zen ST cinebench scores...Zen is 50% higher than 8350.

Let us assume for an instant that is true. Take the next graph

multiply the FX-8350 score and you obtain 144 points, which puts Zen just where many of us predicted or expected years ago. Where are your promises now?

Forget haswell, on par with skylake is coming...

JK says they will be better by skylake. My source says even if they fall short of predictions they will be "on par".

Jim Keller has had tons of time in the industry working on projects since K7...if he says it will be better, I have no reason to doubt. Though, like those I know at AMD, I consider it a win if they are within 5%.

Jim Keller said, "AMD are on track to catch Intel in high performance cores"

I said: Jim Keller expects them to be ahead by skylake, my source says he (conservatively) expects they will be about even by skylake, though he expected there would be within 5% difference in some workloads just because of differences in designs, and that would be fine...

Well assuming that is true (Zen as released will score 144 points in ST, and at relatively modest clock speeds if the info we're getting are accurate), then Zen will be neigh on level pegging with Intel for IPC, albeit at a bit of clock speed disadvantage. I mean the 4790k gets that score purely via higher clock rate, and we also know Skylake scores barely higher than Haswell. Whichever way you slice it that means Zen is a *huge* improvement from a competitive standpoint than at any point during the 'dozer' generation.

I personally would consider a zen ST cinebench score of 144 a big win, certainly anywhere in the 140+ region is fine. Assuming that kind of gain holds true across other single thread applications AMD should actually be a decent position (as they stand a chance of really challenging Intel in many of the MT benchmarks- certainly a 6 core Zen part would be considerably faster overall than intel's quad core i7s).

Lets hope this is the case (I'd argue this is probably best case, even an ST score in the 120 - 130 range would be an abnormally large generation jump in single thread performance). If it is- then going Intel or AMD would be an actual choice for enthusiasts and professional system builders again. That is what we all need- no AMD won't be dominant aka the A64 days, but they will be close enough that those without a massive bias will actually have cause to consider them outside the few niche areas they are competing in today.

1)
Therefore all of us that said Zen would be between Ivy Bridge and Haswell were right. Whereas the guy with "inside" info, the guy with "contacts" at AMD, the guy that said "forget Haswell Zen will be faster than skylake because Jim Keller said so" was wrong. But recall all the info he has been posting about Zen (AVX512, 2x256bit FMAC, "redesigned CMT", 3AGUs,...) was false.

2)
Let us assume that the 144 points are accurate instead another false claim. We are comparing a still inexistent 14nm chip to older 22nm chips. 144 is just the score of the Ivy Bridge-E chip. Whereas Zen is a huge improvement over Piledriver and company, Zen is not enough for AMD to get significant market share and money back.

3)
Skylake DT scores barely higher than Haswell DT when you compare normal Skylake with Haswell refresh, because the Devyls Canyon chips did hit higher frequencies from a mature node. Skylake refresh will increase clocks again.

4)
The natural competitor for Zen 8C/16T is Broadwell-E and Skylake-E. And both increase clocks, IPC, and number of cores. The 8C/16T Broadwell-E would be about 10% faster than the 8C/16T Haswell-E.

5)
CB15 must be obtaining a huge amount of performance from Zen, but it is barely extresing the execution units of Haswell and latter microarchitectures. In CB15 Haswell is barely faster than Ivy because the software is using half of the 256bit units in Haswell. When you recompile for Haswell and use the new instructions then Haswell is ~70% faster than Ivy at same clocks. And Skylake will bring another 70% with the jump to 512bit units.

https://www.pugetsystems.com/labs/articles/Haswell-Floating-Point-Performance-493/

Of course, 99% of desktop users will not benefit from anything of that, but workstation, server, HPC people do.

6)
I agree with you Zen is a huge improvement. It is very much what I have been saying during last two years. What I was a bit tired was of all the hype and nonsense about Zen. Faster than Skylake, the new K8, Intel engineers are morons, only Keller knows how to design a chip, and so on, day after day after day during two years.

