• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD Polaris architecture – GCN 4.0

Lol, that's a dream mate. AMD need to work hard just to get back to 30% / 40% market share within a year.

AMD have done what they set out to do though, and that was to generate income from areas other than just traditional desktop PC. All these console wins are a boon to them. Console sales have been really good this gen so it's only good for AMD. I believe if they can get the next launch of desktop PC GPU's right, they could get back to 35% market share in a year+. They would have to be very very good though. That aside it looks like AMD will make some noise in the desktop CPU space with Zen and also the server side, so things are def on the up and up.

Well they are on the good route. In 2015 Q3/Q4 they gained marketshare, and in 2016 so fat they are kicking NV in the rearend nicely, so i can imagine they will get back some more. Also a good base to start the new lineup and DX12 era.
 
Someone spotted this on AT forums:

http://i.imgur.com/6pA79OG.jpg

6pA79OG.jpg



https://www.youtube.com/watch?v=7Scm6boqNeo&feature=youtu.be&t=23m28s

Silverforce11 said:
This is from a seminar from ALL the players involved in HBM tech, especially those designing the PHY, stacking, with names like eSilicon, Amkor & SK Hynix that some of you may recognize from HBM1 pioneers.

Since we can be confident Polaris is 14nm FF + GDDR5...

Looks like Vega is due to tape out in May, and will be ready in September. Note they are talking about their end, getting the HBM2 on the interposer with TSV linked to the chip, so the finish stack is their "tape out".

They also claim to do a 16nm FF + HBM2... ready in December though. Not sure what that is, potentially a GK104 "refresh" for 2017?

I assume the 14NM LPP is Vega??

What is the 16NM HBM2 chip??
 
Well they are on the good route. In 2015 Q3/Q4 they gained marketshare, and in 2016 so fat they are kicking NV in the rearend nicely, so i can imagine they will get back some more. Also a good base to start the new lineup and DX12 era.

gained market share in 2015?
year on year they were down to 21% from 24% and quarter to quarter they were 21% from 19%

there's not been any new products or pricing really so I don't see how they'll have suddenly gained market share in the last few months
 
gained market share in 2015?
year on year they were down to 21% from 24% and quarter to quarter they were 21% from 19%

there's not been any new products or pricing really so I don't see how they'll have suddenly gained market share in the last few months

They gained marketshare in the last few quarters:
https://jonpeddie.com/press-release...sed-in-q315-amd-gained-market-share-nvidia-lo

Q3 2015 they went up from 18% to 18.8% it appears.

https://jonpeddie.com/press-release...gpu-shipments-increased-2.4-from-last-quarter

In Q4 2015 AMD gained nearly 7% desktop marketshare from Nvidia and lost 1.3% in laptops. Most of the Nvidia mobile laptop GPU increases came from Intel.

I know some are desperate for AMD to fail but even with an old product stack they are starting to get past the worst of the problems they have.

Polaris looks like it will be competitive in laptops so they will probably start eating marketshare from Nvidia in that area.
 
Last edited:
No, this is talking about Making HBM on 28HPC, 14nm LPP and 16nm FF+ nodes. Nothing to do with chips used with HBM.

Having both 14nm and 16nm means they are making designs that can be produced on both nodes.

Look at the dates. The GP100 is using TSMC for chip production and interposer production. It makes me wonder why they said early 2017 for the GP100 is more down to the HBM production dates.

Even the late 2016 launch for Vega might be to do with that too.
 
Look at the dates. The GP100 is using TSMC for chip production and interposer production. It makes me wonder why they said early 2017 for the GP100 is more down to the HBM production dates.

Even the late 2016 launch for Vega might be to do with that too.

None of what is in that chart is to do with Nvidia or AMD, it is purely to do with the production of HBM and HBM2 test chips then full scale designs on the nodes stated.

For the first lots of HBM2, they would more than likely be using a mixture of Samsung and hynix 28nm HBM2 parts. Although the Hynix HBM2 is not coming till Q3 16.

The entire presentation is based around HBM.
 
None of what is in that chart is to do with Nvidia or AMD, it is purely to do with the production of HBM and HBM2 test chips then full scale designs on the nodes stated.

For the first lots of HBM2, they would more than likely be using a mixture of Samsung and hynix 28nm HBM2 parts. Although the Hynix HBM2 is not coming till Q3 16.

The entire presentation is based around HBM.

Which also intrinsically means we know the minimum dates of availability of Vega and the GP100. It states that 14LPP evaluation is complete and 28NM and 16NM evaluation is NOT.
 
They gained marketshare in the last few quarters:
https://jonpeddie.com/press-release...sed-in-q315-amd-gained-market-share-nvidia-lo

Q3 2015 they went up from 18% to 18.8% it appears.

https://jonpeddie.com/press-release...gpu-shipments-increased-2.4-from-last-quarter

In Q4 2015 AMD gained nearly 7% desktop marketshare from Nvidia and lost 1.3% in laptops. Most of the Nvidia mobile laptop GPU increases came from Intel.

