[H]ardOCP Editor Claims Internal Problems for AMD - Polaris Underwhelming and RTG Seeking to Leave AMD For Intel

[don_svetlio]

4 minutes ago, Derangel said:

 

GDDR5X supply issues will be sorted out eventually. I'd argue that AMD is also shooting themselves in the foot by having nothing new on the high-end this year. The earliest we can expect Vega is probably Q2 of next year, if not later. Leaving Nvidia nearly a year to sort out supply issues and get something else (like a 1080ti and potentially a price drop) ready to counter AMD as soon as Vega launches. The mid-range is very important, but even with kind of having the best mid-range options (especially in the $300 and under segments) for a few years AMD has still taken a drastic loss in market share. AMD's mid-range and mobile options are all over the place, but it's not helping. If AMD can capture that $300-$350 market that Nvidia currently dominates then they will probably see their market share return. It'll be interesting to see how Polaris 10 does against the 1070.

Erm? Nvidia haven't had a good mid-range option I'd argue ever since the 660/660 Ti - the reason AMD sales are low is because people just buy Nvidia blindly on mass. It's not a hardware issue - it's a marketing one.

[Derangel]

2 minutes ago, don_svetlio said:

Erm? Nvidia haven't had a good mid-range option I'd argue ever since the 660/660 Ti - the reason AMD sales are low is because people just buy Nvidia blindly on mass. It's not a hardware issue - it's a marketing one.

Well considering I specifically said that AMD has had the best mid-range options for the last few years I'm not sure what you're correcting.

[don_svetlio]

1 minute ago, Derangel said:

Well considering I specifically said that AMD has had the best mid-range options for the last few years I'm not sure what you're correcting.

You were talking about recapturing

Anyhow, the 700$ price point of the 1080 is barely profitable - most people can't afford/aren't willing to purchase such a product simply due to cost alone

[Trixanity]

3 minutes ago, Derangel said:

 

GDDR5X supply issues will be sorted out eventually. I'd argue that AMD is also shooting themselves in the foot by having nothing new on the high-end this year. The earliest we can expect Vega is probably Q2 of next year, if not later. Leaving Nvidia nearly a year to sort out supply issues and get something else (like a 1080ti and potentially a price drop) ready to counter AMD as soon as Vega launches. The mid-range is very important, but even with kind of having the best mid-range options (especially in the $300 and under segments) for a few years AMD has still taken a drastic loss in market share. AMD's mid-range and mobile options are all over the place, but it's not helping. If AMD can capture that $300-$350 market that Nvidia currently dominates then they will probably see their market share return. It'll be interesting to see how Polaris 10 does against the 1070.

They are currently gaining market share but the biggest gains are by far mobile. Nvidia chips has been the go-to option for most due to the great success of Optimus (and AMD's counterpart being incredibly poor for many years leading to big losses in market share). Vega was slated for Q1 2017 but rumors say it has been moved up to October 2016. Vega will be a chip design to sell in the 1080 Ti price slot, so doesn't really matter if Nvidia responds to that.

 

I'm quite certain Polaris 10 slots in below 1070. Perhaps if Sapphire gets free reins, they can produce a custom PCB with a larger TDP and therefore higher clock to smack the 1070 but otherwise I doubt it. It's not meant to go that high in performance design-wise.

 

In either case, AMD is aiming to displace GP106 which would mean very large gains in the mobile market. I think a cutdown Vega (11?)could compete with GP104 and then full Vega (10?) for GP102, since GP100 seems to be missing. 

[Notional]

8 minutes ago, Derangel said:

 

GDDR5X supply issues will be sorted out eventually. I'd argue that AMD is also shooting themselves in the foot by having nothing new on the high-end this year. The earliest we can expect Vega is probably Q2 of next year, if not later. Leaving Nvidia nearly a year to sort out supply issues and get something else (like a 1080ti and potentially a price drop) ready to counter AMD as soon as Vega launches. The mid-range is very important, but even with kind of having the best mid-range options (especially in the $300 and under segments) for a few years AMD has still taken a drastic loss in market share. AMD's mid-range and mobile options are all over the place, but it's not helping. If AMD can capture that $300-$350 market that Nvidia currently dominates then they will probably see their market share return. It'll be interesting to see how Polaris 10 does against the 1070.