 

[juanrga]

I have seen FX8350 versus Zen ST cinebench scores...Zen is 50% higher than 8350.

Let us assume for an instant that is true. Take the next graph

multiply the FX-8350 score and you obtain 144 points, which puts Zen just where many of us predicted or expected years ago. Where are your promises now?

Forget haswell, on par with skylake is coming...

JK says they will be better by skylake. My source says even if they fall short of predictions they will be "on par".

Jim Keller has had tons of time in the industry working on projects since K7...if he says it will be better, I have no reason to doubt. Though, like those I know at AMD, I consider it a win if they are within 5%.

Jim Keller said, "AMD are on track to catch Intel in high performance cores"

I said: Jim Keller expects them to be ahead by skylake, my source says he (conservatively) expects they will be about even by skylake, though he expected there would be within 5% difference in some workloads just because of differences in designs, and that would be fine...

I would be absolutely fine with scores in line with the 4960x or 5960x, especially considering the fact that the higher intel chips are there because of their increased clock speeds.

That is probably going to fall in line with Broadwell-E 8 core parts as well...which is just fine, too.

Yes yes, the previous "forget Haswell faster than Skylake is coming" now changes to "in line with Broadwell-E 8 core", despite your own data (assuming it is real) shows a 14nm Zen performing like 22nm Ivy Bridge-E on a score is obviously favoring AMD.

Broadwell-E increases IPC, clocks, and number of cores compared to Haswell-E. I am 100% convinced I will see how an i7-6950X runs circles around a 8C/16T Zen, despite all the hype about Keller.

 

[-Fran-]

Yes yes, the previous "forget Haswell faster than Skylake is coming" now changes to "in line with Broadwell-E 8 core", despite your own data (assuming it is real) shows a 14nm Zen performing like 22nm Ivy Bridge-E on a score is obviously favoring AMD.

Broadwell-E increases IPC, clocks, and number of cores compared to Haswell-E. I am 100% convinced I will see how an i7-6950X runs circles around a 8C/16T Zen, despite all the hype about Keller.

You are full of poison today, Juan. Bad day?

Also, even if it's not at the 6950X levels (which was kind of disappointing IMO), given the correct pricing, it will mean squat.

I don't understand why you're singling out 8350, when there's been different speculation tendencies from each front. Some wilder than others, but at least inside this thread I haven't read anything that can be considered "wild", including your own (which, I still don't know where you get your sources from).

Cheers!

 

[juanrga]

No problem with personal speculation. But problem with certain people that claims to have information from inside AMD, and it is shown latter that the information posted was false. This has been happening during more than two years for anything AMD related.

To be clear. One thing is someone speculating if Zen will have AVX512 support or not. Another thing is someone claiming in your face that AMD said him that Zen supports AVX512 and ten days latter the Zen patch arises in GCC and there is no support for AVX512 on Zen, and he silently moves over that and makes another claim about Zen "AMD said me", which is latter shown to be false and then moves again and makes a new bogus claim "AMD promised me"...

 

[-Fran-]

No problem with personal speculation. But problem with certain people that claims to have information from inside AMD, and it is shown latter that the information posted was false. This has been happening during more than two years for anything AMD related.

To be clear. One thing is someone speculating if Zen will have AVX512 support or not. Another thing is someone claiming in your face that AMD said him that Zen supports AVX512 and ten days latter the Zen patch arises in GCC and there is no support for AVX512 on Zen, and he silently moves over that and makes another claim about Zen "AMD said me", which is latter shown to be false and then moves again and makes a new bogus claim "AMD promised me"...

Ok, fair enough. Thanks for taking the time to explain.