I know some are desperate for AMD to fail but even with an old product stack they are starting to get past the worst of the problems they have.

Polaris looks like it will be competitive in laptops so they will probably start eating marketshare from Nvidia in that area.

https://jonpeddie.com/publications/add-in-board-report/

Nope. The report you linked to said that AMD's shipments increased by 7%, not that they gained 7% market share, they actually gained less than 2% marketshare
 
gained market share in 2015?
year on year they were down to 21% from 24% and quarter to quarter they were 21% from 19%

there's not been any new products or pricing really so I don't see how they'll have suddenly gained market share in the last few months

They gained marketshare in the last few quarters:
https://jonpeddie.com/press-release...sed-in-q315-amd-gained-market-share-nvidia-lo

Q3 2015 they went up from 18% to 18.8% it appears.


https://jonpeddie.com/press-release...gpu-shipments-increased-2.4-from-last-quarter

In Q4 2015 AMD gained nearly 7% desktop marketshare from Nvidia and lost 1.3% in laptops. Most of the Nvidia mobile laptop GPU increases came from Intel.


I know some are desperate for AMD to fail but even with an old product stack they are starting to get past the worst of the problems they have.

Polaris looks like it will be competitive in laptops so they will probably start eating marketshare from Nvidia in that area.

https://jonpeddie.com/publications/add-in-board-report/

Nope. The report you linked to said that AMD's shipments increased by 7%, not that they gained 7% market share, they actually gained less than 2% marketshare

So they basically gained cumalative marketshare over Q3 2015 and Q4 2015 so going by your figures. Thanks for agreeing with me.

Plus it is also interesting how most of the last two quarters appear to be desktop cards.
 
Last edited:
If going from 18% to 21% is a gain of 7% of marketshare (and anything to be proud of) then well... Yeah ok

7% gain of a low number is still a low number. Going from 18.8 to 21 is actually a gain of nearly 12%, but it's not a gain of 12% market share.

Also, how do nvidia gain marketshare "from intel". Nvidia's might go up and at the same time intel go down, but as intel only make apu's nd nvidia only make discrete, Nvidia can't actually take marketshare from Intel because an nvidia equipped laptop still needs an APU. Intel didn't lose a sale to Nvidia (though the reverse can be true).
 
Last edited:
Well they are on the good route. In 2015 Q3/Q4 they gained marketshare, and in 2016 so fat they are kicking NV in the rearend nicely, so i can imagine they will get back some more. Also a good base to start the new lineup and DX12 era.

andybird123 said:
gained market share in 2015?
year on year they were down to 21% from 24% and quarter to quarter they were 21% from 19%

there's not been any new products or pricing really so I don't see how they'll have suddenly gained market share in the last few months

They gained marketshare in the last few quarters:
https://jonpeddie.com/press-release...sed-in-q315-amd-gained-market-share-nvidia-lo

Q3 2015 they went up from 18% to 18.8% it appears.

https://jonpeddie.com/press-release...gpu-shipments-increased-2.4-from-last-quarter

In Q4 2015 AMD gained nearly 7% desktop marketshare from Nvidia and lost 1.3% in laptops. Most of the Nvidia mobile laptop GPU increases came from Intel.

I know some are desperate for AMD to fail but even with an old product stack they are starting to get past the worst of the problems they have.

Polaris looks like it will be competitive in laptops so they will probably start eating marketshare from Nvidia in that area.

https://jonpeddie.com/publications/add-in-board-report/

Nope. The report you linked to said that AMD's shipments increased by 7%, not that they gained 7% market share, they actually gained less than 2% marketshare

They gained marketshare in the last few quarters:
https://jonpeddie.com/press-release...sed-in-q315-amd-gained-market-share-nvidia-lo

Q3 2015 they went up from 18% to 18.8% it appears.


https://jonpeddie.com/press-release...gpu-shipments-increased-2.4-from-last-quarter

In Q4 2015 AMD gained nearly 7% desktop marketshare from Nvidia and lost 1.3% in laptops. Most of the Nvidia mobile laptop GPU increases came from Intel.


I know some are desperate for AMD to fail but even with an old product stack they are starting to get past the worst of the problems they have.

Polaris looks like it will be competitive in laptops so they will probably start eating marketshare from Nvidia in that area.



So they basically gained cumalative marketshare over Q3 2015 and Q4 2015 so going by your figures. Thanks for agreeing with me.

Plus it is also interesting how most of the last two quarters appear to be desktop cards.

If going from 18% to 21% is a gain of 7% of marketshare (and anything to be proud of) then well... Yeah ok

7% gain of a low number is still a low number. Going from 18.8 to 21 is actually a gain of nearly 12%, but it's not a gain of 12% market share.

Also, how do nvidia gain marketshare "from intel". Nvidia's might go up and at the same time intel go down, but as intel only make apu's nd nvidia only make discrete, Nvidia can't actually take marketshare from Intel because an nvidia equipped laptop still needs an APU. Intel didn't lose a sale to Nvidia (though the reverse can be true).