Vega in Q2 2017 or later? Source please!

 

Not only has Vega been said to be out about december 2016 to January 2017, newest rumours says October 2016. So how can Q2 2016 be Q2 2017?

 

Nvidia's P100 with HBM will either be end of this year or start of next. This is probably why we see a GP102 instead with GDDR5x.

[Derangel]

1 minute ago, Notional said:

Vega in Q2 2017 or later? Source please!

 

Not only has Vega been said to be out about december 2016 to January 2017, newest rumours says October 2016. So how can Q2 2016 be Q2 2017?

 

Nvidia's P100 with HBM will either be end of this year or start of next. This is probably why we see a GP102 instead with GDDR5x.

Is it? Last I heard we were looking at sometime in 2017. If its going to hit in Oct or, latest, Jan then that does change things considerably.

[don_svetlio]

2 minutes ago, Notional said:

Vega in Q2 2017 or later? Source please!

 

Not only has Vega been said to be out about december 2016 to January 2017, newest rumours says October 2016. So how can Q2 2016 be Q2 2017?

 

Nvidia's P100 with HBM will either be end of this year or start of next. This is probably why we see a GP102 instead with GDDR5x.

P100 isn't coming to consumers - GP102 is the 1000$ Titan with GDDR5X

[Trixanity]

1 minute ago, don_svetlio said:

P100 isn't coming to consumers - GP102 is the 1000$ Titan with GDDR5X

Yup, that's why it's called P and not GP (Just a note for those unfamiliar with the naming schemes). There will be no Geforce variant of the Pascal 100 chip (at least not any time soon. Maybe in a year or so?). The yields must be so bad that they can't afford to sell it to consumers. I have to wonder at Nvidia's choice of GDDR5X instead of HBM2 for their Titan/Ti cards.

[Derangel]

5 minutes ago, Trixanity said:

Yup, that's why it's called P and not GP (Just a note for those unfamiliar with the naming schemes). There will be no Geforce variant of the Pascal 100 chip (at least not any time soon. Maybe in a year or so?). The yields must be so bad that they can't afford to sell it to consumers. I have to wonder at Nvidia's choice of GDDR5X instead of HBM2 for their Titan/Ti cards.

GDDR5X will be fast enough for games, at the moment, but it is a weird decision. It does paint a picture of them rushing production to have something out right now. The FE cooler being clearly poorly designed does nothing to dispel that image. If Nvidia is planning to charge over a grand for the Titan/Ti variant, using GDDR5X is going to be a black mark on them.

[Notional]

10 minutes ago, Derangel said:

Is it? Last I heard we were looking at sometime in 2017. If its going to hit in Oct or, latest, Jan then that does change things considerably.

The October launch time is a rumour, the end of 2016/start 17 is the timeframe AMD presented on their slides.

10 minutes ago, don_svetlio said:

P100 isn't coming to consumers - GP102 is the 1000$ Titan with GDDR5X

I know, only Geforce (GP) cards are for consumers. But the P100 chips still has a launch date, that is relevant due to the size of the chip and inclusion og HBM2. I do wonder if they will make Quadro/Tesla cards of them.

[Derangel]

1 minute ago, Notional said:

The October launch time is a rumour, the end of 2016/start 17 is the timeframe AMD presented on their slides.

I know, only Geforce (GP) cards are for consumers. But the P100 chips still has a launch date, that is relevant due to the size of the chip and inclusion og HBM2. I do wonder if they will make Quadro/Tesla cards of them.

Ah. The slides I just looked at when double checking some stuff on AMD's statements about Polaris still had Vega towards 2017, though I could have just been looking at it wrong.

