AMD To Chase Market Share Over High End With First Polaris Offerings

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.
I think that market perception is very important here. The 1080 is a beast and the 1080 TI and titan XI is going to be even more powerful. The mid range segment is important but the upper tier cards drive the overall perception of a companies ability to provide cutting edge technology. Nvidia is winning a PR war here, just like Intel has. Most people don't sit down and look over the real benchmarks. The few 1080 cards that will actually be sold are nothing in comparison with the huge number of 1060 cards that will be sold in mainstream package rigs (Ie laptops). But the 1080 cards will drive the purchase decision for the scores of 1060 mobile because the sticker says Nvidia and Intel inside on it (remember that Nvidia is entirely discrete) and that must make it better. Integrated units are a lost cause. The Integrated game is going to ultimately be slugged out between AMD and Intel with no clear winner in the segment due to the good enough mindset of mainstream purchasers and Intel market dominance in the CPU category. All I can say is that they better get their flagship out fast and it had better go to to toe with Pascal if they are going to maintain viability in the discrete graphics area.
 

Underrating has absolutely nothing to do with it.

The magnetic core used in the main transformer either can or cannot deliver the power, there is next to no room for 'maybe' in-between. For the output caps, your either have or don't have sufficient capacity to handle fast transients. There is very little middle-ground for compromises there either. A decent quality PSU like the Earthwatt 380AD meets those requirements, lesser quality PSUs don't.
 


Usually the PSU would shut off from a protection before the theoretical power limit is actually reached, which usually makes it negligible. Or as the guys told me over on Jonnyguru, it'd be a rare scenario where the power limit is actually ever reached, since units will either:

1) Burn
2) Shut off from protection

Before it's reached. And by time it is reached, the low voltage is more of a concern than anything; but by then UVP would have kicked in.
http://www.jonnyguru.com/forums/showthread.php?t=13311

To quote McSteel:
Pretty much the only way to get to the power peak and dropoff is using artificial load and overriden protections.
 
"AMD doesn't want to see its next architecture eat into sales of its current high-end products just yet..."

What sales? Furies have never sold more than a fraction of what the 980 Ti's alone have, and thinking that after 1070 anyone in their right mind would ever consider getting a Fury is just crazy.

Starting out with the mainstream parts might be a good idea, but if Polaris can't at least match Pascal in terms of architectural efficiency and features, ultimately it won't matter what they put out first, Nvidia will claim that category too in a matter of months. What we've heard so far isn't very encouraging at all unfortunately, and it's kinda hard to see AMD pull out a rabbit out of its hat after years of doing the exact opposite.
 
Some people said that amd made the right decision chasing after mainstream (midrange) segment first. That probably true if nvidia not launching their x60 part (and below) until the end of this year or next. bulk of the sales were made on this segment but this is also the segment where most people care about bang for buck. Means most of them will not going to jump to polaris the moment the cards are coming out. I heard that the NDA on Polaris will end on june 29th. So hard launch probably happen by then or in early july. if nvidia launch their 1060 in august of even in september most people in this price range are very likely to wait for nvidia offering first before making final decision. Plus nvidia have the brand image. 970 and 960 still outsell amd 390s and 380s.

Honestly i was thinking why can't amd just release polaris right now. They have been showing and hyping polaris since january. Raja Kodduri in one of his interview even say that they were confident to be ahead of competitor this time around.
 
1060 is bigger and more expensive than cheapest polaris chips. They are not in the same category... 1060 is supposed to be 300$ device, so it would be twice the cheapest polaris. And thus it should be much faster too...
But who knows. Bot will be out soon enough. Polaris seems to be about efficiency so mobile parts may be the main focus!
 
"AMD doesn't want to see its next architecture eat into sales of its current high-end products just yet..."

What sales? Furies have never sold more than a fraction of what the 980 Ti's alone have, and thinking that after 1070 anyone in their right mind would ever consider getting a Fury is just crazy.

Starting out with the mainstream parts might be a good idea, but if Polaris can't at least match Pascal in terms of architectural efficiency and features, ultimately it won't matter what they put out first, Nvidia will claim that category too in a matter of months. What we've heard so far isn't very encouraging at all unfortunately, and it's kinda hard to see AMD pull out a rabbit out of its hat after years of doing the exact opposite.

Pascal seems to be little more than Maxwell at 16nm. I'd say AMD has good chance of beating that soundly, especially with DX12/Vulkan.