I would like to point out that, even if "AMD tells" someone, that implies underneath a person working from AMD, which may or may not actually know what he/she is talking about, is spreading false/incorrect/inaccurate information. Personally, I don't doubt 8350s good will on communicating what he is being told, but it's a shame he might be being fed incorrect information (as you claim). That can happen to anyone, I'd say. So, please, cut him some slack on this one and let's go back to our good old speculating/rumoring 😛

All that being said, from the press release, I can confidently say 2 things. AMD has faith in Zen, even if they are not making a huge display of that in public. My image from Miss Lisa and Mr Raja is they are not sales people like Jen and their enthusiasm is harder to read. This is so subjective it's hard to make any tangible conclusion out of it, but it at least shows a bit more light at the end of the tunnel. And the other one, butter hands Raja Koduri is a funny unfunny-man.

Anyone else noticed something similar or different?

Cheers!

EDIT: Forgot the second point XD

 

[gamerk316]

The buildup to the Zen launch honestly looks a lot like the buildup BD had, so I say it's more a function of AMDs PR department then anything technical about the chip. Still predicting Haswell level performance.

 

[jdwii]

The buildup to the Zen launch honestly looks a lot like the buildup BD had, so I say it's more a function of AMDs PR department then anything technical about the chip. Still predicting Haswell level performance.

Boy oh boy not nearly as bad as that, i know i expect sandy-ivy bridge single threaded performance even made a bet about it(awhile back), the real question is price i have my guesses but i wouldn't expect the 8 core to be more than an baseline 6 core Intel CPU. I also do not expect 250$ 8 core zen CPUs at launch either unless they are severally low binned.

 

[8350rocks]

No problem with personal speculation. But problem with certain people that claims to have information from inside AMD, and it is shown latter that the information posted was false. This has been happening during more than two years for anything AMD related.

To be clear. One thing is someone speculating if Zen will have AVX512 support or not. Another thing is someone claiming in your face that AMD said him that Zen supports AVX512 and ten days latter the Zen patch arises in GCC and there is no support for AVX512 on Zen, and he silently moves over that and makes another claim about Zen "AMD said me", which is latter shown to be false and then moves again and makes a new bogus claim "AMD promised me"...

You are taking things that were said back in 2014-2015, and you are overlooking words like "expectations", and "should". Some of those things were said before AMD even had their hands on the process.

You also forget that I had the codename for the new chip when you were still calling K12. I also knew that they were not going to use GloFo's process before the announcement came from Samsung. I knew this chip was going to be on 14nm finfet.

The speculation about 22nm FDSOI was before I was told that Zen was moving away from SOI completely.

Additionally, expectations can be tempered as a project goes on. One thing is certain though, I was told what expectations were, and those can change.

As for performance, I expect that the IPC will be in line, once the process matures, and they make a few core advancements, it will be competitive. The margins should be quite good as well, even if they only get $350-400 per chip for the flagship part in consumer space. This will give them a huge push back into servers.

 

[-Fran-]

I remember when the Phenom II 965C came out, it was priced at 400-ish, not being the top performer of the group (QX9850 or something at the time?). I would expect, given corrected for inflation, that Zen would be 550-ish at launch. If it performs in line with Haswell, then it might even be priced above the i7 6700K.

I know my numbers are way high, but if they are aiming the -E series, they will charge similar prices.

Cheers!

 

[salgado18]

I remember when the Phenom II 965C came out, it was priced at 400-ish, not being the top performer of the group (QX9850 or something at the time?). I would expect, given corrected for inflation, that Zen would be 550-ish at launch. If it performs in line with Haswell, then it might even be priced above the i7 6700K.

I know my numbers are way high, but if they are aiming the -E series, they will charge similar prices.

Cheers!

I bet they will price it where each processor matches Intels in performance, then lower by a notch to be competitive. If the top 8-core Zen matches the top 6-core Broadwell-E in performance (which is priced at $617), then it will probably cost around $600. If it stays between 6-core and 8-core B-E, priced $1089 and $617 respectively, it will cost around $800, and so on.