[don_svetlio]

6 minutes ago, Trixanity said:

Yup, that's why it's called P and not GP (Just a note for those unfamiliar with the naming schemes). There will be no Geforce variant of the Pascal 100 chip (at least not any time soon. Maybe in a year or so?). The yields must be so bad that they can't afford to sell it to consumers. I have to wonder at Nvidia's choice of GDDR5X instead of HBM2 for their Titan/Ti cards.

 

2 minutes ago, Derangel said:

GDDR5X will be fast enough for games, at the moment, but it is a weird decision. It does paint a picture of them rushing production to have something out right now. The FE cooler being clearly poorly designed does nothing to dispel that image. If Nvidia is planning to charge over a grand for the Titan/Ti variant, using GDDR5X is going to be a black mark on them.

Not to mention the high TDPs of the GP102/P100 chips - 250-300W is nothing to overlook as that's pretty much a record for a single chip

[SurvivorNVL]

I love how I was included in the original post.  Intel-Radeon, the dream.

[Sparviero]

Zen on the horizon, many large contracts providing stable funding

 

then from what people say in this thread nvidia jumped the gun and didn't take the time to introduce a new architecture that makes full use of the 16nm node and the first cards with gddrx will perform at a lower value than in a few months. 

 

AMD was too soon with HBM 1 or at least should have let the AIBs have a free hand. So it seems better they wait an entire 30 to 60 days. I don't get the doom and gloom at this point. In a month I may be willing to take a sack clothe and join the club. 

[Derangel]

3 minutes ago, SurvivorNVL said:

I love how I was included in the original post.  Intel-Radeon, the dream.

It would be interesting, that's for sure. Not sure if it would be good for the market or not, but it would be interesting to see.

 

1 minute ago, Sparviero said:

Zen on the horizon, many large contracts providing stable funding

 

then from what people say in this thread nvidia jumped the gun and didn't take the time to introduce a new architecture that makes full use of the 16nm node and the first cards with gddrx will perform at a lower value than in a few months. 

 

AMD was too soon with HBM 1 or at least should have let the AIBs have a free hand. So it seems better they wait an entire 30 to 60 days. I don't get the doom and gloom at this point. In a month I may be willing to take a sack clothe and join the club. 

If Zen can manage to live up to the hype and actually be competitive then AMD will be in a much better position. Sadly I have zero faith left in AMD's CPU division. They've had so many missteps over the last several years that its hard to get excited for something new.

[don_svetlio]

1 minute ago, Derangel said:

It would be interesting, that's for sure. Not sure if it would be good for the market or not, but it would be interesting to see.

 

If Zen can manage to live up to the hype and actually be competitive then AMD will be in a much better position. Sadly I have zero faith left in AMD's CPU division. They've had so many missteps over the last several years that its hard to get excited for something new.

Keep in mind - the main force behind zen was Jim Keller - the guy who was in charge of Athlon X2 - something Intel failed to beat for several years.

[SurvivorNVL]

2 minutes ago, Derangel said:

It would be interesting, that's for sure. Not sure if it would be good for the market or not, but it would be interesting to see.

 

If Zen can manage to live up to the hype and actually be competitive then AMD will be in a much better position. Sadly I have zero faith left in AMD's CPU division. They've had so many missteps over the last several years that its hard to get excited for something new.

I have some faith in AMD, but I'd rather see an all out war between nVidia and Intel with all of AMD's assets.  Especially given in the future - Intel could possibly acquire nVidia, and then there's no stopping them.  So, the preferable outcome is AMD's death, Radeon in Intel's hands, and AMD64 / a new x86 license in nVidia's hands.  Just imagine a Volta APU.  Yeah.  If nVidia got the CPU division it'd likely be two to four years, but then again, we would all be with Intel CPU's for that time if Zen was a flop anyway, so, no real change.

That and I can just imagine the fun on the market if there was the announcement, "Intel to buy AMD's Radeon Technologies Group."  Just be a giddy day for shareholders.

[Derangel]

3 minutes ago, don_svetlio said:

Keep in mind - the main force behind zen was Jim Keller - the guy who was in charge of Athlon X2 - something Intel failed to beat for several years.

