Samsung and Global Foundries will produce AMD's Next Gen Greenland GPU and Zen CPU

[AtenRa]

Nice one Dresdenboy, but I would also like to say that AMD sold the highest volume in Q3 2015 to Sony and MS since both of them were getting ready for the holiday season (Q4). So AMD already knew how much better 2015 was since the end of Q3.

 

[ShintaiDK]

That chart is not that useful. Using arguments based on comparing a year with Q4 data included with a year w/o Q4 data unfortunately shows the level of understanding that seasonality as Q4 easily has 2x to 3x the previous quarters' units sold. This is what I say about the smartest algorithms fed with mediocre sensor data: employing the GIGO principle - Garbage In Garbage Out.

*vgchartz removed*


XBox One should also be just estimated as in the Ars Technica chart due to MS policy (which is an understandable move).

You should know vgchartz have nothing to do with reality. They are based purely on guesswork.

 

[Dresdenboy]

You should know vgchartz have nothing to do with reality. They are based purely on guesswork.

Yeah, cool! I expected exactly this reaction (genetic fallacy again). I was too lazy to post my prediction, but now I'm happy.

 

[Techhog]

Yeah, cool! I expected exactly this reaction (genetic fallacy again). I was too lazy to post my prediction, but now I'm happy.

I normally wouldn't agree with him, but he's right. That site isn't reliable at all.

 

[DrMrLordX]

A bad excuse is a bad excuse.

Whose excuse?

Just accept the reality and move on.

I will when someone presents reality.

Your precious AMD isn't selling well.

Not selling well compared to what? And I don't own shares or even buy their dGPUs so . . . whose AMD is this again?

Evidence of sales comes from someone actually showing how many units of a particular product sold in a particular quarter (or over some other time span). Those sales can better be put in context when juxtaposed with production data for the same units. The Steam survey just shows you what's in use. It doesn't tell you what's actually selling right now . . . again, where's the 980 Ti? According to recent reports, shelves are running short of 980 Tis so I have to assume that they're selling quite well relative to production levels. Not that the Steam survey would tell me that . . .

I fully expect that cards like the 390x and 390 aren't selling well thanks to people willingly paying for 970s instead regardless of whether or not that choice makes any sense, and thanks to the 290 and 290x being so close in specs already. But the Steam survey doesn't actually tell me that.

 

[Phynaz]

Nice one Dresdenboy, but I would also like to say that AMD sold the highest volume in Q3 2015 to Sony and MS since both of them were getting ready for the holiday season (Q4). So AMD already knew how much better 2015 was since the end of Q3.

Still looking for the $300M profit AMD was supposed to make from the consoles.

 

[TurtleCrusher]

And here we have it again. Intel shareholders derailing another AMD thread to the point of it being unreadable and unrelated.

Thanks for ruining the forums.

 

[III-V]

And here we have it again. Intel shareholders derailing another AMD thread to the point of it being unreadable and unrelated.

Thanks for ruining the forums.

The problem exists on the other side of the fence too.

 

[ultimatebob]

And here we have it again. Intel shareholders derailing another AMD thread to the point of it being unreadable and unrelated.

Thanks for ruining the forums.

I'm not an Intel or AMD shareholder. I'm not a fanboy of either company, either. I am concerned in AMD's long term viability as a company, though, and I think that everyone should at this point.

Thinking that Greenland and Zen are magically going to fix all of AMD's problems is just really naive.

 

[AtenRa]

Thinking that Greenland and Zen are magically going to fix all of AMD's problems is just really naive.

No, but they are in the right direction. Especially with ZEN they will start selling High Margins SERVER parts again after 4-5 years.

 

[Arachnotronic]

And here we have it again. Intel shareholders derailing another AMD thread to the point of it being unreadable and unrelated.

Thanks for ruining the forums.

You have no problem with those who are constantly boosting AMD, but anybody who is skeptical of this company is an Intel shareholder/shill?

Methinks Scholzpdx doth protest too much

 

[ShintaiDK]

No, but they are in the right direction. Especially with ZEN they will start selling High Margins SERVER parts again after 4-5 years.

No, you hope they will. Just like you did countless times before.

We heard for ages now that the next AMD product is going to fix it and turn it in the right direction. But it only goes south every time.

There was no limit on what Carrizo couldn't do either. It turned out to be the biggest APU flop yet. The PC division will end around a 50% decline in 2015. And it will continue so in 2016.

 

[Arachnotronic]

No, but they are in the right direction. Especially with ZEN they will start selling High Margins SERVER parts again after 4-5 years.

Only if they're competitive. They might be, but it's far from certain at this point. Remember, AMD had big plans to recapture server share with its Construction cores too. Terramar, anybody?

http://www.xbitlabs.com/news/cpu/di...ancels_10_Core_20_Core_Server_Processors.html

 

[AtenRa]

Only if they're competitive. They might be, but it's far from certain at this point. Remember, AMD had big plans to recapture server share with its Construction cores too. Terramar, anybody?

http://www.xbitlabs.com/news/cpu/di...ancels_10_Core_20_Core_Server_Processors.html

Yeap, no 22nm available no server parts. Simple as that.

But this time they will have a nice 14nm FF node and a new mArch so things are way different than before.

 

[III-V]

Wonder how IBM's 22 nm would have compared.

 

[Arachnotronic]

Yeap, no 22nm available no server parts. Simple as that.

But this time they will have a nice 14nm FF node and a new mArch so things are way different than before.

