Polaris not so bright [Hardocp]From ATI to AMD back to ATI? A Journey in Futility

[linkgoron]

Wait, he said this?

Yes, that is a direct quote:
https://hardforum.com/threads/from-...-in-futility-h.1900681/page-8#post-1042319107

After asking for a source, and stating that the A9 doesn't show such a reduction, but only a 8% reduction he said:

No I do not. It is a number that has been thrown around to me by engineers over the past year. Obviously it all comes down to transistor design, right? You threw out an argument, I gave you a reply. If you do not like my answer, I am good with that.

As stated, I don't really know anything about fabrication, so you guys can tell me how right/wrong he is.

 

[DooKey]

Kyle is wrong there. Difference is pretty minimal, same metal pitches for 14/16nm but 14nm has slightly tighter gate pitch.

He's not a process engineer it appears.

 

[AnandThenMan]

As stated, I don't really know anything about fabrication, so you guys can tell me how right/wrong he is.

He's talking out his you know what basically. Quite surprised he actually said what he did honestly.

 

[krumme]

Kyle shows why its no easy task to have and build good technical marketing. Amd better leave that to the oem. To darn difficult. Lol.

Idea for job interview case. How to build good relations with Kyle.

 

[CakeMonster]

I seem to remember an incident from 10+ years ago where [H] made a big whiny deal with an article about not receiving a review sample (obviously they were not considered big/relevant enough at the time. Anyone remember this? Its too long ago for me to recall the product but I think it was a GPU.

 

[krumme]

So any comments from people that participated are now moot since, you know, gravy train?

Looks boring to me. Good he evaded that by his crazy reviews.

He can stay impartial to amd then. Might consider doing the same on nv and all the oem. Can save a lot of gif gold medals and easy traffic on their www server

 

[crisium]

Oh my. Kyle has lost his mind. 232mm P10 was meant to compete with 314mm GP104.

Wow.

It's the Pitcairn of this gen, and the GP104 is the GK104. How can he not see this? The die ratio between the two is nearly identical.

The fab process means P10 is late. You can criticize AMD for this choice.

HBM2 for both variants of Vega, rather than just the higher chip, means Vega is especially late. So, equivalent to AMD launching 7800 first and 7900 series had to wait. You can criticize AMD for this choice, since it's arguable that GDDR5X could have been tried. AMD is ceding the top end performance to Nvidia for quite a time from this choice.

But he really thinks 232mm is supposed to compete with 314mm. And that AMD always intended it to do so.

Tech journalist.

 

[Silverforce11]

The only significants things I gathered from that article is that AMD is in trouble (nothing new) and Intel might buy RTG. If the latter is true, then that could be a huge boon for PC gaming with Radeon GPU's getting made at Intel's fabs. Then again, by the time that happens it'd be 4+ years and Intel's manufacturing lead will likely have eroded more so than right now.

That's surely no more than some of us here have been saying that RTG is spun off to be bought by Apple, Samsung or Intel. Heck, that's what I said last year.

Is it a possibility? Heck yes. Does it matter? No. If any of these other companies get ATI, we would have REAL GPU wars on our hands. It's not even about consumer GPUs, but HPC accelerators, and the consumer stuff just comes on the side as a consequence.

Ultimately it doesn't matter because it's just random speculation. AMD is clawing back marketshare, and their latest and greatest in Zen, Polaris, Vega is about to arrive. Their share prices went from $1.86 to ~$4.20 last I checked on an up trajectory. With all their semi-custom wins and deals, they are clearly on the rebound.

If [H] made this rant a year ago, it would be more believable that a struggling AMD could sell off RTG.. but it ain't happening now.

 

[Owls]

we'll see how things play out in a couple months

Kyle will either end up getting ridiculed or vindicated

he's definitely not going to get any freebies from AMD anytime soon though

wow hardocp, that dumpster fire of a website's arrogant fatty still hasn't kicked the bucket yet? people still care what he has to say?


Insulting others is not allowed here. This is a technical forum
Markfw900

 

[ReignQuake]

I seem to remember an incident from 10+ years ago where [H] made a big whiny deal with an article about not receiving a review sample (obviously they were not considered big/relevant enough at the time. Anyone remember this? Its too long ago for me to recall the product but I think it was a GPU.

I seem to remember something but forgive me if this isn't it. [H] allegedly posted their own benchmarks for a GPU and allegedly they were blacklisted for it. They allegedly claimed that they were ordered to only post benchmarks that made the GPU look the same or better than the competitors product?