I believe that's the strategy AMD will use, position themselves where the competition is at. But I don't believe they will fight too much for price, like go $50 lower to $550 or like that.

Source of prices and other stuff for Broadwell-E:
http://www.tomshardware.com/reviews/intel-core-i7-broadwell-e-6950x-6900k-6850k-6800k,4587.html

 

[juanrga]

No problem with personal speculation. But problem with certain people that claims to have information from inside AMD, and it is shown latter that the information posted was false. This has been happening during more than two years for anything AMD related.

To be clear. One thing is someone speculating if Zen will have AVX512 support or not. Another thing is someone claiming in your face that AMD said him that Zen supports AVX512 and ten days latter the Zen patch arises in GCC and there is no support for AVX512 on Zen, and he silently moves over that and makes another claim about Zen "AMD said me", which is latter shown to be false and then moves again and makes a new bogus claim "AMD promised me"...

Ok, fair enough. Thanks for taking the time to explain.

I would like to point out that, even if "AMD tells" someone, that implies underneath a person working from AMD, which may or may not actually know what he/she is talking about, is spreading false/incorrect/inaccurate information.

Are you aware of the seriousness of your comments? If someone at AMD has spread misinformation about their future products this is motive for a lawsuit. The excuse of "I didn't know enough" or "I was inaccurate", doesn't exempt from legal responsibilities. On the other case, if they weren't the source of the misinformation, they could sue you for defamation.

 

[jimmysmitty]

The buildup to the Zen launch honestly looks a lot like the buildup BD had, so I say it's more a function of AMDs PR department then anything technical about the chip. Still predicting Haswell level performance.

I wouldn't say that bad. And I hope AMD fired their entire PR department and got half smart people because yea, BDs PR and marketing was just bad. From poor comparisons (they would only cherry pick and compare what looked good) to just bad pricing and information in general.

Haswell equivalent performance is a BARE minimum Zen has to meet to be competitive, in my eyes anyways.

The buildup to the Zen launch honestly looks a lot like the buildup BD had, so I say it's more a function of AMDs PR department then anything technical about the chip. Still predicting Haswell level performance.

Boy oh boy not nearly as bad as that, i know i expect sandy-ivy bridge single threaded performance even made a bet about it(awhile back), the real question is price i have my guesses but i wouldn't expect the 8 core to be more than an baseline 6 core Intel CPU. I also do not expect 250$ 8 core zen CPUs at launch either unless they are severally low binned.

A $250 8 core would be near impossible unless it was just THAT bad and I doubt they could be as bad as BD.

 

[juanrga]

You are taking things that were said back in 2014-2015, and you are overlooking words like "expectations", and "should". Some of those things were said before AMD even had their hands on the process.

We can discern between genuine expectations and nonsense. That is the reason why all what you wrote is disproved.

You also forget that I had the codename for the new chip when you were still calling K12. I also knew that they were not going to use GloFo's process before the announcement came from Samsung. I knew this chip was going to be on 14nm finfet.

The speculation about 22nm FDSOI was before I was told that Zen was moving away from SOI completely.

Nice attempt to rewrite the history! You were during days disputing that that K12 wasn't ARM. I corrected you multiple times. Moreover, I was the first one that leaked the codename Zen for the x86 chip here. You asked me from where I got the codename. Weeks latter Rory Read announced Zen to public audiences.

You said that they would use SOI. I explained you why not. I explained you why would use FinFET and why FinFET was a superior technology to SOI. I gave arguments, formulae and graphs. You negated all, did repeat the same hype abut SOI, and pretended I was completely ignorant about process nodes.

Latter AMD confirmed the use of 14nm FINFET.

Additionally, expectations can be tempered as a project goes on. One thing is certain though, I was told what expectations were, and those can change.

