AMD Godavari APUs to be launched in May, says report

[spl]

There's been more than one CPU engineer at Intel who has anonymously expressed their frustration that the top guys at Intel (I don't have a link but Google it) for not investing in R&D on new ideas they have for pushing performance forward because there simply isn't any need to invest the money when they're already in the lead. That would suggest that Intel has a great many ideas up their sleeve to pull out when needed (i.e. when AMD regains the lead).

Course it's entirely possible that these 'anonymous engineers' are in fact PR guys in disguise trying to make people (the public and AMD) think exactly that when the reality is that Intel have nothing. Personally though I'm more inclined to take this at face value (makes sense after all - why spend loads on gaining a huge lead when a fairly small lead is sufficient?).

[Pleiades]

Not my only comment on this article. I've actually contributed something, if you bothered to read. Your comments are (as is typical of you) useless and say nothing. Infact I'm not sure I've ever seen a comment from you that was more than a useless one-liner. Like somebody pointed out, that's your right. But why even bother?

Yes you contributed your usual drivel, as expected. I'm sorry you get so upset about other people's opinions. Perhaps it's best you don't use the internet 'cos you might come across quite a few posts you don't like. Oh, and take your tablets.

On T: I'd love AMD to produce something exciting but that hasn't happened for yonks. Lisa Su is really up against it. Wouldn't be surprised if they were bought out by Samsung or a Middle Eastern SWF next year - and maybe that would be a good thing, having the chance of a massive cash injection for much needed R&D. Just hoping 390 will jumpstart them a bit - waiting to replace my 280x.

Course Chinese companies would be lining up to invest if they were allowed to by the US Govt.

[DanceswithUnix]

There's been more than one CPU engineer at Intel who has anonymously expressed their frustration that the top guys at Intel (I don't have a link but Google it) for not investing in R&D on new ideas they have for pushing performance forward because there simply isn't any need to invest the money when they're already in the lead. That would suggest that Intel has a great many ideas up their sleeve to pull out when needed (i.e. when AMD regains the lead).

Course it's entirely possible that these 'anonymous engineers' are in fact PR guys in disguise trying to make people (the public and AMD) think exactly that when the reality is that Intel have nothing. Personally though I'm more inclined to take this at face value (makes sense after all - why spend loads on gaining a huge lead when a fairly small lead is sufficient?).

That sounds a bit worrying for Intel, the best engineers tend to be confident in their abilities (that is good) but they don't generally whine in public about how they are being held back. If they are any good they hand their notice in and move to somewhere more interesting where they are not held back.

The only Intel architect I have followed with any interest is Andy "crazy" Glew, a P6 architect who used to do some interesting posts in the comp.arch usenet newsgroup around the P6 days. Seems he now works for MIPS: https://www.linkedin.com/in/andyglew

[spl]

Oh, and take your tablets.

You're not as funny as you think you are. Not even close.

Grow up, you silly little child.

[spl]

That sounds a bit worrying for Intel, the best engineers tend to be confident in their abilities (that is good) but they don't generally whine in public about how they are being held back. If they are any good they hand their notice in and move to somewhere more interesting where they are not held back.

The only Intel architect I have followed with any interest is Andy "crazy" Glew, a P6 architect who used to do some interesting posts in the comp.arch usenet newsgroup around the P6 days. Seems he now works for MIPS: https://www.linkedin.com/in/andyglew

All I can think there is comes down to Intel's famously 'deep pockets' - throw enough money at their best engineers to make absolutely certain they don't go elsewhere! With all the pessimism surrounding AMD's financial future (or realism?) I guess they're not gonna be at risk of losing engineers in that direction, but ARM and MIPS (like this Crazy Andy) could be a risk I guess!

[DanceswithUnix]

All I can think there is comes down to Intel's famously 'deep pockets' - throw enough money at their best engineers to make absolutely certain they don't go elsewhere! With all the pessimism surrounding AMD's financial future (or realism?) I guess they're not gonna be at risk of losing engineers in that direction, but ARM and MIPS (like this Crazy Andy) could be a risk I guess!

Right now ARM, Apple, Samsung, Qualcomm and MIPS are the places to be where all the interesting genuine innovation is happening. I can see any of these companies being more aspirational than Intel.

[spl]

Right now ARM, Apple, Samsung, Qualcomm and MIPS are the places to be where all the interesting genuine innovation is happening. I can see any of these companies being more aspirational than Intel.

Too bad none do anything in x86! Very comfortable situation for Intel. Maybe something like the rumoured Samsung acquisition of AMD will make things a little less comfortable for them :-)

[DanceswithUnix]

Too bad none do anything in x86! Very comfortable situation for Intel. Maybe something like the rumoured Samsung acquisition of AMD will make things a little less comfortable for them :-)

It has often been stated that Intel needs AMD to avoid being labelled a monopoly provider. If AMD start making enough money from selling ARM chips that they exit the x86 market then Intel are in serious legislative trouble. Intel may actually be quite worried.

[CAT-THE-FIFTH]

People seem to forget that AMD is constrained by what TSMC and GF can put out and despite the fact that Intel has much deeper pockets and spends more on process node development than their competitors AFAIK,they still are hitting issues too. Sadly,most of the market is dedicated towards processes for low power smaller chips or larger chips which run at lower clockspeeds,and AMD does not have the volume to push them to make specialist processes - the same processes for Zen and the K12 are probably going to be used for the next gen AMD or Nvidia cards(or a derivative of them).

This is probably the main limitation for AMD ATM - the maturity of the nodes they will be using for their new designs in terms of yields(cost),clockspeed(performance) and voltages required(power consumption) and leakiness(cooling).