Did I type Allegedly enough guys?

 

[Silverforce11]

Speaking about the P11 demo, we also know that it ran at >2x perf/W of 950 even 6 months ago.

R9 480X could match 1070, and run at 86C for $299. Yet, it would still live up to Polaris being "slower and hotter". For the rest of us, it would be a smashing deal.

Frankly, I'm not seeing how bad it can get really? At worst, 480X is 2304/1266 and will beat 390X by 10% and thus tie with Fury non-X. Or 10% slower than 1070. Even if all they did was port Hawaii to 14nm, and nothing else, it would still get there at 150W.

So, worst case is 10% slower than 1070, 10% lower perf/W, $300. Still a good deal!

Right. And being affordable with that level of performance makes all the difference. It will allow custom AIB cards to run cool, quiet, have great perf/w and still be cheap enough for the masses.

Best case scenario though? 2560 SP @ 1.5Ghz, a lot of performance for mainstream prices.

 

[Silverforce11]

Just wondering, what are people expectations on cards for all tiers vs Nvidias lineup? What will be vs 1070 and 1080 if Vega is vs 1080ti? Will there be an empty spot in AMDs lineup? Or will the "high end" Vega have something vs the 1080?

Vega 11 and 10.

The smaller one is GP104's competitor. The bigger one is 1080Ti's competitor. These won't be cheap due to bigger die sizes + HBM2.

Polaris 11 and 10.

The bigger one is GP106's competitor. The smaller one is GP107's competitor. These can be cheap because small die size + low TDP/PCB complexity + GDDR5.

 

[crisium]

^How he doesn't see that this is how things are meant to line up is beyond me.

If Kyle were criticizing small Vega being extra late for using HBM2 instead of GDDR5X I'd understand.

AMD's was caught with its pants down to be sure. Their process is late. HBM2 is even later. But unless he has _insider knowledge_ of P10 being more than 10-15% slower than a 1070, I do not see performance being an issue.

AMD did not intend P10 to compete with 1080 and is not having a heart attack because it is going to be slower. "AMD has a problem on its hands, as both these products have come up significantly short of where these were supposed to land." Utterly doubtful. Unless AMD also intended the 7870 to compete with the 680, in which cause man you think they would have learned their lesson.

 

[SolMiester]

I can already hear the fanboys ripping into my thoughts here yelling about the HardOCP's NVIDIA and Intel bias. This editorial is based on what I hear throughout the industry from insiders that know what is going on inside of AMD. AMD is not in a good place and there is no shortage of current and recently departed employees that are eager to share their take on what's going on. Lisa Su has lost command of her ship, and Koduri is about to put it on the beach and hope he can make it to shore in his Radeon Technologies Group lifeboat.

Well, he was right there!..LOL

 

[Silverforce11]

@crisium
If you still think AMD is late, I suggest you read their finance investor interview with JPMorgan. On track for mid-year launch, volume availability starting Q3.

http://seekingalpha.com/article/397...al-technology-media-telecom-conference?page=2

With what @zlatan said about inviting top developers for an E3 launch back in Feb, and everything else AMD have said, it's quite clear they intend to launch at this time period and it hasn't been changed, move forward or delayed.

 

[Silverforce11]

Well, he was right there!..LOL

Lisa Su has lost command of her ship, and Koduri is about to put it on the beach and hope he can make it to shore in his Radeon Technologies Group lifeboat.

That's just utter stupid and propaganda.

You know what's real though?

AMD's stock prices. $4.4 and climbing. Not a sign of their leadership in-fighting or losing the ship about to beach it and bail.

 

[csbin]

http://www.hardocp.com/article/2016/05/27/from_ati_to_amd_back_journey_in_futility#.V0jmluRf2Uk


LOL

Let’s start with where we are currently. Full disclosure - HardOCP was not invited to this weekend’s launch in Macau as AMD PR has made a decision to no longer brief this site with the rest of the industry. That’s well within AMD’s rights to do, but that is telling as well. The fact that the launch is in Macau is also very telling. AMD has a reputation of holding GPU launches in exotic locations when trying to obscure a deficient product that doesn’t cut the mustard ( Anyone remember Tunis? ) – this is a Chris Hook special and something AMD’s current head of marketing has been doing for years. While you will no doubt hear AMD wax poetic next week on the merits of Polaris 10 and Polaris 11 and 14nm goodness the reality will be quite different.

 

[ZGR]

Can we crowdfund Kyle a Fury Nano?