We already knew those weren't genuine expectations, but baseless exaggerations or just plain nonsense, because we have some knowledge of physics, tech and economy. This is the reason why we are right whereas all those 'expectations' have gone to the basket.

I am not going to publish all the misinformation you have posted since I joined this forum before Steamroller. I will only focus on what you said about Zen:

FDSOI process ---> false
4.0--4.1GHz --> false
4ALU+3AGU--> false
2x256bit FMAC --> false
"faster than skylake" --> false
"redesigned CMT architecture" --> false
12 or more cores for desktop --> false
AVX-512 support --> false
Bigger than 8mm² -->false
.....

During years you are posting misinformation about anything related to AMD. You claim the information is coming from "inside AMD". You even shared with us pics with John Byrne. We analyze your information and find that it is false. We explain why. You attack us by saying the truth. Official information is finally published and confirms what we said, your information was false. Then you suddenly forget what you said, whereas pretend to spin history of forum by 180º, not only ignoring what you said, but pretending to attribute to you predictions done by the rest of us.

Some time passes and the cycle start again with another new bach of false information supposedly from "inside AMD"
What I don't understand is why mods allow you to do this. Bye!

I will finish with a quote from a poster has gone:

MD, Juan and 8350Rocks just keep throwing predictions, the only difference between them is that Juan smear it into everyone face's when he's right, and he usually is... while 8350Rocks seems to fail most of his predictions due to his AMD insider constantly giving him massively wrong info.

 

[ComputerSecurityGuy]

Considering that Zen looks to be on par with Haswell or slightly faster from what AMD says and we indicate, I might expect a six core chip at the $300 and possibly higher price range. It depends on where the Broadwell-E chips settle really, because Zen is probably going to be directly competing with Broadwell-E- or maybe Skylake-E.

 

[-Fran-]

Are you aware of the seriousness of your comments? If someone at AMD has spread misinformation about their future products this is motive for a lawsuit. The excuse of "I didn't know enough" or "I was inaccurate", doesn't exempt from legal responsibilities. On the other case, if they weren't the source of the misinformation, they could sue you for defamation.

Have you ever read the TV when they put on screen "the opinions invested in this show might not reflect what the channel believes in or thinks"? I meant it like that. Plus, there is the whole thing about keeping your sources secret and taking everything with a grain of salt. Miss Lisa Su could be sending me emails about their products and be wrong about them after a meeting they told her "no, we can't really do that". That is why roadmap change and all that, Juan. You're taking the point *way* too seriously, lol.

The only things you can believe fully are official statements. And even then, there is room for doubt!

In any case, I'll drop it now. I suggest you also drop the "he said, I said". It's becoming annoying and doesn't add anything positive to the discussion.

Cheers!

 

[juanrga]

Have you ever read the TV when they put on screen "the opinions invested in this show might not reflect what the channel believes in or thinks"? I meant it like that. Plus, there is the whole thing about keeping your sources secret and taking everything with a grain of salt. Miss Lisa Su could be sending me emails about their products and be wrong about them after a meeting they told her "no, we can't really do that". That is why roadmap change and all that, Juan. You're taking the point *way* too seriously, lol.

The only things you can believe fully are official statements. And even then, there is room for doubt!

In any case, I'll drop it now. I suggest you also drop the "he said, I said". It's becoming annoying and doesn't add anything positive to the discussion.

Cheers!

That wasn't my point! Since you use the example of the roadmap I will use that as illustration of which is my point. One thing is when a company (Intel, AMD, Nvidia, IBM, Apple,...) presents officially a roadmap about releasing a 10nm chip in 2018 and adds to that roadmap a footnote disclaimer that says that roadmaps can change without notice. They add this disclaimer because they depend on the Foundry roadmap. If the foundry expectations fail and the process node is delayed six months then the chips will be delayed. This kind of uncertainty is acceptable and it is not my point.

