Vega = RX Vega

[AliG]

Amazing, at $3, you are 4.30x, that is 330% profit if my calculation is correct, I hope you bought as much as you could. Ryzen really helped the shareholders, yes get out while you can.

Already sold at $14.50, was more than half up to 6-figures worth lol. I have a hunch stock is about to go way down based off the 1080 Ti being faster than Titan X (and more power efficient), as well as the typical disappointment that comes from reviews showing a $500 CPU isn't more powerful than 6950X. May buy back once the dust settles with the expectation that earnings will speak for themselves - but at much lower share quantity.

Whoever attended/saw both events ( AMD's and nVIDIA's ) must be wondering WTF is wrong with AMD.

And to all the "AMD will bring pre-high end performance value"... nVIDIA just ruined those dreams with the 1080 at 499 ( which could slip lower in case AMD releases Vega for less ).

Damn.

Yeah that's why I got out. The combination of all the hand waiving (wtf were they thinking with all the VR stuff and the constant heavy metal intros?) and 1080 Ti being very legit makes me think you're going to see a big drop today.

[Mechanical Man]

Why would AMD distract Ryzen hypetrain with something else, specially when Ryzen NDA breaks off tomorrow and that would be another distraction again. Capsaicin also is not consumer event.

Its just stupid to expect them to release performance numbers on product coming in Q2/17.

I think just Radeon Vega would have been better than having RX in it. Maybe go with Ryzen like naming, "Vega 7, Vega 5, Vega 3".

[kromosto]

I am not concern about what amd said or not said but price cut of 1080 will hurt them a lot.

[AliG]

Why would AMD distract Ryzen hypetrain with something else, specially when Ryzen NDA breaks off tomorrow and that would be another distraction again. Capsaicin also is not consumer event.

Its just stupid to expect them to release performance numbers on product coming in Q2/17.

I think just Radeon Vega would have been better than having RX in it. Maybe go with Ryzen like naming, "Vega 7, Vega 5, Vega 3".

I don't think you're taking the market into account. The stock is incredibly risky at its current valuation; people wanted assurances that Vega would compete given Nvidia was announcing the 1080 Ti.

Giving out promising Vega information would have further boosted the Ryzen launch. The lack of any substance, compounded by all the random bs they did (like that "spicy" tidbit that it'll be retail as RX Vega), is very telling.

[BenchZowner]

People are building new rigs now with Ryzen... most of them accompany their purchase with a graphics card.
nVIDIA's move makes things very tempting. And plenty of people are now speculating that AMD might have a flop with Vega and that's why they haven't disclosed a single performance figure.

[Mechanical Man]

I don't think you're taking the market into account. The stock is incredibly risky at its current valuation; people wanted assurances that Vega would compete given Nvidia was announcing the 1080 Ti.

Giving out promising Vega information would have further boosted the Ryzen launch. The lack of any substance, compounded by all the random bs they did (like that "spicy" tidbit that it'll be retail as RX Vega), is very telling.

In any case nVidia timing TI on same day they could have not won it. They gave some more with showing hbcc in action. And really, capsaicing is not consumer event, try to get that.

And what comes about stock price, no real investor needed any half assed performance figures that could have been given now. Its just what consumers wanted. nVidia never gives performance numbers three months before launch, there is real reasons for it, same goes for AMD. its just stupid to think otherwise.

[Mechanical Man]

People are building new rigs now with Ryzen... most of them accompany their purchase with a graphics card.
nVIDIA's move makes things very tempting. And plenty of people are now speculating that AMD might have a flop with Vega and that's why they haven't disclosed a single performance figure.

You really think it would make any difference? We dont know if Vega launch is march or june, most likely its june. Nobody would wait for that long to get the final performance. Those who would, will wait it anyway (they have good enough gpu now).

[BenchZowner]

The percent of patient people in the gaming and PC enthusiast category is... pretty low you know.

Not much can be done now, they should focus on launching earlier not taking their time.
Their release schedule has been terrible for the past 2 years, aren't you tired of them trying to catch up with nVIDIA like 1 year later with every new piece ?

[Mechanical Man]

The percent of patient people in the gaming and PC enthusiast category is... pretty low you know.

Not much can be done now, they should focus on launching earlier not taking their time.
Their release schedule has been terrible for the past 2 years, aren't you tired of them trying to catch up with nVIDIA like 1 year later with every new piece ?