1.)In your desperate attempts at purchase justification,you first misrepresented what the first poster said,who clearly stated 2H 2015.

2.)Then said they could never have had increased sales in the last few months.

3.)Then I showed data which debunked your claim.

4.)Then you desperately attempted to change the argument to include the whole of 2015,when the point you were trying to debunk was 2H 2015.

5.)Then you tried to debunk what I said and agreed AMD gained marketshare over the last two quarters.

6.)Brings us back full circle as you just agreed with what I and the first poster said.

You have not bothered even trying to post any information about the AMD GPUs in question or even where they have been placed. I won't be answering any more about this since you have just bored me and I will report subsequent posts to a mod.

Look for more attention from somebody else.

OK,we are done here.
 
Last edited:
No. They said that AMD were kicking nvidia's behind in Q1/16. I said I don't see how. There aren't any figures for Q1/16. I didn't say "never" because we don't have figures yet.

/facepalm

The article you posted actually says;

"Desktop discrete GPUs increased 6.69% from last quarter,"

It does not say they gained 7% market share, it says shipments.

I don't find 2% to be significant, its also nowhere near the 7% you're trying to claim. Shipments and marketshare are not the same thing.

I even pointed out that they did gain 2%, my remark was that I don't find 2% significant in the face of a year over year loss.
 
Last edited:
Which also intrinsically means we know the minimum dates of availability of Vega and the GP100. It states that 14LPP evaluation is complete and 28NM and 16NM evaluation is NOT.

No it doesn't. All it means is that they have final HBM designs evaluated on Samsungs 14nm LPP but not on TSMC's 16nm FF+ node.

It has nothing to do with how soon AMD or Nvidia will have parts prepared on the respective nodes.
 
No. They said that AMD were kicking nvidia's behind in Q1/16. I said I don't see how. There aren't any figures for Q1/16. I didn't say "never" because we don't have figures yet.

/facepalm

The article you posted actually says;

"Desktop discrete GPUs increased 6.69% from last quarter,"

It does not say they gained 7% market share, it says shipments.

I don't find 2% to be significant, its also nowhere near the 7% you're trying to claim. Shipments and marketshare are not the same thing.

I even pointed out that they did gain 2%, my remark was that I don't find 2% significant in the face of a year over year loss.

I said they gained marketshare in the 2H of 2015, and when i said they are kicking NVs behind in 2016 i meant in games and i assume they will increase further in Q1 as NV getting a beating while AMD going strong in DX12 (and even in DX11 games they are running nice lately)
I have already seen in forums that ppl don't recommend maxwell cards to buy, instead they say "if you buy now the amd cards are more future proof, or wait for the new cards at summer..."

Also the point was not really the amount they gained, but instead tge fact they gained as you said without any new card release or other changes.
 
No it doesn't. All it means is that they have final HBM designs evaluated on Samsungs 14nm LPP but not on TSMC's 16nm FF+ node.

It has nothing to do with how soon AMD or Nvidia will have parts prepared on the respective nodes.

The presentation is from the companies onboard with HBM1 and HBM2.

Which means HBM2 on 14NM LPP will be the first to enter production,with first availability being in September 2016. The TSMC produced HBM2 says December 2016. 28NM HBM2 is not even as close to production readiness as the other two and probably is a designed as a lower cost alternative for the next generation midrange cards.

Nvidia has themselves stated Q1 2017 for large scale GP100 availability and AMD a vague "late" 2016 for Vega. These seem to follow when HBM2 is meant to enter full scale production.

This probably means we won't see an HBM2 equipped card in any sort of large scale production for cards until November or December at the earliest I suspect,especially if September is when the first available chips will trickle out.

I expect a lot of noise from Nvidia and AMD about their HBM2 equipped cards but we won't see much if any shipping in quantity for quite a while now.

I think all the people expecting large scale HBM2 production by the beginning of this year were being premature - that was barely after HBM1 hit mainstream and why AMD probably bothered with HBM1 in the first place.
 
Last edited:
The presentation is from the companies onboard with HBM1 and HBM2.

Which means HBM2 on 14NM LPP will be the first to enter production,with first availability being in September 2016. The TSMC produced HBM2 says December 2016. 28NM HBM2 is not even as close to production readiness as the other two and probably is a designed as a lower cost alternative for the next generation midrange cards.

Nvidia has themselves stated Q1 2017 for large scale GP100 availability and AMD a vague "late" 2016 for Vega. These seem to follow when HBM2 is meant to enter full scale production.

I think all the people expecting large scale HBM2 production by the beginning of this year were being premature - that was barely after HBM1 hit mainstream and why AMD probably bothered with HBM1 in the first place.

Samsung has been producing 20nm class 4GB HBM2 modules since the begging of this year. Creating designs for 14 and 16nm straight away is just future proofing. HBM2 is already in production.

The later release of GP100 is due to dire yield for such a big part this early on, considering they also need a fully working 1200mm interposed. And even with Q1 17, that is still only to select HPC partners and Universitys.

You can even check the press releases on Samsung's own website.
 
Back
Top Bottom