My point is another. My point is what happens when someone claims to have "inside info" from company X and claims as a "fact" that company X will release a 4nm chip the next year that will have 50% higher IPC than the faster chip and that will be clocked at 5.5GHz base. Everyone with some knowledge of foundries roadmaps, of process node technology, and of microarchitecture know that the "fact" is just a lie because 4nm technology will be not ready next year, and the rest is pure nonsense. My point is about the guys that are spreading misinformation and nonsense on the forums but pretending to sell them as "facts" from "inside AMD".

Any bet in some few months we will reading again another batch of misinformation, this time about Zen+, from the same individual? I am advising in advance.

P.S.: The "I said you said" is needed when one person is trying to rewrite the history of the forum.

 

[juanrga]

Considering that Zen looks to be on par with Haswell or slightly faster from what AMD says and we indicate, I might expect a six core chip at the $300 and possibly higher price range. It depends on where the Broadwell-E chips settle really, because Zen is probably going to be directly competing with Broadwell-E- or maybe Skylake-E.

We know that Zen will be slower than Haswell. We know from analysis of both microarchitectures and process nodes. The unknown is the pricing policy. Will AMD sell 8C Zen at 6C Intel prices or will target Intel 8C? I expect the first option, but don't reject that AMD can push the second. Anyone here recall the initial pricing of the FX-9000 series?

 

[8350rocks]

You are taking things that were said back in 2014-2015, and you are overlooking words like "expectations", and "should". Some of those things were said before AMD even had their hands on the process.

We can discern between genuine expectations and nonsense. That is the reason why all what you wrote is disproved.

You also forget that I had the codename for the new chip when you were still calling K12. I also knew that they were not going to use GloFo's process before the announcement came from Samsung. I knew this chip was going to be on 14nm finfet.

The speculation about 22nm FDSOI was before I was told that Zen was moving away from SOI completely.

Nice attempt to rewrite the history! You were during days disputing that that K12 wasn't ARM. I corrected you multiple times. Moreover, I was the first one that leaked the codename Zen for the x86 chip here. You asked me from where I got the codename. Weeks latter Rory Read announced Zen to public audiences.

You said that they would use SOI. I explained you why not. I explained you why would use FinFET and why FinFET was a superior technology to SOI. I gave arguments, formulae and graphs. You negated all, did repeat the same hype abut SOI, and pretended I was completely ignorant about process nodes.

Latter AMD confirmed the use of 14nm FINFET.

Additionally, expectations can be tempered as a project goes on. One thing is certain though, I was told what expectations were, and those can change.

We already knew those weren't genuine expectations, but baseless exaggerations or just plain nonsense, because we have some knowledge of physics, tech and economy. This is the reason why we are right whereas all those 'expectations' have gone to the basket.

I am not going to publish all the misinformation you have posted since I joined this forum before Steamroller. I will only focus on what you said about Zen:

FDSOI process ---> false
4.0--4.1GHz --> false
4ALU+3AGU--> false
2x256bit FMAC --> false
"faster than skylake" --> false
"redesigned CMT architecture" --> false
12 or more cores for desktop --> false
AVX-512 support --> false
Bigger than 8mm² -->false
.....

During years you are posting misinformation about anything related to AMD. You claim the information is coming from "inside AMD". You even shared with us pics with John Byrne. We analyze your information and find that it is false. We explain why. You attack us by saying the truth. Official information is finally published and confirms what we said, your information was false. Then you suddenly forget what you said, whereas pretend to spin history of forum by 180º, not only ignoring what you said, but pretending to attribute to you predictions done by the rest of us.

Some time passes and the cycle start again with another new bach of false information supposedly from "inside AMD"
What I don't understand is why mods allow you to do this. Bye!

I will finish with a quote from a poster has gone:

MD, Juan and 8350Rocks just keep throwing predictions, the only difference between them is that Juan smear it into everyone face's when he's right, and he usually is... while 8350Rocks seems to fail most of his predictions due to his AMD insider constantly giving him massively wrong info.