Concentrating on mainstream is most likely at least as much a push for laptops and next gen consoles (which would be finalizing spec-levels, if not actual specs, right about now) as it is a push for mid-range desktops. Personally I hope AMD all the best, and will most likely continue as their customer. Both GPUs and CPUs benefit from healthy competition (which has been lacking in the CPU-side for some time now).
 


Yes, my reply was a simplified version of what you said. My point being saying you can run a R9 380 on a 350w PSU is generally BS. You would need a very specific combination of lower power processor and VERY high quality 350ish watt PSU for it to work reliably at full load. Most 350 watt PSU's aren't going to be able to handle that load (there are very few low wattage high quality PSU's out there) reliably at all. The PC will boot, but running a 3D Mark or a game is going to end in a reboot.
 
You would need a very specific combination of lower power processor and VERY high quality 350ish watt PSU for it to work reliably at full load. Most 350 watt PSU's aren't going to be able to handle that load (there are very few low wattage high quality PSU's out there) reliably at all. The PC will boot, but running a 3D Mark or a game is going to end in a reboot.

Most people overestimate how much power they're actually using. If you have a crap 350W no-name unit that's probably a rebadged 300W model, yeah you might have problems. It probably would burn up on it's own, given time. But let's say you have a halfway decent brand-name 350W PSU. Nothing fancy, just a reputable non-exploding model. No special CPU, just any run of the mill cheap i3 or Pentium dual core. You're not made of money, and you want to toss in an inexpensive but relatively powerful GPU. Maybe a friend sold you his old card half price. Is a vanilla, relatively stock-clocked, non-X 380 going to work? Sure.

http://www.hardocp.com/article/2016/01/18/asus_strix_r9_380_directcu_ii_oc_video_card_review/10#.V0sO6Y-cGUk

The WHOLE SYSTEM power under load was 294W AT THE WALL. Now before you go "ooh that's cutting it too close for me", look at the test setup on page 2 of that review - he's testing with an i7 3770K @ 4.6Ghz. Most of your affordable Haswell dual-cores are around 55W, and will draw substantially less power.

While *I* would have more headroom than that for a number of reasons, including overclocking and future upgrades, it's not far-fetched.
 


I'm not disagreeing with you, I am very aware with the power requirements, heck I monitor the power draw of my system from the wall (PS AMD says an R9 280 needs a 750w PSU, I run 2 in crossfire and it draws 650w from the wall max). I am very familiar with this, the reason I made my comment however is there are not very many 350w PSU's on the market these days that are of the quality that I would say fit within that envelope. Off the top of my head I can name 3, the Antec Earthwatts 380, and the Seasonic S12II 350 and SSR-350. When it came time to buy a PSU for my home server (which uses like 120w tops) I ended up with a Seasonic G-series 450w because I wanted a great quality gold PSU (and I do prefer Seasonic) but there are nearly no other choices. Even in the review of the PSU said most quality PSU makers don't bother with low wattage because its not worth the development costs.

I'm sure a 350w PSU can run it, however MOST 350's cannot, most enough that I call BS on it. And saying things like that here is dangerous, as people sometimes read this stuff and assume the 350w PSU that came with their Dell can run their new R9 380 with a couple of molex adaptors. It won't, we see it every day here.
 
That's a nice spin to put on their total inability to compete with Nvidia. AMD's has been totally blindsided this time. Maybe they should renumber Polaris 10 to Radeon 460 to save face, Polaris 11 can be Radeon 440.
 
I think it's too early to say that AMD won't be able to compete with Nvidia. At the high end, they're definitely not going to be competitive over the next 6 months (at best, maybe much longer), but they're gunning for the mid-range, high volume market, which they have an advantage in because of their console business. If they can convince the Dells of the world to include their cards instead of Nvidia's, nobody *here* is going to be all that impressed, but there's revenue to be made. It's tricky because the mid-range market seems to be shrinking on account of integrated graphics getting better, but since they're one of the players in the integrated market as well, I'm sure they took that into account in their market analysis. I'll be getting a GTX 1080, but I do hope AMD can get some punches in and stay competitive, because I'd hate to see what Nvidia would become if they were the only game in town.
 


I agree, the ultra enthusiast market like us isn't what keeps these companies alive, nor is it cards like the Fury X, 980ti or Titan X. Its the high margin bread and butter that benefits from the tech/R&D of the expensive stuff. At this point AMD needs that (and based on their stock doubling in value it DOES look promising for them so far).