Even look at Kaveri for example - it was meant to be released one year earlier than it eventually was. This was down to many factors,including the fact that the process AMD was using did not deliver the clockspeeds they wanted - they missed the CPU clockspeed by 8% and the IGP clockspeed by 25% and the CPU section of Kaveri tends to not Turbo all the time.

Godavari is what Kaveri for desktop should have been at launch.

[DanceswithUnix]

I think Intel have lost some of their manufacturing sheen recently. I keep reading things like https://www.semiwiki.com/forum/conte...d-finfets.html

Samsung are doing very well, TSMC seem bullish about their 10nm process, GloFo look to be coming out of the process rut and people are showing an interest in FD SOI 22nm which might actually be a nice place for enthusiast CPUs to go.

[kingpotnoodle]

At least Carrizo will really up their game in mobile, which is where more money is made.

I'm not so sure about that, mobile devices might be high volume but are being driven by the low cost segment at the moment, I don't think Atom makes much profit for Intel and ARM SoCs are quite cheap.

[DanceswithUnix]

I'm not so sure about that, mobile devices might be high volume but are being driven by the low cost segment at the moment, I don't think Atom makes much profit for Intel and ARM SoCs are quite cheap.

That is going to be interesting to watch. Carrizo-L should get good design wins on the cheap laptop front, having a pin compatible Carrizo that can go in the same chassis is the nice bit. I just hope people take the bait as I am fed up with seeing mini laptops that look interesting until I realise they aren't going to be appreciably faster than my E350 based 11.6" machine

[GuidoLS]

People seem to forget that AMD is constrained by what TSMC and GF can put out and despite the fact that Intel has much deeper pockets and spends more on process node development than their competitors AFAIK,they still are hitting issues too. Sadly,most of the market is dedicated towards processes for low power smaller chips or larger chips which run at lower clockspeeds,and AMD does not have the volume to push them to make specialist processes - the same processes for Zen and the K12 are probably going to be used for the next gen AMD or Nvidia cards(or a derivative of them).

This is probably the main limitation for AMD ATM - the maturity of the nodes they will be using for their new designs in terms of yields(cost),clockspeed(performance) and voltages required(power consumption) and leakiness(cooling).

Even look at Kaveri for example - it was meant to be released one year earlier than it eventually was. This was down to many factors,including the fact that the process AMD was using did not deliver the clockspeeds they wanted - they missed the CPU clockspeed by 8% and the IGP clockspeed by 25% and the CPU section of Kaveri tends to not Turbo all the time.

Godavari is what Kaveri for desktop should have been at launch.

Which is why some people (mainly me, at least here) have said over and again why it was foolish for AMD to have purchased ATI instead of doing complete refurbs or new builds on their on fabs like they had intended to do. Not that it's cheap, but it's why Intel has, and always will have, the advantage. They own their own fabrication sites, and they have the bankroll to make the adjustments needed.

[GuidoLS]

It has often been stated that Intel needs AMD to avoid being labelled a monopoly provider. If AMD start making enough money from selling ARM chips that they exit the x86 market then Intel are in serious legislative trouble. Intel may actually be quite worried.

I think the EU would have a tough time with this one - a company voluntarily dropping out of a market segment in favor of another segment of the same market does not a monopoly make. Afterall, ARM chips are still cpu's, and many are capable of running either the same or close variants of the same software. It can be argued, quite easily, that the tablet market is far larger, and possibly more important (on the consumer side) than the PC market is. And neither Intel nor AMD has any real market presence there.

Personally, I think the days of AMD being the foil to Intel being a monopoly are done and over with, courtesy of Arm and Samsung.

[GuidoLS]

I think Intel have lost some of their manufacturing sheen recently. I keep reading things like https://www.semiwiki.com/forum/conte...d-finfets.html

Samsung are doing very well, TSMC seem bullish about their 10nm process, GloFo look to be coming out of the process rut and people are showing an interest in FD SOI 22nm which might actually be a nice place for enthusiast CPUs to go.

Color me confused - why would anyone be concerned over a blog that basically says they don't understand why anyone would be happy over a real world 50% improvement, instead of an on-paper potential of 75%, in regards to a 14nm transistor vs a 16nm transistor? And even then, the commentary was more on a blog post someone else made over on an investment site than on the actual state of the tech.

Sites like semiwiki are good for light socializing with colleagues, but the odds of any real insider info is very slim - nda's and job security trump talking shop on an open forum every day of the week.

[DanceswithUnix]

Color me confused - why would anyone be concerned over a blog that basically says they don't understand why anyone would be happy over a real world 50% improvement, instead of an on-paper potential of 75%, in regards to a 14nm transistor vs a 16nm transistor? And even then, the commentary was more on a blog post someone else made over on an investment site than on the actual state of the tech.

Sites like semiwiki are good for light socializing with colleagues, but the odds of any real insider info is very slim - nda's and job security trump talking shop on an open forum every day of the week.

The point is that it used to be nothing but universal acceptance that Intel was years ahead in their fab technology. Now, non crazy people are saying that isn't so. That particular page argues that TSMC currently have better transistors than Intel, a couple of years ago that would have been laughable.

I think the EU would have a tough time with this one - a company voluntarily dropping out of a market segment in favor of another segment of the same market does not a monopoly make. Afterall, ARM chips are still cpu's, and many are capable of running either the same or close variants of the same software. It can be argued, quite easily, that the tablet market is far larger, and possibly more important (on the consumer side) than the PC market is. And neither Intel nor AMD has any real market presence there.

Personally, I think the days of AMD being the foil to Intel being a monopoly are done and over with, courtesy of Arm and Samsung.

I think you are absolutely right, but after all the litigation of the past I expect someone will argue that AMD were driven out of a duopoly by Intel.