AMD might launch its NAVI 10 next-gen GPU by June

#5661632 Posted on: 04/19/2019 09:47 PM

---

For particular set of reasons to which I will not go into here, I consider Vega architecture an fail with one redeeming quality which is compute. But this "failure" in form of Vega 7 has quite interesting comparison to RTX 2080.
2080 which you can buy at about 10% higher price than Vega 7 has ~3% more transistors and ~3% higher clock.
On 1080p Vega 7 is about 12% slower than 2080. On 1440p, they are about equal.
If those 2 values are adjusted to performance per transistor count per clock, Vega 7 is only 6% behind on 1080p and gets ahead on 1440p.

Then one can say that there are only following technological differences which actually have relevance:
- Vega 7 has around 30% higher compute power
- RTX 2080 eats around 30% less energy
- RTX 2080 has RTX, VRS, ...

But from general gaming perspective they are quite equal. So where does it leave Navi 7nm and Turing 7nm?
Well, 1st for Turing. As @las stated, it was meant to be 7nm. But would it be really much stronger? 2080 Ti has already 18.6 Billion transistors. That would not be any cheaper on 7nm. It would be more power efficient, and maybe clock bit higher as result. But would people buy 22~24 Billion transistor graphics card costing another 40% more than 2080 Ti? Nothing else could be expected since it would be same architecture.

Now there is Navi. It is not Vega, no GCN, it is GPU where AMD did learn from their mistakes (or lack of foresight, which was pretty bad).
What is biggest variable for Navi? With GCN, AMD changed multiple times number of transistors per building block drastically. (Vega 64 has same count of all building blocks as Fury X, but 40% more transistors... Sadly... Clock to Clock it does no better in games than Fury X.)

So, Navi is kind of wild card till we see actual stats of 1st few different chips. Then we can approximate its transistor cost per building block in comparison to latest GCN. Power consumption will tell us a lot too in terms of actual size limits. Will AMD spend its transistor budget worse than with Vega or better than Fury X? And there is a lot of space between.
= = = =

So, did AMD spent all those years building something new to end up with same power efficiency as Vega? With same transistor cost per building block? I do not think so.

#5661634 Posted on: 04/19/2019 10:16 PM

---

So just because they release on a holiday means that if its lacking in performance this means that a reviewer will give it a bad review because its a waste of their time?

If ANY reviewer knowingly gives a bad review just because its not insanely impressive or it doesn't beat a GTX2080Ti by 50% dont deserve to be reviewers! Its their job, they chose it, they should always be professional and truthful.

I hope you really don't actually feel you need to explain the concept of professional integrity to me, as if it might be something I'm not already keenly aware of.

Let me put it this way, and after four posts in this thread, I'll then let it go.

If I worked at AMD, and had Dr. Su's attention, I'd tell her the "cuteness" associated with launching a 7nm product on the 7th day of the 7th month means exactly ZERO in terms of the making a great first impression with the reviewers. You know, those people who are going to, in turn, create the first impression regarding this product with our potential customers. The "7" gimmick is just that; it's absolutely meaningless and will not likely be the cause of even a single sale.

I'd offer that lifting a launch embargo on a Sunday, and here in the U.S. a Sunday that is likely part of a mini-vacation that some people have had circled on the calendar since January, probably might not be the best possible plan. I'm not suggesting reviewers would deliberately put their thumb on the scale due to this, but the tone of the words and phrases that someone chooses to frame an evaluation can be crucial.

I'm just of the opinion that the reviewer at Hardware Extravaganza Deluxe would rather be in a lounger on the patio, beer in hand, watching a baseball game instead of enduring yet another iteration of the Far Cry 5 benchmark, while listening to his or her spouse ask, "Are you going to be done soon? You've been at it all day."

Human nature being what it is, the demeanor of the resulting review might not be as neutral as it should be. And that would be both unfortunate, and unnecessary. Navi should be afforded every opportunity to stand or fall based solely on it's own merits, and not potentially hobbled, even a tiny bit, due to a silly date for lifting a review embargo. Since I don't work for AMD, I don't have any insight on when the reviews will go "live" any more than anyone else has. I'm simply saying Sunday, July 7th, certainly wouldn't be my first choice.