If its not ready, its not ready. There is no point releasing half year ahead, it does not do any good business wise. I'm not tired of waiting, I don't do that, i buy what ever is best for me at the time of the purchase (when I have the need). Naturally I do hope AMD gets also Vega right (Ryzen looks like pretty good product, tomorrow we will know), that would do only good for us consumers.

I do have my hopes up becouse Sampsa wrote in end of the Titan X article "N?ht?v?ksi j??, kuinka paljon NVIDIAn tarvitsee puristaa tehoja irti GP102-grafiikkapiirist?, jotta se pystyy kilpailemaan Vegan kanssa." 13.2.2017

That translates to: "It remains to be seen how much nVidia will squeece more performance out of GP102 graphics chip (as TI variant), so that it can conpete with Vega".

[Tim]

Why would AMD distract Ryzen hypetrain with something else, specially when Ryzen NDA breaks off tomorrow and that would be another distraction again. Capsaicin also is not consumer event.

Because Vega complements Ryzen. They go hand in hand when it comes to marketing. There are a lot of people who want AMD/AMD in their rigs. It would not be a distraction to the hypetrain, but an extra push in the right direction. Consumers, investors need to believe in AMD again, in all their products.

And really, capsaicing is not consumer event, try to get that.

Yet it's consumers and then some that talk all about it!

[Ket]

Lets not count Vega out. While Vega itself I don't find all that interesting (or VR for that matter), I am interested in what all the Vega GPUs that don't make the "big Vega" cut become. No doubt a whole new RX line to put the semi-failure of Polaris (solid, but dodgy OC-ability hurts it a fair bit) out to pasture.

That said, I have noticed the latest batches of RX480s are clocking at around what I initially said they would - 1.45GHz or so, so maybe not all is lost for Polaris quite yet.

Still though, I hardly see how coming out and saying officially what the name of Vega will be along with a design could be called good info.. all somewhat predictable really. Surely AMD could of told people some peliminary clock speeds or something for something of even tangible interest? I mean really, the "announcement", wasn't any announcement at all. AMDs failure of a PR department at its finest again I see

[StAndrew]

Lets not count Vega out. While Vega itself I don't find all that interesting (or VR for that matter), I am interested in what all the Vega GPUs that don't make the "big Vega" cut become. No doubt a whole new RX line to put the semi-failure of Polaris (solid, but dodgy OC-ability hurts it a fair bit) out to pasture.

That said, I have noticed the latest batches of RX480s are clocking at around what I initially said they would - 1.45GHz or so, so maybe not all is lost for Polaris quite yet.

Still though, I hardly see how coming out and saying officially what the name of Vega will be along with a design could be called good info.. all somewhat predictable really. Surely AMD could of told people some peliminary clock speeds or something for something of even tangible interest? I mean really, the "announcement", wasn't any announcement at all. AMDs failure of a PR department at its finest again I see

I agree; I don't think Vega will be a terrible launch; depending on price, I think matching or even beating out the GTX1080 would keep AMD in the game. I don't know what they hoped to accomplish at Capsaicin and cream...

[AliG]

Still though, I hardly see how coming out and saying officially what the name of Vega will be along with a design could be called good info.. all somewhat predictable really. Surely AMD could of told people some peliminary clock speeds or something for something of even tangible interest? I mean really, the "announcement", wasn't any announcement at all. AMDs failure of a PR department at its finest again I see

That was my concern. They knew Nvidia had a big disclosure planned, but then just presented a poor man's hype machine about VR instead. Juxtaposing this to the relative transparency of Ryzen/Polaris, but obliquity of Bulldozer, does not leave me with confidence.

In the end I got my nut off the stock and am planning on investing most of it back into my usual allocations of index funds. Given Ryzen and Vega doing well were already priced in, I have a hard time believing they have more room for growth baring massive midrange sales.

[Aleki]

I'm going to put on my conspiracy hat for a moment and see if this all makes sense..
Based on what people have said, Vega is supposed to beat a 1080 in performance, but they didnt mention relative performance to a 1080Ti, correct?
What if the "spicy tidbit" at the end was supposed to be a contingent announcement? Like, what if they were looking to see what NV had in their hand before revealing that the card is capable of "x/y/z"?
Considering they are still months away from release, it is possible AMD is playing something close to its chest(much like when they made the claim that Zen was a 40% IPC improvement, and we all assumed that figure was compared to bulldozer IPC).
It isnt a far-fetched idea to think AMD will trickle more information nearing VEGA's release in order to keep the hype train moving

[AliG]

Doubt it.

