Technology Leadership conference. Vega coming within the next two Q's & more

[Elixer]

http://wsw.com/webcast/pc34/amd/index.aspx Lisa Su talks about lots of stuff... (unless search is still broken, I didn't find this link mentioned before)

Some of the highlights from this Technology Leadership conference audiocast...

More technology licensing.
Vega coming within the next two Q's (I assume that means full ramp in '17, sporadic availability in '16?)
Polaris yields are NOT meeting demand... trying to ramp faster.
She once again stated they are working with TSMC as well. (Just consoles, or maybe Vega?)
Zen on track. (I assume full ramp in '17 as she was saying in her earnings report)
They do got 3 SoC wins, if it isn't nintendo, then whom?
She skirts around the nintendo question. They got a good relationship with them she says. (Is it possible they got more than one machine coming, like a main console, and another, portable version?)
VR...Playstation VR, lower end VR... blah blah blah.

 

[frozentundra123456]

http://wsw.com/webcast/pc34/amd/index.aspx Lisa Su talks about lots of stuff... (unless search is still broken, I didn't find this link mentioned before)

Some of the highlights from this Technology Leadership conference audiocast...

More technology licensing.
Vega coming within the next two Q's (I assume that means full ramp in '17, sporadic availability in '16?)
Polaris yields are NOT meeting demand... trying to ramp faster.
Zen on track. (I assume full ramp in '17 as she was saying in her earnings report)
They do got 3 SoC wins, if it isn't nintendo, then whom?
She skirts around the nintendo question. They got a good relationship with them she says. (Is it possible they got more than one machine coming, like a main console, and another, portable version?)
VR...Playstation VR, lower end VR... blah blah blah.

Two quarters from now could actually be Q1 2017, since this is Q3 2016 already.

Anyway, I hope at some point we get actual sales figures on 1070/1080 and 1060/48x. Would shed more light on the actual demand for the cards vs limited supply, or obviously, probably a bit of both. Not sure it is legitimate to blame GF for the supply issues (which I assume is the meaning of your thread title), demand may just be very high. More questionable is the suitability of their 14nm process for high frequency applications.

 

[boozzer]

they need to release for the chrismas shopping season. at least show up with a few thousand cards. Q1 2017 release doesn't do much. that just screams super bad timing.

 

[IllogicalGlory]

3 wins - Scorpio, Neo, Slim. No need for Nintendo.

 

[Arachnotronic]

Next two quarters...we're in Q3 now, so we're looking at a Q1 2017 launch, as expected.

NVIDIA literally has the high end of the market all to itself with GP104/GP102 for that time.


Trolling/off-topic on an AMD thread, we don't need to hear about Nvidia here
Markfw900

 

[Arachnotronic]

Two quarters from now could actually be Q1 2017, since this is Q3 2016 already.

Anyway, I hope at some point we get actual sales figures on 1070/1080 and 1060/48x. Would shed more light on the actual demand for the cards vs limited supply, or obviously, probably a bit of both. Not sure it is legitimate to blame GF for the supply issues (which I assume is the meaning of your thread title), demand may just be very high. More questionable is the suitability of their 14nm process for high frequency applications.

NVIDIA reports financial results later today. The revenue numbers (most recent QTR and guidance for following qtr) will tell you everything you want to know about Pascal sales.

 

[raghu78]

Lisa Su confirms they are using TSMC 16nm along with GF 14nm. Thats good news. I hope Vega is built at TSMC 16FF+ as TSMC has the best foundry FINFET process and more than a decade of experience with manufacturing big GPUs with die size > 500 sq mm. AMD has also used TSMC 16FF+ to build Xbox One S so that basically serves as a process pipe cleaner for Vega.

http://www.eurogamer.net/articles/digitalfoundry-2016-inside-xbox-one-s-tech-interview

AMD need TSMC 16FF+ along with a significantly improved Vega architecture to compete with GP102 / GP104. Polaris is nowhere near competitive against Pascal in terms of perf/watt. TSMC 16FF+ has superior transistor performance than 14LPP+. Moreover GF has a pretty pathetic implementation of Samsung 14LPP. Since AMD is going to be power limited in trying to compete with GP102 / GP104. Thus going with TSMC 16FF+ for Vega is their best chance of competing with Nvidia GP102 / GP104.