#5661656 Posted on: 04/20/2019 01:26 AM

---

You missed something here, it makes a huge deal for marketing. It makes the products easier to sell, sure its a gimmick and means nothing in terms of the product. But for creating a buzz for it, it makes perfect sense to release it on this day. AMD need all the publicity they can get after the long 10 year span that Intel has slapped them into the ground.

If any reviewer knocks it because it stopped them from going to the beach on the holiday is not worth listening to.

As long as AMD get the chips into reviewers hands before hand I see no issue with the release date rumoured.

#5661661 Posted on: 04/20/2019 02:16 AM

---

Not every reviewer is based in the US, so why exactly is a July 7 release a problem?

#5661768 Posted on: 04/20/2019 05:59 PM

---

Just to clarify what I meant, AMD graphic card launches have a history of being marred by one thing or another. The Radeon VII with it's somewhat broken release driver, the deafening roar of the reference R9 290X in "Uber" mode, the Fury X and it's apparently less-than-silent liquid cooling, and the RX 480 with the controversy of the single six-pin power socket being "marginal" are a few examples. It seems like it's always AMD providing the axe-handle for their critics to use to beat them up with. I thought that when they finally ditched the blower-style cooler on the Radeon VII, it was a sign that the lunatics were no longer running the asylum at RTG. But, reportedly, the three radial fan arrangement wasn't what it should have been from an acoustic standpoint.

The examples I listed above are, I believe, widespread perceptions held by the buying public. I'm certain there are any number of people who have actually owned, and used, these products that would steadfastly insist those perceptions are inaccurate. Just like the, "Radeon drivers are complete garbage" reputation that always leaves me totally mystified, as the last driver issue I can recall was when my first Radeon, a 9700 Pro, and Half-Life 2, were both all shiny and new. And, while I feel certain it's unwarranted, that perception never seems to go away. But, this isn't a thread devoted to that discussion.

When you're chasing a formidable competitor not only for market share, but for mind share as well, your product launches need to be as perfect as you can possibly manage. Near flawless. A well-performing graphics card, neither loud or toasty, with a ready-to-rock Day One driver package, and plenty of stock on the shelves at a price that screams, "Add to Cart." When AMD elected to release Vega 64 quite literally on the heels of Threadripper, and over a weekend, some hardware reviewers, (people who do indeed make money, but aren't necessarily paid), were left in a rather stressed-out frame of mind. I'm hoping AMD took note of that.

Maybe they are indeed drama-queens, but with as much as AMD has riding on Navi being well-received by the people performing the independent evaluations, I'm simply hoping they won't schedule their review sample delivery and subsequent embargo lift so as to tempt U.S.-based reviewers as to feeling they were deprived of time with family and friends, and to be in a cranky, bomb-throwing state of mind.

Wow...I guess you missed the horrifically marred launch of nVidia's RTX products to date... Live under a rock, eh? nVidia couldn't even ship a usable *demo* of the RTX stuff when they shipped them--just a lame 2d videoclip rendered on much more expensive hardware by professional FX people in anything *but* real time. And DLSS is an ongoing debate. From my point of view, nVidia PR has always been way, way over the top--which is just one of several reasons nVidia's products don't appeal to me. It doesn't bother me that you obviously think nVidia is the way to go, heck, no. But try and get your facts at least somewhat straight on what the competition is doing. The fact is that if the FTC in the states or the EUC in Europe knew much of anything about the proper marketing of GPUs (likely, they don't even care) nVidia would be having to rescind a few of its marketing claims--such as "real-time ray tracing," etc. The phrase is actually an oxymoron.

If the mainstream version of Navi sells for < $300 and is 1.5x-2x as powerful as the RX-590, it will be a massive hit for AMD, no doubt about it, and I'm all in. My thought is that the RX-590 is just a stop gap on the way to mainstream Navi. As far as "bang-for-the-buck" goes, whether it's a CPU or a GPU, nobody does it better than AMD.