I'm already somewhat skeptical that Ryzen actually is as impressive across the board as it is with Cinebench, simply because of the AVX2 limitations, their very selective usage of benchmarks that don't use those instructions and the launch prices. We'll find out for sure tomorrow, but the leak from the Iranian site already seems to confirm my suspicions there.

If Vega was a big winner compared even just 1080, I think they would have given out more teaser info.

[LordEC911]

So people in this thread, many of whom are widely praised for their tech/hardware knowledge, believe Nvidia made moves and dropped prices out of the goodness of their heart?
What world are we living in?

[AliG]

You drop prices to squeeze the bottom out of your competition.

Nvidia pulls much better yeild per GPU than AMD right now, and have gotten away with overcharging for a good 2 years now. Then introducing the 1080 Ti yesterday is the same tactic they did with the 980 Ti: you leverage your competitive advantage to gain a stranglehold on the market.

AMD has to now reportedly deliver something faster than Titan X now if they want to keep in the $700 price range. I just don't think they can deliver that based off yesterday's debacle.

[LordEC911]

You drop prices to squeeze the bottom out of your competition.

Nvidia pulls much better yeild per GPU than AMD right now, and have gotten away with overcharging for a good 2 years now. Then introducing the 1080 Ti yesterday is the same tactic they did with the 980 Ti: you leverage your competitive advantage to gain a stranglehold on the market.

AMD has to now reportedly deliver something faster than Titan X now if they want to keep in the $700 price range. I just don't think they can deliver that based off yesterday's debacle.

??? They needed something to do with binned GM200... and when they released it, they didn't shift the rest of their lineup by ~$100.

AMD doesn't necessarily need something faster than Titan XP to price it up there...

I think people are over-reacting to a presentation aimed at Devs and interestingly enough... heavily underestimating Vega when everything we know about it shows them going in the right direction. Yes, I understand the proof is in the pudding but there isn't any indications of Vega being a flop other than people connecting nonexistent dots.

[StAndrew]

??? They needed something to do with binned GM200... and when they released it, they didn't shift the rest of their lineup by ~$100.

AMD doesn't necessarily need something faster than Titan XP to price it up there...

I think people are over-reacting to a presentation aimed at Devs and interestingly enough... heavily underestimating Vega when everything we know about it shows them going in the right direction. Yes, I understand the proof is in the pudding but there isn't any indications of Vega being a flop other than people connecting nonexistent dots.

I don't think the TI is made from binned chips. They make mention of silicon changes and tweaks for better efficiency and performance.

[AliG]

I don't think the TI is made from binned chips. They make mention of silicon changes and tweaks for better efficiency and performance.

The Ti chips are typically just cut down versions of the full Titan X versions. Nvidia castrates the double precision compute shaders to protect their deep learning market, then deactivates 1 shader cluster for good measure and sells a server chip as a high end gamer.

My guess is the extra clocks come from a combination of improved manufacturing process and the deactivation of shaders (especially the double precision logic).

[Ket]

Its easy to be skeptical of AMD with the massive failure it became for 5-7years but lets put a few things in perspective here.

1. The RX470 and RX480.

Anyone care to name a better bang for your buck card than the 4GB RX470? No? Thought not. This card is by far and away the ultimate price/performance card out there right now, and as far as OC potential goes.. its ok, not stellar, but you'll get perhaps another 10% from it. In the UK anyone can go and pick one up for 150 (why in the hell does the pound sign STILL not work in these forums?) for that money you can't deny how good the card is.

Now, the RX480. I said this card could be a dark horse way back and IMO its become exactly that. Newer batches having it OC right on what I said theoretically it would be capable of 1.4GHz - 1.5GHz kind of range with a average OC being 1.4GHz -1.45GHz and even the odd report of almost 1.6GHz. The 8GB RX480 in the UK can be picked up for 185 and thats with a code for Doom. Certainly not bad at all.

So, right here we can see that teething issues aside AMD have got the upper-mid range segment nailed pretty well. I'd say the extra vRAM of the 480, or 470 if you find one at a bargain price, combined with DX12 and Vulkan performance that seems to continuously gradually creep up makes either card a better choice for lifespan than any 6GB 1060. The RX470/480 has plenty of performance to challenge competing cards, AMD are proving by doing that they are getting things right again, despite some stumbles on the way.