 

[Elixer]

AMD need TSMC 16FF+ along with a significantly improved Vega architecture to compete with GP102 / GP104. Polaris is nowhere near competitive against Pascal in terms of perf/watt. TSMC 16FF+ has superior transistor performance than 14LPP+. Moreover GF has a pretty pathetic implementation of Samsung 14LPP. Since AMD is going to be power limited in trying to compete with GP102 / GP104. Thus going with TSMC 16FF+ for Vega is their best chance of competing with Nvidia GP102 / GP104.

It was never mentioned what the problem is with GF, AFAIK, the 14LPP should be the exact same, they are using the same equipment. I highly doubt we will get any info either, GF is private, and I bet they have a NDA with AMD.

That wafer agreement is the noose around AMD's neck that forces them to use GF, so, it remains to be seen if Zen / Vega or other parts can be made elsewhere. I sure hope it is possible, since I don't think anyone here thinks GF is up to the task.

 

[Glo.]

I suppose using TSMC for Vega architecture is because it was TSMC that AMD worked with on Interposer GPUs.
All of the packaging is done by Amkor, but I think it would be logical to use this manufacturer for another set of Interposer GPUs especially on new, smaller node.

Now, we can start the hype train. Again.

 

[Arachnotronic]

Now, we can start the hype train. Again.

Probably not a good idea, "hype trains" have a tendency to go off the rails, even if the product itself is objectively good ("good" mind you takes into account both performance and price).

The best thing for consumers to do is to keep expectations in check and wait for benchmarks from third party review sites. That way, one can objectively judge a product based not on expectations (anybody can imagine any product with any performance they'd like) but on what is actually presented.

 

[linkgoron]

If indeed Vega is TSMC, it'll be a great opportunity to see just how much GF is to blame with Polaris not hitting the required clocks (or at least perf/watt). Somehow, I have a feeling that people will be disappointed, but I wouldn't mind being wrong.

 

[railven]

Now, we can start the hype train. Again.

Probably not a good idea, "hype trains" have a tendency to go off the rails, even if the product itself is objectively good ("good" mind you takes into account both performance and price).

Yeah, I don't think AMD can survive another hype train. Seems I still got a soft spot for the old ATI fragments. Recent stroll around other forums, AMD is not experiencing the rebellion/revolution their marketing team was promoting.


The sooner they announce Zen/Vega to be coming from TSMC, the overall better perception for the two products if you ask me. GloFlo is not a selling endorsement.

EDIT:

NVIDIA reports financial results later today. The revenue numbers (most recent QTR and guidance for following qtr) will tell you everything you want to know about Pascal sales.

I don't own any stock in NV, nor do I personally care how they end up. But this is a blood bath. Intel on one side NV on the other. I don't understand how anyone thinks AMD is going to compete.

I'm starting to feel like Big Vega will just barely compete with GTX 1080, let alone GP102 and whatever comes of it for consumers.

 

[Madpacket]

I think you may be suprised as to how many RX 470/480's are actually selling. Failure can be measured in many ways.

 

[ViRGE]

they need to release for the chrismas shopping season. at least show up with a few thousand cards. Q1 2017 release doesn't do much. that just screams super bad timing.

HD 7970 launched shortly after Christmas, and it did okay. I agree that launching before Christmas is better, but for high-end gaming it's not as seasonal as the overall CE market.

 

[RussianSensation]

Next two quarters...we're in Q3 now, so we're looking at a Q1 2017 launch, as expected.

Perfect. Leaves at minimum 5-6 months for each RX 480 purchased in August to pay for itself to finance free Vegas.

On the positive side, even if Vega launches in Q1 2017, NV still has nothing worth upgrading to for most GTX980Ti users. At the very least it should encourage NV to release a 1080Ti (or w/e they'll call it) that slots between a $650-700 1080 and a $1200 Titan XP. Right now the upgrade path for GTX980Ti users is basically a next gen upper mid-range Pascal GTX1080 (25% faster max OC vs. OC) or a card nearly twice the cost of what they paid for a 980Ti. Doesn't sound like a healthy high-end market for the non-1%ers. For that reason alone, Vega succeeding is in the best interest of a lot of gamers.