I said they were aiming for 12 cores in HEDT, but they would not do less than 8. -> right

I said they were aiming for 4 GHz base clock -> right (they were, but are not able to get there afaik)

I said 4 ALU + 2 AGU -> right

I said they were looking to compete with Skylake. -> right

I did initially say that I suspected it may be redesigned CMT, though, I later said that it would be SMT, so you get partial credit.

I did say they "should" use 22nm FD-SOI, and later I was informed that it would be the 14nm finfet samsung process, which I was unable to disclose while under NDA.

I asked you where you heard "Zen" because I was under NDA, and was told I was unable to disclose that information before AMD let it out. So, I inquired where you heard it to determine who might have leaked it, or if it was "official".

 

[USAFRet]

I see 2 things in recent comments on this thread:

1. Opinions. Not facts, because they cannot yet exist for products that do not yet exist
2. Attitudes. Please tone them down

If anyone has a difference of opinion on 1 or 2, you can address them to me.

 

[-Fran-]

We know that Zen will be slower than Haswell. We know from analysis of both microarchitectures and process nodes. The unknown is the pricing policy. Will AMD sell 8C Zen at 6C Intel prices or will target Intel 8C? I expect the first option, but don't reject that AMD can push the second. Anyone here recall the initial pricing of the FX-9000 series?

I would imagine that depends on what they will try to project with Zen. Like... If they put it in "-E" territory, then the expectations on price *and* performance will be there as well. Also, I am curious why they'll release the Consumer versions first (IIRC)... I know they want to get some market share back, but commercially speaking doesn't make much sense... Could it be they already have a lot of harvested dies or something? Maybe they already sampled Zen for some people, but haven't made an official lunch?

Cheers!

 

[juanrga]

We know that Zen will be slower than Haswell. We know from analysis of both microarchitectures and process nodes. The unknown is the pricing policy. Will AMD sell 8C Zen at 6C Intel prices or will target Intel 8C? I expect the first option, but don't reject that AMD can push the second. Anyone here recall the initial pricing of the FX-9000 series?

I would imagine that depends on what they will try to project with Zen. Like... If they put it in "-E" territory, then the expectations on price *and* performance will be there as well. Also, I am curious why they'll release the Consumer versions first (IIRC)... I know they want to get some market share back, but commercially speaking doesn't make much sense... Could it be they already have a lot of harvested dies or something? Maybe they already sampled Zen for some people, but haven't made an official lunch?

The plan seems to be Zen CPU for HEDT and APUs for mainstream users, just as Intel does. Didn't AMD state clearly that octo-core Zen CPU is aimed at the enthusiast market?

AMD has confirmed at last investor conference that the launch schedule is desktop CPU first, server CPU next, and APU last. No confirmation of dates still.

 

[-Fran-]

The plan seems to be Zen CPU for HEDT and APUs for mainstream users, just as Intel does. Didn't AMD state clearly that octo-core Zen CPU is aimed at the enthusiast market?

AMD has confirmed at last investor conference that the launch schedule is desktop CPU first, server CPU next, and APU last. No confirmation of dates still.

Now that you mention it, yes they did mention that. Zen is being "up" marketed towards "enthusiasts". I wonder if that includes "prosumer" market as well... I guess it does.

I'd like to know if they are going to include the same stuff they used for Carrizo to manage thermals. I remember that thing works amazingly. Since now "Turbo" is something we all expect to be inside a CPU, then the more reason to have something even better than Carrizo. I would imagine they do have it, but I haven't seen any marketing slides talking about Turbo... I guess because that would imply talking about speeds as well? Oh welp. Maybe they have and I've missed them.

In any case, the MSRP is going to be the deal breaker for many. I'm not sure I'd pay $500+ for a CPU... I paid $150 for my 2700K and it still is a beastly CPU. Oh man... The future can't come soon enough, haha.

Cheers!