 

[C@mM!]

Intel considering licensing Radeon tech because Nvidia is asking a lot more to renew their deal with Intel has been an ongoing rumor for a while now.

Its not so much licensing Radeon tech, its licensing patents so they don't get sued into oblivion.

I doubt AMD would license its actual tech to Intel, unless it stood to make a fucktonne of money out of it. AMD's entire CPU\GPU strategy hinges on APU's, so needs both sides of the business. Hell, it was the reason AMD\ATi merged in the first place. And anything that risks that won't be considered.

As for the rumour of RTG breaking away and being bought by Intel, keep dreaming.

 

[monstercameron]

Can we crowdfund Kyle a Fury Nano?

I'd donate if that'd stop the whining...

 

[crisium]

@crisium
If you still think AMD is late, I suggest you read their finance investor interview with JPMorgan. On track for mid-year launch, volume availability starting Q3.

http://seekingalpha.com/article/397...al-technology-media-telecom-conference?page=2

With what @zlatan said about inviting top developers for an E3 launch back in Feb, and everything else AMD have said, it's quite clear they intend to launch at this time period and it hasn't been changed, move forward or delayed.

I don't mean late to their own plans - but to Nvidia's. It's all relative. Make no mistake, launching after Pascal is not desirable for AMD. Even when they launch first for 5000, 6000, 7000, series they can't secure 50% market share.

 

[topmounter]

I don't mean late to their own plans - but to Nvidia's. It's all relative. Make no mistake, launching after Pascal is not desirable for AMD. Even when they launch first for 5000, 6000, 7000, series they can't secure 50% market share.

Sure, but AMD can't control when nV decides to paper launch their cards. AMD just needs to stick to their game plan and execute. The 1070 still has a couple weeks before even the FE cards are available.

 

[VulgarDisplay]

Does anyone actually believe amd won an apple contract with a hot, loud, and power hungry chip? The only thing apple cares about is battery life and decent performance and if amd was significantly behind nvidia in what matters to them they would not win that contract.

This guy has lost his marbles.

Sent from my XT1575 using Tapatalk

 

[jpiniero]

Does anyone actually believe amd won an apple contract with a hot, loud, and power hungry chip? The only thing apple cares about is battery life and decent performance and if amd was significantly behind nvidia in what matters to them they would not win that contract.

Apple isn't using nVidia moreso because of the lawsuit than anything else.

 

[JDG1980]

I don't mean late to their own plans - but to Nvidia's. It's all relative. Make no mistake, launching after Pascal is not desirable for AMD. Even when they launch first for 5000, 6000, 7000, series they can't secure 50% market share.

I think there's a real chance that Nvidia advanced the time frame for the GP104 launch to beat AMD's Polaris to market. AMD isn't late; it was Nvidia that released earlier than expected. And they certainly deserve credit for being able to manage such an aggressive timeline. But AMD is still on schedule. They've been saying all along that Polaris will be released mid-year and it looks like that's going to be literally true - NDA lift (and presumably concurrent launch) on June 29, can't get much closer than that.

I don't buy Kyle's claims about Polaris being less efficient and having worse performance than expected, and frankly I think that's just the result of butthurt for being declined a Macau invite, and isn't based on any inside sources at all. We've already been given some pretty good indications of what to expect and I don't see that changing. Polaris 11 is aimed at laptops and will offer "console-level" (read: Pitcairn-class) performance at low wattage (40W-60W). Desktop versions are more of a question mark, but might come close to Tonga performance if clocks go high enough and the memory bottleneck doesn't hold it back too much. Polaris 10 is being pitched as a "VR Ready" card at a lower price. That means at least Hawaii-class performance (anything lower fails SteamVR's "VR Ready" test, which was specifically cited in this context). Pricing will be whatever AMD thinks is competitive. The biggest question mark is how high Polaris 10 can go on the top SKU. I wouldn't be surprised if it managed to match Fury X/980Ti/1070 at ~180W, but also wouldn't be surprised or disappointed if it fell a bit short of that.

As so many others have pointed out, it was disingenuous of Kyle to suggest that Polaris 10 was meant to compete with the much bigger GP104. I do think it's possible that the smaller Vega chip (formerly Greenland) was supposed to be released sooner, but ended up delayed by HBM2 issues. If AMD's plans went awry, that's probably where it happened. I suspect if they had to do it over again they would have given Vega a 384-bit GDDR5X bus for quicker time-to-market, and saved the next-gen HBM for the big SKU to compete with GP102.