True. I'd love for my pessimism to be wrong. The first computer I built was with a Barton core 2500+, I'd actually love to have an AMD system again. It just hasn't been the best option for me since Intel released the C2D.

[Derangel]

2 minutes ago, SurvivorNVL said:

I have some faith in AMD, but I'd rather see an all out war between nVidia and Intel with all of AMD's assets.  Especially given in the future - Intel could possibly acquire nVidia, and then there's no stopping them.  So, the preferable outcome is AMD's death, Radeon in Intel's hands, and AMD64 / a new x86 license in nVidia's hands.  Just imagine a Volta APU.  Yeah.  If nVidia got the CPU division it'd likely be two to four years, but then again, we would all be with Intel CPU's for that time if Zen was a flop anyway, so, no real change.

That and I can just imagine the fun on the market if there was the announcement, "Intel to buy AMD's Radeon Technologies Group."  Just be a giddy day for shareholders.

Samsung would probably be a better fit for AMD's CPU division. Samsung can be a shitty company, but they have essentially limitless resources and can spend billions a year in R&D to make something. Samsung with the ability to make desktop processors would be some serious competition.

[SurvivorNVL]

Just now, Derangel said:

Samsung would probably be a better fit for AMD's CPU division. Samsung can be a shitty company, but they have essentially limitless resources and can spend billions a year in R&D to make something. Samsung with the ability to make desktop processors would be some serious competition.

The FTC would never let Samsung buy them.  More like is Intel and nVidia splitting up AMD.  Apple, IBM, or Microsoft(god help us).  Apple would truly be limitless, and then they could patent troll with all of AMD's resources.

[Derangel]

1 minute ago, SurvivorNVL said:

The FTC would never let Samsung buy them.  More like is Intel and nVidia splitting up AMD.  Apple, IBM, or Microsoft(god help us).  Apple would truly be limitless, and then they could patent troll with all of AMD's resources.

IBM wouldn't be able to find the capital required to buy AMD's CPU division and all of it's patents. Apple would be interesting as well.

[Castdeath97]

1 hour ago, SurvivorNVL said:

I have some faith in AMD, but I'd rather see an all out war between nVidia and Intel with all of AMD's assets.  Especially given in the future - Intel could possibly acquire nVidia, and then there's no stopping them.  So, the preferable outcome is AMD's death, Radeon in Intel's hands, and AMD64 / a new x86 license in nVidia's hands.  Just imagine a Volta APU.  Yeah.  If nVidia got the CPU division it'd likely be two to four years, but then again, we would all be with Intel CPU's for that time if Zen was a flop anyway, so, no real change.

That and I can just imagine the fun on the market if there was the announcement, "Intel to buy AMD's Radeon Technologies Group."  Just be a giddy day for shareholders.

That would be a worst case scenario with AMD dying, there is no way they would just die now. All they need to do is beat the 950, 960 and come up with a good GPU just under the 1070. It's not like they need a magical solution, they have beaten Nvidia before in the Mid tier in terms of performance.

[The Benjamins]

3 hours ago, WereCat said:

He got over 1900MHz on the Founders card and he couldnt go higher because of thermals and almost 2100MHz on the EVGA card with the reference PCB with DEFAULT voltage. Now I know that the voltage changes thanks to GPU Boost but this was on default BIOS as well. You can most likely push this even further with custom BIOS on the reference card with custom cooler.

yes but my point is a stock power limit the card will throttle on normal boost. and he showed that in the video. 

[ApolloFury]

The GTX 1080 and 1070 was as expected and been happening for years, how is that even a surprise? I am actually disappointed that it doesn't even have HBM2, I really want fast memory for 4K gaming.

[WereCat]

55 minutes ago, ApolloFury said:

The GTX 1080 and 1070 was as expected and been happening for years, how is that even a surprise? I am actually disappointed that it doesn't even have HBM2, I really want fast memory for 4K gaming.

Memory speed doesent matter as long as it is not bottleneck. And GDDR5X is plentifull for GTX 1080 so far.