Um, Bulldozer & lineage were the "new arch" that was supposed to fix the problems of the prior gen, so maybe we should wait and see before assuming "new" means "competitive & great"?

Also that 14FF node seems good for handhelds and other mobile form factors, but I have yet to see any >2.2GHz parts built on it. Let's wait and see how good this node is in server/HEDT parts before proclaiming that it will close the node gap between Intel and AMD.

 

[Dresdenboy]

You should know vgchartz have nothing to do with reality. They are based purely on guesswork.

It's called estimation, projection, interpolation, extrapolation, update. Wanna try an extended Kalman filter on that?

JPR, Mercury Research, and others do the same, with the same or different sources. And so do analysts. Yet some take their words like those of a new prophet, while at the same time for many stocks, some say "sell", some say "buy", some say "hold". How can they contradict eachother while analyzing the same thing?

But it is known that the human mind without preparation/training isn't well suited to handle probabilities and estimations. Some know that, most don't. That's why there is always crying, when some estimates don't correlate with reality. So VGChartz has been discussed all the time due to game sales numbers, which weren't fully published before by the distributors, or differences to other estimators and published sales numbers. A waste of time, like 95% of any forum.

 

[ShintaiDK]

Second time you post on the same quote. Not going to make vgchartz any better. And the financials dont back your statement.

 

[frozentundra123456]

And here we have it again. Intel shareholders derailing another AMD thread to the point of it being unreadable and unrelated.

Thanks for ruining the forums.

Why do you automatically assume anyone who refutes a pro-AMD post is an Intel shareholder. Maybe they just want to bring "balance to the force", so to speak.

Personally, I have no financial stake in Intel *or* AMD. I have an intel cpu and an AMD gpu in my main system. But I certainly take issue with a lot of the statements made in these forums by both AMD and Intel "supporters".

 

[mikk]

Vgchartz is unreliable as hell.

 

[myocardia]

Also that 14FF node seems good for handhelds and other mobile form factors, but I have yet to see any >2.2GHz parts built on it. Let's wait and see how good this node is in server/HEDT parts before proclaiming that it will close the node gap between Intel and AMD.

I don't want to seem like I'm disagreeing with you, because I do not, but you haven't forgotten the new market for servers with smaller/slower, yet efficient cores that Intel created fairly recently, when they introduced the Xeons based on their Atom CPUs, have you? If AMD could get anywhere near 50% of that market alone, it would be a financial godsend for them. Then they would only need to completely replace their BoD, and they might actually become profitable. Hey, I said might!

 

[Dresdenboy]

Vgchartz is unreliable as hell.

That depends on the used and available data sources. And these might be different for individual games and consoles. If we want to wait for the exact sales numbers, we should stop discussing the consoles in this Samsung/GF thread and open another thread in a couple of weeks, as Q4 sales numbers - even according to VGChartz - are an important data point for a yearly decline discussion.

 

[itsmydamnation]

Um, Bulldozer & lineage were the "new arch" that was supposed to fix the problems of the prior gen, so maybe we should wait and see before assuming "new" means "competitive & great"?

The funny thing is, and every time i bring it up the IDF goes silent. The CON cores actually did this. What they did wrong is design the overall core architecture in the wrong direction. The Entire CPU pipeline got modernized and the limitations of the stars cores removed.

Now look at the areas it looks like amd hurt the most,
miss predict penalty
int unit balance, not enough int execution/branch for high IPL situations
shared fetch
FP latency (only FMA units with ~5 cycle execution times)
L1D size
L1I association + its shared
L2 latency
L1-L2 cache policy/ WCC random oddness that seems to randomly blow out

None of those got touched/fixed for CON cores because it likely involved massive core redesign.

Look at the things that got continuous improvement in CON core development:
decode ( uop cache etc)
predictors/ prefetches
load store unit
scheduling
paging/tlb etc

Then when you hear AMD say:
that they are reusing existing logic where it makes sense.
that a Zen core is wider(integer) then both STARS and CON.
that Zen will have an all new cache design.
that Zen will clocks closer to CON then it does to CAT.

You put all of that together and this seems pretty logical:
ZEN:
evolution:
decode
predictors/ prefetches
load store unit
scheduling
paging/tlb etc
RPF
evolution but functional changed:
integer execution (increased width)
Fetch (becomes dedicated per core)
New:
L1D
L1I
cache/interconnect
FP unit

Now obviously there will be areas here that are completely wrong were AMD thinks that they have maxed the gains and need to go in a different direction (i think integer unit organisation will be one) but broadly speaking this is what i think you will see.

Put all the together im very positive in AMD delivering a strong CPU core as there is actually a strong foundation to build off.

 

[Dresdenboy]

Second time you post on the same quote. Not going to make vgchartz any better. And the financials dont back your statement.

Nope, I just finished watching a good movie, and felt, I might add something more to it, as my thoughts, which were related to this, accumulated over time.

It is as good/bad as it is. Anything else with exact data in a couple of weeks.

 

[cytg111]

Maybe they just want to bring "balance to the force", so to speak.

If you read this forum and go thinking "hmm an unusual amount of pro AMD activity here" you are wearing differently coloured shades than me.

Take the keller threads, the zen hasnt taped out yet threads, and see the scrutiny that ie. wccftech is under, as a source, by certain posters depending on what story it portraits(pro/anti amd/intel), the double standards are so obvious that you HAVE to assume a person is invested somehow, either that or some psychological phenomenon/disorder is at play.