But the hope from us was (and still is in the next 6 months) that AMD pushes the high end envelope as well to keep Nvidia on their toes. Nobody benefits if when it comes to high end gaming GPU's theres only 1 game in town. In fact even for Nvidia it will begin to devalue their product as it becomes commoditized and work against them. Competition is needed for survival.
 
The worry is that Nvidia being the undisputed king of the high-end means people will associate them with quality, and buyers of lower-end cards will buy Nvidia ones even if AMD outperforms them in that segment.
This... just this. Sadly this is one of the biggest reasons why nVidia has been selling so much more than AMD for the last few years.
 
The problem with amd is their r&d is non-specific. Nvidia has gpu's, amd has gpus, cpus, mobos, mobile stuff, consoles and a host of other stuff they have to stick their limited budget to. Trying to make all that happen and still try and remain competitive is not an easy task and is guaranteed to end up with problems in time management. That means delays and releases of tech before it's time. I've always liked amd, but in this case, a little fat trimming would go a long way. Imho.
 

And their semi-custom SoC/Console ASIC unit is the only part of AMD currently turning a profit but that unit would not exist without the GPU and CPU R&D. AMD does not reinvent the whole thing for each customer, they reuse almost everything from their existing stock, the only extra R&D on top is the glue-logic between those parts but most of it is also done based on existing libraries. AMD axing their only profitable unit to "streamline" their design process would be suicidal.

Nvidia does CPUs too: the Tegra line would not exist without Nvidia investing cash in CPUs, albeit ARM-based ones.
 
NVidia die size:
There seems to be a lot of confusion on this. I do NOT any direct information, however I just thought I would point out there are TWO methods for making GPU's:

1) disable the defective areas on the larger chips, and

2) design smaller chips for the same GPU model

Basically they will likely still make small chips for the low and mid-range products, then simply make use of the defective GPU's rather than throwing them in the garbage.

AMD does NOT necessarily have an advantage either if NVidia's chips can overclock higher because NVidia can make smaller chips to do the same thing as AMD's but get more chips per wafer.

It's hard to say how the higher frequency will work though because its going to come down to the thermal and power limitations of mobile designs.

The biggest concern I have for AMD is whether they support the VR optimizations that NVidia does which can boost FPS by up to 60%. To be stamped "VR READY" you need GTX970 performance regardless of this boost capability or not (because not all games will have this feature implemented).

A GTX970 costs as little as $260USD right now!!

That means a Polaris GPU for $150 to $200 that meets or exceeds the GTX970 performance might be possible!
 
So why people thinking there is advantage to amd in the mid range due to their console business? What the relation of console with pc market? And it is not easy for AMD to regain the market share on laptop. It all depends on how much R&D they willing to pour into product development. It is not as simple as selling their gpu to those laptop maker. Amd for their part also have to spend R&D integrating their gpu into those laptop design. This is the main reason why amd almost none existent in discrete gpu laptop market depite they were dominating the segment back in 2010. Back in 2012 Rory Read decided not deal with OEM business unless OEM want to sell AMD product in large volume.
 
I always was under the impression Polaris 10 was mid range not high end :/ with Polaris 11 covering the lowend/low power. vega being 490/x meaning high end!
 
Well actually when AMD first unveil polaris early on january all we know polaris will be the code name for amd gpu. And they also mention polaris will be 2016 gpu. Vega will be 2017 gpu. For many that tells us that polaris will cover from high end to low end. Vega was expected to be the ultra high end part that will be competing with nvidia big gun (1080ti). Hence many expect polaris will compete with nvidia offering from top to bottom this year with both dug it out their big chip next year (GP100 vs Vega). The expectation is for polaris to compete head to head with nvidia GP104 in term of performance.

Personally i want to know if amd will be able to rival nvidia power efficiency or not (just for educational purpose).
 
Well actually when AMD first unveil polaris early on january all we know polaris will be the code name for amd gpu. And they also mention polaris will be 2016 gpu. Vega will be 2017 gpu. For many that tells us that polaris will cover from high end to low end. Vega was expected to be the ultra high end part that will be competing with nvidia big gun (1080ti). Hence many expect polaris will compete with nvidia offering from top to bottom this year with both dug it out their big chip next year (GP100 vs Vega). The expectation is for polaris to compete head to head with nvidia GP104 in term of performance.

Personally i want to know if amd will be able to rival nvidia power efficiency or not (just for educational purpose).

There's now been rumors of Vega being moved to this year thanks to things going better than expected with HBM2. We'll see. What is certain, is that we will see a 480 in the next couple of days.
 
Status
Not open for further replies.