2. Zen. (Ryzen sounds crap, wtf AMD )

The current gorilla in the room. Lets not forget this is a brand new architecture that is a colossal 52% IPC improvement. No matter how you slice that, it will compete with intel, and kick a large chunk of intels overpriced CPUs in quite nicely. AVX is a little questionable, but its fair to say Zens performance on this instruction set is still going to be good, which is perfectly fine for pretty much everybody. Also remember new UEFIs, platform drivers, software updated and optimised for Zen, are all going to only but improve things further.

Theres no such thing as a perfect architecture, all things being the same generation something will always perform worse on one architecture than it does on another. This is not unusual and certainly doesn't mean the performance when the CPU comes up against its "weaker" abilities will be bad.

The only real difference the "X" models seem to have are a better cooler, the CPU automatically clocking higher by itself, and possibly better binning than non-X models, suggests the general rule for Zen could likely be X models hit 4.6GHz - 4.8GHz, while non-X models hit 4.3GHz - 4.5GHz, only needing voltage in the ballpark of what "X" models need for 4.6GHz - 4.8GHz. Thats not bad at all, particularly if those 8 core monsters are the ones hitting those speeds which evidence would suggest is very possible as it looks like CPUs that don't make the R7 grade just get cut down to the R5 grade. Its anyones guess how that might effect OCing as the CPUs won't be native Hex or Quads, but my guess would be it won't effect them much.

Some stumbles (maybe) here and there but AMD are showing again by doing that they have something that will rival competing CPUs, and lets be honest, I don't think a single one of us will pay much attention to slightly lower IPC in exchange for noticeable extra cores/threads for the same money intel are asking for quads with 4/8 threads.


3. Consoles

Somehow, AMD figured out how to make money off of a modified Jaguar based APU for XB and PS4. Providing hardware for 2 out of the 3 console manufacturers is not only going to give them cash to make even better stuff, but come on, AMD are making money off of something none of us would be caught dead with. They deserve credit for that.


All of this is just a brief breakdown to highlight one thing; its easy to be skeptical of a company that looked like it was minutes away from sinking entirely but this emergency pumps and all hands on deck with buckets have done the job. AMD are back, they have been smart, got some solid stuff with better days to come, and a consistent source of cashflow in their back pocket providing hardware for XB and PS consoles.

[StAndrew]

The Ti chips are typically just cut down versions of the full Titan X versions. Nvidia castrates the double precision compute shaders to protect their deep learning market, then deactivates 1 shader cluster for good measure and sells a server chip as a high end gamer.

My guess is the extra clocks come from a combination of improved manufacturing process and the deactivation of shaders (especially the double precision logic).

Thought there was some silicon tweaks to run the 11GB/s memory but I guess I misread that section.

[Mechanical Man]

SNIP

Well said there. I don't know why anyone is dissapointed for amd being little behind ipc wise on games, when just few months ago the gap looked a lot bigger.

And if the performance for watt is what it looks like, Ryzen will be hit in server market, that is what matters even more for the bottom line of the AMD.

Vega looks like good gpu and having the right direction, but it really is too much to ask any accurate performance numbers from chip that is not ready and will not be launched in many more months. I have not seen anyone demanding that from nvidia. Just looks like fanboyism and double standards.

[AliG]

Well said there. I don't know why anyone is dissapointed for amd being little behind ipc wise on games, when just few months ago the gap looked a lot bigger.

And if the performance for watt is what it looks like, Ryzen will be hit in server market, that is what matters even more for the bottom line of the AMD.

Vega looks like good gpu and having the right direction, but it really is too much to ask any accurate performance numbers from chip that is not ready and will not be launched in many more months. I have not seen anyone demanding that from nvidia. Just looks like fanboyism and double standards.

Depends on the application though. The results Fugger put out show that AVX512 makes an enormous difference in certain server applications (like 4x faster per core difference).

Kinda reminds me of the old Athlon 64 Opeterons vs Netburst based Xeons. The Xeons would actually be quite a bit faster if the software was recompiled to utilize SSE3, but anything that was memory dependent tended to be faster (and far more power efficient) on the Opterons.

[Ket]

That just all comes back to different architecture strengths and weaknesses I mentioned, Zen will do well, and I'd be surprised if Zen+ doesn't support AVX512. The real question I have right now is just how fast can Zen become, theres a awful lot of flexibility in its design. It could well be the GCN of the CPU world.