NVIDIA literally has the high end of the market all to itself with GP104/GP102 for that time.

Good for them and their investors -- achieving 53% higher gross margins than Apple by increasing prices 2-3X per each GPU tier since 2012 and having majority of PC gamers buy into/defend this marketing strategy is impressive! JHH has outdone even SJ. At the same time I don't feel bad waiting 6 months to skip a $650 GTX1080 or a $1200 Titan XP when the competition is offering 32" 4K IPS FreeSync monitors for $420 and Vega will be free by the time it comes out. NV gamers get the bragging rights they always wanted by buying NV Pascal they would have bought anyway. Also, good times for brand agnostic PC gamers who don't play 3DMark Leaderboards and have fast enough rigs to wait it out 6-7 months. Sounds like we finally reached a market "equilibrium" where both sides are happy with the end result.

I think you may be suprised as to how many RX 470/480's are actually selling. Failure can be measured in many ways.

Not according to Steam hardware survey and Amazon best sellers list, although I have personally purchased no RX 480 cards on Amazon and haven't run a Steam information query on any RX 480 cards.

HD 7970 launched shortly after Christmas, and it did okay. I agree that launching before Christmas is better, but for high-end gaming it's not as seasonal as the overall CE market.

What's interesting, at least based on my observation, is that in the past many PC gamers used to upgrade at or around the launch of key AAA games. This has not been true for RX 470/480/1060/1070/1080/Titan XP. There are no killer AAA games that slaughter video cards and make gamer die for an upgrade like Far Cry, the original Doom, Metro, Crysis, BF, etc. BF1 isn't out, Mankind Divided isn't out, Star Citizen isn't out, etc. and yet all of these GPUs are selling extremely well. Most of the big games are coming out starting in October and beyond. Historically speaking Q2 was generally weak for GPU sales but NV has proven this to be wrong. The times are changing or high-end GPU sales are picking up major steam in 3rd world countries.

More questionable is the suitability of their 14nm process for high frequency applications.

That's a great point. Polaris 10/11's perf/watt was mediocre for gaming after all the AMD hype. It's pretty hard to compete when you start off with terrible perf/watt against the competition and then have to deal with the fact that the competitor's cards Pascal line-up can hit 2-2.1Ghz on air while yours top out at 1.4Ghz. That means if Vega can only manage 1.4Ghz clocks, it will have a 40%+ GPU clock speed deficit against the eventual GP102 / GTX1080Ti consumer SKU. Let's hope AMD actually went with TSMC for Vega and that they spent some time optimizing the architecture for higher clocks than Polaris 10. It's pretty remarkable for AMD to have such a gargantuan GPU clock advantage (i.e. it's literally almost 2 generations ahead of AMD) while also having a 1-2 generation lead in perf/watt.

 

[boozzer]

HD 7970 launched shortly after Christmas, and it did okay. I agree that launching before Christmas is better, but for high-end gaming it's not as seasonal as the overall CE market.

but it will help sales for sure. holiday season people spend like crazy.

 

[Rifter]

Probably not a good idea, "hype trains" have a tendency to go off the rails, even if the product itself is objectively good ("good" mind you takes into account both performance and price).

The best thing for consumers to do is to keep expectations in check and wait for benchmarks from third party review sites. That way, one can objectively judge a product based not on expectations (anybody can imagine any product with any performance they'd like) but on what is actually presented.

This would be true except most review sites are biased as well towards one side or the other.

 

[Glo.]

Probably not a good idea, "hype trains" have a tendency to go off the rails, even if the product itself is objectively good ("good" mind you takes into account both performance and price).

The best thing for consumers to do is to keep expectations in check and wait for benchmarks from third party review sites. That way, one can objectively judge a product based not on expectations (anybody can imagine any product with any performance they'd like) but on what is actually presented.

That was completely Ironic. We know what happened with Polaris launch.